(Methods in Molecular Biology) George Karlin-Neumann, Francisco Bizouarn - Digital PCR - Methods and Protocols-Humana Press (2018)

Methods in
Molecular Biology 1768
George Karlin-Neumann
Francisco Bizouarn Editors
Digital PCR
Methods and Protocols
METHODS IN MOLECULAR BIOLOGY
Series Editor
John M. Walker
School of Life and Medical Sciences
University of Hertfordshire
Hatfield, Hertfordshire, AL10 9AB, UK
For further volumes:

http://www.springer.com/series/7651
Digital PCR
Methods and Protocols
Edited by
George Karlin-Neumann and Francisco Bizouarn

Digital Biology Center
Bio-Rad Laboratories, Pleasanton, CA, USA
Editors
George Karlin-Neumann Francisco Bizouarn
Digital Biology Center Digital Biology Center
Bio-Rad Laboratories Bio-Rad Laboratories
Pleasanton, CA, USA Pleasanton, CA, USA
ISSN 1064-3745 ISSN 1940-6029 (electronic)

Methods in Molecular Biology
ISBN 978-1-4939-7776-5 ISBN 978-1-4939-7778-9 (eBook)
https://doi.org/10.1007/978-1-4939-7778-9
Library of Congress Control Number: 2018937639
© Springer Science+Business Media, LLC, part of Springer Nature 2018

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is
concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction
on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation,
computer software, or by similar or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply,
even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations
and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to
be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty,
express or implied, with respect to the material contained herein or for any errors or omissions that may have been made.
The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Printed on acid-free paper
This Humana Press imprint is published by the registered company Springer Science+Business Media, LLC part of
Springer Nature.
The registered company address is: 233 Spring Street, New York, NY 10013, U.S.A.
Preface
The ability to detect and manipulate a single, specific DNA or RNA molecule via the
polymerase chain reaction (PCR) has revolutionized our understanding of biology and the
possibilities for influencing it. This remarkable technique has been successively honed over
the past two decades to improve upon its ability to detect and quantify targets of interest.
Digital PCR is both one of the oldest and newest implementations of PCR and with the
recent technical innovation of scalable, droplet-based digital PCR, brings to the practice the
greater ease of use, precision, specificity and reproducibility needed to further advance many
areas of investigation and clinical practice. This volume will provide guidance and illustrate
how it is helping to make deeper inroads in:
Infectious disease;
Evolution of cancer and response to treatment;
Genome structural variation and associated phenotypes;
Prevalence of somatic mosaicism;
Genome editing and cell therapy;
Non-invasive blood monitoring of fetus, organ or tumor status;
Environmental monitoring; and
Food testing for pathogens and GMO’s.
For these and other diverse applications, it details optimal experimental design for
achieving the user’s needs for precision and sensitivity, assay design considerations for
various target types and sample types, insights into data analysis and interpretation, and it
reveals other related benefits of sample partitioning such as target size determination and
linkage measurements for haplotyping. This volume should prove useful for a wide range of
specialists including geneticists, molecular biologists, virologists, immunologists, oncolo-
gists, pathologists—those involved in basic and translational research, as well as clinicians
and diagnosticians—and those interested in applied and environmental sciences.
Pleasanton, CA, USA George Karlin-Neumann

Francisco Bizouarn
v
Contents
Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v
Contributors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi
PART I BACKGROUND CONCEPTS

1 Entering the Pantheon of 21st Century Molecular Biology
Tools: A Perspective on Digital PCR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2 Basic Concepts and Validation of Digital PCR Measurements . . . . . . . . . . . . . . . . 11
Leonardo Pinheiro and Kerry R. Emslie
3 Fundamentals of Counting Statistics in Digital PCR:
I Just Measured Two Target Copies–What Does It Mean? . . . . . . . . . . . . . . . . . . . 25
Svilen Tzonev
4 Control Materials and Digital PCR Methods for Evaluation
of Circulating Cell-Free DNA Extractions from Plasma . . . . . . . . . . . . . . . . . . . . . . 45
Alexandra S. Whale, Ana Fernandez-Gonzalez, Alice Gutteridge,
and Alison S. Devonshire
PART II ABSOLUTE QUANTIFICATION
5 Multiplex Droplet Digital PCR Protocols for Quantification

of GM Maize Events . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
David Dobnik, Bjørn Spilsberg, Alexandra Bogožalec Košir,
Dejan Štebih, Dany Morisset, Arne Holst-Jensen, and Jana Žel
6 Using Droplet Digital PCR to Detect Coinfection of Human
Herpesviruses 6A and 6B (HHV-6A and HHV-6B) in
Clinical Samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
Ashley Vellucci, Emily C. Leibovitch, and Steven Jacobson
7 Biomarkers in Cerebrospinal Fluid: Analysis of Cell-Free
Circulating Mitochondrial DNA by Digital PCR. . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
Petar Podlesniy and Ramon Trullas
8 Testing of General and Human-Associated Fecal Contamination
in Waters. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
Yiping Cao, Meredith R. Raith, and John F. Griffith
PART III COPY NUMBER VARIATION

9 Analyzing Copy Number Variation with Droplet Digital PCR . . . . . . . . . . . . . . . . 143
Avery Davis Bell, Christina L. Usher, and Steven A. McCarroll
10 Assessing HER2 Amplification in Plasma cfDNA . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
Isaac Garcia-Murillas and Nicholas C. Turner
vii
viii Contents
11 Detection and Quantification of Mosaic Genomic DNA

Variation in Primary Somatic Tissues Using ddPCR: Analysis
of Mosaic Transposable-Element Insertions, Copy-Number
Variants, and Single-Nucleotide Variants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
Bo Zhou, Michael S. Haney, Xiaowei Zhu, Reenal Pattni, Alexej Abyzov,
and Alexander E. Urban
PART IV RARE MUTATION AND RARE ALLELE DETECTION

12 Monitoring of Response and Resistance in Plasma of
EGFR-Mutant Lung Cancer Using Droplet Digital PCR . . . . . . . . . . . . . . . . . . . . 193
Yanan Kuang, Allison O’Connell, Adrian G. Sacher,
Nora Feeney, Ryan S. Alden, Geoffrey R. Oxnard,
and Cloud P. Paweletz
13 Detection of Cancer DNA in Early Stage and Metastatic
Breast Cancer Patients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
Arielle J. Medford, Riaz N. Gillani, and Ben Ho Park
14 Droplet Digital PCR for Minimal Residual Disease Detection
in Mature Lymphoproliferative Disorders . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
Daniela Drandi, Simone Ferrero, and Marco Ladetto
15 Quantitation of JAK2 V617F Allele Burden by Using the
QuantStudio™ 3D Digital PCR System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
Elena Kinz and Axel Muendlein
16 Novel Multiplexing Strategies for Quantification of Rare
Alleles Using ddPCR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275
Miguel Alcaide and Ryan D. Morin
17 Identification and Use of Personalized Genomic Markers
for Monitoring Circulating Tumor DNA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303
Yilun Chen, Anthony M. George, Eleonor Olsson, and Lao H. Saal
18 Single Color Multiplexed ddPCR Copy Number Measurements
and Single Nucleotide Variant Genotyping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323
Christina M. Wood-Bouwens and Hanlee P. Ji
19 A Universal Droplet Digital PCR Approach for Monitoring
of Graft Health After Transplantation Using a Preselected SNP Set . . . . . . . . . . . 335
Julia Beck, Michael Oellerich, and Ekkehard Schütz
20 Detection and Quantification of HDR and NHEJ Induced by Genome
Editing at Endogenous Gene Loci Using Droplet Digital PCR . . . . . . . . . . . . . . . 349
Yuichiro Miyaoka, Steven J. Mayerl, Amanda H. Chan,
and Bruce R. Conklin
21 DNA Methylation Analysis Using Droplet Digital PCR. . . . . . . . . . . . . . . . . . . . . . 363
Ming Yu, Tai J. Heinzerling, and William M. Grady
PART V GENE EXPRESSION AND RNA QUANTIFICATION
22 Simultaneous Quantification of Multiple Alternatively

Spliced mRNA Transcripts Using Droplet Digital PCR . . . . . . . . . . . . . . . . . . . . . . 387
Bing Sun and Yun-Ling Zheng
Contents ix
23 Using Droplet Digital PCR to Analyze Allele-Specific

RNA Expression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 401
Nolan Kamitaki, Christina L. Usher, and Steven A. McCarroll
24 Very Low Abundance Single-Cell Transcript Quantification
with 5-Plex ddPCRTM Assays. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 423
George Karlin-Neumann, Bin Zhang, and Claudia Litterst
25 Quantification of Circulating MicroRNAs by Droplet Digital PCR. . . . . . . . . . . . 445
Manuela Ferracin and Massimo Negrini
26 Droplet Digital PCR for Absolute Quantification
of Extracellular MicroRNAs in Plasma and Serum: Quantification
of the Cancer Biomarker hsa-miR-141 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 459
Maria D. Giraldez, John R. Chevillet, and Muneesh Tewari
PART VI OTHER USES OF PARTITIONING
27 Droplet Digital™ PCR Next-Generation Sequencing Library

QC Assay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 477
Nicholas J. Heredia
28 Phasing DNA Markers Using Digital PCR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 489
John Regan and George Karlin-Neumann
29 ddTRAP: A Method for Sensitive and Precise Quantification
of Telomerase Activity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 513
Andrew T. Ludlow, Dawne Shelton, Woodring E. Wright,
and Jerry W. Shay
30 Highly Efficient and Reliable DNA Aptamer Selection Using
the Partitioning Capabilities of ddPCR: The Hi-Fi SELEX Method . . . . . . . . . . . 531
Aaron Ang, Eric Ouellet, Karen C. Cheung, and Charles Haynes
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 555
Contributors
ALEXEJ ABYZOV Department of Health Sciences Research, Center for Individualized

Medicine, Mayo Clinic, Rochester, MN, USA
MIGUEL ALCAIDE Department of Molecular Biology and Biochemistry, Simon Fraser
University, Burnaby, BC, Canada
RYAN S. ALDEN Lowe Center for Thoracic Oncology, Dana-Farber Cancer Institute, Boston,
MA, USA; Department of Medicine, Brigham and Women’s Hospital and Harvard
Medical School, Boston, MA, USA
AARON ANG Michael Smith Laboratories, Department of Chemical and Biological
Engineering, University of British Columbia, Vancouver, BC, Canada; Centre for Blood
Research, University of British Columbia, Vancouver, BC, Canada
JULIA BECK Chronix Biomedical GmbH, Göttingen, Germany
AVERY DAVIS BELL Department of Genetics, Harvard Medical School, Boston, MA, USA;
Program in Medical and Population Genetics, Broad Institute of MIT and Harvard,
Cambridge, MA, USA; Stanley Center for Psychiatric Research, Broad Institute of MIT
and Harvard, Cambridge, MA, USA
FRANCISCO BIZOUARN Digital Biology Center, Bio-Rad Laboratories, Pleasanton, CA, USA
YIPING CAO Source Molecular Corporation, Miami, FL, USA; Southern California Costal
Water Research Project Authority, Costa Mesa, CA, USA
AMANDA H. CHAN Gladstone Institute of Cardiovascular Disease, San Francisco, CA, USA
YILUN CHEN Translational Oncogenomics Unit, Division of Oncology and Pathology,
Department of Clinical Sciences, Lund University, Lund, Sweden
KAREN C. CHEUNG Centre for Blood Research, University of British Columbia, Vancouver,
BC, Canada; Department of Electrical and Computer Engineering, University of British
Columbia, Vancouver, BC, Canada
JOHN R. CHEVILLET Institute for Systems Biology, Seattle, WA, USA
BRUCE R. CONKLIN Gladstone Institute of Cardiovascular Disease, San Francisco, CA,
USA; Department of Medicine, University of California San Francisco, San Francisco,
CA, USA; Department of Cellular and Molecular Pharmacology, University of California
San Francisco, San Francisco, CA, USA
ALISON S. DEVONSHIRE Molecular and Cell Biology Team, LGC, Teddington, UK
DAVID DOBNIK Department of Biotechnology and Systems Biology, National Institute
of Biology, Ljubljana, Slovenia
DANIELA DRANDI Department of Molecular Biotechnologies and Health Sciences,
Hematology Division, University of Torino, Torino, Italy
KERRY R. EMSLIE National Measurement Institute, Lindfield, NSW, Australia
NORA FEENEY Belfer Center for Applied Cancer Science, Dana-Farber Cancer Institute,
Boston, MA, USA
ANA FERNANDEZ-GONZALEZ Molecular and Cell Biology Team, LGC, Teddington, UK
MANUELA FERRACIN Department of Experimental, Diagnostic and Specialty Medicine—
DIMES, University of Bologna, Bologna, Italy
SIMONE FERRERO Department of Molecular Biotechnologies and Health Sciences,
Hematology Division, University of Torino, Torino, Italy
xi
xii Contributors
ISAAC GARCIA-MURILLAS Breast Cancer Now Research Centre, The Institute of Cancer
Research, London, UK
ANTHONY M. GEORGE Translational Oncogenomics Unit, Division of Oncology and
Pathology, Department of Clinical Sciences, Lund University, Lund, Sweden
RIAZ N. GILLANI Johns Hopkins School of Medicine, Baltimore, MD, USA; Boston Children’s
Hospital, Boston, MA, USA
MARIA D. GIRALDEZ Department of Internal Medicine, University of Michigan, Ann
Arbor, MI, USA
WILLIAM M. GRADY Clinical Research Division, Fred Hutchinson Cancer Research Center,
Seattle, WA, USA; Department of Medicine, University of Washington School of Medicine,
Seattle, WA, USA
JOHN F. GRIFFITH Southern California Costal Water Research Project Authority, Costa
Mesa, CA, USA
ALICE GUTTERIDGE University College London Cancer Institute, London, UK
MICHAEL S. HANEY Department of Psychiatry and Behavioral Sciences, Stanford Center for
Genomics and Personalized Medicine, Stanford University School of Medicine, Palo Alto,
CA, USA; Program on Genetics of Brain Function, Department of Genetics, Stanford
Center for Genomics and Personalized Medicine, Palo Alto, CA, USA
CHARLES HAYNES Michael Smith Laboratories, Department of Chemical and Biological
Engineering, University of British Columbia, Vancouver, BC, Canada; Department of
Electrical and Computer Engineering, University of British Columbia, Vancouver, BC,
Canada
TAI J. HEINZERLING Clinical Research Division, Fred Hutchinson Cancer Research Center,
Seattle, WA, USA
NICHOLAS J. HEREDIA Digital Biology Center, Bio-Rad Laboratories, Pleasanton, CA, USA
ARNE HOLST-JENSEN Section of Virology, Norwegian Veterinary Institute, Oslo, Norway
STEVEN JACOBSON Neuroimmunology Branch, Viral Immunology Section, National
Institute of Neurological Disorders and Stroke (NINDS), National Institutes of Health,
Bethesda, MD, USA
HANLEE P. JI Division of Oncology, Department of Medicine, Stanford University School of
Medicine, Stanford, CA, USA; Stanford Genome Technology Center, Stanford University,
Palo Alto, CA, USA
NOLAN KAMITAKI Program in Medical and Population Genetics, Broad Institute of MIT
and Harvard, Cambridge, MA, USA; Stanley Center for Psychiatric Research, Broad
Institute of MIT and Harvard, Cambridge, MA, USA; Department of Genetics, Harvard
Medical School, Boston, MA, USA
GEORGE KARLIN-NEUMANN Digital Biology Center, Bio-Rad Laboratories, Pleasanton,
CA, USA
ELENA KINZ Vorarlberg Institute for Vascular Investigation and Treatment (VIVIT),
Feldkirch, Austria
ALEXANDRA BOGOŽALEC KOŠIR Department of Biotechnology and Systems Biology, National
Institute of Biology, Ljubljana, Slovenia; Jožef Stefan International Postgraduate School,
Ljubljana, Slovenia
YANAN KUANG Belfer Center for Applied Cancer Science, Dana-Farber Cancer Institute,
Boston, MA, USA
MARCO LADETTO Division of Hematology, A.O. SS Antonio e Biagio e Cesare Arrigo,
Alessandria, Italy
Contributors xiii
EMILY C. LEIBOVITCH Neuroimmunology Branch, Viral Immunology Section, National

Bethesda, MD, USA; Institute for Biomedical Sciences, School of Medicine and Health
Sciences, Washington, DC, USA
CLAUDIA LITTERST Digital Biology Center, Bio-Rad Laboratories, Pleasanton, CA, USA
ANDREW T. LUDLOW Department of Cell Biology, UT Southwestern Medical Center, Dallas,
TX, USA
STEVEN J. MAYERL Gladstone Institute of Cardiovascular Disease, San Francisco, CA, USA
STEVEN A. MCCARROLL Department of Genetics, Harvard Medical School, Boston, MA,
USA; Program in Medical and Population Genetics, Broad Institute of MIT and Harvard,
ARIELLE J. MEDFORD Johns Hopkins School of Medicine, Baltimore, MD, USA;
Massachusetts General Hospital, Boston, MA, USA
YUICHIRO MIYAOKA Gladstone Institute of Cardiovascular Disease, San Francisco, CA,
USA; Regenerative Medicine Project, Tokyo Metropolitan Institute of Medical Science,
Tokyo, Japan
RYAN D. MORIN Department of Molecular Biology and Biochemistry, Simon Fraser
University, Burnaby, BC, Canada
DANY MORISSET Department of Biotechnology and Systems Biology, National Institute
of Biology, Ljubljana, Slovenia
AXEL MUENDLEIN Vorarlberg Institute for Vascular Investigation and Treatment (VIVIT),
Feldkirch, Austria
MASSIMO NEGRINI Laboratory for Technologies of Advanced Therapies (LTTA),
Department of Morphology, Surgery and Experimental Medicine, University of Ferrara,
Ferrara, Italy
ALLISON O’CONNELL Belfer Center for Applied Cancer Science, Dana-Farber Cancer
Institute, Boston, MA, USA
MICHAEL OELLERICH Institute for Clinical Pharmacology, University Medical Center
Göttingen, Göttingen, Germany
ELEONOR OLSSON Translational Oncogenomics Unit, Division of Oncology and Pathology,
ERIC OUELLET Michael Smith Laboratories, Department of Chemical and Biological
Engineering, University of British Columbia, Vancouver, BC, Canada; Centre for Blood
Research, University of British Columbia, Vancouver, BC, Canada
GEOFFREY R. OXNARD Lowe Center for Thoracic Oncology, Dana-Farber Cancer Institute,
Boston, MA, USA; Department of Medicine, Brigham and Women’s Hospital and
Harvard Medical School, Boston, MA, USA
BEN HO PARK Department of Oncology, Johns Hopkins University, Baltimore, MD, USA;
Sidney Kimmel Comprehensive Cancer Institute, Johns Hopkins University, Baltimore,
MD, USA
REENAL PATTNI Department of Psychiatry and Behavioral Sciences, Stanford Center for
CLOUD P. PAWELETZ Belfer Center for Applied Cancer Science, Dana-Farber Cancer
Institute, Boston, MA, USA
LEONARDO PINHEIRO National Measurement Institute, Lindfield, NSW, Australia
xiv Contributors
PETAR PODLESNIY Neurobiology Unit, Institut d’Investigacions Biomèdiques de Barcelona,

Consejo Superior de Investigaciones Cientı́ficas (CSIC), Barcelona, Spain; Centro de
Investigacion Biomédica en Red sobre Enfermedades Neurodegenerativas (CIBERNED),
Barcelona, Spain
MEREDITH R. RAITH Southern California Costal Water Research Project Authority,
Costa Mesa, CA, USA
JOHN REGAN Digital Biology Center, Bio-Rad Laboratories, Pleasanton, CA, USA
LAO H. SAAL Translational Oncogenomics Unit, Division of Oncology and Pathology,
ADRIAN G. SACHER Lowe Center for Thoracic Oncology, Dana-Farber Cancer Institute,
Boston, MA, USA; Department of Medicine, Brigham and Women’s Hospital and
Harvard Medical School, Boston, MA, USA
EKKEHARD SCHÜTZ Chronix Biomedical GmbH, Göttingen, Germany
JERRY W. SHAY Department of Cell Biology, UT Southwestern Medical Center, Dallas,
TX, USA
DAWNE SHELTON Digital Biology Center, Bio-Rad Laboratories, Pleasanton, CA, USA
BJØRN SPILSBERG Section of Virology, Norwegian Veterinary Institute, Oslo, Norway
DEJAN ŠTEBIH Department of Biotechnology and Systems Biology, National Institute of
Biology, Ljubljana, Slovenia
BING SUN Georgetown University, Washington, DC, USA
MUNEESH TEWARI Comprehensive Cancer Center, University of Michigan, Ann Arbor, MI,
USA; Department of Internal Medicine, University of Michigan, Ann Arbor, MI, USA;
Department of Biomedical Engineering, University of Michigan, Ann Arbor, MI, USA;
Biointerfaces Institute, University of Michigan, Ann Arbor, MI, USA; Center
for Computational Medicine and Bioinformatics, University of Michigan, Ann Arbor,
MI, USA
RAMON TRULLAS Neurobiology Unit, Institut d’Investigacions Biomèdiques de Barcelona,
Consejo Superior de Investigaciones Cientı́ficas (CSIC), Barcelona, Spain; Centro de
Investigacion Biomédica en Red sobre Enfermedades Neurodegenerativas (CIBERNED),
Barcelona, Spain; Institut d’Investigacions Biomèdiques August Pi i Sunyer (IDIBAPS),
Barcelona, Spain
NICHOLAS C. TURNER Breast Cancer Now Research Centre, The Institute of Cancer
Research, London, UK; Breast Unit, The Royal Marsden Hospital, London, UK
SVILEN TZONEV Digital Biology Center, Bio-Rad Laboratories, Pleasanton, CA, USA
ALEXANDER E. URBAN Department of Psychiatry and Behavioral Sciences, Stanford Center
for Genomics and Personalized Medicine, Stanford University School of Medicine, Palo
Alto, CA, USA; Program on Genetics of Brain Function, Department of Genetics, Stanford
CHRISTINA L. USHER Department of Genetics, Harvard Medical School, Boston, MA, USA;
Program in Medical and Population Genetics, Broad Institute of MIT and Harvard,
ASHLEY VELLUCCI Neuroimmunology Branch, Viral Immunology Section, National
Bethesda, MD, USA
ALEXANDRA S. WHALE Molecular and Cell Biology Team, LGC, Teddington, UK
Contributors xv
CHRISTINA M. WOOD-BOUWENS Division of Oncology, Department of Medicine, Stanford

University School of Medicine, Stanford, CA, USA; Stanford Genome Technology Center,
Stanford University, Palo Alto, CA, USA
WOODRING E. WRIGHT Department of Cell Biology, UT Southwestern Medical Center,
Dallas, TX, USA
MING YU Clinical Research Division, Fred Hutchinson Cancer Research Center, Seattle,
WA, USA
JANA ŽEL Department of Biotechnology and Systems Biology, National Institute of Biology,
Ljubljana, Slovenia
BIN ZHANG Digital Biology Center, Bio-Rad Laboratories, Pleasanton, CA, USA
YUN-LING ZHENG Georgetown University, Washington, DC, USA; Cancer Prevention and
Control Program, Lombardi Comprehensive Cancer Center, Georgetown University
Medical Center, Washington, DC, USA
BO ZHOU Department of Psychiatry and Behavioral Sciences, Stanford Center for Genomics
and Personalized Medicine, Stanford University School of Medicine, Palo Alto, CA, USA;
Program on Genetics of Brain Function, Department of Genetics, Stanford Center for
Genomics and Personalized Medicine, Palo Alto, CA, USA
XIAOWEI ZHU Department of Psychiatry and Behavioral Sciences, Stanford Center for
Part I
Background Concepts
Chapter 1
Entering the Pantheon of 21st Century Molecular Biology

Tools: A Perspective on Digital PCR
Abstract
After several decades of relatively modest use, in the last several years digital PCR (dPCR) has grown to
become the new gold standard for nucleic acid quantification. This coincides with the commercial availabil-
ity of scalable, affordable, and reproducible droplet-based dPCR platforms in the past five years and has led
to its rapid dissemination into diverse research fields and testing applications. Among these, it has been
adopted most vigorously into clinical oncology where it is beginning to be used for plasma genotyping in
cancer patients undergoing treatment. Additionally, innovation across the scientific community has
extended the benefits of reaction partitioning beyond DNA and RNA quantification alone, and demon-
strated its usefulness in evaluating DNA size and integrity, the physical linkage of colocalized markers, levels
of enzyme activity and specific cation concentrations in a sample, and more. As dPCR technology gains in
popularity and breadth, its power and simplicity can often be taken for granted; thus, the reader is reminded
that due diligence must be exercised in order to make claims not only of precision but also of accuracy in
their measurements.
Key words Digital PCR, dPCR, Droplet digital PCR, ddPCR, qPCR, Real-time PCR, Reproducibil-
ity, Precision, Accuracy, Rare mutation detection, Copy number variation, DNA, RNA, Absolute
quantification, Control materials, Partitioning, Linkage
1 Introduction
After a 25 year gestation period, digital PCR (dPCR) has finally

emerged as an indispensable tool in the research world and is
headed toward a pivotal role in the clinic [1–7]. It joins a host of
other technologies developed over the past 4 decades—such as
Sanger sequencing, DNA and protein gel electrophoresis, real-
time PCR, DNA microarrays, and next-generation sequencing
(NGS)—which continue to help drive forward much of the prog-
ress of current molecular biology research and testing. Each new
technology brings to the table particular strengths and limitations.
What had not been adequately served before the explosive
adoption of digital PCR in the last several years was the need for
more precise, specific and reproducible quantification of DNA and
George Karlin-Neumann and Francisco Bizouarn (eds.), Digital PCR: Methods and Protocols, Methods in Molecular Biology,
vol. 1768, https://doi.org/10.1007/978-1-4939-7778-9_1, © Springer Science+Business Media, LLC, part of Springer Nature 2018
3
4 George Karlin-Neumann and Francisco Bizouarn
RNA from diverse biological sample types. Although its earlier

maturing cousin, real-time quantitative PCR (qPCR), has also
enabled the sensitive detection and quantification of nucleic acids,
it is more suited to relative than to absolute quantification, does not
readily detect rare single-nucleotide mutations, and is more sensi-
tive to factors affecting amplification efficiency such as assay design
and PCR inhibitors. These factors affecting its robustness and
interlab reproducibility have combined to slow its advance into
clinical use [8–10]. Conversely, the endpoint reactions occurring
in digital PCR partitions (be they prefabricated microchambers or
de novo formed droplets) enable the robust amplification and
absolute digital quantification of target sequences without the
need for a standard curve—critical to measurements in environ-
mental, clinical, and other sample types [11–14]; permit the highly
specific detection of rare mutations in the midst of a large back-
ground of wildtype sequence—crucial to the evaluation of hetero-
geneous tumor samples and liquid biopsy specimens [1, 15–17],
viruses [18–20], and somatic mosaicism [21]; and provide the high
precision needed to distinguish copy number alleles dispersed
throughout the human genome—key to the understanding of
their role in genome structure and disease [22, 23].
For these and other reasons, digital PCR has become the new
gold standard for nucleic acid quantification to which other meth-
ods compare themselves [24–27]. The methods in this volume have
been selected to illustrate its application to a range of experimental
questions organized by commonality of purpose. After an intro-
ductory part on “Background Concepts” in digital PCR (more on
that below), examples are given in Part II of its adoption for
“Absolute Quantification” of GMO contamination in foodstocks
(Dobnik et al., Chapter 5); of viral variants of herpesvirus in differ-
ent clinical sample types (Vellucci et al., Chapter 6); of mitochon-
drial DNA levels in the CSF as a biomarker for dementia patients
(Podlesniy and Trullas, Chapter 7); and in water-quality testing for
fecal contamination (Cao et al., Chapter 8).
Limitations in the achievable precision of gene amplification
measurements have hampered investigations of their role in various
diseases and in cancer [28–30]. Part III offers protocols for study-
ing “Copy Number Variation.” Davis et al. (Chapter 9) explain how
assays can be designed and evaluated for the study of copy number
alleles and their potential association with phenotypes of interest.
Combining the challenges of liquid biopsy in breast cancer recur-
rence with evaluation of copy number variation, Garcia-Murillas
and Turner (Chapter 10) describe how to assess whether amplifica-
tion of Her2 is likely causal to disease progression. Zhou and
colleagues (Chapter 11) illustrate the power of dPCR for studying
somatic mosaicism of amplified genomic sequences, SNPs, and
transposable elements in primary human tissue and iPSCs, a
Digital PCR: A Tool for the 21st Century 5
phenomenon that was little recognized before the availability of

high precision dPCR technology.
Part IV (“Rare Mutation/Allele Detection”) reflects the inten-
sity and inventiveness with which digital PCR has been applied to
the study of rare mutations or foreign alleles accessible in the blood,
in both plasma cell-free DNA (cfDNA) and in cells. Kuang et al.
(Chapter 12) illustrate the use of dPCR for detection and monitor-
ing of both sensitizing and resistance mutations in liquid biopsy of
non-small-cell lung cancer, while Medford and colleagues
(Chapter 13) describe a method for preamplification of cfDNA
from early-stage breast cancer patients to assess whether residual
disease is still present. To assess minimal residual disease in multiple
lymphoid malignancies, Drandi and her team (Chapter 14) present
methods for both the identification of patient-specific rearrange-
ments and their conversion to droplet digital™ PCR (ddPCR™)
assays for serial blood monitoring. The workflow for use of the
QuantStudio™ 3D dPCR system is described by Kinz and Muen-
dlein (Chapter 15) for measuring JAK2 V617F allele burden, prev-
alent in myeloproliferative diseases. Alcaide and Morin
(Chapter 16) tackle the common challenge in liquid biopsy of
having limited cfDNA from a patient sample and wanting maximal
sensitivity for multiple mutations. They describe a suite of multiplex
assay design strategies for achieving this where they minimize the
number of probes used and hence cost required. In some design
strategies they screen for different mutations but do not distinguish
among them, in others, they do discriminate between the specific
mutation(s) that may be present. Chen and coworkers (Chapter 17)
detail an approach to achieve early detection of occult metastases in
breast cancer by monitoring patient-specific rearrangements with
ddPCR fusion assays in serial bloods. They derive these trunk
mutations from NGS sequencing of tumor tissue and convert
them to digital PCR assays for highly sensitive monitoring of
potential recurrence.
In keeping with the desire for flexibility, speed, and low cost for
detection and monitoring of CNVs and SNPs, Wood and Ji
(Chapter 18) demonstrate how to multiplex with the probe-less
EvaGreen dye-binding chemistry in droplet digital PCR. This
method requires only the use of specially designed allele-specific
primers and entirely avoids the cost and time needed for probe
synthesis. In a non-cancer application, Beck and colleagues
(Chapter 19) illustrate the use of the liquid biopsy paradigm for
the monitoring of transplant health. Treating the transplanted
organ as a “genome transplant,” they identify which of a prese-
lected set of high minor allele-frequency SNPs can be used to
quantify turnover of the donor tissue within the recipient, and
thus its state of health. In yet another variation of digital PCR for
the detection and quantification of rare alleles, Miyaoka and his
coworkers (Chapter 20) describe the use of ddPCR for monitoring
of gene editing experiments where they are able to simultaneously

quantify the frequency of both homologous (HDR) and nonho-
mologous (NHEJ) repair events. Part IV concludes with an exam-
ple of epigenetic analysis where Yu et al. describe MethylLight
ddPCR (Chapter 21), the adaptation of bisulfite methylation anal-
ysis to dPCR to enable the study of “field cancerization” in solid
tumors.
Gene expression and RNA quantification studies (Part V) have
also benefited from the increased precision and absolute quantifica-
tion of dPCR. In Chapter 22, Sun and Zheng address the need for
better understanding of the role of alternative splicing in the regu-
lation of telomerase activity and function. They illustrate a general
method for simultaneously measuring the levels of the four major
splice variants of the human telomerase gene in a single droplet
digital PCR reaction. Kamitaki and colleagues also leverage the
high precision of dPCR in Chapter 23 to enable measurement of
allele-specific expression differences between two alleles in an indi-
vidual differing by only a single SNP. These subtle SNP effects on
expression may explain the role of these variants on phenotypes
with which they have been associated in genome-wide association
studies. In Chapter 24, Karlin-Neumann et al. illustrate the appli-
cation of ddPCR to the study of single cell expression, especially
useful where quantifying very low abundance transcripts is impor-
tant. To increase the power of the method, they present a strategy
for probe-multiplexing (or “radial” multiplexing) which enables
them to quantify five different targets within a single PCR reaction,
or by splitting the cDNA from each cell into halves, as many as nine
or ten genes per cell rapidly and with absolute quantification. The
two remaining chapters in this part provide guidance in quantifying
miRNAs from serum or plasma using either EvaGreen probeless
chemistry (Ferracin and Negrini, Chapter 25) or probe-based Taq-
man chemistry (Giraldez et al., Chapter 26).
Although absolute quantification of DNA and RNA is the
predominant use of digital PCR systems, Part VI (“Other Uses of
Partitioning”) reveals some less anticipated uses for which parti-
tioning can be deployed. Besides the quantification of nucleic acids,
partitioning can also provide information on DNA fragment size as
illustrated by Heredia et al. (Chapter 27) for characterization of
NGS library preparations. Regan and colleague (Chapter 28) dem-
onstrate how the physical linkage of two genetic markers can be
tested by whether or not assay signals for each are colocalized in
partitions more often than would be expected by chance alone.
They also describe a simple method for preparation of large DNA
satisfactory for testing linkage over as much as 150 or even 200Kb.
The detection and quantification of telomerase activity by Ludlow
et al. (Chapter 29) reveals a more general paradigm for translating
non-DNA measurements (e.g., enzyme activities that can act on
nucleic acid templates) into a DNA readout which can be quantified
via digital PCR. Finally, the act of partitioning a reaction into many
smaller reactions can greatly improve the signal-to-noise ratio over

bulk reactions and has been innovatively leveraged by Ouillet and
colleagues (Chapter 30) to develop a highly efficient droplet-based
SELEX method (Hi-Fi SELEX) which accomplishes in 3 or
4 rounds of selection over several weeks what normally requires
~8 rounds and several months, and achieves the identification of
high affinity aptamers.
Although not included in this volume due to their more recent
publication, additional clever applications have been developed by
the world’s dPCR user community. Among these is the simulta-
neous quantification of a protein and its corresponding mRNA level
in single mammalian cells jointly using a digital proximity ligation
assay and an RT-ddPCR reaction on a crude cell lysate [31]. Perhaps
even more unexpected, Cheng et al. [32] have been able to quantify
picomolar concentrations of Ag(I) and Hg(II) through their dose-
dependent effects in permitting extension of mismatched primers
annealed to their nearly complementary DNA templates! Again,
the presence and amount of an initial non-DNA analyte is trans-
lated into a quantifiable DNA signal. So it is highly probable that
even more unexpected applications will develop from the funda-
mental ability to partition samples.
Returning to the opening Part I on “Background Concepts,” it
is important to remember that despite the seductive appeal of
rendering absolute measurements of the numbers of target copies
in a sample, it is still necessary to be critical of our experimental
methods before drawing conclusions from their results, whether for
research or testing purposes. The quantitative nature of digital PCR
should not make us complacent about the non-uniform structure
of genomic targets, i.e., there are regions and contexts of genomic
sequences that can mask their accessibility to PCR reactions result-
ing in greatly under-quantified measurements, presumably due to
secondary structure. This may be mitigated by alternative assay
designs, reducing the size of the genomic target fragments or
modified reaction conditions. That said, the onus is on the experi-
menter to validate the accuracy of their results and not to confuse
precision and reproducibility with truth.
Similarly, the beauty and power of digital PCR should not allow
us to forget that there are other hurdles that lie between our
experimental results and the questions we seek to answer. As an
example, Whale and colleagues (Chapter 4) describe in Part I the
use of sample extraction controls that are important when trying to
calculate back from a dPCR measurement to the number of tumor
DNA sequences present in a milliliter of blood. Similarly, Pinheiro
and Emslie (Chapter 2) explain the concept of digital PCR and
what factors influence the final measurement obtained so that we
can perform these experiments with appropriate vigilance and with
suitable run controls. Lastly, in applications where detection of rare
species is common, it is crucial to understand the limit of detection
of our assay and whether or not we are confident we have detected a

species of interest. Tzonev (Chapter 3) provides us a comprehen-
sive overview of test performance descriptors such as sensitivity and
specificity, sources of variability and uncertainty in molecular
counting and how a measured false positive rate influences the
limit of detection achievable.
The active ferment of dPCR investigations in thousands of labs
throughout the world is also evidenced by the increasing number of
dPCR system implementations that are being devised in academic
labs (e.g., SlipChip, ref. 33) and also making their way to commer-
cial availability. At the time we began assembling this collection of
methods, we aimed to include representative peer-reviewed proce-
dures using all of the commercially available dPCR systems. At that
time, this comprised Fluidigm’s Biomark, Bio-Rad’s QX100/200,
RainDance’s RainDrop, and Thermo-Fisher’s QuantStudio 3D
systems, although the majority of published studies were heavily
weighted toward the QX100/200 platform and thus most of the
contributions in this volume were based on it. Though we were
unable to obtain contributions from users of either the RainDance
or Fluidigm systems, in part due to the smaller user base of the
former and the diminished use of the latter, we trust that the overall
strategies described in these chapters will be generally applicable to
various dPCR platforms present and future. In that vein, since
commencement of this project, three new systems have recently
become commercially available from Formulatrix (the Constella-
tion™) and JN Medsys (the Clarity™), both with fixed chambers,
and from Stilla (the Naica™, with droplets). Suffice it to say, dPCR
research and testing is growing at an ever-expanding rate with an
increasing variety of implementations, and promises to become an
enduring and incisive tool in the arsenal of the molecular biologist.
References
1. Hindson BJ, Ness KD, Masquelier DA, 3. Miliaras N (2014) Digital PCR comes of age.
Belgrader P, Heredia NJ, Makarewicz AJ, Genetic Engineering & Biotechnology News
Bright IJ, Lucero MY, Hiddessen AL, Legler 34(4):14–16. https://doi.org/10.1089/gen.
TC, Kitano TK, Hodel MR, Petersen JF, Wyatt 34.04.08
PW, Steenblock ER, Shah PH, Bousse LJ, 4. Perkel J (2014) The digital PCR revolution.
Troup CB, Mellen JC, Wittmann DK, Erndt Science 344(6180):212–214. https://doi.
NG, Cauley TH, Koehler RT, So AP, Dube S, org/10.1126/science.344.6180.212
Rose KA, Montesclaros L, Wang S, Stumbo 5. Pinheiro & Emslie, Chapter 2, this MIMB
DP, Hodges SP, Romine S, Milanovich FP, volume
White HE, Regan JF, Karlin-Neumann GA,
Hindson CM, Saxonov S, Colston BW (2011) 6. Sacher AG, Paweletz C, Dahlberg SE, Alden
High-Throughput Droplet Digital PCR Sys- RS, O’Connell A, Feeney N, Mach SL, J€anne
tem for Absolute Quantitation of DNA Copy PA, Oxnard GR (2016) Prospective validation
Number. Anal Chem 83(22):8604–8610. of rapid plasma genotyping for the detection of
https://doi.org/10.1021/ac202028g EGFR and KRAS mutations in advanced lung
cancer. JAMA Oncol 2(8):1014–1022.
2. Baker M (2012) Digital PCR hits its stride. Nat https://doi.org/10.1001/jamaoncol.2016.
Methods 9(6):541–544. https://doi.org/10. 0173
1038/nmeth.2027
7. Mellert H, Foreman T, Jackson L, Maar D, (6):991–994. https://doi.org/10.1373/

Thurston S, Koch K, Weaver A, Cooper S, clinchem.2012.197855
Dupuis N, Sathyanarayana UG, Greer J, 14. Dingle TC, Sedlak RH, Cook L, Jerome KR
Hahn W, Shelton D, Stonemetz P, Pestano (2013) Tolerance of droplet-digital PCR vs
GA (2017) Development and clinical utility of real-time quantitative PCR to inhibitory sub-
a blood-based test service for the rapid identi- stances. Clin Chem 59(11):1670–1672.
fication of actionable mutations in non-small https://doi.org/10.1373/clinchem.2013.
cell lung carcinoma. J Mol Diagn 19 211045
(3):404–416. https://doi.org/10.1016/j. 15. Oxnard GR, Paweletz CP, Kuang Y, Mach SL,
jmoldx.2016.11.004 O’Connell A, Messineo MM, Luke JJ,
8. Whale AS, Huggett JF, Cowen S, Speirs V, Butaney M, Kirschmeier P, Jackman DM,
Shaw J, Ellison S, Foy CA, Scott DJ (2012) J€anne PA (2014) Noninvasive detection of
Comparison of microfluidic digital PCR and response and resistance in EGFR-mutant lung
conventional quantitative PCR for measuring cancer using quantitative next-generation gen-
copy number variation. Nucleic Acids Res 40 otyping of cell-free plasma DNA. Clin Cancer
(11):e82. https://doi.org/10.1093/nar/ Res 20(6):1698–1705. https://doi.org/10.
gks203 1158/1078-0432.CCR-13-2482
9. Hindson CM, Chevillet JR, Briggs HA, Galli- 16. Garcia-Murillas I, Schiavon G, Weigelt B,
chotte EN, Ruf IK, Hindson BJ, Vessella RL, Ng C, Hrebien S, Cutts RJ, Cheang M,
Tewari M (2013) Absolute quantification by Osin P, Nerurkar A, Kozarewa GJA,
droplet digital PCR versus analog real-time Dowsett M, Reis-Filho JS, Smith IE, Turner
PCR. Nat Methods 10(10):1003–1005. NC (2015) Mutation tracking in circulating
https://doi.org/10.1038/nmeth.2633 tumor DNA predicts relapse in early breast
10. Whale AS, Devonshire AS, Karlin-Neumann G, cancer. Sci Transl Med 7(302):302ra133.
Regan J, Javier L, Cowen S, Fernandez- https://doi.org/10.1126/scitranslmed.
Gonzalez A, Jones G, Redshaw N, Beck J, aab0021
Berger A, Combaret V, Kjersgaard ND, 17. Janku F, Huang HJ, Fujii T, Shelton DN,
Davis L, Fina F, Forshew F, Andersen RF, Madwani K, Fu S, Tsimberidou AM, Piha-
Galbiati S, Hernańdez AG, Haynes CA, Paul SA, Wheler JJ, Zinner RG, Naing A,
Janku F, Lacave R, Lee J, Mistry V, Pender A, Hong DS, Karp DD, Cabrilo G, Kopetz ES,
Pradines A, Proudhon C, Saal L, Stieglitz E, Subbiah V, Luthra R, Kee BK, Eng C, Morris
Ulrich B, Foy CA, Parkes H, Tzonev S, Hug- VK, Karlin-Neumann GA, Meric-Bernstam F
gett JF (2017) International interlaboratory (2017) Multiplex KRASG12/G13 mutation
digital pcr study demonstrating high reproduc- testing of unamplified cell-free DNA from the
ibility for the measurement of a rare sequence plasma of patients with advanced cancers using
variant. Anal Chem 89(3):1724–1733. droplet digital polymerase chain reaction. Ann
https://doi.org/10.1021/acs.analchem. Oncol 28(3):642–650. https://doi.org/10.
6b03980 1093/annonc/mdw670
11. Cao Y, Raith MR, Griffith JF (2015) Droplet 18. Strain MC, Lada SM, Luong T, Rought SE,
digital PCR for simultaneous quantification of Gianella S, Terry VH, Spina CA, Woelk CH,
general and human-associated fecal indicators Richman DD (2013) Highly precise measure-
for water quality assessment. Water Res ment of HIV DNA by droplet digital PCR.
70:337–349. https://doi.org/10.1016/j. PLoS One 8(4):e55943. https://doi.org/10.
watres.2014.12.008 1371/journal.pone.0055943
12. Nadauld L, Regan JF, Miotke L, Pai RK, Long- 19. Persaud D, Gay H, Ziemniak C, Chen YH,
acre TA, Kwok SS, Saxonov S, Ford JM, Ji HP Piatak M Jr, Chun TW, Strain M, Richman D,
(2012) Quantitative and sensitive detection of Luzuriaga K (2013) Absence of detectable
cancer genome amplifications from formalin HIV-1 viremia after treatment cessation in an
fixed paraffin embedded tumors with droplet infant. N Engl J Med 369(19):1828–1835.
digital PCR. Transl Med (Sunnyvale) 2 https://doi.org/10.1056/NEJMoa1302976
(2):1–12. https://doi.org/10.4172/2161- 20. Leibovitch EC, Brunetto GS, Caruso B,
1025.1000107 Fenton K, Ohayon J, Reich DS, Jacobson S
13. Belgrader P, Tanner SC, Regan JF, Koehler R, (2014) Coinfection of human herpesviruses
Hindson BJ, Brown AS (2013) Droplet digital 6A (HHV-6A) and HHV-6B as demonstrated
PCR measurement of HER2 copy number by novel digital droplet PCR assay. PLoS One 9
alteration in formalin-fixed paraffin-embedded (3):e92328. https://doi.org/10.1371/jour
breast carcinoma tissue. Clin Chem 59 nal.pone.0092328
21. Abyzov A, Mariani J, Palejev D, Zhang Y, 27. Low H, Chan SJ, Soo GH, Ling B, Tan EL
Haney MS, Tomasini L, Ferrandino AF, (2017) Clarity™ digital PCR system: a novel
Rosenberg Belmaker LA, Szekely A, platform for absolute quantification of nucleic
Wilson M, Kocabas A, Calixto NE, Grigorenko acids. Anal Bioanal Chem 409(7):1869–1875.
EL, Huttner A, Chawarska K, Weissman S, https://doi.org/10.1007/s00216-016-0131-
Urban AE, Gerstein M, Vaccarino FM (2012) 7
Somatic copy number mosaicism in human 28. Aldhous MC, Abu Bakar S, Prescott NJ,
skin revealed by induced pluripotent stem Palla R, Soo K, Mansfield JC, Mathew CG,
cells. Nature 492(7429):438–442. https:// Satsangi J, Armour JA (2010) Measurement
doi.org/10.1038/nature11629 methods and accuracy in copy number varia-
22. Handsaker RE, Van Doren V, Berman JR, tion: failure to replicate associations of beta-
Genovese G, Kashin S, Boettger LM, McCar- defensin copy number with Crohn’s disease.
roll SA (2015) Large multiallelic copy number Hum Mol Genet 19(24):4930–4938. https://
variations in humans. Nat Genet 47 doi.org/10.1093/hmg/ddq411
(3):296–303. https://doi.org/10.1038/ng. 29. Liu S, Yao L, Ding D, Zhu H (2010) CCL3L1
3200 copy number variation and susceptibility to
23. Davis JM, Searles VB, Anderson N, Keeney J, HIV-1 infection: a meta-analysis. PLoS One 5
Dumas L, Sikela JM (2014) DUF1220 dosage (12):e15778. https://doi.org/10.1371/jour
is linearly associated with increasing severity of nal.pone.0015778
the three primary symptoms of autism. PLoS 30. Usher CL, Handsaker RE, Esko T, Tuke MA,
Genet 10(3):e1004241. https://doi.org/10. Weedon MN, Hastie AR, Cao H, Moon JE,
1371/journal.pgen.1004241 Kashin S, Fuchsberger C, Metspalu A, Pato
24. Guttery DS, Page K, Hills A, Woodley L, CN, Pato MT, McCarthy M1, Boehnke M,
Marchese SD, Rghebi B, Hastings RK, Luo J, Altshuler DM, Frayling TM, Hirschhorn JN,
Pringle JH, Stebbing J, Coombes RC, Ali S, McCarroll SA (2015) Structural forms of the
Shaw JA (2015) Noninvasive detection of acti- human amylase locus and their relationships to
vating estrogen receptor 1 (ESR1) mutations SNPs, haplotypes and obesity. Nat Genet 47
in estrogen receptor-positive metastatic breast (8):921–925. https://doi.org/10.1038/ng.
cancer. Clin Chem 61(7):974–982. https:// 3340
doi.org/10.1373/clinchem.2015.238717 31. Albayrak C, Jordi CA, Zechner C, Lin J, Bich-
25. Thress KS, Brant R, Carr TH, Dearden S, sel CA, Khammash M, Tay S (2016) Digital
Jenkins S, Brown H, Hammett T, quantification of proteins and mRNA in single
Cantarini M, Barrett JC (2015) EGFR muta- mammalian cells. Mol Cell 61(6):914–924.
tion detection in ctDNA from NSCLC patient https://doi.org/10.1016/j.molcel.2016.02.
plasma: A cross-platform comparison of lead- 030
ing technologies to support the clinical devel- 32. Cheng N, Zhu P, Xu Y, Huang K, Luo Y,
opment of AZD9291. Lung Cancer 90 Yang Z, Xu W (2016) High-sensitivity assay
(3):509–515. https://doi.org/10.1016/j. for Hg (II) and Ag (I) ion detection: A new
lungcan.2015.10.004 class of droplet digital PCR logic gates for an
26. Gu J, Zang W, Liu B, Li L, Huang L, Li S, intelligent DNA calculator. Biosens Bioelec-
Rao G, Yu Y, Zhou Y (2017) Evaluation of tron 84:1–6. https://doi.org/10.1016/j.bios.
digital PCR for detecting low-level EGFR 2016.04.084
mutations in advanced lung adenocarcinoma 33. Shen F (2017) SlipChip device for digital
patients: a cross-platform comparison study. nucleic acid amplification. Methods Mol Biol
Oncotarget. https://doi.org/10.18632/ 1547:123–132. https://doi.org/10.1007/
oncotarget.18866. [Epub ahead of print] 978-1-4939-6734-6_10
Chapter 2
Basic Concepts and Validation of Digital PCR Measurements

Leonardo Pinheiro and Kerry R. Emslie
Abstract
Use of digital polymerase chain reaction (dPCR) technology is rapidly growing and diversifying into a range
of areas in life science. The release of dPCR commercial systems has facilitated access, leading to recognition
of the potential advantages compared to previous quantitative PCR technologies, and the scope for novel
applications. The capability of dPCR to deliver unprecedented levels of precision, accuracy, and resolution
in quantification of nucleic acids has triggered a strong interest by academia and the life sciences industry in
use of this technology as a molecular diagnostic tool. However, the performance of dPCR, as for a
“classical” PCR assay, essentially still relies on enzyme-based amplification of nucleic acid using specific
reagents and instrumentation. This chapter describes basic concepts, key properties, and important factors
to consider for the verification and validation of dPCR measurements.
Key words Digital PCR, Nucleic acids quantification, Validation
1 Introduction
Use of dPCR technology is undoubtedly on the rise considering the

rapid increase in number of peer-reviewed publications and public-
ity over the past few years. A Pubmed search (http://www.ncbi.
nlm.nih.gov/pubmed) of the words “digital PCR” identified about
20 publications in 2012, while for the period between January
2013 and October 2015 there were more than 300 publications.
The main drivers behind increased use of dPCR have been devel-
opment of commercial systems offering reduced costs and more
flexible workflow, and recognition by the broader scientific com-
munity of the potential advantages of the technology and the wide
scope of its application.
Recent developments in application of dPCR technology have
been far reaching and diverse, including detection of various patho-
gens [1–4], monitoring food safety [5, 6] and water quality [7, 8],
and studying microbial ecology [9]. However, due to the high level
of precision afforded by dPCR and consequent ability to detect
small differences in amount of nucleic acid molecules, it has mostly
11
12 Leonardo Pinheiro and Kerry R. Emslie
been used in clinical research including development and detection

of biomarkers [10–12], detection of genome amplification status
and genetic variants in cancer patients [13–17], genetic screening
[18], genome editing [19, 20], detection of single nucleotide
changes and copy number alterations in stem cells [21, 22], and
monitoring graft-derived DNA and chimeras in transplant recipi-
ents [23, 24]. As a result, leading research groups and the life
sciences industry are developing dPCR technology based tests
with the aim of utilizing these tests for molecular diagnosis. For
translation of a dPCR-based test from research laboratory to clinical
laboratory setting, the test must satisfy requirements of regulatory
authorities which include validation and/or verification
studies [25].
For over 20 years, real-time quantitative PCR (qPCR) [26, 27]
has been the main technology used for quantitative measurement of
specific targets within nucleic acid samples. However, despite
thousands of publications reporting on clinical applications for
qPCR and a number of PCR tests cleared by regulatory authorities,
we are yet to see widespread use of routine quantitative nucleic
acids measurement in clinical molecular diagnostics. One reason for
this limited uptake of qPCR in a diagnostic setting may be the
difficulty encountered in reproducing qPCR results reliably.
qPCR is an analog measurement based on monitoring exponential
increase in fluorescence signal after each PCR cycle. The point at
which the fluorescence signal crosses an intensity threshold is called
the quantification or threshold cycle (Cq). Target nucleic acid
concentration in a sample is determined using a calibration curve
which fits the logarithm of nucleic acid concentration for a dilution
series of calibrant to the Cq value for each calibrant dilution. To
ensure qPCR data is reproducible over time and between labora-
tories, the calibrant should demonstrate appropriate stability and
homogeneity and be traceable back to an international or higher-
order DNA reference material. However, these DNA reference
materials are only available for a limited number of target
sequences. In addition, accurate qPCR measurements rely on
good amplification efficiency for both calibrant and sample, with
minimal effects from sample matrix and inhibitory substances. In
contrast, dPCR does not require a calibration curve and its linear
digital signal read out has the potential to provide more precise data
than that obtained from the exponential amplification signal of
qPCR. These key factors give dPCR a technical advantage over
qPCR.
The dPCR workflow involves random distribution of PCR mix
containing target nucleic acid, primers, probe, and mastermix
across a large number of uniformly sized partitions, such that
some partitions contain no nucleic acid template and others contain
one or more template copies. Following PCR amplification parti-
tions containing target DNA can be distinguished from partitions
Concepts and Validation of Digital PCR 13
containing no target DNA through their fluorescence signal. The

target nucleic acid copy number per partition is derived based on
the number of positive partitions and the total number of partitions
using a Poisson model.
As for qPCR, dPCR essentially relies on enzyme-based amplifi-
cation of nucleic acid using specific reagents and instrumentation.
In the absence of assay optimization and validation, either reduced
assay performance or complete failure of an assay is possible. For
dPCR to realize its full potential as a quantitative molecular diag-
nostic tool, carefully designed validation studies are necessary and
sufficient understanding of sources of variation and bias that can
lead to incorrect results or poor reproducibility of data is needed.
This chapter aims to describe basic concepts, key properties of
dPCR and to highlight important factors to consider for verifica-
tion and validation of dPCR measurements.
2 Development of dPCR
2.1 Limiting Limiting dilution consists of diluting a sample to such a degree that,
Dilution—Key to the when divided among several individual assays, the analyte of inter-
Development of dPCR est will be present in some but not all of the assays. The principle
behind limiting dilution originated from work developed during
the first-world war for the enumeration of bacterial growth for
infection management purpose [28]. Limiting dilution followed
by PCR was first used to quantify the number of human immuno-
deficiency proviral molecules [29]. Soon after, a similar methodol-
ogy was described for quantification of leukaemic cells [30]. The
first study describing the general applicability of limiting dilution
for quantification of DNA target molecules was published in 1992
[31]. These initial studies used a small number of standard PCR
replicates (less than 30) to generate results using the limiting
dilution principle. At that time, development and commercializa-
tion of qPCR technology [26, 27] offered a practical and easy
option of PCR based nucleic acid quantification and discouraged
further development of limiting dilution to quantify nucleic acids.
In 1999, limiting dilution was applied across a much larger number
of replicates using a commercially available 384-well plate format to
quantify disease-associated mutations in colorectal cancer patients.
The authors then coined the term “digital PCR” [32]. However,
the method was still too cumbersome to compete with qPCR and
was not widely adopted at the time. By 2002, advances in micro-
fluidics engineering allowed development of microfluidic chips
amenable to rapid, simple, and scalable fabrication [33]. This
turned dPCR into a relatively simple and practical method of
accurate nucleic acid quantification and the first publications
using commercial microfluidic devices describing the potential of
dPCR technology started to appear [34–37].
2.2 Commercial In 2006, Fluidigm Corporation launched a commercial dPCR

dPCR Systems platform (Biomark™) based on the company’s trademark
integrated fluidics technology using array chips comprising
12 panels of 765 partitions. In 2009, it released a second chip
comprising 48 panels of 770 partitions which increased throughput
and decreased costs of using the technology. Also in 2009, Life
Technologies launched the OpenArray® system which utilizes small
plates perforated with over 3000 micron-size through-holes that
retain PCR mix through surface tension. In 2013, they introduced
high density nanofluidic chips carrying up to 20,000 data points
with simplified workflow and reduced cost. In 2014, Formulatrix
launched the Constellation system which utilizes a proprietary
96-well microfluidic system with each well containing 496 cham-
bers where individual PCR assays take place.
The launch of droplet-based dPCR commercial systems led to
more widespread uptake of dPCR technology. The workflow of
droplet dPCR involves generation of a water-in-oil emulsion in
which aqueous droplets comprise the PCR mix. The fine-tuned
titration of PCR and oil chemistry can produce thousands or
millions of nanoliter or picoliter size uniform droplets which can
withstand fast flow through microfluidic channels and thermal
cycling without bursting or coalescing. By the end of 2011
Bio-Rad Laboratories released the QX100 droplet dPCR system
which utilizes probe based chemistry for detection. In the following
year the QX200 droplet dPCR system was launched which utilizes
both probe based and intercalating dye chemistries for detection.
Both systems generate approximately 20,000 nanoliter-sized dro-
plets from each PCR mix at a lower cost compared to early array
chip-based systems. Also in 2012, RainDance Technologies
launched the RainDrop® system, a dPCR platform which generates
millions of picoliter-sized droplets from each PCR mix.
The various dPCR platforms can differ in workflow, partition
number, partition volume and sample throughput. In addition, the
proportion of the PCR mix which is analyzed by dPCR varies
between platforms. In some systems, a portion of the PCR mix is
retained in the microfluidics and not partitioned. Partitions may
also be excluded from data analysis if they fail to meet predefined
specifications. These properties may have particular advantages and
disadvantages depending on the application. Before making a deci-
sion on purchase of a dPCR instrument, consideration should be
given to the required sensitivity, precision, dynamic range and
throughput for the intended application, as well as per sample
operating costs.
3 Basic Concepts of dPCR
3.1 The Workflow The workflow of dPCR is generally fairly simple. In chip-based
systems, PCR mix is evenly distributed across prefabricated parti-
tions normally with the aid of loading instruments. The loaded chip
is placed onto a thermal cycler for PCR amplification. After thermal
cycling, partitions containing target molecules (positive partitions)
are distinguished from partitions with no target molecules (nega-
tive partitions) by imaging fluorescence intensity for each partition
in the chip using a camera. Based on a normalized fluorescence
intensity signal, a threshold is applied to assign partitions as either
positive or negative. Some chip-based systems also collect real-time
data, thus providing a Cq value for each partition. In droplet-based
systems, partitions are generated dynamically by the user: PCR mix
and oil are typically pipetted into wells of a cartridge and then the
cartridge is transferred to specifically designed instruments in which
a water-in-oil emulsion is generated. The droplet emulsion is trans-
ferred to either a well in a 96 well plate or a tube in a strip of tubes
and is then placed into a thermal cycler for PCR amplification. After
thermal cycling, the plate or tubes are transferred to a droplet
reading instrument where droplets are aspirated and fast-flow
streamed past a fluorescence detector, in a manner similar to a
flow cytometer, for measurement of fluorescence in each droplet.
Based on the fluorescence signal, a threshold is applied to assign
droplets to either the positive or negative droplets population.
3.2 Estimating Copy Poisson statistics, which take into account the probability of any
Number given partition containing zero, one or more than one molecule,
are used to estimate the average number of molecules per partition
[36]. If the average partition volume, VP, is known, Poisson statis-
tics can also be used to estimate the concentration of target nucleic
acid molecules in the PCR mix using Eq. 1.

NP 1
½Target DNA ¼ ln 1 copy number per unit volume
NT VP
ð1Þ
where [Target DNA] is the number of target DNA molecules per
unit volume of PCR mix, NP is the number of positive partitions,
and NT is the total number of partitions. Each component in the
equation needs to be measured accurately to avoid compromising
the accuracy of the nucleic acid concentration estimate.
3.3 Dynamic Range The theoretical dynamic range of dPCR is limited by the number of
and Precision partitions analyzed (Fig. 1). Because the dPCR partitioning process
follows a binomial distribution, the dynamic range extends beyond
the number of partitions analyzed [36] although precision is com-
promised at both low and high ends of the dynamic range
A. B.
20 10
16 8
12 6
CV (%)
CV (%)
8 4
4 2
0 0
10 100 1,000 10,000 100,000 0 20 40 60 80 100
Percent positive partitions
Number molecules per PCR
Fig. 1 Variability arising from subsampling and partitioning of sample in relation to (a) the number of target
molecules and (b) the percent of positive partitions for digital PCR assays containing 1000 (······), 10,000 (----),
and 100,000 (—) partitions. CV, coefficient of variation
[37, 38]. The uniformity of partition size can also impact on the
dynamic range and precision of dPCR. Monodisperse droplets are
required for the highest precision. Methods that use polydisperse
partition sizes will have lower precision since larger droplets will
have a higher likelihood of containing multiple targets than smaller
droplets in the same sample. This, in turn, will impact on the
Poisson statistics calculation since this calculation is based on the
assumption that the partitions are of equivalent volume. Assuming
uniform partition size a typical 20,000 partition dPCR assay pro-
vides a theoretical dynamic range over orders of magnitude and up
to 105 molecules [38]. One commercially available dPCR instru-
ment is capable of producing millions of partitions, thus extending
the dynamic range to over six orders of magnitude [39]. Assuming
sufficient biological sample is available, a large dynamic range can
be useful for detecting rare target molecules in a background of
nontarget molecules since nontarget molecules will be distributed
across a large number of partitions, thus reducing the possibility
that they will interfere with detection of rare target molecules [40].
The theoretical precision of dPCR can be estimated using
Poisson statistics and improves with increasing partition number,
provided that analysis is undertaken within the optimal window of
the dynamic range (Fig. 1). The level of precision and consequent
resolving power offered by dPCR is typically better than can be
achieved using qPCR [41] or other technologies such as fluores-
cence in situ hybridization [42], and is useful for applications such
as copy number variation measurements. However, theoretical pre-
cision is not always reflected in experimental precision observed
from analyzing replicates since the Poisson model is only one
component in a range of factors that will combine to reflect
observed experimental precision. In general, if the number of
partitions is less than 1000, the Poisson model copy number esti-
mate will have a coefficient of variation of at least 4% and can be one
of the main factors contributing to the observed imprecision,
whereas if the number of partitions is more than 10,000 the coeffi-
cient of variation can be less than 1.5%, so other factors may
become the dominant sources of imprecision (Fig. 1).
3.4 Partitioning and dPCR has a lower susceptibility to inhibition than qPCR for inhibi-
Inhibitory Effects tory substances commonly found in clinical [43, 44], food [5], and
environmental [7] samples. In conventional qPCR, inhibitors can
reduce amplification efficiency and increase the Cq value [45] with
a resultant impact on accurate quantification of the target molecule.
In contrast, moderate levels of inhibitor in dPCR can often be
tolerated without compromising dPCR accuracy, since accuracy
relies on the ability to distinguish negative and positive partitions
after end-point PCR, and this is often possible even when efficiency
is slightly reduced. Also the possibility of inhibitory effects due to
cross-reactivity or “cross talk” observed multiplex qPCR assays is
much reduced by the partioning process in dPCR. However, if
present at high enough concentrations, inhibitory substances can
still reduce amplification efficiency to the point where some parti-
tions are misclassified, resulting in an inaccurate estimate of the
original template concentration. As the susceptibility of different
PCR assays to inhibition is variable and the mechanism of interfer-
ence can differ between inhibitory substances [7, 46], it is impor-
tant to empirically assess potential inhibitory effects by carefully
designed control experiments prior to establishing dPCR methods.
3.5 Partitioning and Random distribution of target nucleic acids molecules across parti-
Molecule Segregation tions is necessary for accurate quantification since this is a funda-
mental assumption of the Poisson statistics model. The number of
target molecules in the sample will be underestimated following
nonrandom partitioning as, for instance, when target sequences are
linked or in tandem since linked copies will colocate in the same
partition. Restriction digest treatment of samples prior to partition-
ing can be used to separate linked target sequences [47]. For that,
prior knowledge of restriction enzymes that do not cut the target
sequence is necessary and an evaluation of the effect of the restric-
tion enzyme and buffer on subsequent dPCR assay performance
should be undertaken.
Conversely, potential overestimation of copy number can occur
if a portion of the DNA molecules are single-stranded. The pres-
ence of single strands of nucleic acid during partitioning can result
in up to a twofold increase in concentration estimate as the single
strand conformation can produce up to twice as many indepen-
dently segregated amplifiable templates [48].
3.6 The Partition To calculate copy number concentration using dPCR (see Eq. 1),
Volume Effect the average copy number per partition is divided by partition
volume. Three properties of partition volume can affect accuracy
of dPCR measurements: accuracy of the manufacturer-specified
partition volume, variability in average partition volume between
replicate dPCR measurements (interreaction variability) and varia-
bility in partition volume within a single dPCR replicate (intrareac-
tion variability). The partition volume in both currently available
chip and droplet-based dPCR technologies can vary slightly from
manufacturer’s stated volume [37, 38] and, if uncorrected, will bias
copy number concentration estimates [49]. Both interreaction and
intrareaction variability in partition volume will affect precision
and, to a lesser extent, accuracy of dPCR concentration estimates
[50]. Variability in partition volume will contribute to observed
technical precision and is one factor which will influence how
closely aligned technical precision is with the theoretical precision
estimated using Poisson statistics.
The effect of partition volume variation is minimized for copy
number ratio measurements between two assays conducted in a
duplex dPCR format. However, partition volume variability is
important for clinical applications requiring absolute quantification
of nucleic acids [11, 51].
4 Validation of dPCR Assays
4.1 Validation, The term verification is defined as “provision of objective evidence

Verification, and that a given item fulfills specified requirements” while validation is
Measurement defined as “verification, where the specified requirements are ade-
Uncertainty Definitions quate for an intended use” [52]. In the context of dPCR measure-
ments, verification confirms that specified performance properties
of the measurement system (e.g., a dPCR assay targeting a specific
gene sequence) are achieved, whereas validation establishes that the
dPCR assay is suitable for the intended application.
Approval from regulatory authority generally requires a test to
be validated for a specific purpose using specified equipment,
reagents, and controls. If any specifications are altered or modified,
it may be necessary to revalidate the modified measuring system to
ensure it is still adequate for its intended use or “fit-for-purpose.”
Validated performance characteristics may need to be verified at
regular intervals to demonstrate that specified performance proper-
ties continue to be met, for example with different batches of
reagents.
During analytical validation, it is important to identify and
quantify sources of analytical variation and potential bias. The
combined impact of these sources is best captured as an estimate
of measurement uncertainty. Measurement uncertainty is defined as
“non-negative parameter characterizing the dispersion of the
quantity values being attributed to a measurand, based on the

information used” [51]. More simply it is defined as the “an esti-
mate of the range of values within which the true value lies”. The
process of estimating measurement uncertainty for quantitative
DNA measurements produced by qPCR has been reviewed [53]
and many of the principles described in this review also apply to
dPCR measurements.
The following section briefly discusses important factors to
consider when assessing the performance of dPCR assays.
4.2 Considerations To systematically identify factors that may affect dPCR assays per-
for Validation of dPCR formance, each step in the method should be assessed (Fig. 2).
Assays A dPCR method normally involves preparing the sample (i.e.,
extracting nucleic acid), setting up PCR mix, partitioning the
PCR mix, amplifying target by thermal cycling, counting the num-
ber of partitions, classifying partitions into positive and negative
populations based on end point fluorescence signal and estimating
target sequence concentration in the reaction using Poisson statis-
tics and the known partition volume (see Eq. 1). Sample prepara-
tion and reaction set up are similar to those required for other
quantitative PCR methods. Many factors which can introduce vari-
ation in these two steps are likely to affect both dPCR and qPCR.
Control materials such as a positive extraction control or a spike-in
control can be used to assess extraction efficiency for absolute
quantification. However, care should be taken in selection of the
positive control as inhibitors can have different effects depending
on the assays. For dPCR of genomic DNA, physical methods such
as ultrasonication [24] and shredding [54] to fragment DNA have
been successfully used to facilitate partitioning and efficient ampli-
fication of target sequences, as well as the use of aptly chosen
restriction endonuclease digestion. From this point onward in the
method, the partitioning process and use of binary end point
fluorescence signal detection for quantification distinguishes
dPCR from other PCR methods and introduces different potential
sources of bias and variation to the end result.
The preparation of dPCR mix involves pipetting and mixing
the sample containing nucleic acid into a master mix typically
comprising buffer, enzyme, primers, and probe. Pipetting impreci-
sion introduced at this stage are inevitable, but can be reduced by
Target Reaction Endpoint

Partitions N T, N P , V p
DNA/RNA Mix Fluorescence
sampling pipetting partition number of poor or non-specific proportion of (+)

misclassification
variation variation volume partitions amplification partitions
Fig. 2 Diagram representing different steps involved in the digital PCR workflow. Potential factors capable of
introducing variability to the end measurement result are highlighted by arrows between the different steps in
the workflow. Adapted from Jacobs et al. [49]
using calibrated pipettes and larger volumes, and for very high
accuracy measurements, gravimetric dilutions using a calibrated
balance [38]. Sample heterogeneity and, in the case of very low
template DNA concentrations, stochastic effects will also impact on
precision of replicates.
Primers and probe sequences that are highly selective to the
target nucleic acid sequence are critical for performance and repro-
ducibility of dPCR measurements. Since it is usually impractical to
recover dPCR amplicons for confirmatory analysis, specificity of a
set of oligonucleotides primers and probe should initially be
checked by qPCR using control materials, if available, and/or
nucleotide sequencing of amplicons produced from qPCR assays.
Accuracy is a primary performance characteristic for validation
of any quantitative analytical method. Accurate quantification of
nucleic acid molecules by dPCR relies on correct classification of
partitions into negative and positive populations since the propor-
tion of positive partitions is an integral component of the equation
used to determine copy number concentration (Eq. 1). A study
using mathematical simulation to assess variance components in
dPCR concluded that incorrect classification of partitions is one
of the largest contributors to inaccuracy of nucleic acid quantifica-
tion results [50]. Therefore, a key objective in optimization of a
dPCR assay is to maximize fluorescence amplitude difference
between negative and positive partitions, and to minimize the
number of partitions with intermediate fluorescence intensity. In
practice, this can be achieved by optimizing efficiency of the dPCR
assay.
While dPCR is more tolerant of suboptimal amplification effi-
ciency than qPCR, the risk of misclassifying partitions, specifically
false negative classification, is reduced with a sensitive, efficient
dPCR assay. Ideally, reference materials at known copy number
concentration should be used for assay optimization, to optimize
the number of cycles and to determine the appropriate fluorescence
threshold. In the absence of suitable reference materials, two or
more assays spanning across the target sequence region can be used
to cross-check that optimal conditions have been reached and the
assay is not being compromised by DNA secondary structure
effects Thermal cycling conditions should be optimized using an
annealing temperature gradient to determine conditions for maxi-
mum separation between fluorescence levels of negative and posi-
tive partitions. This can be followed by a denaturation temperature
gradient, particularly in the case of high GC content templates.
Optimization of primer and probe concentrations can be under-
taken using similar approaches used for qPCR. Optimization of
primer concentration is also required for dPCR assays using inter-
calating dye chemistry. If optimal qPCR conditions have previously
been established, these should be verified using dPCR, since
conditions optimized using qPCR are not always optimal for

dPCR. In some cases, redesign of assays may be required.
After optimization, partitions containing at least one target
molecule are expected to generate a positive fluorescence signal,
while partitions without target molecule are expected to generate a
negative signal. However, in practice, both false positive and false
negative signal can be produced from dPCR assays. The threshold
setting will determine false positive and false negative rates. A small
proportion of false positives will have a large impact on accuracy of
assays with low number of target molecules, while a small propor-
tion of false negatives will have a large impact on accuracy of assays
containing high numbers of target molecules.
Limit of detection (LoD) and limit of blank (LoB) are also two
important performance parameters to be established for validation
of dPCR measurements. In brief, LoD and LoB are, respectively,
defined as “the lowest amount of analyte that can detected” and
“the highest measurement result that is likely to be observed for a
blank sample” with respective probability stated [55]. The general
approach to determine LoB involves determination of the distribu-
tion of values from multiple measurements of blank samples and for
LoD involves a series of measurements of samples containing very
low levels of analyte. The 95th percentile of distribution of the
blank signal typically determines the LoB. The LoD will be greater
than the LoB [55]. Applied to dPCR, LoD could be defined as the
smallest number of target nucleic acid molecules that is statistically
distinguishable from the background or negative control. One
approach to estimating LoD and LoB for a dPCR assay has been
experimentally demonstrated in a study to determine lower limits of
detection for cancer-related mutations [56].
Verification and monitoring of instrument performance is
important for implementation of dPCR as a routine diagnostic.
Instrument related factors can introduce bias to results even when
assay protocol is followed correctly. For example, accurate and
reproducible temperature parameters during amplification are crit-
ical for reproducible results. Inaccurate thermal cycling tempera-
tures, due to poor uniformity of temperature across the 96 wells of
a thermocycler heating block, can result in no dPCR amplification
or low amplification efficiency in wells where optimal temperatures
are not reached.
5 Concluding Remarks
dPCR has increasingly been proved as a “ground-breaking” tech-

nology for nucleic acid quantification. Recent developments in
commercial dPCR systems made the technology accessible and
affordable, and as an immediate consequence the number of users
and diversity of application has progressed at staggering pace. As
evident by the published literature from both academia and the

industry, the main focus of interest in dPCR technology is in clinical
diagnostics applications. However, the translation of a dPCR based
test from research laboratory to clinical laboratory setting will
require the test to satisfy validation requirements from regulatory
authorities.
For dPCR to realize its full potential as a quantitative molecular
diagnostic tool, carefully designed validation studies, and sufficient
understanding of sources of variation, bias, and associated uncer-
tainty that can lead to incorrect results or poor reproducibility of
dPCR measurements data are necessary.
Acknowledgments
We thank Kate Griffiths and Somanath Bhat for their inputs while
reviewing the manuscript.
References
1. Strain MC, Lada SM, Luong T et al (2013) https://doi.org/10.1016/j.foodchem.2014.

Highly precise measurement of HIV DNA by 10.138
droplet digital PCR. PLoS One 8(4):e55943. 7. Racki N, Dreo T, Gutierrez-Aguirre I et al
https://doi.org/10.1371/journal. (2014) Reverse transcriptase droplet digital
pone.0055943 [doi] PONE-D-12-26722 [pii] PCR shows high resilience to PCR inhibitors
2. Yang R, Paparini A, Monis P et al (2014) Com- from plant, soil and water samples. Plant Meth-
parison of next-generation droplet digital PCR ods 10(1):42. https://doi.org/10.1186/
(ddPCR) with quantitative PCR (qPCR) for s13007-014-0042-6
enumeration of cryptosporidium oocysts in 8. Cao Y, Raith MR, Griffith JF (2015) Droplet
faecal samples. Int J Parasitol 44 digital PCR for simultaneous quantification of
(14):1105–1113. https://doi.org/10.1016/j. general and human-associated fecal indicators
ijpara.2014.08.004 for water quality assessment. Water Res
3. Pond MJ, Nori AV, Patel S et al (2015) Perfor- 70:337–349. https://doi.org/10.1016/j.
mance evaluation of automated urine micros- watres.2014.12.008
copy as a rapid, non-invasive approach for the 9. Kim TG, Jeong SY, Cho KS (2014) Compari-
diagnosis of non-gonococcal urethritis. Sex son of droplet digital PCR and quantitative
Transm Infect 91(3):165–170. https://doi. real-time PCR for examining population
org/10.1136/sextrans-2014-051761 dynamics of bacteria in soil. Appl Microbiol
4. Devonshire AS, Honeyborne I, Gutteridge A Biotechnol 98(13):6105–6113. https://doi.
et al (2015) Highly reproducible absolute org/10.1007/s00253-014-5794-4
quantification of mycobacterium tuberculosis 10. Day E, Dear PH, McCaughan F (2013) Digital
complex by digital PCR. Anal Chem 87 PCR strategies in the development and analysis
(7):3706–3713. https://doi.org/10.1021/ of molecular biomarkers for personalized med-
ac5041617 icine. Methods (San Diego, Calif) 59
5. Morisset D, Stebih D, Milavec M et al (2013) (1):101–107. https://doi.org/10.1016/j.
Quantitative analysis of food and feed samples ymeth.2012.08.001
with droplet digital PCR. PLoS One 8(5): 11. Jennings LJ, George D, Czech J et al (2014)
e62583. https://doi.org/10.1371/journal. Detection and quantification of BCR-ABL1
pone.0062583 fusion transcripts by droplet digital PCR. J
6. Floren C, Wiedemann I, Brenig B et al (2015) Mol Diagn 16(2):174–179. https://doi.org/
Species identification and quantification in 10.1089/hgtb.2013.131 10.1016/j.
meat and meat products using droplet digital jmoldx.2013.10.007
PCR (ddPCR). Food Chem 173:1054–1058.
12. Regan JF, Kamitaki N, Legler T et al (2015) A reprogrammed induced pluripotent stem cells.
rapid molecular approach for chromosomal Nat Commun 6:6318. https://doi.org/10.
phasing. PLoS One 10(3):e0118270. https:// 1038/ncomms7318
doi.org/10.1371/journal.pone.0118270 23. Beck J, Bierau S, Balzer S et al (2013) Digital
13. Nadauld L, Regan JF, Miotke L et al (2012) droplet PCR for rapid quantification of donor
Quantitative and sensitive detection of cancer DNA in the circulation of transplant recipients
genome amplifications from formalin fixed par- as a potential universal biomarker of graft
affin embedded tumors with droplet digital injury. Clin Chem 59(12):1732–1741.
PCR. Transl Med (Sunnyvale, Calif) 2(2). https://doi.org/10.1373/clinchem.2013.
https://doi.org/10.4172/2161-1025. 210328
1000107 24. George D, Czech J, John B et al (2013) Detec-
14. Sanmamed MF, Fernandez-Landazuri S, tion and quantification of chimerism by droplet
Rodriguez C et al (2015) Quantitative cell- digital PCR. Chimerism 4(3):102–108.
free circulating BRAFV600E mutation analysis https://doi.org/10.4161/chim.25400
by use of droplet digital PCR in the follow-up 25. Halling KC, Schrijver I, Persons DL (2012)
of patients with melanoma being treated with Test verification and validation for molecular
BRAF inhibitors. Clin Chem 61(1):297–304. diagnostic assays. Arch Pathol Lab Med 136
https://doi.org/10.1373/clinchem.2014. (1):11–13. https://doi.org/10.5858/arpa.
230235 2011-0212-ED
15. Heredia NJ, Belgrader P, Wang S et al (2013) 26. Higuchi R, Dollinger G, Walsh PS et al (1992)
Droplet digital PCR quantitation of HER2 Simultaneous amplification and detection of
expression in FFPE breast cancer samples. specific DNA sequences. Biotechnology 10
Methods 59(1):S20–S23. https://doi.org/10. (4):413–417
1016/j.ymeth.2012.09.012 27. Heid CA, Stevens J, Livak KJ et al (1996) Real
16. Garcia-Murillas I, Lambros M, Turner NC time quantitative PCR. Genome Res 6
(2013) Determination of HER2 amplification (10):986–994
status on tumour DNA by digital PCR. PLoS 28. McCrady MH (1915) The numerical interpre-
One 8(12):e83409. https://doi.org/10. tation of fermentation-tube results. J Infect Dis
1371/journal.pone.0083409 17(1):183–212. https://doi.org/10.2307/
17. Beaver JA, Jelovac D, Balukrishna S et al 30083495
(2014) Detection of cancer DNA in plasma of 29. Simmonds P, Balfe P, Peutherer JF et al (1990)
patients with early-stage breast cancer. Clin Human immunodeficiency virus-infected indi-
Cancer Res 20(10):2643–2650. https://doi. viduals contain provirus in small numbers of
org/10.1158/1078-0432.ccr-13-2933 peripheral mononuclear cells and at low copy
18. Barrett AN, McDonnell TC, Chan KC et al numbers. J Virol 64(2):864–872
(2012) Digital PCR analysis of maternal plasma 30. Brisco MJ, Condon J, Sykes PJ et al (1991)
for noninvasive detection of sickle cell anemia. Detection and quantitation of neoplastic cells
Clin Chem 58(6):1026–1032. https://doi. in acute lymphoblastic leukaemia, by use of the
org/10.1373/clinchem.2011.178939 polymerase chain reaction. Br J Haematol 79
19. Weber ND, Stone D, Sedlak RH et al (2014) (2):211–217
AAV-mediated delivery of zinc finger nucleases 31. Sykes PJ, Neoh SH, Brisco MJ et al (1992)
targeting hepatitis B virus inhibits active repli- Quantitation of targets for PCR by use of lim-
cation. PLoS One 9(5):e97579. https://doi. iting dilution. BioTechniques 13(3):444–449
org/10.1371/journal.pone.0097579
32. Vogelstein B, Kinzler KW (1999) Digital PCR.
20. Miyaoka Y, Chan AH, Judge LM et al (2014) Proc Natl Acad Sci U S A 96(16):9236–9241
Isolation of single-base genome-edited human
iPS cells without antibiotic selection. Nat 33. Thorsen T, Maerkl SJ, Quake SR (2002)
Methods 11(3):291–293. https://doi.org/ Microfluidic large-scale integration. Science
10.1038/nmeth.2840 298(5593):580–584. https://doi.org/10.
1126/science.1076996
21. Cai J, Miao X, Li Y et al (2014) Whole-genome
sequencing identifies genetic variances in 34. Ottesen EA, Hong JW, Quake SR et al (2006)
culture-expanded human mesenchymal stem Microfluidic digital PCR enables multigene
cells. Stem Cell Reports 3(2):227–233. analysis of individual environmental bacteria.
https://doi.org/10.1016/j.stemcr.2014.05. Science 314(5804):1464–1467. https://doi.
019 org/10.1126/science.1131370. 314/5804/
1464 [pii]
22. Gao S, Zheng C, Chang G et al (2015) Unique
features of mutations revealed by sequentially 35. Warren L, Bryder D, Weissman IL et al (2006)
Transcription factor profiling in individual
hematopoietic progenitors by digital RT-PCR. 46. Dingle TC, Sedlak RH, Cook L et al (2013)
Proc Natl Acad Sci U S A 103 Tolerance of droplet-digital PCR vs real-time
(47):17807–17812 quantitative PCR to inhibitory substances. Clin
36. Dube S, Qin J, Ramakrishnan R (2008) Math- Chem 59(11):1670–1672. https://doi.org/
ematical analysis of copy number variation in a 10.1373/clinchem.2013.211045
DNA sample using digital PCR on a nanoflui- 47. Hindson BJ, Ness KD, Masquelier DA et al
dic device. PLoS One 3(8):e2876 (2011) High-throughput droplet digital PCR
37. Bhat S, Herrmann J, Armishaw P et al (2009) system for absolute quantitation of DNA copy
Single molecule detection in nanofluidic digital number. Anal Chem 83(22):8604–8610.
array enables accurate measurement of DNA https://doi.org/10.1021/ac202028g [doi]
copy number. Anal Bioanal Chem 394 48. Bhat S, McLaughlin JL, Emslie KR (2011)
(2):457–467 Effect of sustained elevated temperature prior
38. Pinheiro LB, Coleman VA, Hindson CM et al to amplification on template copy number esti-
(2012) Evaluation of a droplet digital polymer- mation using digital polymerase chain reaction.
ase chain reaction format for DNA copy num- Analyst 136(4):724–732. https://doi.org/
ber quantification. Anal Chem 84:1003–1011. 10.1039/c0an00484g [doi]
https://doi.org/10.1021/ac202578x 49. Corbisier P, Pinheiro L, Mazoua S et al (2015)
39. Zhong Q, Bhattacharya S, Kotsopoulos S et al DNA copy number concentration measured by
(2011) Multiplex digital PCR: breaking the digital and droplet digital quantitative PCR
one target per color barrier of quantitative using certified reference materials. Anal Bioanal
PCR. Lab Chip 11(13):2167–2174. https:// Chem 407(7):1831–1840. https://doi.org/
doi.org/10.1039/c1lc20126c 10.1007/s00216-015-8458-z
40. Pekin D, Skhiri Y, Baret JC et al (2011) Quan- 50. Jacobs BK, Goetghebeur E, Clement L (2014)
titative and sensitive detection of rare muta- Impact of variance components on reliability of
tions using droplet-based microfluidics. Lab absolute quantification using digital PCR.
Chip 11(13):2156–2166. https://doi.org/ BMC Bioinformatics 15:283. https://doi.
10.1039/c1lc20128j [doi] org/10.1186/1471-2105-15-283
41. Whale AS, Huggett JF, Cowen S et al (2012) 51. Hindson CM, Chevillet JR, Briggs HA et al
Comparison of microfluidic digital PCR and (2013) Absolute quantification by droplet dig-
conventional quantitative PCR for measuring ital PCR versus analog real-time PCR. Nat
copy number variation. Nucleic Acids Res 40 Methods 10(10):1003–1005. https://doi.
(11):e82. https://doi.org/10.1093/nar/ org/10.1038/nmeth.2633
gks203 52. JCGM (2008) JCGM 200:2012 international
42. Belgrader P, Tanner SC, Regan JF et al (2013) vocabulary of metrology–basic and general
Droplet digital PCR measurement of HER2 concepts and associated terms (VIM), 3rd
copy number alteration in formalin-fixed paraf- edn. BIPM, Sèvres Cedex, France
fin-embedded breast carcinoma tissue. Clin 53. Griffiths KR, Burke DG, Emslie KR (2011)
Chem 59(6):991–994. https://doi.org/10. Quantitative polymerase chain reaction: a
1373/clinchem.2012.197855 framework for improving the quality of results
43. Nixon G, Garson JA, Grant P et al (2014) and estimating uncertainty of measurement.
Comparative study of sensitivity, linearity, and Anal Methods 3:2201–2211
resistance to inhibition of digital and nondigital 54. Yukl SA, Kaiser P, Kim P et al (2014) Advan-
polymerase chain reaction and loop mediated tages of using the QIAshredder instead of
isothermal amplification assays for quantifica- restriction digestion to prepare DNA for drop-
tion of human cytomegalovirus. Anal Chem 86 let digital PCR. BioTechniques 56
(9):4387–4394. https://doi.org/10.1021/ (4):194–196. https://doi.org/10.2144/
ac500208w 000114159
44. Sedlak RH, Kuypers J, Jerome KR (2014) A 55. Clinical and Laboratory Standards Institute
multiplexed droplet digital PCR assay performs (2004) Protocols for determination of limits
better than qPCR on inhibition prone samples. of detection and limits of quantitation;
Diagn Microbiol Infect Dis 80(4):285–286. approved guideline, vol vol 24. CLSI, Wayne,
https://doi.org/10.1016/j.diagmicrobio. OA, USA. Contract No.: EP17
2014.09.004 56. Milbury CA, Zhong Q, Lin J et al (2014)
45. Huggett JF, Novak T, Garson JA et al (2008) Determining lower limits of detection of digital
Differential susceptibility of PCR reactions to PCR assays for cancer-related gene mutations.
inhibitors: an important and unrecognised Biomol Detect Quantif 1:8–22
phenomenon. BMC Res Notes 1:70
Chapter 3
Fundamentals of Counting Statistics in Digital PCR: I Just

Measured Two Target Copies–What Does It Mean?
Svilen Tzonev
Abstract
Current commercially available digital PCR (dPCR) systems and assays are capable of detecting individual
target molecules with considerable reliability. As tests are developed and validated for use on clinical
samples, the need to understand and develop robust statistical analysis routines increases. This chapter
covers the fundamental processes and limitations of detecting and reporting on single molecule detection.
We cover the basics of quantification of targets and sources of imprecision. We describe the basic test
concepts: sensitivity, specificity, limit of blank, limit of detection, and limit of quantification in the context of
dPCR. We provide basic guidelines how to determine those, how to choose and interpret the operating
point, and what factors may influence overall test performance in practice.
Key words Statistics, Counting, Sensitivity, Specificity, Subsampling, Limit of detection, Limit of
blank, Poisson distribution, False positive, False negative, Performance characteristics
1 Introduction
Digital PCR (dPCR) technology promises to change how genomic

markers are detected and quantified [1, 2]. It has been demon-
strated to effectively count amplifiable targets at single molecule
resolution. This is possible because partitioning isolates the targets,
while end-point amplification allows highly reliable detection of the
reaction components in each partition.
This chapter will focus on the fundamental concepts of count-
ing statistics as they apply to detection and quantification of targets
in dPCR tests. We will cover basic detection and quantification,
describe critical test performance parameters, and provide recom-
mendations for choosing the test operating point.
We will first cover detection of single species and discuss detec-
tion and counting of generic target molecules.
25
26 Svilen Tzonev
2 Counting Molecules
In digital PCR the PCR reaction is first partitioned into indepen-

dent, usually equal size, volumes (droplets or chambers on a device)
[3]. During a standard thermocycling protocol, amplifiable geno-
mic target molecules will be replicated and their number will
increase exponentially. In a common format, hydrolysis probes
matching specific amplicon counterparts are used. The probes
have quencher and dye molecules attachments. When the polymer-
ase enzyme cleaves the hybridized probes it releases unquenched
dye molecules into the reaction. Thus, the concentration of free dye
in the partition will increase exponentially with the PCR cycles. At
end-point, any partition that contained at the beginning an amplifi-
able target will contain detectable high concentration of fluorescent
dye and can be detected as “positive.” Any partition that did not
contain a target will remain “negative.” In practice, we have to deal
with damaged or otherwise poorly PCR-accessible target molecules
that may not amplify from cycle 1, nonspecific probe hybridization
that may lead to false positives, polymerase errors that may lead to
false positive or false negatives, and other complications. We will
cover such cases later.
First we will focus on an “ideal dPCR counting machine” that
can detect with 100% certainty and no confusion the presence or
absence of a target molecule in a partition.
3 Subsampling
Let us first consider a test that detects a single species of target

molecules. Even a perfect counting machine, which makes no mis-
takes of any kind, has to deal with the stochastic nature of many
molecular phenomena. In a typical experiment or test, we want to
measure the concentration of a particular biomolecule as it is
in vivo. In order to do this, we have to use a subsample of the
whole—a draw of blood or a tissue biopsy. We measure the sub-
sample in vitro to estimate the state of the whole. In the following
we will use the liquid biopsy paradigm to illustrate the basic
concepts.
A hypothetical patient has 5 L of blood. If we draw 10 ml of
blood at a time, we are subsampling 0.2% of the whole blood
volume. Let us say that there are 5000 copies of the target mole-
cules of interest in the 5 L of blood. This corresponds to a concen-
tration of 1 target/1 ml of blood. In any 10 ml draw, we expect to
collect on average 10 target molecules. Assuming perfect sample
processing with no losses of any kind, we would expect to load on
average 10 targets in our machine, which would detect all of them.
However, any individual 10 ml draw may contain 8 or 11 or
Counting Statistics 27
5 L of 10ml 8 copies
blood (0.2%) present
11 copies
present
5000 expect 10 copies

copies 10 copies present
present
3 different draws
Fig. 1 Subsampling process. A patient has 5 L of blood in which 5000 molecules

of interest are present. Any 10 ml draw of blood represents 0.2% of the total
blood volume. We would expect to see on average 10 target molecules.
However, in any actual 10 ml drawn, we may see different numbers of targets
molecules due to the stochastic nature of molecule distribution in the 5 L of
blood
10 actual target molecules. Our ability to make confident state-

ments about the whole is still limited by the subsampling problem.
See Fig. 1.
How variable is the number of actual targets per draw? When
we are drawing a tiny fraction of the whole and, as long as we can
assume that the targets are all independent and equally likely to be
sampled, we can describe the probability to draw exactly k actual
targets with the Poisson distribution:
Probability ðto draw k molecules; given expected number molecules is μÞ
e μ
¼ μk
k!
where μ is the expected or average number drawn; in our case it is
10. Note that k must be an integer—we may draw 0, 1, 2, etc.
molecules, while μ can be a noninteger number—the average may
be 1.5 molecules. Figure 2 illustrates the Poisson distribution for
μ ¼ 3, 10, and 100. As we can see, when μ is small, there is
significant variability of how many targets will be subsampled in a
draw relative to the expected number. Note how in the case of three
expected molecules, there is about 5% probability that the draw
would contain no targets at all! This observation will form the basis
of the fundamental bound on the limit of detection, this is dis-
cussed later. Note also that for large μ the relative uncertainty
shrinks as the distribution becomes more concentrated around
the average. It also becomes more like the familiar normal or
Gaussian distribution.
28 Svilen Tzonev
Probability
0.25
0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14

0.20
0.15
0.10
0.05
0.00
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
0
10
Number of molecules drawn
0.05
0.04
0.03
0.02
0.01
0.00
80 90 100 110 120
Fig. 2 Poisson distribution. Poisson distribution for three different values of the expected number of molecules
μ
drawn, μ ¼ 3, 10 or 100. Probability (to draw k molecules, given expected number molecules is μ ¼ μk ek ! ).
The height of each bar for particular integer k represents the probability to draw k molecules. Note how for
small μ, the distribution shape is skewed and becomes increasingly more symmetrical and concentrated
around the maximum for larger values of μ. For each case, the probability to draw exactly zero molecules is
also shown
4 Multiple Occupancy and Partitioning
The original concept of dPCR assumed a limiting dilution regime,

i.e., each partition would not contain more than one target with
any significant probability. Most of current commercial systems
[4–7] allow robust measurement of target copies even when multi-
ple molecules can occupy the same partition with a significant
probability. Quantification is possible under the assumption that
the targets are allocated to a partition independently and with equal

probability. (Such assumptions would be violated if the targets are
in some way “entangled” or if the partitions are of different
volumes.) It turns out that the appropriate statistical framework is
also based on the Poisson distribution, since each partition is
similar to a blood draw from the previous example—each partition
is a small subsample of a larger whole. For specific details see [8, 9];
here we will only focus on how the target concentration is
estimated.
A positive partition may have contained one or more targets.
There is no way to be certain. A negative partition, on the other
hand, must have contained exactly 0 targets, otherwise it would be
positive. Once we have counted all partitions (positive and nega-
tive), Ntot, and all negative partitions, Nneg, we can use this formula
to estimate the average target occupancy per partition, λ:

N neg
λ ¼ ln ðN tot Þ ln N neg ¼ ln
N tot
The total number of targets, T, present at the beginning of the
reaction is then simply
T ¼ λ N tot
The value of λ calculated in this way is the “most likely value.”
We cannot be sure of the exact number in reality since we do not
know the precise distribution of targets into partitions.
To arrive at the target concentration, we divide λ by the parti-
tion volume:
λ
½T ¼
V partition
Consider a given reaction mix with a fixed number of targets in
it. If we perform multiple thought partitioning experiments we can
understand the variability of the results around the most likely
values. For each partitioning, depending on the exact (stochastic)
pattern of target distribution into the partitions, we may see differ-
ent number of negative partitions. This would lead us to estimate
slightly different values of the target copies. We are again subject to
the inevitable stochastic nature of how molecules “choose” their
partitions.
Figure 3 illustrates the concept of partitioning uncertainty. We
have partitioned the same reaction volume with the same number
of molecules, T, three times into the same number of partitions of
equal volume. Since the molecules are distributed stochastically
into the partitions, the patterns in each realization will be different.
The observed number of partitions with no molecules will also vary
between realizations. In fact, the number of actual negative parti-
tions is a variable that also follows a Poisson distribution, given the
expected number of negative partitions. We can then use similar
30 Svilen Tzonev
20 negatives 17 negatives 19 negatives

Fig. 3 Partitioning uncertainty. A thought experiment where a reaction volume with fixed number of molecules
is partitioned three times. Exact location of each molecule in the resulting partitions is stochastic. This leads to
a different number of partitions with no molecules at all—negative partitions are highlighted. This will result in
somewhat different “most likely estimate” of the number of molecules in the reaction. The number of negative
partitions is itself a variable that follows a Poisson distribution
math as above to describe how uncertain the estimate of λ is going

to be even when we know the exact number of molecules, T, in the
total volume.
5 Total Poisson Uncertainty
So far, we saw how subsampling and partitioning variability limit

our ability to estimate precisely the number of target molecules
in vivo even with a perfect counting machine. When we combine
the two sources of variability, we arrive at the frequently described
dPCR uncertainty curve, shown also Chap. 2. When only a small
number of targets is drawn (provided we have enough partitions),
the total uncertainty is dominated by the subsampling processes. In
other words, it does not matter if we put, five actual targets in
20,000 or 20,000,000 partitions, we will know with high confi-
dence that there were five targets in the reaction we measured and
approximately five targets per blood draw, if these were repeated
multiple times. When the number of targets increases relative to the
number of partitions (average occupancy λ is larger), at some point
there will be too few negative partitions. Since the actual number
of negative partitions is a stochastic variable, the uncertainty of λ
increases and, thus, the uncertainty of T.
Figure 4 illustrates the contributions from subsampling and
partitioning errors to the total Poisson statistical error. For low
occupancy numbers, λ, (number of targets is much less than the
number of partitions), total uncertainty is dominated by subsam-
pling. While for large occupancy numbers (number of targets much
larger than the number of partitions), total uncertainty is domi-
nated by partitioning. The lowest overall uncertainty is reached at
λopt 1.6. For 10,000 partitions even when λ ¼ 6, the coefficient of
5
4
3
CV, %
2
Partitioning error
1
Total Poisson error
Subsampling error
0
0 1 2 3 4 5 6
Average targets per partition, λ
Fig. 4 Total Poisson Error ¼ Subsampling + Partitioning Errors, expressed as %

CV, as a function of average targets per partition, λ. The black curve shows the
subsampling error. The blue curve shows the total Poisson error. The distance
between the curves is due to the partitioning error. For small λ, the subsampling
error dominates, while the partitioning error is small. For large λ, the
subsampling error is small, while the partitioning error dominates. The total
Poisson error is minimized at λ_opt 1.6. Values for %CV are calculated for
10,000 partitions. Increasing the number of partitions brings all curves lower
(smaller %CV) but does not change the general shapes. Optimum stays at 1.6
targets per partition
variation of λ and T is ~3%, i.e., sufficiently low to allow accurate

and precise quantification in most cases.
A few words about the concept of confidence intervals. When
we estimate experimentally some unknown variable we talk about a
point estimate and a confidence interval around it. The point
estimate is the most likely value of the variable; this is what is usually
reported. A 95% confidence interval, around the point estimate,
describes the range between a minimum and a maximum value of
the variable, between which the true value of it will lie 95% of the
time. In other words, if we knew the exact value of the variable and
repeated the estimation experiment many times, 95% of the time it
will indeed lie in the stated confidence interval. Most commercial
dPCR systems report both point value estimates and 95% confi-
dence intervals for target copies and concentrations.
6 False Positives and False Negatives
So far, we have discussed an idealized case where our perfect system

makes no mistakes of any kind. In practice, we have to deal with
false positive and false negative reporting. In the simplest case, we
can define a false positive or negative at the level of the partitions—a
32 Svilen Tzonev
partition that should have been detected as a negative is reported as

a positive and vice versa. The rates of such false reporting per
partition define a false positive rate, FPR, and a false negative rate,
FNR. Which of these is more important depends on the test and
how much we care about each kind of error.
For example, when false (positive and negative) rates are small
and there is a small number of targets, the FNR is usually not
important in absolute terms, as we only have a few true positive
partitions that may be converted into a false negative. In this case,
the FPR is more significant as there is a large number of true
negative partitions, each of which may turn into a false positive.
In the opposite case—a large number of targets—it is the positive
partitions that are numerous and thus FNR matters more, in abso-
lute terms, than the FPR per partition.
In both cases, the prevalence of the state or condition measured
plays a role in determining which false rate is more important. This
is true at the patient or sample level as well, in the sense of the
prevalence of the condition we are trying to measure with our test.
There are cases where the danger of misdiagnosis at the clinical level
influences greatly the relative importance of FPR and FNR. For
example, in a prenatal screening test for Down syndrome, most
patients are expected to have normal pregnancies. Due to the large
number of individuals screened, even a small test FPR will produce
a large number of false positive results that would require additional
procedures. On the other hand, a false negative result is also very
damaging as the overall cost to the patient is significant, since the
fetus would be incorrectly classified as normal.
We can think of false positive and false negative rates at the level
of a well, test (multiple wells) or sample (if multiple tests are
performed). It is important to keep in mind which of these con-
cepts is being used. The combination of the sample preparation
protocol, detection assay and dPCR measurement system will be
subject to the limitations imposed by both stochastic molecular
processes and the FP/FN detection rates.
A few common examples of processes that can lead to false
results include:
l Nonspecific primers and/or probe hybridization
l Polymerase errors
l Fluorescent dust particles
l Contamination and cross-contamination
l Inappropriate thresholding algorithms
In practice, these should be recognized and controlled at the
assay design level, periodic equipment maintenance, robust labora-
tory practices, and frequent specific process monitoring.
7 Basic Test Concepts and Their Interpretation in dPCR
We distinguish two major types of tests—qualitative and quantita-

tive. A qualitative test can only categorize an unknown sample as
positive or negative (in some case there is also a third, uncalled
category). A quantitative test will produce a value for the measur-
and of interest in an unknown sample when possible or perhaps
produce an out-of-reportable-range result. For dPCR, quantifica-
tion is most often as target counts or copies in the test or concen-
tration of target copies in the test. There may be other derivative
measures that are based on combinations of such measurements,
for example a ratio of two concentrations or similar.
dPCR essentially always produces a direct estimate of the target
copies in the reaction. Thus, it is by nature quantitative. To call a
sample negative or positive, an analytical call cutoff value is
applied—any sample with detected copies higher than or equal to
the cutoff is declared as positive; any sample with fewer copies than
the cutoff is declared as negative. In some cases, there may be a
“grey zone” where no call is produced. We will focus on the simpler
binary version of sample positive/negative determination.
Three fundamental parameters determine the performance of a
given dPCR test: FPR, FNR, and the choice of analytical cutoff
value (if needed for a qualitative test).
Next, we will define additional important test descriptors and
discuss how they are influenced by these fundamental parameters in
the context of counting assays in dPCR. Looking ahead at Fig. 5,
discussed below, will help the reader visualize the concepts.
8 Sensitivity
For a qualitative test the sensitivity is defined as the probability that a

truly positive (as determined by a prior test or clinical criterion)
sample is measured as positive by the test. 100% sensitivity means
that all truly positive samples do come out as positive by the test.
Sensitivity may be less than 100% when the FNR per test is higher
than zero or if the call cutoff value is set too high and truly positive
samples are called as negatives. Intuitively, if we challenge the test
with strongly positive or strongly negative samples, we can achieve
better performance (See below on the concept of limit of detection).
9 Specificity
For a qualitative test, specificity is defined as the probability that a

truly negative sample is measured as negative by the test. Again,
100% specificity means that the test is never wrong when negative
34 Svilen Tzonev
a
Call cut-off
NEGATIVE POSITIVE
Probability LOD LOD
negative control positive controls
level level
0 Detected positive events per test

level
Fig. 5 Relationship between sensitivity, specificity, and call cutoff (a). Classic description and definition of
these test concepts. Horizontal axis—signal measured by the test (in dPCR this is detected targets/events).
Vertical axis—probability of a particular level to be measured. Left curve—the test response to a negative/
blank control, Right curves—the test response to two positive samples at different levels. Depending where
we choose the value of the call cutoff, the areas under the curves to the right and to the left of the cutoff
determine the values of specificity (β level) and sensitivity (α level), respectively. Sensitivity ¼ 1 α.
Specificity ¼ 1 β. Moving the cutoff to the right will decrease β and increase α , thus, increase specificity
but decrease sensitivity. The opposite is true when moving the cut-off to the left. The value of the measurand
for which α0 ¼ 5%, determines the 95% confidence level limit of detection of the test (indicated as the peak of
the corresponding positive control curve, or most likely value). For increased Sensitivity, α1 < 5%, the
corresponding LOD will be greater than the LOD at 95%. When we use blank samples for the negative control,
the cutoff value corresponds to the limit of blank at 1β level. (b). Same concepts as they apply in dPCR.
Continuous curves are replaced by histograms as dPCR reports integer values for the targets detected. Blue:
histogram of possible copies for a negative sample, Red: histogram of possible copies for a positive sample.
Moving the cutoff value is only possible in discrete ways, which results in jumps of possible values of
sensitivity and specificity (α and β levels not shown)
samples are measured. Any value higher than zero for the FPR per
test will lead to specificity that is less than 100%. Alternatively, a
cutoff that is set too low, causing truly negative samples to be called
as positives, will also cause specificity of less than 100%.
Figure 5 illustrates the inherent trade-off between sensitivity
and specificity. Note how by moving the cutoff value, one changes
both the specificity and sensitivity of a given test, increasing one
while decreasing the other. These change in opposite directions.
For tests that produce integer values (molecules detected in test)
cutoff values can only be discrete, which leads to a set of possible
combinations of sensitivity/specificity values.
10 Limit of Blank (LOB)
The limit of blank, LOB, at a particular confidence level, usually

95%, is a critical concept. The formal definition, according to CLSI
guidelines [10], is the maximum value of the analyte that may be
reported 95% of the time when we measure a true blank sample. In
dPCR we report counts of target molecules, thus the LOB must be
an integer—0, 1, 2, etc. counts per test. When we say that the LOB
for a test is 1 copy, we mean that 95% of the time when we measure
a negative/blank sample we will report at most one copy. This is
equivalent to stating that we will report 2, 3, or more copies at most
5% of the time (See Fig. 5).
The LOB is closely related to the specificity of the test and the
underlying FPR per partition or per test. When we measure a blank
sample, each partition should be reported as negative. However,
when the FPR per partition is not 0, some truly negative partitions
may be reported as false positives. In most tests we will still report
0 positive partitions and, thus, 0 target copies. But for some sam-
ples, we may report 1, 2 or more copies. These will all be false
positive counts.
The robust way to experimentally measure LOB of a test is to
run multiple blank samples, record the number of targets reported
and rank order the results. When we draw a line at the 95% percen-
tile, the value where this happens is the 95% LOB. It is good
practice to verify the assumed Poisson distribution of the counts
by fitting a model. Theory predicts that these counts will also follow
a Poisson distribution, as long as each negative partition may be
independently misreported as positive. The average number of false
positive target counts per test will equal the FPR per partition times
the number of partitions per test. If the counts per test do not
appear to follow a Poisson distribution, we may still use the LOB
number, but need to be aware there are other effects playing a role,
other than unreliable detection of negative partitions. These may
point to additional assay variability or inadequate lab practices.
36 Svilen Tzonev
11 Limit of Detection (LOD)
The limit of detection, LOD, at a particular confidence level,

frequently 95%, is another critical concept. The formal definition,
according to the CLSI guidelines, is the minimum level of the
analyte in a sample that will be reported as detected with this
same 95% probability. The LOD level depends on the analytical
call cutoff threshold and FPR and FNR values. An increase of the
FPR per partition will lead to an increase of the false positives and
will cause the system to report on average higher counts than truth,
pushing the apparent LOD lower if the cutoff is held constant (see
below). An increase of the FNR has the opposite effect—it may
cause positive partitions to appear as negative, effectively reporting
a lower value for the analyte than truth. This will increase the LOD,
if we hold the cutoff threshold constant.
The lowest possible value for the call cutoff is 1 copy per test,
i.e., if we see a single copy of the target we will call the sample
positive. Even if FPR and FPN are exactly zero, we already saw how
subsampling puts a fundamental limit on what target levels we can
detect confidently. Recall when we expect to draw on average three
target copies, there is a 5% probability that any individual draw will
contain no copies at all—there would be nothing to detect! This
observation sets a critical level of three copies as the natural LOD of
molecular counting assays like in dPCR when subsampling is a
factor (see exceptions in “Finite Subsampling” section below).
This limitation applies to any other counting based methods, like
versions of next generation sequencing, that detect individual tar-
get molecules. Choosing a higher level for the call cutoff will
produce a higher LOD value. Table 1 illustrates the LOD 95%
values for different cutoff values.
A word of caution, the reader should not confuse the concepts
of Sensitivity and Positive Predictive Value (PPV) of a test. The
Table 1
LOD 95% values for different test cutoff values (copies per test). FPR ¼ 0,
FNR ¼ 0
Call cutoff (positive sample) LOD 95% copies

copies per test per test
1 3.00
2 4.74
3 6.30
4 7.75
5 9.15
Sensitivity, again, relates to the probability that the test will call a
positive sample as positive, while PPV inverts the logic—it defines
the probability that a sample that is called positive by the test is
indeed positive in reality.
12 Limit of Quantification (LOQ)
For quantitative tests the limit of quantification (LOQ) of a test is

defined as the minimum level of an analyte that can be measured
within a predefined level of uncertainty. The uncertainty is typically
expressed by the standard deviation or the coefficient of variation
based on multiple measurements of a given sample at the LOQ
level. The preferred imprecision level depends on the situation,
typical numbers include 20%, 35%, or higher and depend on the
requirements of the application.
By definition, the values of the three parameters must be
ordered
LOB < LOD LOQ
For any test the LOD must be higher than the LOB, and the
LOQ cannot be lower than the LOD. At the lowest LOD achiev-
able, three copies, the minimal possible coefficient of variation is
1
CV min ¼ pffiffiffi 58%
3
In practice, the observed CV at this level will usually be higher

as other sources of variability will inevitably add to this theoretical
minimum. There are dPCR tests with remarkably low levels of
LOB, LOD, and LOQ, which reach very close to the theoretical
limits of single molecule detection levels.
13 Why and How Do These Performance Characteristics Matter
The concepts discussed above describe the overall performance of a

test of interest. This is expressed in the context of “absolute truth”
on the analytical side or a prior method of measurement of what
matters on the clinical side. For example, if a patient has ten viral
particles per 10 ml of blood (for a total of 5000 in the circulation),
is this clinically significant? Should this be viewed as a positive or a
negative patient? Typically, one chooses the clinical cutoff level on a
population basis based on criteria relevant to patient health or
prognosis.
The relationships between the test performance characteristics
and the clinical cutoff level determine the ultimate utility of the test.
38 Svilen Tzonev
Table 2
Clinical and test concepts meaning and context
When known
Applies to Determines How known or chosen relative to test
Clinical Patient Is a patient considered clinically Based on external Before the test
cutoff population positive or negative clinical knowledge is performed
Test/call Test When the test reports positive or Chosen to satisfy test Before the test
cutoff negative result requirements is performed
LOB Test Maximum value (95% Determined during test Before the test
probability) that may be development and is performed
reported on a blank sample validation
LOD Test Level of analyte that can be Determined during test Before the test
reliably (95% probability) development and is performed
detected by the test validation
Test result Sample + Test What the test measured for a By performing a test on After the test is
given sample a sample performed
The meaning of and the context of selected concepts are summar-

ized in Table 2.
In general, a test is suitable for determining patient status
when its LOD matches the clinical cutoff and does not produce too
many false positive results. But there may be cases where higher
specificity is preferred and the LOD differs from the clinical cutoff.
Table 2 also illustrates why it may be acceptable to report below a
stated LOD: the LOD is what we know about the test before we
have measured any particular sample. The test result is what we
know about the combination of test and sample—after we have
performed a measurement (see “Reporting below LOD” below).
14 Effects of False Positives, False Negatives, and Call Cutoffs
When a test has a finite (nonzero) FPR or FNR values, the values of
LOB, LOD, and LOQ will be different than when the FPR and
FNR are zero. In general, higher rates of false positive partitions
will lead to higher values of LOB for a fixed specificity. Since the
cutoff values must be integers, in practice there is a quantization
effect in this relationship.
As an illustration, let us consider first a test with zero FPR and
FNR. In this case the LOB will be zero copies per test at 100%
specificity level, the LOD will be three copies per test at the 95%
sensitivity level for a call cutoff of one copy per test. (The LOB will
of course also be zero at 95% specificity level.) The LOQ could be
11 copies per test at 30% CV, or perhaps, five copies per test at 45%
CV. These are the best possible performance characteristics of any
test based on counting single molecules, no matter what technology

is used.
Now let us increase the FPR to 0.01 copies per test. The LOB is
zero copies but at 99% specificity level, the LOD will be 2.99 copies
at 95% sensitivity. As we increase further the FPR per test, we can
keep LOB at zero but will lose specificity; at the same time LOD
will keep going lower for a fixed sensitivity level. We are still keeping
the cutoff value at one copy per test. Refer to Fig. 5b. At some value
of the FPR per test, the specificity may become unacceptably low
and we may be forced to choose a different cutoff—to the next
integer up. When we do this, we regain higher specificity, but have
to jump to a different zone for the LOD value—as a result, there
will be a significant change in the LOD value. This happens because
the specificity, sensitivity, and cutoff values are interrelated, and in
digital we must use an integer value for the cutoff. This “integer-
ness” requirement leads to specific zones of operation for any
digital test.
Table 3 illustrates the critical values of various parameters for
selected specificity and sensitivity levels and the zones defined by
the choice of cutoff level. Here is how to read it: The first column
contains the value of the call cutoff. Column 2 contains the maxi-
mum value of the FP counts per test allowed so we can operate at
95% specificity. Column 3 contains the corresponding value for the
LOD at 95% sensitivity (given the cutoff value and the max FPR).
All units are copes per test. The last column repeats the numbers
from Table 1 and illustrates the LOD level if we were operating at
exactly zero FPR.
Let us look at the row for cutoff of 1. If we need to have
specificity of at least 95% and use this cutoff, our test cannot have
FPR of higher than 0.05 copies per test. At this FPR level, the LOD
at 95% sensitivity will be 2.94 copies per test.
Table 3
Critical levels of cutoff value, FPR, and LOD for selected specificity and
sensitivity values
Spec 95%, Sens 95%
Call cutoff max FPR copies LOD 95%

(positive sample) per test LOD 95% at 0 FPR
1 0.05 2.94 3.00
2 0.36 4.39 4.74
3 0.82 5.48 6.30
4 1.37 6.38 7.75
5 1.97 7.19 9.15
40 Svilen Tzonev
For a test with FPR of 0.1 copies per test, in order to keep
specificity above 95%, we must choose a cutoff of 2. For this cutoff,
maximum FPR is 0.36 copies per test and the LOD at this FPR will
be 4.39 at 95% Sensitivity (row 2). Actual specificity for a test with
0.1 FPR and cutoff of 2 copies will be higher than 95% and we can
estimate the LOD 95% by subtracting 0.1 from 4.74 to get 4.64.
The careful reader will notice that the values of columns 2 and
3 add up to the value of column 5 for any given cutoff value (subject
to rounding effects). Intuitively, this roughly translates to the state-
ment that “all positive calls are either a true or a false positive for any
given cutoff level.”
Our last example will be a test with a FPR of one copy per test
and also negligible FNR. Since we cannot choose a cutoff between
three and four, we have to go to four. A sample is positive when we
see at least four copies. In this case the LOB is three copies (at a
level higher than 95% but we have no choice) per test. The 95%
LOD is approximately 6.75 copies per test (7.75–1.00). In this
framework, we call a sample positive when we measure four copies
and we will correctly call a sample as positive when the expected
number of copies in our test is 6.75 as drawn from the larger whole.
We need to emphasize once more that these relationships
between the critical values are governed by fundamental counting
statistics considerations. As long as we are trying to ascertain some-
thing about the sample in vivo based on small representative sub-
sample drawn from it, they will always apply. Subsampling at the
molecular level is governed by Poisson statistics and no technology
or approach can do better. dPCR technology and its commercial
implementation do allow us to reach very close to these physical
limitations, indeed.
15 Additional Considerations
15.1 Selecting We saw that the value of call cutoff, sensitivity, and specificity are
the Operating Point interrelated. How one chooses where to operate depends on the
requirement of the test: Is it more acceptable to call false positives
or false negatives? A commonly used scheme is to first select the
specificity level required. Together with the measured FPR (and
related LOB), this will determine the choice for the analytical cutoff
value. Then we evaluate the sensitivity required, which will finally
determine the LOD level for the test.
Referring to Fig. 5a, we first measure the distribution of results
on blank samples (left curve). We choose desired specificity, (1 β),
which sets the value of the call cutoff (area under the curve to the
right of the cutoff). We next select the desired Sensitivity, (1 α).
To do so, measure multiple replicates of samples at different levels
of positivity to generate a family of response curves (see Fig. 5a).
The curve (i.e., level of positivity) for which α% of the curve area is
to the left of the cutoff value determines the LOD at α%. Again, for
quantized measurements, we may be forced to settle on particular
values for specificity and sensitivity, since the cutoff must be an
integer.
15.2 Finite There are circumstances where subsampling is minimal and we

Subsampling therefore do not subsample a very small fraction of the whole. For
example, if experiments are performed on a single cell level and
most of the biological material is actually assayed. In such cases we
will not suffer the subsampling penalties to the same level. We need
to deal with hypergeometric distributions instead of Poisson, but
such details go beyond the scope of this chapter. In the extreme
case where we do measure the whole, “what we see is what was
there” principle applies and the only uncertainty would come from
partitioning variability and any dead volume in the system.
15.3 Nonstandard We described above the cases where pure molecular stochasticity is
Variability involved—the factors we know will be always present and that we
cannot control. In practice, there may be other sources of variability
that could extend uncertainty. Issues like inadequate lab practices,
contamination, reagent quality, and intermittent system perfor-
mance may invalidate the rules described here. Regular monitoring
of predicted versus measured behavior is important. This is accom-
plished by running controls with any unknown samples and peri-
odic evaluation of performance on such controls. Monitoring the
LOB on negative controls and verification of a Poisson distribution
of the false positives is helpful as well.
15.4 Multitarget A very common type of diagnostic tests is where a reference target is
and Related-Target measured together with the biomarker of interest. For example, the
Considerations number of wild-type genome copies detected is used as normaliza-
tion for the number of mutant target copies. In this case a natural
quantity of interest is the Minor Allele Frequency (MAF), the ratio
of (mutant)/(mutant + wild type) copies. Clearly, samples with a
higher number of wild-type copies may show a higher number of
false mutant copies. The FPR per wild-type copy, then, becomes
the natural metric to watch. Careful estimation and monitoring of
this rate is important to develop and apply appropriate cutoff
criteria. The concepts of LOB, LOD, and LOQ, as described
above, are applicable for the MAF value. In many cases, the LOD
when expressed in MAF space, will still be determined by the
availability of actual mutant copies. As we saw, the critical value is
three mutant copies for 95% confidence, which, when divided by
the wild-type copies present, will establish the critical level for the
MAF at 95% confidence.
When a large number of wild-type copies are present, the
cluster of partitions with both wild-type and mutant copies, i.e.,
double-positives, becomes more diffuse and it may be harder to
42 Svilen Tzonev
select appropriate amplitude threshold values. One technique that

can minimize the false calls is to use only pure mutant partitions as a
reliable source of mutant copies and compare against the double-
negative partitions’ count, although this will somewhat diminish
the sensitivity of the test.
15.5 Reporting Should we report a result which falls below the LOD? The nature of
Below LOD dPCR allows us to do this with confidence, provided that certain
criteria are met.
Let us first consider what the LOD means. This is what we
know about the detection capabilities of our test before we have
performed an experiment on an unknown sample. At this stage we
usually know nothing about the sample—this is why it is called
unknown. After we have performed the measurement, we know
more about the sample. The LOD applies to the test, while actual
reported level applies to the sample.
Because dPCR essentially counts individual molecules, when
we know that the FPR is sufficiently low, we may reliably call sample
levels below formal LOD. Consider a test with FPR less than 0.05
per test and, thus, with 95% LOB of zero target copies. If we detect
one or two targets, we are 95% confident that these are real and can
be reported with such confidence, even though this is below the
95% LOD of three copies per test. Operating at very low FPR is
always desirable from a performance guarantee perspective and
allows one to report confidently below formal test LOD levels.
15.6 Increasing Many of the limitations described above arise due to low levels of
the Measured Volume target molecule counts. If we subsample a small volume of the
whole, we will always be limited by the molecular stochasticity
processes. One way to counteract that is to increase the subsam-
pling volume so we do not operate so close to the fundamental limit
of three copies. In other words, increasing the expected number of
molecules to be detected will always help to improve test performance.
However, if preamplification is used, one has to watch for potential
biases and nonuniformity and accept the likely loss of absolute
quantification—a major benefit for dPCR.
16 Conclusions
Digital PCR technology and available commercial systems and

assays have reached the point where they are becoming mainstream
and preferred approaches when most precise and sensitive detection
of genomic targets is required. With increased availability of such
solutions, the field needs to understand the issues that are specific
to counting statistics so better tests can be developed, validated,
and put into practice for the benefit of science and patients alike.
While we cannot beat the fundamental limits due to molecular

stochasticity, dPCR technology allows us to operate at the best
possible mode.
Acknowledgments
Many thanks to my Bio-Rad colleagues George Karlin-Neumann,

Dianna Maar, Xitong Li, Lucas Frenz, and Francisco Bizouarn for
multiple suggestions on content and clarity of this chapter.
References
1. Sykes PJ, Neoh SH, Brisco MJ et al (1992) from the plasma of colorectal cancer patients.
Quantitation of targets for PCR by use of lim- Clin Chem 2013(59):1722–1731
iting dilution. BioTechniques 13(3):444–449 7. Madic J, Zocevic A, Senlis V et al (2016)
2. Vogelstein B, Kinzler KW (1999) Digital PCR. Three-color crystal digital PCR. Biomolecular
Proc Natl Acad Sci U S A 96(16):9236–9241 Detection and Quantification 2016
3. Baker M (2012) Digital PCR hits its stride. Nat (10):34–46. https://doi.org/10.1016/j.bdq.
Meth 9:541–544 2016.10.002
4. Bhat S, Herrmann J, Armishaw P et al (2009) 8. Dube S, Qin J, Ramakrishnan R (2008) Math-
Single molecule detection in nanofluidic digital ematical analysis of copy number variation in a
array enables accurate measurement of DNA DNA sample using digital PCR on a nanoflui-
copy number. Anal Bioanal Chem 394 dic device. PLoS One 3(8):e2876
(2):457–467 9. Whale AS, Huggett JF, Tzonev S (2016) Fun-
5. Hindson BJ, Ness KD, Masquelier DA et al damentals of multiplexing with digital PCR.
(2011) High-throughput droplet digital PCR Biomolecular Detection and Quantification.
system for absolute quantitation of DNA copy https://doi.org/10.1016/j.bdq.2016.05.002
number. Anal Chem 83(22):8604–8610. 10. CLSI (2012) Evaluation of detection capability
https://doi.org/10.1021/ac202028g for clinical laboratory measurement proce-
6. Taly V, Pekin D, Benhaim L, Kotsopoulos SK dures; approved guideline - second edition.
et al (2013) Multiplex picodroplet digital PCR CLSI document EP17-A2. Clinical and Labo-
to detect KRAS mutations in circulating DNA ratory Standards Institute, Wayne, PA
Chapter 4
Control Materials and Digital PCR Methods for Evaluation

of Circulating Cell-Free DNA Extractions from Plasma
Alexandra S. Whale, Ana Fernandez-Gonzalez, Alice Gutteridge,
and Alison S. Devonshire
Abstract
Cell-free DNA is an accessible source of genetic material found naturally in plasma that could be used in
many diagnostic applications. Translation of cfDNA analysis methods from research laboratories into the
clinic would benefit from controls for monitoring the efficiency of patient sample purification and for
quality control of the whole workflow from extraction through to analysis. Here we describe two types of
control materials that can be “spiked” into plasma samples to monitor and evaluate different aspects of the
workflow. The first control material is an internal control that enables evaluation of extraction efficiency,
fragment size bias, and sample inhibition. The second control material serves as a parallel quality control
material for measurement of specific genetic targets such as tumor mutations.
Key words Cell-free DNA, cfDNA, Calibration, Control, Digital PCR, Droplet digital PCR, dPCR,
ddPCR, Efficiency, Extraction, Plasma, Standardization
1 Introduction
Cell-free DNA (cfDNA) is non-cellular fragmented DNA that is

found naturally in body fluids such as the plasma. The vast majority
of cfDNA originates from apoptotic whole blood cells and other
normal living cells and is less than 200 bp in size [1–4]. However,
the shedding of cfDNA can also derive from other tissue compo-
nents such as fetus [5], tumor [6, 7], or graft organs [8, 9]. This
presents a tremendous opportunity for human health research and
diagnostic applications as it provides an accessible source of genetic
material that in the past may have been difficult to obtain. Genetic
alterations that can be detected include point mutations [10],
single nucleotide polymorphisms [11, 12] and copy number varia-
tions [13]. Furthermore, the methylation status of cfDNA can also
be measured and this has considerable potential for guiding treat-
ment decisions [14, 15].
45
46 Alexandra S. Whale et al.
Although cfDNA was first isolated from plasma 60 years ago

[16], several reports have highlighted a number of factors which
may be hindering the translation of cfDNA into a routine analyte
for clinical diagnostics [17–21]. These include a lack of appropriate
controls for pre-analytical stages and for standardization of the
whole process from extraction through to analysis, thereby making
it difficult to compare results from different studies [17]. Such
differences can be caused by choice of extraction and quantification
method, the reporting metric used or the storage condition of the
sample [18, 19]. Additional considerations include the assessment
of PCR inhibition and fragment size bias [20] – important con-
siderations for analysis of biological and clinical matrices given the
short fragment size of cfDNA. Furthermore, cfDNA has low abun-
dancy with concentrations in the region of 1.8 to 44 ng/mL plasma
reported [18]. This coupled with the fact that only a proportion of
the cfDNA originates from the fetus, tumor, or graft organ being
evaluated compounds the difficulty in the analytical measurement
and the inherent variation associated with low level detection [22].
In this chapter we describe two different types of control
material that are “spiked” into plasma samples before cfDNA
extraction to enable the investigator to monitor different aspects
of the analytical workflow:
1. Internal extraction control: DNA fragments of 115 bp,
461 bp, and 1448 bp generated by restriction digestion of
the “ADH” plasmid, which lacks homology to human and
mammalian sequences [20, 22, 23]. This control is suitable
for monitoring sample-specific extraction efficiency, and iden-
tifying fragment size bias effects and the presence of inhibitor
carryover from the extraction process (Fig. 1).
2. Whole process quality control material: fragmented human
genomic DNA (gDNA), to mimic cfDNA that is found in vivo,
with a specified abundance of the target genetic alteration to
enable whole process monitoring from extraction efficiency
through to analytical performance of the assay (Fig. 2).
To generate the data for this chapter, we used the QIAamp
Circulating Nucleic Acid kit (Qiagen P/N 55114) and have
included details specific to this kit. However, the above controls
enable evaluation of the performance of the extraction process and
so are not specific for a particular extraction method. A list of the
currently available commercial kits developed for, or commonly
applied to cfDNA extraction from plasma, can be found in
Table 1. Specific details on how to perform a dPCR experiment
have been omitted as these are covered in other chapters in this
book (Chaps. 2 and 3) and elsewhere [20, 24].
cfDNA Measurement by Digital PCR 47
Fig. 1 Workflow schematic for the internal extraction control. The fragmented ADH plasmid can be used to
assess two different aspects of the cfDNA extraction: an internal extraction control that is “spiked” into the
plasma sample to evaluate the extraction efficiency and fragment size bias (red workflow), and as an inhibition
control that is spiked into the reaction mix (blue workflow). The different reactions are shown in the dPCR box
with template volumes given as referenced in the main protocol in a total reaction volume of 20 μL
2 Materials
2.1 Internal 1. The restriction digested pSP64 poly(A) ADH plasmid (see
Extraction Control Note 1) containing various fragment lengths including
115 bp, 461 bp, and 1448 bp (Fig. 3a) (see Note 2) is available
upon request at two dilutions: approximately 2.9 103ADH
copies/μL and 2.9 104ADH copies/μL (see Note 3) in
non-human carrier (see Note 4). Store in 50 μL aliquots at
20 C to reduce freeze–thaw effects and vortex upon thawing
for 10 s to mix.
2. Oligonucleotides for ADH-115, ADH-461, and ADH-1448
assays (Table 2).
Fig. 2 Workflow schematic for the whole process quality control material. The whole process quality control
material can be diluted to different concentrations and then “spiked” into plasma; one dilution (S3) is shown in
the workflow as used to generate Fig. 5. The different reactions are shown in the dPCR box with template
volumes given as referenced in the main protocol in a total reaction volume of 20 μL
2.2 Whole Process 1. Human cfDNA Reference Standards at approximately 50 ng/μ

Quality Control L (Horizon Diagnostics) (see Note 5). The example described
Material in this chapter is gDNA heterozygous for KRAS G12C (allelic
frequency of 50% verified by dPCR) prepared by sonication to a
mean fragment size of ~160 bp (Fig. 3b). (Horizon Diagnos-
tics cfDNA Reference Standards currently available: Multiplex I
(8 targets) (P/N HD780) and BRAF V600E (P/N HD781)).
Table 1
Commercially available kits for cfDNA extraction from plasma
Manufacturer and part Purification Elution volume

Kit name number method Plasma volume (mL) (μL) Other matrices Citation
PME free-circulating DNA Extraction Analytik Jena Polymer beads 5 50 Serum, urine
Kit 845-IR-0003010
NextPrep-Mag™ cfDNA Bioo Scientific Magnetic beads <1 mL–3 mL 12–36 Serum
Isolation Kit
FitAmp™ Plasma/Serum DNA Epigentek Column 0.5 8–20 Serum and body fluids [20, 25–28]
Isolation Kit P-1004
EpiQuik Circulating Cell-Free Epigentek Magnetic beads 0.5 20 Serum

DNA Isolation Kit P-1064
Nucleospin XS kit Macherey Nagel Column 0.24–0.72 5–30 Serum, bronchial lavage [19, 20, 25, 29, 30]
740900
Plasma/Serum Cell-Free Norgen Biotek Corp. Column 1–4 25 Serum [30]

Circulating DNA Purification 55600
Plasma/Serum Circulating DNA Norgen Biotek Corp. Slurry 0.4–2 100 Serum [30–32]
Isolation 51200
Chemagic Circulating NA Kit special PerkinElmer Magnetic beads 1–4 60–100 [30, 33]
CMG-1096
Maxwell® RSC ccfDNA Plasma Kit Promega Magnetic beads 0.2–1 50

AS1480
QIAamp Circulating Nucleic Acid kit QIAGEN Column 1–5 20–150 Serum and urine [20, 25, 30, 34–36]
55114
QIAamp DNA Blood Mini kit QIAGEN Column 0.2 (single 50–200 Whole blood, body fluids [25, 36–39]
51104 loading)
MagMAX™ Cell-Free DNA Isolation Thermo Fisher Scientific Magnetic beads 0.5–10 15–50 Liquid samples (serum, plasma)
Kit
cfDNA Measurement by Digital PCR
Quick-cfDNA™ Serum & Plasma Kit Zymo Research Column 10 35–50 Serum, amniotic fluid, and
D4076 cerebrospinal fluid
As of March 2016, these were the commercially available kits; however, this list is unlikely to be exhaustive. At the time of compilation, new kits were being brought to market on a
49
regular basis and so may not be included in this table. For the same reason, at the time of this publication to our knowledge, not all of the kits had been cited in peer-review publications
Fig. 3 Example of electropherograms showing the fragment size of the extraction controls. Control samples
were analyzed using the 2100 Bioanalyzer with the DNA 1000 series II kit (Agilent). The lower and upper
markers are indicated as LM and UM, respectively. (a) Electropherogram of the internal extraction control
showing the peaks corresponding to the six fragments generated from the restriction digest of the linearized
ADH plasmid (67, 115, 461, 530, 1448, and 1889 bp). The corresponding ADH assays that target the different
fragment sizes are indicated with an asterix (*). (b) Electropherogram of the human cfDNA Reference
Standards fragmented to a mean size of 160 bp
2. Non-human carrier RNA. For the example given in this study,

yeast tRNA (Sigma-Aldrich P/N R3508 at 10 ng/mL)
was diluted in 1 TE to a final concentration of 5 ng/μL
(see Note 4).
3. 1 Tris–EDTA (TE), buffer pH 8.0, (molecular biology
grade), for example Fluka P/N 93283.
Table 2
ADH assay oligonucleotide sequences
ADH [Final
fragment [20 stock] reaction]
Assay name size Oligonucleotide sequence (50 -30 ) (μM) (μM)
ADH-115 115 bp F: GGGCCGAGCGCAGAA 18 0.9
R: ACTCTAGCTTCCCGGCAACA 18 0.9
P: HEX-TGGTCCTGCAACTTTATCCG 5 0.25
CCTCC-BHQ1
ADH-461 461 bp F: TTGAGAGTGTTGGAGAAGGAGTGA 18 0.9
(Adhβ) R: CGGTAAAGATCGGCAACACA 18 0.9
P: FAM-TCTTCAGCCAGGAGATC-MGB 4 0.2
ADH-1448 1448 bp F: TGAACCCGAAAGACCATGACA 18 0.9
(Adhδ) R: CCCACCATCCGTCATCTCA 18 0.9
P: FAM-CCAATTCAACAGGTGATC- 4 0.2
MGB
It is recommended that all primers and probes are HPLC-purified. For the example given in this study, the fluorophore is
given in the table; however, these probes have been published conjugated with a number of different fluorophore dyes
depending on the experimental set up [20, 22, 23]. Key: bp base pairs, F forward primer, R reverse primer, P probe, HEX
Hexachlorofluorescein, FAM 6-carboxyfluorescein, BHQ1 black hole quencher 1, MGB minor groove binder (Life
Technologies)
4. Specific assay to detect the genetic alteration present in whole

process quality control material. In this chapter, the Pri-
mePCR™ ddPCR™ Mutation Assays (Bio-Rad): KRAS WT
for p.G12C, Human (P/N 10031249, dHsaCP2000008) and
KRAS p.G12C, Human (P/N 10031246, dHsaCP2000007)
were used in a duplex reaction to quantify the KRAS G12C
mutation.
2.3 cfDNA Extraction 1. Blood samples collected in EDTA K2 vacutainers followed by

from Plasma plasma separation using standard methods (see Note 6). Pre-
pare 1.1 mL plasma aliquots and freeze at 80 C (see Note 7).
2. 0.5 mL and 1.5 mL microfuge tubes (see Note 8).
3. Vortex mixer.
4. Centrifuge and microfuge.
5. Extraction reagents or kit (Table 1). For the experimental
workflow illustrated in this book chapter, the QIAamp®
Circulating Nucleic Acid Kit (Qiagen P/N 55114) was used.
2.4 Digital PCR 1. Digital PCR system: the worked examples in this chapter were
Analysis performed using the QX200™ Droplet Digital™ PCR System
with the QuantaSoft™ analysis program (Bio-Rad). This pro-
tocol can be adapted for other digital PCR platforms.
2. Compatible mastermix for the selected digital PCR system:

ddPCR™ Supermix for Probes (Bio-Rad P/N 186-3010)
and ddPCR™ Supermix for Probes (No dUTP) (Bio-Rad
P/N 186-3024) were used to generate the results shown in
this chapter for the internal extraction control and whole pro-
cess quality control, respectively.
3. Nuclease-free water (molecular biology grade), for example,
Ambion™ (P/N AM9937).
3 Methods
3.1 Internal This protocol is designed for the assessment of extraction efficiency,
Extraction Control: presence of downstream inhibitors and fragment size bias. This
Fragmented ADH protocol is described for extraction of cfDNA from 1 mL plasma,
Plasmid however, the principles hold for smaller or larger volumes of
plasma. A schematic illustrating the work-flow and the different
uses for the internal extraction control is provided (Fig. 1). Spike-in
volumes can be scaled up or down according to the volume of
plasma processed.
3.1.1 Spiking 1. Thaw the required 1.1 mL plasma aliquots at room tempera-
and Extraction of Internal ture and maintain on ice thereafter (see Note 7).
Extraction Control from 2. Perform an additional centrifugation step at 1000 g for
Plasma 10 min at 4 C (see Note 9).
3. Remove 1 mL of plasma supernatant into a fresh tube for the
first stage of the extraction protocol (a 50 mL tube in the case
of the QIAamp® Circulating Nucleic Acid Kit).
4. Add 35 μL of the diluted fragmented ADH plasmid (2.9 103
copies/μL stock) per 1 mL plasma to generate the “spiked
plasma” sample (see Note 10). Vortex to mix. Ensure that
remaining plasmid is kept for dPCR analysis as the “spike
only” control.
5. Prepare a “plasma only” control for each plasma sample con-
taining 1 mL of plasma but no fragmented ADH plasmid (see
Note 11).
6. Follow preferred extraction method (examples given in
Table 1) according to the manufacturer’s specifications (see
Note 12).
7. Elute cfDNA in 35 μL elution buffer (see Note 10).
8. Store eluted cfDNA at 4 C for up to 1 month and 20 C for
longer term.
Table 3
Preparation of reactions for analysis with the ADH assays
ADH-461and ADH-115
duplex assay ADH-1448 assay
Component Unit [stock] 1 [work] Volume (μL) 1 [work] Volume (μL)

Nuclease-free water 3.00 4.00
ddPCR™ Supermix for probes 2 1 10.00 1 10.00
ADH-461 assay (FAM) 20 1 1.00 – –
ADH-1448 assay (FAM) 20 – – 1 1.00
ADH-115 assay (HEX) 20 1 1.00 – –
cfDNA extract 5.00 5.00
Total 20.00 20.00

For the setup, each ADH assay was prepared as a 20 stock and stored in aliquots at 20 C to reduce freeze–thaw
effects. ADH assays were conjugated to FAM or HEX fluorophore dyes as indicated
Table 4
Preparation of reactions for inhibition analysis with the ADH-461 assay
ADH-461 assay
Component Unit [stock] 1 [work] Volume (μL)

Nuclease-free water 3.50
ddPCR™ Supermix for probes 2 1 10.00
ADH-461 assay (FAM) 20 1 1.00
Fragmented ADH plasmid c/μL 2.9 10 4
0.50
cfDNA extract (plasmid only) 5.00
Total 20.00
3.1.2 Digital PCR 1. Prepare each reaction for dPCR as per manufacturer’s instruc-
Analysis of Internal tions using 5 μL of template per 20 μL reaction for each ADH
Extraction Control assay as shown in Table 3 (see Note 13).
2. Prepare triplicate reactions (see Note 14) for each sample
extract from “spiked plasma” and corresponding “plasma
only” control (if applicable).
3. Prepare additional triplicate reactions (see Note 16) for the
“spike only” control (no extraction performed) and
no-template controls (NTCs) using water in place of template.
4. To test for inhibitors carried over from the extraction process,
prepare additional reactions with the fragmented ADH plasmid
(2.9 104ADH copies/μL stock) added directly to the reac-
tion mix as shown in Table 4. For these reactions, only the
“plasma only” samples can be analyzed. Prepare additional

reactions for the “inhibition control” using water in place of
cfDNA template (will only contain fragmented ADH plasmid
from the reaction mix).
5. Partition the reaction as per manufacturer’s instructions.
6. Perform PCR using the following cycling conditions: Enzyme
activation for 10 min at 95 C, 40 cycles of denaturation for
30 s at 94 C and annealing/extension for 1 min at 60 C,
followed by signal stabilization for 10 min at 98 C and then
hold at 4 C. Follow manufacturer’s instructions for the
cycling, for example, using the QX200™ Droplet Digital™
PCR System requires all temperature ramping to be set at
2 C/s.
7. Analyze the cycled partitions and ensure the thresholds are set
correctly to separate the positive and negative partitions
(Figs. 4a–c) (see Note 15).
8. Check that the “plasma only” controls (if applicable) and
NTCs have no positive partitions (see Note 16).
9. Calculate the concentration (copies/μL) for each “spiked
plasma” and “spike only” samples for the three ADH assays
using the digital platform software or previously published
equations for digital PCR analysis [23, 40].
10. Calculate the extraction efficiencies for the different ADH
fragment sizes (see Notes 17 and 18) using the following
equation:
Calculation of extraction efficiency of the fragmented ADH
plasmid

ADH spiked plasma
Extraction efficiency ð%Þ ¼ ð1Þ
ADH spike only
11. Plot the extraction efficiencies of the three ADH fragments and
perform ANOVA statistical analysis to identify any fragment
size bias (Fig. 4d).
12. For the inhibition analysis, calculate the difference in concen-
tration of the “plasma only” samples using the reactions con-
taining the fragmented ADH plasmid, “plasma only inhibition
test” (Table 4) with the “inhibition control” using the follow-
ing equation:
Calculation of inhibition

ADH -461plasma only inhibition test
Relative inhibition ¼ ð2Þ
½ADH -461inhibition control
Fig. 4 Example of results obtained using the internal extraction control. Plasma was extracted using the
QIAamp® Circulating Nucleic Acid Kit. The QX200™ Droplet Digital™ PCR System with the QuantaSoft™
analysis program (Bio-Rad) was used to generate the data. (a–c) Examples of the QuantaSoft™ generated
one-dimensional scatter plots of the ADH assays for the “spiked” plasma and spike only samples. The x-axis
shows the partition count (event number) with the y-axis showing the fluorescent amplitude of the probe. The
horizontal pink line represents the threshold that defines the positive (above) and negative (below) partitions.
Positive partitions are either blue (FAM probe) or green (HEX probe) while negative partitions are grey. (d) Bar
graph showing the extraction efficiency of each assay as calculated with Eq. 1. The error bars represent the
standard deviation between triplicate dPCR measurements
3.2 Whole Process This protocol illustrates how the whole process from extraction
Quality Control through to dPCR analysis can be tested across a range of concen-
Material trations of the analyte (in the example, the analyte is the KRAS
G12C sequence found in plasma). A schematic illustrating the
work-flow for the whole process quality control material is provided
(Fig. 2). As with the protocol described in Sect. 3.1, this protocol is
written for extraction of cfDNA from 1 mL plasma, however, the
Table 5
Preparation of a dilution series of cfDNA Reference Standards from stock solution
[cfDNA Standard]
[cfDNA [cfDNA Standard] (KRAS G12C Sample Diluent Final
Dilution Standard] (KRAS G12C copies/mL Dilution volume volume volume
ID (ng/μL) copies/μL)a plasma)b factor (μL) (μL) (μL)
Stock 50 ~4250
S4 2.3 195 6840 21.3 25.0 507 432.3
S3 0.47 37 1306 5.0 100 400 400
S2 0.094 9 312 5.0 100 400 400
S1 0.019 2 69 5.0 100 400 500
Avoid using small volumes (<10 μL) in dilution series as precision of pipetting is lower for small volumes. Mix each
dilution by vortexing and brief centrifugation to pool contents before subsequent dilution. Sufficient diluted cfDNA
Reference Standard should be prepared for spiking into all plasma extractions and for the “spike only”dPCR analysis
a
As measured by dPCR. bBased on addition of 35 μL cfDNA Reference Standard/mL plasma
principles hold for smaller or larger volumes of plasma and spike-in

volumes can be scaled up or down according to the volume of
plasma processed.
3.2.1 Spiking 1. Prepare a dilution series of cfDNA Reference Standards in

and Extraction of cfDNA 5 ng/μL non-human carrier RNA. An example of a dilution
Reference Standards from series is given in Table 5 that is suitable for testing the linearity
Plasma of extraction efficiency and analytical sensitivity of cfDNA
extractions. Store the dilutions in 50 μL aliquots at 4 C for
short term (<1 month) and 20 C for longer term.
2. Thaw a 1.1 mL plasma aliquot at room temperature and main-
tain on ice thereafter (see Note 7).
3. Perform additional centrifugation step at 1000 g for 4 min at
4 C (see Note 8).
4. Remove 1 mL of plasma supernatant into a fresh tube for the
first stage of the extraction protocol (a 50 mL tube in the case
of the QIAamp® Circulating Nucleic Acid Kit).
5. Add 35 μL of the diluted cfDNA Reference Standard (Table 5)
to 1 mL plasma in a fresh 1.5 mL tube (see Note 10). Vortex to
mix and pulse down. Ensure that remaining cfDNA Reference
Standard dilutions are kept for analysis as the “cfDNA stan-
dard” control.
6. Prepare a “plasma only” control for each plasma sample
containing 1 mL of plasma but no cfDNA Reference Standard
(see Note 19).
Table 6
Preparation of reactions for analysis with the PrimePCR™ ddPCR™ Mutation Assays (Bio-Rad)
Component Unit [stock] 1 [working] Volume (μL)

ddPCR™ Supermix for probes (no dUTP) 2 1 10.00
PrimePCR KRAS p.G12C (FAM) 20 1 1.00
PrimePCR KRAS WT for p.G12C (HEX) 20 1 1.00
cfDNA extract 5.00
Total 20.00
“The PrimePCR assays were conjugated to FAM (6-carboxyfluorescein) or HEX (Hexachlorofluorescein)” fluorophore
dyes as indicated
7. Follow preferred extraction method (examples given in

Table 1) according to the manufacturer’s specifications (see
Note 12).
8. Elute cfDNA in 35 μL elution buffer (see Note 10).
9. Store eluted cfDNA at 4 C for up to 1 month and 20 C for
longer term.
3.2.2 Digital PCR 1. Prepare each reaction for dPCR as per manufacturer’s instruc-
Analysis of cfDNA tions using 5 μL of template per 20 μL reaction using the assay
Reference Standards to detect the genetic alteration present in human cfDNA Ref-
erence Standards. In the example, a duplex reaction containing
the assays for detecting the KRAS WT and KRAS G12C
sequence were used (Table 6).
2. Prepare triplicate reactions for each sample extract from
“spiked plasma” and the corresponding “plasma only” control
(see Note 14).
3. Prepare additional triplicate reactions to analyze the cfDNA
Reference Standard (“cfDNA standard”) dilutions
(no extraction performed) and no template controls (NTCs)
using water in place of template.
4. Partition the reaction as per manufacturer’s instructions.
5. Perform PCR using the cycling conditions recommended by
the manufacturer for the PrimePCR assay: Enzyme activation
for 10 min at 95 C, 40 cycles of denaturation for 30 s at 94 C
and annealing/extension for 1 min at 55 C, followed by signal
stabilization for 10 min at 98 C and then hold at 4 C. For all
stages the ramp rate is set at 2 C/s.
Fig. 5 Example of results obtained using the Human KRAS G12C cfDNA Reference Standard as a whole
process quality control material. Plasma was extracted using the QIAamp® Circulating Nucleic Acid Kit. The
QX200™ Droplet Digital™ PCR System with the QuantaSoft™ analysis program (Bio-Rad) was used to
generate the data. (a–d) Examples of the QuantaSoft™ generated two-dimensional scatter plots of the (a)
NTC, (b) “plasma only,” (c) “spiked” plasma (dilution S3 as detailed in Table 5) and (d) cfDNA Reference
Standard for the duplex KRAS G12C assay (also dilution S3). The fluorescent amplitude for each probe is
shown on the x-axis (WT) and y-axis (G12C). The vertical and horizontal pink lines represent the two thresholds
that define the positive and negative partitions. Positive partitions are either blue (G12C; FAM probe) or green
(WT; HEX probe) while negative partitions are grey. Double positive partitions, containing both G12C and WT
sequences are orange. (e–f) Analysis of the extraction of the four dilutions of the KRAS G12C cfDNA Reference
Standard (as given in Table 5). Six replicate extractions were performed for each dilution with a single dPCR
measurement. (e) Correlation analysis demonstrated good linearity down to log 2.5 (312 copies/mL plasma;
equivalent to 45 copies per dPCR). The limit of detection (LOD) for the whole process was approximately log
1.8 (69 copies/mL plasma; equivalent to nine copies per dPCR) with KRAS G12C undetectable in 1 of 6 extracts
analyzed. (f) Graph showing the extraction efficiency (bars) plotted on the left y-axis as calculated with Eq. 2
with the associated precision (red diamonds) plotted on the right y-axis
6. Analyze the cycled partitions and ensure the thresholds are set
correctly to separate the positive and negative partitions
(Figs. 5a–d) (see Note 15).
7. Check the NTCs have no positive partitions (Fig. 5a).
8. Check the “plasma only” controls for positive partitions that
target the analyte of interest. In the example, the “plasma only”
control is from a normal healthy blood donor and so should
have no positive partitions for the KRAS G12C point
mutation. The KRAS WT sequence will be present in the

endogenous cfDNA (see Notes 16 and 20) (Fig. 5b).
9. Calculate the concentration (copies/μL) for each “plasma
only,” “spiked plasma” (Fig. 5c), and “spike only” samples
(Fig. 5d) for the KRAS G12C and KRAS WT using the digital
platform software or previously published equations for digital
PCR analysis [23, 40]. The observed concentration of the
spiked KRAS G12C DNA can be expressed as copies per mL
plasma by multiplying the concentration (copies/μL) by the
elution volume (35 μL) (Fig. 5e).
10. Calculate the extraction efficiency using the following equa-
tions (see Note 20):
Calculation of extraction efficiency of mutant KRAS G12C
sequences

KRAS G12C spiked plasma
Extraction efficiency ð%Þ ¼ ð3Þ
½KRAS G12C cfDNA standard
Calculation of extraction efficiency of KRAS WT sequences

KRAS WT spiked plasma KRAS WT plasma only
Extraction efficiency ð%Þ ¼
½KRAS WT cfDNA standard
ð4Þ
11. Plot the extraction efficiencies of the plasma samples (Fig. 5f).
4 Notes
1. The pSP64 poly(A) ADH plasmid is a 4510 bp plasmid that

contains a portion of the Arabidopsis thaliana alcohol dehy-
drogenase (ADH) gene cloned into the pSP64 poly(A) plasmid
[22]. The cloned ADH sequence is nucleotides 774 to 2274 of
the GenBank sequence M12196 and encodes the “landsberg”
allele.
2. Fragmentation was achieved as described in Devonshire et al.,
2014 [20]. Briefly, 1.5 μg of pSP64 poly(A) ADH was linear-
ized with 10 units of BglI in NEBuffer 2 in a 50 μL reaction at
37 C for 2 h. Fragmentation of 0.75 μg (25 μL) of the
linearization reaction was performed with 10 units AlwNI,
2 units BsrDI and 5 μg BSA in NEBuffer 2 in a 50 μL reaction
at 37 C for 1 h followed by 65 C for 1 h and enzyme
inactivation at 80 C for 20 min. All restriction endonucleases
and buffers were purchased from NEB.
3. The concentration of the fragmented ADH plasmid is approxi-

mately 15 ng/μL as determined by fluorimetry (Qubit® 2.0
fluorometer with the dsDNA HS Assay Kit (Invitrogen™)).
1 ng of the fragmented ADH plasmid is approximately
2.015 108 copies.
4. Alternative non-human carrier molecules can be used, how-
ever, the investigator will need to verify that no cross-reactivity
exists between the analytical assay and carrier molecules. The
available fragmented ADH plasmid is diluted in sonicated
salmon sperm DNA (sssDNA) (Agilent P/N 201190 at
10 μg/μL) to a final concentration of 50 ng/μL.
5. A list of the currently available cfDNA Reference Standards can
be found at https://www.horizondiscovery.com/reference-
standards/our-formats/cfdna.
6. Plasma from normal healthy individuals in the described proto-
col was purchased from SeraLab Ltd. (UK), which was
prepared using standard procedures as follows. Whole blood
from normal healthy individuals was drawn into Vacutainer
tubes containing anticoagulants (EDTA) and mixed pursuant
to BD specifications and allowed to sit at room temperature for
30 min. Plasma was separated by refrigerated centrifugation at
1303 g for 20 min and then removed by aspiration, ensuring
no Red Blood Cell contamination, into a fresh tube prior to
shipment. Shipment should be on dry ice and plasma stored at
80 C upon arrival. When thawed, plasma should be a clear
liquid. A cloudy appearance indicates protein or cell lysis con-
tamination; such samples should be discarded.
7. Plasma is a biohazard as it may contain infectious agents. It is
advisable to handle all plasma samples in a Class II Biological
Safety Cabinet until lysis or proteinase treatment in buffers
containing chaotropic agents is complete. The risk of spillages
can be reduced whilst mixing by sealing tubes with Parafilm or
in double bags. Plasma should be homogenized by mixing on a
rotator at 4 C for 30 min prior to aliquoting.
8. Wherever possible it is recommended to use tubes suitable for
low concentration nucleic acid samples, for example Lo-bind®
tubes (Eppendorf).
9. A further centrifuge step at 1000 g for 10 min at 4 C to
remove any residual precipitated material or cellular debris can
also be performed [25].
10. For the experiment set up described in this chapter, 35 μL of
the diluted fragmented ADH plasmid was spiked into each
plasma sample extraction to give a total of 1 105 copies per
extraction (2.9 103 copies 35 μL ¼ 1 105 copies). This
protocol is developed so that the volume spiked into the plasma
is the same as the elution volume. This enables a more
straightforward calculation of the extraction efficiency and

recovery yield. Different spike and elution volumes can be
used, but these need to be accounted for in the calculations.
It is advisable to avoid using small volumes (<10 μL) as the
precision of pipetting is lower for small volumes.
11. “Plasma only” extractions are included as a control for cross-
contamination at the extraction stage. “Plasma only” extracts
can also be used for inhibition testing (Table 4).
12. Several of the example methods in Table 1 are readily scalable
and can extract cfDNA from 0.2 mL up to 10 mL plasma. This
protocol is developed using 1 mL plasma with 35 μL of internal
control material extracted using the QIAamp® Circulating
Nucleic Acid Kit and eluted 35 μL; the volume and concentra-
tion of the internal control material and elution volumes
should be scaled for larger/smaller volumes accordingly.
13. There are three ADH assays that target the different fragment
sizes of the fragmented ADH plasmid: ADH-115 (115 bp
fragment), ADH-461 (461 bp fragment), ADH-1448
(1448 bp fragment). To evaluate the fragment size bias of the
extraction method all three assays can be used in parallel. The
example set up described here shows a duplex reaction with the
ADH-115 and ADH-461 assays and a separate reaction with
the ADH-1448 assay. These assays can be used in duplex or
uniplex reactions with changes in the conjugated fluorophore,
however, investigators will need to validate any changes made
to the published assays.
14. It is preferable to prepare triplicate reactions as technical repli-
cates (set up three separate reactions with mastermix and tem-
plate) instead of preparing a triple volume reaction and
aliquoting into three separate reaction wells as this will capture
the technical error of the experiment set up in addition to the
instrument error.
15. It is important to set the thresholds that separate the positive
and negative partitions globally across all reactions for an assay.
Where possible, it is advisable to classify each droplet into a
category; in the given example, this was achieved by selecting
all wells and using the “cross-hair” tool in the QuantaSoft™
analysis program for the QX200™ Droplet Digital PCR
System.
16. The identification of positive partitions in the NTC or “plasma
only” control for an assay that targets a sequence unique to the
control material indicates either (1) nonspecificity of the assay
or (2) a background source of contamination. Both scenarios
should be investigated by comparison with NTC and “carrier
only” reactions to identify non-specificity or sources of con-
tamination. To determine if the non-specific amplification
originates from the endogenous cfDNA extracted from the

plasma samples, control reactions can be performed using
genomic DNA as a template; presence of positive partitions
indicates non-specific amplification of the ADH assays or assay
to detect the genetic alteration present in the human cfDNA
Reference Standards.
17. As the ADH assays target a sequence present in the fragmented
ADH plasmid, the recovery of the fragmented ADH plasmid
can be calculated directly without subtraction of the “plasma
only” control (see Note 15 to confirm specificity).
18. If the elution volume does not equal the spiked volume then a
modified version of Eq. 1 can be used:

Adh spiked plasma V e
Extraction efficiency ð%Þ ¼
Adh spike only V s
where Ve and Vs are the elution and spiked volumes,

respectively.
19. The presence of positive partitions for the KRAS G12C muta-
tion in the “plasma only” sample may indicate the presence of
the mutation in the endogenous cfDNA. This should be ver-
ified with further analysis of separate plasma extractions.
The false positive rate of the KRAS G12C assay should also
be characterized using gDNA which is 100% wild-type for the
mutation of interest [41], for the example KRAS Wild Type
Reference Standard (Horizon Discovery P/N HD710).
20. Selection of a target (KRAS G12C) that is only present in the
cfDNA Reference Standard allows calculation of the recovery
yield of the spiked DNA without subtraction of the “plasma
only” control. Selection of a target that is present in both the
plasma and cfDNA Reference Standard needs the “plasma
only” control concentration to be subtracted.
Acknowledgment
The work described in this chapter was in part funded by the UK

National Measurement System (NMS) and Innovate UK under the
project “Enabling stratified medicine through cell and tissue refer-
ence standards for minimally invasive cancer testing” (Project
Number: 101862) in partnership with Horizon Discovery. The
remaining funding was from the European Metrology Research
Programme (EMRP) joint research project [SIB54] “Bio-SITrace”
(http://biositrace.lgcgroup.com) which is jointly funded by the
EMRP participating countries within EURAMET and the
European Union. The authors would like to acknowledge
Dr. Karin Schmitt and Dr. Hadas Raveh-Amit at Horizon
Diagnostics for providing the Human cfDNA Reference Standard

material, Dr. Tim Forshew at UCL for support with dPCR testing
of the ADH internal extraction control, and Dr. George Karlin-
Neumann and Dr. Svilen Tzonev at Bio-Rad for their assistance in
the development of ddPCR experiments.
References
1. Stroun M, Lyautey J, Lederrey C, Olson- Hutchison JB, Laurent-Puig P (2013) Multi-
Sand A, Anker P (2001) About the possible plex Picodroplet digital PCR to detect KRAS
origin and mechanism of circulating DNA: mutations in circulating DNA from the plasma
apoptosis and active DNA release. Clinica Chi- of colorectal cancer patients. Clin Chem 59
mica Acta 313(1–2):139–142. https://doi. (12):1722–1731. https://doi.org/10.1373/
org/10.1016/S0009-8981(01)00665-9 clinchem.2013.206359
2. Lui YY, Chik KW, Chiu RW, Ho CY, Lam CW, 11. Beck J, Bierau S, Balzer S, Andag R, Kanzow P,
Lo YM (2002) Predominant hematopoietic Schmitz J, Gaedcke J, Moerer O, Slotta JE,
origin of cell-free DNA in plasma and serum Walson P, Kollmar O, Oellerich M, Schutz E
after sex-mismatched bone marrow transplan- (2013) Digital droplet PCR for rapid quantifi-
tation. Clin Chem 48(3):421–427 cation of donor DNA in the circulation of
3. Jahr S, Hentze H, Englisch S, Hardt D, Fack- transplant recipients as a potential universal
elmayer FO, Hesch RD, Knippers R (2001) biomarker of graft injury. Clin Chem 59
DNA fragments in the blood plasma of cancer (12):1732–1741. https://doi.org/10.1373/
patients: quantitations and evidence for their clinchem.2013.210328
origin from apoptotic and necrotic cells. Can- 12. Snyder TM, Khush KK, Valantine HA, Quake
cer Res 61(4):1659–1665 SR (2011) Universal noninvasive detection of
4. van der Vaart M, Pretorius PJ (2007) The ori- solid organ transplant rejection. Proc Natl
gin of circulating free DNA. Clin Chem 53 Acad Sci U S A 108(15):6229–6234. https://
(12):2215. https://doi.org/10.1373/ doi.org/10.1073/pnas.1013924108
clinchem.2007.092734 13. Shaw JA, Page K, Blighe K, Hava N,
5. Lo YM, Corbetta N, Chamberlain PF, Rai V, Guttery D, Ward B, Brown J,
Sargent IL, Redman CW, Wainscoat JS (1997) Ruangpratheep C, Stebbing J, Payne R,
Presence of fetal DNA in maternal plasma and Palmieri C, Cleator S, Walker RA, Coombes
serum. Lancet (London, England) 350 RC (2012) Genomic analysis of circulating
(9076):485–487. https://doi.org/10.1016/ cell-free DNA infers breast cancer dormancy.
s0140-6736(97)02174-0 Genome Res 22(2):220–231. https://doi.
6. Chen XQ, Stroun M, Magnenat JL, Nicod LP, org/10.1101/gr.123497.111
Kurt AM, Lyautey J, Lederrey C, Anker P 14. Potter NT, Hurban P, White MN, Whitlock
(1996) Microsatellite alterations in plasma KD, Lofton-Day CE, Tetzner R, Koenig T,
DNA of small cell lung cancer patients. Nat Quigley NB, Weiss G (2014) Validation of a
Med 2(9):1033–1035 real-time PCR–based qualitative assay for the
7. Nawroz H, Koch W, Anker P, Stroun M, detection of methylated SEPT9 DNA in
Sidransky D (1996) Microsatellite alterations human plasma. Clin Chem 60(9):1183–1191.
in serum DNA of head and neck cancer https://doi.org/10.1373/clinchem.2013.
patients. Nat Med 2(9):1035–1037 221044
8. Lo YM, Tein MS, Pang CC, Yeung CK, Tong 15. Barault L, Amatu A, Bleeker FE, Moutinho C,
KL, Hjelm NM (1998) Presence of donor- Falcomatà C, Fiano V, Cassingena A,
specific DNA in plasma of kidney and liver- Siravegna G, Milione M, Cassoni P, De
transplant recipients. Lancet (London, Eng- Braud F, Rudà R, Soffietti R, Venesio T,
land) 351(9112):1329–1330 Bardelli A, Wesseling P, de Witt Hamer P,
Pietrantonio F, Siena S, Esteller M, Sartore-
9. Lui YY, Woo KS, Wang AY, Yeung CK, Li PK, Bianchi A, Di Nicolantonio F (2015) Digital
Chau E, Ruygrok P, Lo YM (2003) Origin of PCR quantification of MGMT methylation
plasma cell-free DNA after solid organ trans- refines prediction of clinical benefit from alky-
plantation. Clin Chem 49(3):495–496 lating agents in glioblastoma and metastatic
10. Taly V, Pekin D, Benhaim L, Kotsopoulos SK, colorectal cancer. Ann Oncol. https://doi.
Le Corre D, Li X, Atochin I, Link DR, Griffiths org/10.1093/annonc/mdv272
AD, Pallier K, Blons H, Bouche O, Landi B,
16. Mandel P, Metais P (1948) [Not Available] Vandesompele J, Wittwer CT, Bustin SA
Comptes rendus des seances de la Societe de (2013) The digital MIQE guidelines: mini-
biologie et de ses filiales 142 (3–4):241–243 mum information for publication of quantita-
17. Ilie M, Hofman V, Long E, Bordone O, tive digital PCR experiments. Clin Chem 59
Selva E, Washetine K, Marquette CH, Hofman (6):892–902. https://doi.org/10.1373/
P (2014) Current challenges for detection of clinchem.2013.206375
circulating tumor cells and cell-free circulating 25. Page K, Guttery DS, Zahra N, Primrose L,
nucleic acids, and their characterization in Elshaw SR, Pringle JH, Blighe K, Marchese
non-small cell lung carcinoma patients. What SD, Hills A, Woodley L, Stebbing J, Coombes
is the best blood substrate for personalized RC, Shaw JA (2013) Influence of plasma pro-
medicine? Annals of Translational Medicine 2 cessing on recovery and analysis of circulating
(11):107 nucleic acids. PLoS One 8(10):e77963.
18. Fleischhacker M, Schmidt B (2007) Circulat- https://doi.org/10.1371/journal.pone.
ing nucleic acids (CNAs) and cancer--a survey. 0077963
Biochim Biophys Acta 1775(1):181–232. 26. Iba T, Miki T, Hashiguchi N, Tabe Y, Nagaoka
https://doi.org/10.1016/j.bbcan.2006.10. I (2014) Combination of antithrombin and
001 recombinant thrombomodulin modulates neu-
19. Fleischhacker M, Schmidt B, Weickmann S, trophil cell-death and decreases circulating
Fersching DM, Leszinski GS, Siegele B, Stot- DAMPs levels in endotoxemic rats. Thromb
zer OJ, Nagel D, Holdenrieder S (2011) Meth- Res 134(1):169–173. https://doi.org/10.
ods for isolation of cell-free plasma DNA 1016/j.thromres.2014.04.015
strongly affect DNA yield. Clinica chimica 27. Garcia-Gimenez JL, Sanchis-Gomar F,
acta; international journal of clinical chemistry Lippi G, Mena S, Ivars D, Gomez-Cabrera
412(23–24):2085–2088. https://doi.org/10. MC, Vina J, Pallardo FV (2012) Epigenetic
1016/j.cca.2011.07.011 biomarkers: a new perspective in laboratory
20. Devonshire AS, Whale AS, Gutteridge A, diagnostics. Clinica chimica acta; international
Jones G, Cowen S, Foy CA, Huggett JF journal of clinical chemistry 413
(2014) Towards standardisation of cell-free (19–20):1576–1582. https://doi.org/10.
DNA measurement in plasma: controls for 1016/j.cca.2012.05.021
extraction efficiency, fragment size bias and 28. Egert S, Boesch-Saadatmandi C, Wolffram S,
quantification. Anal Bioanal Chem 406 Rimbach G, Muller MJ (2010) Serum lipid and
(26):6499–6512. https://doi.org/10.1007/ blood pressure responses to quercetin vary in
s00216-014-7835-3 overweight patients by apolipoprotein E geno-
21. Oxnard GR, Paweletz CP, Kuang Y, Mach SL, type. J Nutr 140(2):278–284. https://doi.
O’Connell A, Messineo MM, Luke JJ, org/10.3945/jn.109.117655
Butaney M, Kirschmeier P, Jackman DM, 29. Kirsch C, Weickmann S, Schmidt B, Fleisch-
Janne PA (2014) Noninvasive detection of hacker M (2008) An improved method for
response and resistance in EGFR-mutant lung the isolation of free-circulating plasma DNA
cancer using quantitative next-generation gen- and cell-free DNA from other body fluids.
otyping of cell-free plasma DNA. Clinical can- Ann N Y Acad Sci 1137:135–139. https://
cer research : an official journal of the American doi.org/10.1196/annals.1448.035
Association for Cancer Research 20 30. Mauger F, Dulary C, Daviaud C, Deleuze JF,
(6):1698–1705. https://doi.org/10.1158/ Tost J (2015) Comprehensive evaluation of
1078-0432.ccr-13-2482 methods to isolate, quantify, and characterize
22. Sanders R, Huggett JF, Bushell CA, Cowen S, circulating cell-free DNA from small volumes
Scott DJ, Foy CA (2011) Evaluation of digital of plasma. Anal Bioanal Chem 407
PCR for absolute DNA quantification. Anal (22):6873–6878. https://doi.org/10.1007/
Chem 83(17):6474–6484. https://doi.org/ s00216-015-8846-4
10.1021/ac103230c 31. Keim SA, Kulkarni MM, McNamara K, Ger-
23. Whale AS, Cowen S, Foy CA, Huggett JF aghty SR, Billock RM, Ronau R, Hogan JS,
(2013) Methods for applying accurate digital Kwiek JJ (2015) Cow’s milk contamination of
PCR analysis on low copy DNA samples. PLoS human milk purchased via the internet. Pediat-
One 8(3):e58177. https://doi.org/10.1371/ rics. https://doi.org/10.1542/peds.2014-
journal.pone.0058177 3554
24. Huggett JF, Foy CA, Benes V, Emslie K, Gar- 32. Tritten L, O’Neill M, Nutting C, Wanji S,
son JA, Haynes R, Hellemans J, Kubista M, Njouendoui A, Fombad F, Kengne-Ouaffo J,
Mueller RD, Nolan T, Pfaffl MW, Shipley GL, Mackenzie C, Geary T (2014) Loa Loa and
Onchocerca ochengi miRNAs detected in host 37. Fong SL, Zhang JT, Lim CK, Eu KW, Liu Y
circulation. Mol Biochem Parasitol 198 (2009) Comparison of 7 methods for extract-
(1):14–17. https://doi.org/10.1016/j. ing cell-free DNA from serum samples of colo-
molbiopara.2014.11.001 rectal cancer patients. Clin Chem 55
33. How-Kit A, Tost J (2015) Pyrosequencing ® (3):587–589. https://doi.org/10.1373/
-based identification of low-frequency muta- clinchem.2008.110122
tions enriched through enhanced- ice 38. Keshavarz Z, Moezzi L, Ranjbaran R,
-COLD-PCR. In: Lehmann U, Tost J (eds) Aboualizadeh F, Behzad-Behbahani A,
Pyrosequencing: Methods and Protocols. Abdullahi M, Sharifzadeh S (2015) Evaluation
Springer, Berlin, Germany, pp 83–101 of a modified DNA extraction method for iso-
34. Dawson S-J, Tsui DWY, Murtaza M, Biggs H, lation of cell-free fetal DNA from maternal
Rueda OM, Chin S-F, Dunning MJ, Gale D, serum. Avicenna journal of medical biotech-
Forshew T, Mahler-Araujo B, Rajan S, nology 7(2):85–88
Humphray S, Becq J, Halsall D, Wallis M, 39. Jorgez CJ, Dang DD, Simpson JL, Lewis DE,
Bentley D, Caldas C, Rosenfeld N (2013) Bischoff FZ (2006) Quantity versus quality:
Analysis of circulating tumor DNA to monitor optimal methods for cell-free DNA isolation
metastatic breast cancer. New England Journal from plasma of pregnant women. Genetics in
of Medicine 368(13):1199–1209. https://doi. Medicine: Official Journal of the American
org/10.1056/NEJMoa1213261 College of Medical Genetics 8 (10):615-619.
35. Holmberg RC, Gindlesperger A, Stokes T, doi: 10.109701.gim.0000241904.32039.6f
Lopez D, Hyman L, Freed M, Belgrader P, 40. Dube S, Qin J, Ramakrishnan R (2008) Math-
Harvey J, Li Z (2013) Akonni TruTip((R)) ematical analysis of copy number variation in a
and Qiagen((R)) methods for extraction of DNA sample using digital PCR on a nanoflui-
fetal circulating DNA--evaluation by real-time dic device. PLoS One 3(8):e2876. https://doi.
and digital PCR. PLoS One 8(8):e73068. org/10.1371/journal.pone.0002876
https://doi.org/10.1371/journal.pone. 41. Milbury CA, Zhong Q, Lin J, Williams M,
0073068 Olson J, Link DR, Hutchison B (2014) Deter-
36. Breitbach S, Tug S, Helmig S, Zahn D, mining lower limits of detection of digital PCR
Kubiak T, Michal M, Gori T, Ehlert T, assays for cancer-related gene mutations. Bio-
Beiter T, Simon P (2014) Direct quantification molecular Detection and Quantification 1
of cell-free, circulating DNA from unpurified (1):8–22. https://doi.org/10.1016/j.bdq.
plasma. PLoS One 9(3):e87838. https://doi. 2014.08.001
Part II
Absolute Quantification
Chapter 5
Multiplex Droplet Digital PCR Protocols for Quantification

of GM Maize Events
David Dobnik, Bjørn Spilsberg, Alexandra Bogožalec Košir, Dejan Štebih,
Dany Morisset, Arne Holst-Jensen, and Jana Žel
Abstract
The standard-curve based simplex quantitative polymerase chain reaction (qPCR) has been the gold
standard for DNA target quantification for more than a decade. The large and growing number of
individual analyses needed to test for genetically modified organisms (GMOs) is reducing the cost-
effectiveness of qPCR. Droplet digital PCR (ddPCR) enables absolute quantification without standard
curves, avoids the amplification efficiency bias observed with qPCR, allows more accurate estimations at low
target copy numbers and, in combination with multiplexing, significantly improves cost efficiency. Here we
describe two protocols for multiplex quantification of GM maize events: (1) nondiscriminating, with
multiplex quantification of targets as a group (12 GM maize lines) and (2) discriminating, with multiplex
quantification of individual targets (events). The first enables the quantification of twelve European Union
authorized GM maize events as a group with only two assays, but does not permit determination of the
individual events present. The second protocol enables the quantification of four individual targets (three
GM events and one endogene) in a single reaction. Both protocols can be modified for quantification of any
other DNA target.
Key words Droplet digital PCR, ddPCR, GM maize, GMO, Multiplex, Quantification
1 Introduction
Standard-curve based quantitative PCR (qPCR) is regarded as the

gold standard technology for analysis of genetically modified
organisms (GMOs) due to its sensitivity and robustness. In the
last decade the number of GMOs found on a given market, and
for example authorized in the EU, has steadily increased. Conse-
quently, testing for GMOs must target a large number of sequence
motifs to ensure detection and quantification of all GMOs in con-
sideration can be detected and quantified. Analysis employing sep-
arate event-specific qPCRs is therefore no longer cost efficient. The
use of multiplex PCR can mitigate the problem of cost effective-
ness. However, a multiplex PCR is a complex experimental system
69
70 David Dobnik et al.
where potential interference between oligonucleotides and amplifi-

cation products can produce false positive/negative results. Several
multiplex qPCRs for quantification of GMOs have been developed
and reported, but only two (duplexes) are validated in an interla-
boratory Scheme [1, 2].
The concept of digital PCR (dPCR) was presented already in
1992 [3]. In dPCR the reaction mixture is distributed into a large
number of partitions and at end point of the reaction these parti-
tions are qualitatively scored as positive or negative, based on the
fluorescence signal. Initial DNA concentration is then calculated
using the Poisson distribution [4]. In contrast to qPCR, partial
inhibition of the PCR does not affect dPCR-based quantitation, as
it is only necessary to distinguish partitions positive for the target of
interest from those that are negative in order to calculate the sample
concentration of the target. Two available dPCR technologies,
microfluidic/chip based dPCR and emulsion (droplet) based
dPCR [5], have been tested for applicability for GMO detection
[6–10].
Droplet digital PCR (ddPCR) does not need standard curves
for absolute quantification, thus avoiding the amplification effi-
ciency bias observed with qPCR [7, 8]. Targets even at low copy
numbers can be accurately determined [11] and cost-efficiency can
be significantly improved in combination with multiplexing. The
protocols described below are focused on the ddPCR implementa-
tion, as this technology has already been shown to be suitable for
routine GMO diagnostics using duplex reactions [9] and uses high
number of partitions per well, thus enabling a wide linear range of
the response.
Two different protocols for multiplex quantification of GM
maize events are described below: (1) nondiscriminative multiplex
quantification of GM maize targets as a group, i.e., per ingredient
or plant species (multiplex quantification per ingredient, MQI) and
(2) discriminative multiplex quantification of individual GM events
(multiplex event quantification, MEQ).
In the MQI protocol, all 12 GM maize events authorized in the
EU by April 1, 2015 are quantified simultaneously as a group using
two assays per sample, without possibility of distinguishing individ-
ual targets present in the sample. This protocol is in line with the
EU labeling regulation [12], which states that labeling of a product
is not required, when the combined concentration of all EU
authorized GM events per individual plant species is below 0.9%
and the presence is unintended and technically unavoidable. The
two assays, 4- and 10-plex, also include, besides the 3 and 9 GM
events (see Tables 1 and 2), respectively, a maize endogene (hmgA)
as a target. The inclusion of hmgA is necessary for calculation of
GM maize content relative to all maize content in the sample. The
full performance of these multiplex assays has
Table 1
Primers and probes used in the MQI 4-plex assay
Final Volume of 40 μM
concentration in solution needed for
Target and method reference Name DNA sequence of the oligonucleotide (50 -sequence-30 ) the PCR [nmol/l] 72 μl of PP mix [μl]
hmgA, QT-TAX-ZM-002 Fw-hmgA TTGGACTAGAAATCTCGTGCTGA 900 5.40
R-hmgA GCTACATAGGGAGCCTTGTCCT 900 5.40
P-hmgA HEX-CAATCCACACAAACGCACGCGTA-BHQ-1 180 1.08
Bt11a, Brodmann et al. [13] Fw-Bt11 GCGGCTTATCTGTCTCAGGG 600 3.60
R-Bt11 CAACTGGTCTCCTCTCCGGA 600 3.60
P-Bt11 6-FAM-CGTGTTCCCTCGGATCTCGACATGT- 180 1.08
BHQ-1
NK603, QT-EVE-ZM-008 Fw-NK603 ATGAATGACCTCGAGTAAGCTTGTTAA 900 5.40
R-NK603 AAGAGATAACAGGATCCACTCAAACACT 900 5.40
P-NK603 6-FAM-TGGTACCACGCGACACACTTCCACTC- 180 1.08
BHQ-1
DAS59122, QT-EVE-ZM-012 Fw-DAS59122 GGGATAAGCAAGTAAAAGCGCTC 600 3.60
R-DAS59122 CCTTAATTCTCCGCTCATGATCAG 600 3.60
P-DAS59122 6-FAM-TTTAAACTGAAGGCGGGAAACGACAA 200 1.20
-BHQ-1
Method reference: reference for the method for quantification of event/gene. Methods starting with QT- were validated by the EURL-GMFF (http://gmo-crl.jrc.ec.europa.eu/
gmomethods/)
PP mix: a mix of forward primers, reverse primers, and probes, 6-FAM 6-carboxyfluorescein, BHQ1 black hole quencher 1, HEX hexachloro-6-carboxyfluorescein
a
Construct specific method
Multiplex Droplet Digital PCR Protocols for Quantification of GM Maize Events
71
Table 2
72
Primers and probes used in the MQI 10-plex assay
Volume of 40 μM
Final concentration solution needed for
Target and method reference Name DNA sequence of the oligonucleotide (50 -sequence-30 ) in the PCR [nmol/l] 120 μl of PP mix [μl]
hmgA, QT-TAX- Fw-hmgA TTGGACTAGAAATCTCGTGCTGA 900 9.0
ZM-002 R-hmgA GCTACATAGGGAGCCTTGTCCT 900 9.0
P-hmgA HEX-CAATCCACACAAACGCACGCGTA-BHQ-1 180 1.8
David Dobnik et al.
DAS1507, QT-EVE- Fw-DAS1507 TAGTCTTCGGCCAGAATGG 450 4.5

ZM-010 R-DAS1507 CTTTGCCAAGATCAAGCG 450 4.5
P-DAS1507 6-FAM-TAACTCAAGGCCCTCACTCCG-BHQ-1 90 0.9
GA21, QT-EVE- Fw-GA21 CGTTATGCTATTTGCAACTTTAGAACA 450 4.5
ZM-014 R-GA21 GCGATCCTCCTCGCGTT 450 4.5
P-GA21 6-FAM-TTTCTCAACAGCAGGTGGGT 90 0.9
CCGGGT-BHQ-1
MIR604, QT-EVE- Fw-MIR604 GCGCACGCAATTCAACAG 150 1.5
ZM-013 R-MIR604 GGTCATAACGTGACTCCCTTAATTCT 150 1.5
P-MIR604 6-FAM-AGGCGGGAAACGACAATCTGAT 90 0.9
CATG-BHQ-1
MIR162, QT-EVE- Fw-MIR162 GCGCGGTGTCATCTATGTTACTAG 150 1.5
ZM-022 R-MIR162 TGCCTTATCTGTTGCCTTCAGA 150 1.5
P-MIR162 6-FAM-TCTAGACAATTCAGTACATTAA 90 0.9
AAACGTCCGCCA-BHQ-1
MON810, QT-EVE- Fw-MON810 TCGAAGGACGAAGGACTCTAACGT 150 1.5
ZM-020 R-MON810 GCCACCTTCCTTTTCCACTATCTT 150 1.5
P-MON810 6-FAM-AACATCCTTTGCCATTGCCCAGC-BHQ-1 90 0.9
MON863, QT-EVE- Fw-MON863 TGTTACGGCCTAAATGCTGAACT 450 4.5
ZM-009 R-MON863 GTAGGATCGGAAAGCTTGGTAC 450 4.5
P-MON863 6-FAM-TGAACACCCATCCGAACAAG 90 0.9
TAGGGTCA-BHQ-1
MON88017, QT-EVE- Fw-MON88017 GAGCAGGACCTGCAGAAGCT 150 1.5
ZM-016 R-MON88017 GCCGGAGTTGACCATCCA 150 1.5
P-MON88017 6-FAM-TCCCGCCTTCAGTTTAAACAGAGTC 90 0.9
GGGT-BHQ-1
MON89034, QT-EVE- Fw-MON89034 TTCTCCATATTGACCATCATACTCATT 450 4.5
ZM-018 R-MON89034 CGGTATCTATAATACCGTGGTTTTTAAA 450 4.5
P-MON89034 6-FAM-ATCCCCGGAAATTATGTT-MGBNFQ 180 1.8
T25, QT-EVE- Fw-T25 ACAAGCGTGTCGTGCTCCAC 450 4.5
ZM-011 R-T25 GACATGATACTCCTTCCACCG 450 4.5
P-T25 6-FAM-TCATTGAGTCGTTCCGCCATTGTCG-BHQ- 90 0.9
1
Methods were validated by the EURL-GMFF (http://gmo-crl.jrc.ec.europa.eu/gmomethods/). Method reference: reference for the method for quantification of event/gene
PP mix a mix of forward primers, reverse primers, and probes, 6-FAM 6-carboxyfluorescein, BHQ1 black hole quencher 1, HEX hexachloro-6-carboxyfluorescein, MGB-NFQ
minor groove binding non-fluorescent quencher
73
been reported in Dobnik et al. (2015) [14]. Additionally a MQI

system for quantification of soybean GM lines, has been assessed by
[15].
In the MEQ protocol, in contrast, three GM maize events (see
Table 3) and a maize endogene (hmgA) are quantified in one single
reaction as distinguishable DNA targets (appearing in distinguish-
able droplet clusters). The performance of this multiplex was
recently reported [16]. Note that this multiplex assay targets differ-
ent GMO events than the MQI 4-plex.
A flowchart of experimental steps for both protocols is shown
in Fig. 1. Both of these protocols can be modified for quantification
of any other DNA target by designing new assays and after adjust-
ments of primer and probe concentrations as described in Subhead-
ing 3.
For the purpose of these protocols DNA isolated from different
kinds of samples (food, feed, seed, and plants) can be used. The
extraction technique and sample properties have a crucial influence
on the results of GMO detection and quantification [17]. Efficient
and reliable extraction of nucleic acids is a prerequisite for obtaining
accurate results from molecular analyses such as amplification of
specific DNA or RNA sequences by polymerase chain reaction
(PCR). The extraction method should yield sufficient DNA of
adequate structural integrity and purity, independently of the
matrix to which it is applied [18]. Acceptance criteria applicable
to DNA extraction methods were recently defined in the new
European Network of GMO Laboratories (ENGL) document on
Definition of minimum performance requirements for analytical
methods of GMO testing [19]. Briefly, concentration, yield, struc-
tural integrity, and purity are to be considered. Although these
requirements were prepared for real-time PCR, they can also all
be considered as minimum requirements for ddPCR. Regardless of
the method used for DNA extraction, it is essential to include
quality controls to exclude the possibility of sample contamination
during extraction. For this purpose the use of extraction blank
control and environment control is necessary. These controls are
later on used in PCR, together with positive and no template
control.
2 Materials
2.1 Droplet 1. Droplet Generation Oil for Probes (Bio-Rad, Pleasanton, CA).
Generation 2. Droplet Generator DG8™ Cartridges and Gaskets (Bio-Rad,
Pleasanton, CA).
3. 2 ddPCR™ Supermix for Probes (No dUTP) (Bio-Rad,
Pleasanton, CA).
4. Nuclease-free water (Sigma-Aldrich®, MO) or equivalent.
Table 3
Primers and probes used in MEQ assay
Final Volume of 40 μM
Target and method concentration in solution needed for
reference Name DNA sequence of the oligonucleotide (50 -sequence-30 ) the PCR [nmol/l] 120 μl of PP mix [μl]
hmgA, QT-TAX- Fw-hmgA TTGGACTAGAAATCTCGTGCTGA 300 3
ZM-002 R-hmgA GCTACATAGGGAGCCTTGTCCT 300 3
P-hmgA 6-FAM-CAATCCACACAAACGCACGCGTA-BHQ-1 100 1
MON810, QT-EVE- Fw-MON810 TCGAAGGACGAAGGACTCTAACGT 900 9
ZM-020 R-MON810 GCCACCTTCCTTTTCCACTATCTT 900 9
P-MON810 HEX-AACATCCTTTGCCATTGCCCAGC-BHQ-1 300 3
MON863, QT-EVE- Fw-MON863 TGTTACGGCCTAAATGCTGAACT 900 9
ZM-009 R-MON863 GTAGGATCGGAAAGCTTGGTAC 900 9
P-MON863 6-FAM-TGAACACCCATCCGAACAAGTAGGGTCA-BHQ-1 300 3
DP98140, QT-EVE- Fw-DP98140 GTGTGTATGTCTCTTTGCTTGGTCTT 300 3
ZM-021 R-DP98140 GATTGTCGTTTCCCGCCTTC 300 3
P-DP98140 HEX- CTCTATCGATCCCCCTCTTTGATAGTTTAAACT 200 2
-BHQ-1
Nuclease-free water 63
Method reference: reference for the method for quantification of event/gene. Methods were validated by the EURL-GMFF (http://gmo-crl.jrc.ec.europa.eu/gmomethods/)
PP mix a mix of forward primers, reverse primers, and probes, 6-FAM 6-carboxyfluorescein, BHQ1 black hole quencher 1, HEX hexachloro-6-carboxyfluorescein
75
Fig. 1 Flowchart of experimental steps for both protocols. For both protocols the steps of droplet generation,
PCR, and droplet readout are the same, but the reaction mixture and analysis steps with final reporting of
results are different
5. Primers and probes for each of the two reactions of MQI

protocol are described in Table 1 (4-plex (4P) reaction) and
Table 2 (10-plex (10P) reaction).
6. Primers and probes for MEQ 4-plex protocol are described in
Table 3.
7. DG8™ Cartridge Holder (Bio-Rad, Pleasanton, CA).
8. QX100™/200™ Droplet Generator (Bio-Rad, Pleasanton,
CA).
2.2 PCR 1. Rainin presterilized BioClean LTS tips with filter (see Note 1).
2. Eppendorf™ 96-Well twin.tec™ PCR Plates, Semi-skirted
(Eppendorf, Germany) (see Note 2).
Multiplex Droplet Digital PCR Protocols for Quantification of GM Maize Events 77
3. Pierceable Foil Heat Seal (Bio-Rad, Pleasanton, CA) or

equivalent.
4. Heat Sealer (Eppendorf, Germany) or equivalent.
5. T100™ Thermal Cycler (Bio-Rad, Pleasanton, CA) or equiva-
lent (see Note 3).
2.3 Droplet Readout 1. Droplet Reader Oil (Bio-Rad, Pleasanton, CA).

2. QX100™/200™ Droplet Reader (Bio-Rad, Pleasanton, CA).
3 Methods
3.1 Sample Sampling methodology is not described in this protocol, as it is

Acquisition usually not performed by the laboratory personnel. However, to
get an insight into this process and the procedure of subsampling,
we propose reading of the following document: Guidelines for
sample preparation procedures in GMO analysis [20].
3.2 Sample Most of the samples, including seeds and feed, should be ground
Preparation with a mill enabling good particle size reduction (e.g., Retsch
ZM200 rotor mill). Samples with high fat content are to be cooled
with liquid nitrogen before grinding with a Retsch GM200 knife
mill. Soft samples, such as sausages or tofu, are to be homogenized
with a Bioreba HOMEX 6 homogenizer. For leaf samples, first a
small piece is cut from each individual leaf and all the pieces are then
combined and homogenized as one sample with a FASTprep
instrument (MP Biomedicals) using 15 ml tubes, a ceramic ball,
and quartz sand.
Different procedures can be applied for DNA extraction from
the homogenized sample, but for the purpose of this protocol the
following extraction methods can be used, based on the sample
type [21]:
l NucleoSpin® Food (Macherey-Nagel) kit, which should be used
as described by the manufacturer.
l The cetyltrimethylammonium bromide (CTAB) method with
RNase-A and proteinase-K solutions for removing RNA and
proteins from the sample should be performed as described in
ISO 21571, Annex A.3—Preparation of PCR-quality DNA
using the CTAB-based DNA extraction methods [22]. Because
the CTAB method uses chloroform, the relevant steps should be
performed in a fume hood.
l DNeasy Plant Mini Kit (Qiagen) should be used as described by
the manufacturer to extract DNA from plant leaves.
DNA prepared by following one of these protocols is usually of
such quality that it is suitable for further analysis with ddPCR.
Regardless of the protocol used for DNA extraction, two controls

must be used during the process: (1) the extraction blank control,
where water is used instead of a sample and (2) environment con-
trol, where a tube with a volume of water equal to elution volume of
samples is left opened during the DNA extraction [23].
3.3 Optimization The primers and probes combined into the described multiplex
of Multiplex Assays assays were originally designed for real-time PCR with one primer
pair and probe per event (event specific primer–probe combination,
ESPPC), but the combinations were optimized to achieve reliable
detection and quantification of the targets also in ddPCR.
The first step when combining different ESPPCs into multiplex
is the in silico analysis of interactions between the primers and
probes. For the purpose of the described protocols, this was per-
formed with the software Autodimer [24]. All primers and probes
were aligned with a local alignment algorithm with a score of þ1 for
match and a penalty of 1 for mismatch. The interactions with a
score 9 were considered as the ones potentially causing significant
interference. Thus the ESPPCs with such scores were separated.
Nevertheless, as not all interactions can be predicted by in silico
analysis, the performance of multiplex assays was first experimen-
tally evaluated on DNA samples containing single targets (e.g., in
MQI two of the assays were removed from one group, based on the
experimental results).
As the described multiplex assays were developed for GMO
quantification, their performance (sensitivity, specificity, accuracy,
trueness, . . .) have to be compliant with Minimum Performance
Requirements for Analytical Methods of GMO Testing [19]. For
that reason a thorough in-house validation of the new assay should
be performed (following reported guidelines [19]), including test-
ing the precision/repeatability on serial dilutions to determine
limits of quantification and detection [19].
Additionally, PCR cycling protocol can be optimized (e.g., by
using gradient PCR) to get the best possible separation of droplet
clusters and minimize the “rain” of droplets between clusters. The
term “rain” refers to the droplets with fluorescence amplitude
ranging from explicitly negative and explicitly positive. The definite
origin of rain is not clear, but it was speculated that it occurs due to
delayed onset of PCR, partial inhibition of individual droplets or
damaged positive or negative droplets, what affects the fluorescence
signal [25, 26]. However, for the described protocols the cycling
parameters were not optimized and were identical to those from the
simplex real-time PCR methods. Specific considerations for devel-
opments of new multiplexes falling in line with the two different
protocols described in this chapter are given below.
3.3.1 Nondiscriminative This kind of multiplexing which quantifies a group of targets with-
Multiplex out the possibility to discriminately identify them requires that the
assay is tested for interactions between primers and probes and their
potential influence on the performance (especially false negative or
false positive signals). Additionally, the concentrations of primers
and probes must be adjusted for some individual targets. The
performance of multiplex assay must be evaluated against simplex
reaction on the same DNA sample containing only one target.
Whenever the absolute difference between the results is 25%,
the concentrations of primers and probes must be adjusted, so
that results would fit in this limit.
Increasing the number of targets in the multiplex within one
fluorescent reporter is a major limitation. With the increased num-
ber of targets, the concentration of probes is also increasing, lead-
ing to the increase of the negative background fluorescence. When
the fluorescence amplitude of negative droplets is too high it
reaches the amplitude of positive droplets and the clusters can no
longer be separated. Therefore, to allow appropriate discrimination
between negative and positive droplets the number of individual
assays that can be combined into multiplex is limited. The described
process of optimization can be done for both fluorescence channels
and finally only one group of targets per channel can be quantified
simultaneously.
3.3.2 Discriminative Bio-Rad’s ddPCR devices (QX100 or QX200) enable detection in

Multiplex two fluorescence channels, what makes them ready for at least
duplex quantification (one target in each channel). Because the
fluorescence amplitude measured in negative and positive droplets
after amplification depends on the probe concentration, the multi-
plexing level of ddPCR can be higher. For the MEQ protocol
described below, we have exploited this characteristic and prepared
a quadruplex reaction, where two targets are quantified within one
fluorescence channel. It is at most important, in this kind of multi-
plexing, to optimize primer and probe concentrations to achieve a
clear separation of droplet clusters. With that in mind, assay devel-
opers should test different concentration combinations (e.g., low
probe concentration for one target and high probe concentration
for second target in FAM channel). In the case of low concentra-
tions, the cluster of positive droplets should still be clearly above
the cluster of negative droplets. Examples of different fluorescence
amplitudes when using high or low probe concentrations are pre-
sented in Fig. 2. Note that when optimizing, low concentration
conditions might give amplitude of positive droplets at the same
level as is the cluster of negative droplets in high concentration
conditions (Fig. 2). However, when combining these two together,
the negative and two positive clusters would still be separated, as
the fluorescence amplitude will be summed up.
Fig. 2 Example of the effect of different probe concentrations on the fluorescence amplitude when developing
the MEQ assays. Simplex reactions for hmgA and MON863 (both in FAM channel) were performed on two
dilutions of DNA using high and low probe concentrations and readout is visualized in 1D (a). Both simplex
assays were combined into duplex (both in FAM channel) where hmgA had low probe concentration and
MON863 had high probe concentration (b). Clear separation of negative and different positive clusters can be
observed in 1D view. The same cluster separation can be observed also in 2D view (c)
3.4 Preparation Primer and probe stock solutions should be kept at a final concen-
of Reaction Mixtures tration of 100 μM, stored at 20 C. To reduce freezing–thawing
cycles, smaller volumes (for one time use) of 40 μM aliquots of
working solution should be prepared (diluted in nuclease-free
water). Such aliquots can be stored for 3 months at 20 C.
1. To prepare a primers–probe (PP) mix for MQI, for each of the
assays (4P and 10P) mix the specified volumes of 40 μM work-
ing solutions of individual primers and probes according to the
Tables 1 and 2. For MEQ, prepare a PP mix by mixing the
specified volumes of 40 μM working solutions of individual
primers and probes according to the Table 3 (see Note 4).
2. To prepare the reaction mixture for MQI for each of the assays
(4P and 10P), mix ddPCR™ Supermix for Probes and 4P PP
mix or 10P PP mix, respectively. For MEQ protocol mix
ddPCR™ Supermix for Probes with MEQ PP mix. For one
reaction prepare 17.6 μl of Mastermix composed of 11 μl
ddPCR™ Supermix for Probes and 6.6 μl PP mix (see Note 5).
3. Distribute 17.6 μl of Mastermix into the wells of nuclease-free
8-well strips or nuclease-free 96-well plates (see Note 6).
4. To get final reaction mixture, add 4.4 μl of sample DNA to each
well with Mastermix (see Note 7). It is essential to include
controls in the experiment, and therefore, instead of sample
DNA, a prepared mixture of DNA containing all the targets
(as positive control), nuclease-free water (as no template con-
trol, NTC), a sample of extraction blank control, and
environment control are added to individual wells with Mas-
termix (see Note 8).
5. Briefly mix by pipetting up and down (at least five times) or
vortexing and centrifuge to collect liquid at the bottom of
wells/tubes prior to loading the cartridge (see Note 9).
3.5 Droplet 1. Insert 8-well cartridge for droplet generation into the cartridge
Generation holder.
2. Pipette 20 μl of reaction mixture in each of the middle wells (see
Note 10).
3. Pipette 70 μl of Droplet Generation Oil in each of the lower
wells.
4. Place the gasket over the cartridge holder.
5. Insert the cartridge holder with filled cartridge into the
QX100/200™ droplet generator.
6. After the droplets are generated, carefully transfer 40 μl of
droplet suspension (water-in-oil emulsion) from a cartridge to
a 96-well plate (see Note 11).
7. Cover the plate with Pierceable Heat Seal Foil, seal the plate by
using Heat Sealer and transfer it to a thermal cycler (see Note 12).
3.6 PCR Reaction 1. Insert the sealed 96-well plate into a conventional thermal
cycler (see Note 3).
2. Set the cycling program to:
l 2 min at 50 C.
l 10 min at 95 C.
l 40 cycles of a two-step thermal profile comprising of 15 s at
95 C and 60 s at 60 C (at ramp rate 2.5 C/s).
l 10 min at 98 C (see Note 13).
l Infinite hold at 4 C.
3.7 Droplet Reading 1. When thermal cycling is completed transfer the sealed 96-well
plate into the QX100/200™ Droplet Reader.
2. Start the QuantaSoft software (Bio-Rad).
3. First flush the system by performing the following operation: in
Quantasoft software under “Setup” and “Instrument Rou-
tines” select “Flush System”.
4. Set up new experiment by using “Setup,” “Template,” and
“New,” and then use well-editor to complete the required
information for each well (see Note 14).
5. Save the created experiment.
6. Select “Run” to start the analysis.
7. When prompted, choose appropriate Color compensation
option (FAM/HEX).
8. After the analysis has finished save experiment data and remove
the plate from the machine.
3.8 Data Analysis 1. To access the droplets analysis results in QuantaSoft, use
for Nondiscriminating “Setup,” “Plate Load,” select and open the saved experiment
MQI Testing and select “Analyze.”
2. To distinguish between positive and negative droplets, set the
threshold for each well, for each detector separately. Use NTC
and positive control results as guidance for threshold setting
(see example in Figs. 3 and 4) (see Note 15). An example of
droplet readout on a dilution series of DNA mix is presented in
Fig. 5.
3. Export the results as .csv file, which can be later on opened
and further analyzed in Excel (Microsoft Office) or equivalent
software.
4. Wells with accepted number of droplets <8000 should not be
considered for further analysis (see Note 16)
5. The reaction in each well is considered as positive, if three or
more droplets show the fluorescent signal above the threshold
value (see Note 17).
Fig. 3 Example of a threshold setting for 4P MQI with NTC, positive control, and different samples. Ch1 (a)
represents the fluorescence signal originating from FAM, and Ch2 (b) represents the fluorescence signal
originating from HEX. a and b represent the 1D view of analyzed droplets for transgene and hmgA targets,
respectively. Two replicates for each of the following samples are shown: NTC (C06-D06), positive (þ) control
6. The reaction in each well is considered as negative, if <3

droplets show the fluorescent signal above the threshold value.
7. The target sequence is considered as detected, if for the 4P or
10P assay the reaction is positive, the NTC is negative, and the
positive amplification control (positive DNA target control) is
positive.
8. The target sequence is considered as not detected, if for the 4P
and 10P assay the reaction is negative, the NTC is negative, and
the positive amplification control (positive DNA target con-
trol) is positive.
9. The analysis must be repeated, if the positive amplification
control reaction is negative or the negative controls are
positive.
10. For calculation of GM maize content use the concentration
values (copies/μl of reaction) converted to a concentration of
undiluted DNA (the dilution of the sample is used for this
calculation).
11. Percentage of GM maize in each sample (relative to all maize
content) is calculated using the following equation, where
average is calculated from concentration values of undiluted
DNA (see Table 4 for example of calculation):
average transgene concentration of 4P replicates þ average transgene concentration of 10P replicates
GM maize content½% ¼
average concentration of hmgA for replicates of both assays ð4P and 10PÞ
100
3.9 Data Analysis 1. To access the droplets analysis results in QuantaSoft, use
for Discriminating MEQ “Setup,” “Plate Load,” select and open the saved experiment
Testing and select “Analyze.”
2. QuantaSoft software does not enable setting of more than one
threshold for a given fluorophore, but there are several clusters
that need to be separated (see Fig. 6). Thus two separate
exports must be performed for each reaction of MEQ protocol
(see Note 18 and Fig. 7). Use positive control results as guid-
ance for setting the thresholds. The appropriate clusters for
Fig. 3 (continued) for transgenes (A05-B05), two unknown samples containing GM maize at different
concentrations (A01-B01 and C03-D03), non-GM maize (A09-B09), GM rapeseed (A07-B07), and GM rice
(C08-D08). The two clusters that appear in the FAM channel originate from different targets, as the primer and
probe concentration of some is higher than of the other (but as there are three targets altogether and only two
clusters, they cannot be discriminated. The positive droplets in NTC and non-GM maize sample (in FAM
channel) do not make a reaction positive, because the number of such droplets per sample is less than three,
and thus reactions are considered negative. The threshold should be set just above the cluster of negative
droplets at fluorescence amplitude around 6400 and 2050 in a and b, respectively (note that these values may
vary between the experiments and individual wells). c represents the 2D overlay view of analyzed droplets of
all samples shown in a and b. Both thresholds in c are dividing the clusters in four groups: negative droplets
(bottom left), droplets positive for transgenes (top left), droplets positive for hmgA (bottom right), and droplets
positive for both the transgenes and hmgA (top right)
Fig. 4 Example of a threshold setting for 10P MQI with NTC, positive control, and different samples. Ch1 (a)
represents the fluorescence signal originating from FAM, and Ch2 (b) represents the fluorescence signal
originating from HEX. a and b represent the 1D view of analyzed droplets for transgene and hmgA targets,
respectively. Two replicates for each of the following samples are shown: NTC (G06-H06), positive (þ) control
for transgenes (E05-F05), two unknown samples containing GM maize at different concentrations (E01-F01
and G03-H03), non-GM maize (E09-F09), GM rapeseed (E07-F07), and GM rice (G08-H08). The positive
droplet in GM rapeseed sample (in HEX channel) does not make a reaction positive, because the number of
such droplets per sample is less than three, and thus reactions are considered negative. Due to the close
position of positive and negative clusters with some rain between, the threshold for these reactions was set
with lasso tool in 2D view, always comparing NTC, positive control and sample of interest (c). In c cluster of
negative droplets is marked with “a,” cluster of droplets positive for transgenes with “b,” cluster of droplets
positive for hmgA with “c” and cluster of droplets positive for transgenes and hmgA with “d”
Fig. 4 (continued)
Fig. 5 Example of MQI results on serial dilution of DNA sample positive for 12 GM maize lines. The droplet
readouts in 1D view for 10P and 4P assay are shown on left and right panel, respectively. Dilution factor for
DNA sample used in individual reactions is marked above each lane. Ch1 amplitude represents the response in
FAM for the group of transgene targets and Ch2 amplitude represents the response in HEX for the hmgA target.
For 10P analysis the negative and positive droplet clusters were separated using the lasso tool, and therefore
no threshold is visible on the readout (lower cluster of droplets represents the negative droplets). For 4P
analysis a single threshold was set in each of the fluorescence channels to separate the positive and negative
droplet clusters (positive are above the line)
each of the target must be selected with lasso tool to get groups
as presented in Fig. 7.
3. After selecting targets in one channel, the export of the results
in .csv file must be made. Same process is then repeated for the
other two targets in the second channel (see Note 19).
4. From each export we get the result (concentrations in target
copies per μl of reaction) for two of the targets:
l hmgA concentration ¼ concentration from Ch1 export 1.
l MON863 concentration ¼ concentration from Ch2 export 1.
l DP98140 concentration ¼ concentration from Ch1 export 2.
l MON810concentration ¼ concentration from Ch1 export 2.
5. For final percentage content of each of the three GM event
targets (listed in Table 3) in each sample the following equation
is used, where average is calculated from concentration values
of all replicates per undiluted DNA (see Table 4 for example of
calculation):
average concentration of individual event
GMcontent½% ¼
averageconcentration of hmgA
100
88
Table 4
Examples of calculation of GM content using MQI and MEQ protocols using the formulas provided in the protocol
David Dobnik et al.
Concentration Concentration of undiluted Average

Well Sample Assay Target (cp/μl)a sample (cp/μl) concentration GM content (%)
B01 DNA sample 1 MEQ hmgA 1013 1013
B04 DNA sample 5 MEQ hmgA 196 980 1005
B01 DNA sample 1 MEQ MON810 89 89
B02 DNA sample 1 MEQ MON810 96 96
B03 DNA sample 5 MEQ MON810 16 80 ¼ 90/1005 100
B04 DNA sample 5 MEQ MON810 19 95 90 ¼ 8.96
D01 DNA sample 1 4P MQI hmgA 971 971
D08 DNA sample 5 10P MQI hmgA 171 855 936
D01 DNA sample 1 4P MQI 3 GM maize lines 50 50
D03 DNA sample 5 4P MQI 3 GM maize lines 9.6 48
D04 DNA sample 5 4P MQI 3 GM maize lines 8 40 45
D07 DNA sample 5 10P MQI 9 GM maize lines 7 35 ¼ (45 þ 28)/
936 100
D08 DNA sample 5 10P MQI 9 GM maize lines 6 30 28 ¼7.80
a
Output from Quantasoft software
89
Fig. 6 Example of possible droplet clusters in MEQ in 2D view of analyzed droplets. The cluster in the left
bottom corner contains the fully negative droplets. All other clusters contain the droplets positive for one or
more targets. The combination of positive targets in each cluster is marked with letters A (hmgA), B (MON863),
C (DP98140), and D (MON810). Note that in some clusters there are no positive droplets, although all four
targets were present in the sample. To always have an idea of where the clusters should be positioned, a
positive control containing high concentration of all the targets must be used in each experiment
4 Notes
1. Use of the proposed tips is strongly recommended as others

can have rough tip openings and can lead to shedding of
microscopic plastic debris in the sample reaction, which can
cause shearing of the droplets. Thus, the use of Rainin tips is
also recommended for upstream sample preparation process
and ddPCR reaction setup.
2. The proposed Eppendorf™ 96-Well twin.tec™ PCR Plates
must be used, as the use of other plates can have significant
deleterious interactions with the droplets.
3. The cycler must enable setting of the ramp rate, which should
be set to 2.5 C/s, to ensure the correct temperature of each
droplet during the cycling.
4. The tables give instruction on minimum volume of PP mix that
needs to be prepared to avoid pipetting of small volumes.
Bigger volume of PP mix can be prepared, aliquoted in smaller
volumes (for one time use to avoid several freezing–thawing
cycles) and stored at 20 C. Aliquots can be used in
subsequent experiments for at least 3 months. Concentrations
of primers and probes in the mix were optimized in a way that
all targets are successfully amplified and that the cluster of
positive droplets is clearly separated from the cluster of negative
droplets (based on the fluorescence).
Fig. 7 Selecting the cluster groups in MEQ. After selection of cluster groups with lasso tool for the targets A
and B the results are exported (Export 1). For the second export the cluster groups of targets C and D are
selected (Export 2). See Note 19 for further explanations
5. To ensure proper loading of 20 μl of reaction mixture into the

cartridge without bubbles (which can cause subsequent shear-
ing of droplets), the proposed volume of final reaction mixture
is slightly larger (22 μl) than the volume loaded into the car-
tridge. Additionally, due to small pipetting losses, we also
recommend to prepare a 10% excess volume of Mastermix prior

to distributing to individual sample preparation wells.
6. The amount of Mastermix needed for the whole experiment
depends on the number of samples to be tested. For each
experiment we recommend that individual sample is tested in
two different dilutions (DNA concentrations of 50 and 10 ng/μ
l; 5 dilution between samples is recommended), each in dupli-
cate for each individual assay. For example, for MQI that would
mean one full cartridge (8 wells) for one sample as two assays
(4P and 10P must be run in parallel) and for MEQ two samples
can be tested per cartridge. For controls (see Note 8) we recom-
mend running them in duplicates per plate (altogether 8 wells).
7. It is important to evaluate the concentration, yield, structural
integrity, purity and amplifiability of the extracted DNA to be
used as a sample. This could be performed by spectrophotom-
etry, fluorimetry, capillary/gel-electrophoresis and/or by the
means of assessment of multiplication efficiency of a taxon-
specific target (with real-time PCR). DNA concentration
should be high enough to fulfill the requirements of the proto-
col (at least 50 ng/μl) with yield sufficient to perform all the
tests. Lower concentrations might still be used, but then there
is the possibility of false negative results. The minimum size of
the majority of DNA fragments should be larger than the size
of the amplicon produced by the PCR assay. However, less
fragmented DNA (large DNA fragments) is recommended
for these protocols to reduce the possibility of having breaks
within the amplicon. The presence of inhibitors, when
performing the DNA extraction and purification according to
the instructions, is not problematic for these protocols. When
checking the amplifiability of DNA with real-time PCR, the
slope of the serial dilution curve should be between 3.1 and
3.6. As stated in Note 6, for further ddPCR reactions we
recommend use of two different dilutions of the DNA sample.
8. Inclusion of appropriate controls is necessary for reliable
results. NTC, positive control, environmental control, and
extraction blank control need to be included in duplicates,
filling one whole cartridge. For MQI each DNA sample must
be tested with 4P and 10P assay to enable quantification of all
GM maize events currently authorized in the EU. When a full
96-well plate is to be prepared, for ease of use, we suggest a
scheme of pipetting following the columns as one cartridge
contains eight wells corresponding to the number of wells per
96-well plate column.
9. Ensure that the sample DNA and the Mastermix were mixed
adequately, as the Supermixes are fairly viscous and inadequate
mixing will lead to less reproducible results. When using vor-
texing, use lower speed as vigorous shaking might cause bubble
formation, which can reduce the available volume of the reac-

tion mixture that can be reliably pipetted in the following step.
10. It is essential to avoid bubble formation at the bottom of the
cartridge wells, as they will disturb the droplet formation. It is
also critical to load sample wells first and only then continue
with loading of Droplet Generation Oil for Probes. We recom-
mend the use of special 20 μl aerosol-barrier (filtered) Rainin
tips and pipetting at around 15 angle to the well wall. In case
all wells of the cartridge are not filled with samples, special
buffer must be used (ddPCR™ Buffer Control Kit [Bio-Rad,
Pleasanton, CA]) in the otherwise empty cartridge wells
(or none of your sample wells will form droplets).
11. We suggest using the cooling block as the plate holder, to keep
the reaction mixture cold, if several cartridges are planned for
the experiment. Keeping the plate with droplets cold also
helps to prevent evaporation. Alternatively, this could be done
by covering the wells with caps (which need to be removed prior
to the sealing of the plate). When transferring the droplets from
cartridge to the PCR plate it is essential to use slow pipetting
speed to avoid shearing or coalescing of droplets. The use of
8-channel pipette is recommended. We suggest 5 s aspiration by
holding the pipette in an angle between 30 and 45 and press-
ing the tip near the lower edge of the well, but not pressing
against the bottom in such a way as to reduce the size of the tip
inlet or you will shred your droplets. Dispensing of droplets
should also take around five seconds by holding the pipette
vertically, touching the lower part of the well wall with the tip
(again avoiding pressing the pipette tip against a surface such
that the opening is reduced in size).
12. It is of great importance not to centrifuge the plate containing
the droplets. If using manual Heat sealer, be careful not to press
the sealer to the plate for more than 5 s, as the applied heat could
cause deformation of the plastic and evaporation could occur.
13. Final incubation at 98 C is essential for polymerase deactiva-
tion. If used, the plate does not need to be put into droplet
reader right after the cycling is completed, but can be stored up
to three days at 4 C.
14. The information important for the analysis is: Sample name,
Target name for Target 1 and 2 and Type for Target 1 and 2.
15. There is an option of 1D (individual channel) or 2D (both
channels simultaneously) threshold settings. QuantaSoft soft-
ware also enables automatic threshold detection. For the pur-
pose of this protocol we advise to use 2D chart, as sometimes
the clusters might be shifted and manual threshold setting with
lasso tool must be used. The threshold should be set manually
at the lowest amplitude that captures true negative cluster
droplets of highest fluorescence amplitude, as visualized using

both the fluorescence amplitude vs. event number and the
histogram of events vs. amplitude data streams. At the same
time, by referring to the positive control cluster amplitude and
spread, the threshold should not be drawn so high as to lose
true positive droplets (losing sensitivity), nor so low as to
produce false positive droplets from the negative droplet clus-
ter (losing specificity). Normally the threshold can be set for a
group of wells (with same assay) at the same time, but some-
times slight shifts of fluorescence amplitude can be observed,
and thus we recommend some caution when performing the
threshold setting for higher number of wells to avoid misiden-
tification of positive or negative droplets.
16. The machine is specified to produce up to 20,000 droplets.
However, the actual number recovered and read may vary
mostly dependent on droplet pipetting. We believe the read-
outs below 8000 copies might not be trusted.
17. Through several experiments, we have noticed random occur-
rence of positive droplets in no template controls (see Fig. 3a,
NTC lane and non-GM maize lane). However, the number of
positive droplets was never higher than two per well. Thus, we
use three positive droplets as a cutoff for scoring a well as
positive. At the opposite side, a reaction containing high target
concentration must produce at least four negative droplets, to
enable the calculations with Poisson statistics.
18. The total number of clusters of droplets with this protocol can
be up to 16, where clusters have different combinations of
targets (Fig. 6). Therefore, setting of one threshold per chan-
nel cannot give the result for concentration of individual tar-
gets. It is essential to use a positive control containing all four
targets to be able to determine the cluster groups for the
samples, which may not contain all of the targets. The value
of each threshold separating the cluster groups may vary
between the experiments, and therefore no fixed value can be
offered. The most optimal way to select the individual cluster
groups is by using lasso tool.
19. First export (Fig. 7, upper panel) covers all positive droplets in
FAM fluorescence channel (droplets positive for target A
(hmgA) and droplets positive for target B (MON863) and
droplets positive for both targets). Second export (Fig. 7,
lower panel) covers all positive droplets in HEX/VIC fluores-
cence channel (droplets positive for target C (DP98140) and
droplets positive for target C (MON810) and droplets positive
for both targets. Note that the result for targets A and C in the .
csv file are marked as Ch1 and targets B and D are marked as
Ch2. This does not reflect the actual fluorescent labeling, but is
merely a cause of the multiplex analysis procedure that is dif-

ferent and cannot be performed otherwise. Examples of drop-
let readout of two different samples, negative and positive
control are given in Fig. 8.
Fig. 8 Example of MEQ results with negative and positive control and two GM maize samples in 2D view.
Negative cluster that can be distinguished using negative control (a) is marked with (1) in all panels. Positive
control (b) helps determining the correct position of clusters in the unknown samples. In panel c the MON810
DNA was used and the positive clusters for MON810 target (in HEX channel) are marked with (2). Few droplets
in the clusters marked with (3) are false positive signal for DP98140, which is actually the rain effect between
negative and MON810 clusters. Cluster marked with (4) is considered negative for HEX channel, but represents
positive droplets for hmgA in FAM channel. In panel D the MON863 DNA was used and the positive clusters are
marked with (5), (6), and (7) for hmgA, MON863, and both together, respectively
Fig. 8 (continued)
Acknowledgment
This work was supported by the European Union under grant

agreement no. 613908, (project DECATHLON). Complementary
financial support was given by the Norwegian Research Council
and the Slovenian Research Agency (contract numbers P4-0165
and 1000-15-0105). Consumables for the research resulting in the
MEQ protocol were kindly provided by Bio-Rad.
References
1. Takabatake R, Koiwa T, Kasahara M, on genetically modified food and feed. Off J
Takashima K, Futo S, Minegishi Y, Eur Union L 268:0001–0023
Akiyama H, Teshima R, Oguchi T, Mano J 13. Brodmann PD, Ilg EC, Berthoud H, Herr-
et al (2011) Interlaboratory validation of quan- mann A (2002) Real-time quantitative poly-
titative duplex real-time PCR method for merase chain reaction methods for four
screening analysis of genetically modified genetically modified maize varieties and maize
maize. Food Hyg Saf Sci 52:265–269 DNA content in food. J AOAC Int
2. Waiblinger H-U, Ernst B, Anderson A, Pietsch 85:646–653
K (2007) Validation and collaborative study of 14. Dobnik D, Spilsberg B, Bogožalec Košir A,
a P35S and T-nos duplex real-time PCR Holst-Jensen A, Žel J (2015) Multiplex quan-
screening method to detect genetically modi- tification of 12 European Union authorized
fied organisms in food products. Eur Food Res genetically modified maize lines with droplet
Technol 226:1221–1228 digital polymerase chain reaction. Anal Chem
3. Sykes PJ, Neoh SH, Brisco MJ, Hughes E, 87:8218–8226
Condon J, Morley AA (1992) Quantitation of 15. Demšar T, Gruden K, Milavec M, Štebih D,
targets for PCR by use of limiting dilution. Žel J (2015) Extraction of DNA from different
BioTechniques 13:444–449 sample types – a practical approach for GMO
4. Pinheiro LB, Coleman VA, Hindson CM, testing. Acta Biol Slov 58:61–75
Herrmann J, Hindson BJ, Bhat S, Emslie KR 16. Dobnik D, Štebih D, Blejec A, Morisset D, Žel
(2012) Evaluation of a droplet digital polymer- J (2016) Multiplex quantification of four DNA
ase chain reaction format for DNA copy num- targets in one reaction with Bio-Rad droplet
ber quantification. Anal Chem 84:1003–1011 digital PCR system for GMO detection.
5. Baker M (2012) Digital PCR hits its stride. Nat https://doi.org/10.1038/srep35451
Methods 9:541–544 17. Cankar K, Štebih D, Dreo T, Žel J, Gruden K
6. Demeke T, Gr€a fenhan T, Holigroski M, (2006) Critical points of DNA quantification
Fernando U, Bamforth J, Lee S-J (2014) by real-time PCR—effects of DNA extraction
Assessment of droplet digital PCR for absolute method and sample matrix on quantification of
quantification of genetically engineered genetically modified organisms. BMC Biotech-
OXY235 canola and DP305423 soybean sam- nol 6:37
ples. Food Control 46:470–474 18. Codex Committee on Methods of Analysis and
7. Bhat S, Herrmann J, Armishaw P, Corbisier P, Sampling (2010) Guidelines on performance
Emslie KR (2009) Single molecule detection in criteria and validation of methods for detec-
nanofluidic digital array enables accurate mea- tion, identification and quantification of spe-
surement of DNA copy number. Anal Bioanal cific DNA sequences and specific proteins in
Chem 394:457–467 foods
8. Corbisier P, Bhat S, Partis L, Xie VRD, Emslie 19. EURL-GMFF (2015) Minimum performance
KR (2010) Absolute quantification of geneti- requirements for analytical methods of GMO
cally modified MON810 maize (Zea mays L.) testing
by digital polymerase chain reaction. Anal 20. Berben G, Charels D, Demšar T, Hochegger R,
Bioanal Chem 396:2143–2150 Nardini E, Onori R, Schulze M, Philipp P,
9. Morisset D, Štebih D, Milavec M, Gruden K, Weber T (2014) Guidelines for sample prepa-
Žel J (2013) Quantitative analysis of food and ration procedures in GMO analysis. https://
feed samples with droplet digital PCR. PLoS doi.org/10.2788/738570
One 8:e62583 21. Bogožalec Košir A, Spilsberg B, Holst-Jensen
10. Burns MJ, Burrell a M, Foy C a (2010) The A, Žel J, Dobnik D (2017) Development and
applicability of digital PCR for the assessment inter-laboratory assessment of droplet digital
of detection limits in GMO analysis. Eur Food PCR assays for multiplex quantification of 15
Res Technol 231:353–362 genetically modified soybean lines. Nature Sci
11. Whale AS, Cowen S, Foy C a, Huggett JF Rep 7:8601. https://doi.org/10.1038/
(2013) Methods for applying accurate digital s41598-017-09377-w
PCR analysis on low copy DNA samples. PLoS 22. International Organization for Standardization
One 8:e58177 (2005) ISO 21571:2005 foodstuffs – methods
12. European Commission (2003) Regulation of analysis for the detection of genetically mod-
(EC) no 1829/2003 of the European Parlia- ified organisms and derived products – nucleic
ment and of the Council of 22 September 2003
acid extraction ISO 21571. International 25. Dreo T, Pirc M, Ramšak Ž, Pavšič J, Milavec
Organization for Standardization, Geneva M, Žel J, Gruden K (2014) Optimising droplet
23. International Organization for Standardization digital PCR analysis approaches for detection
(2006) ISO 24276:2006 foodstuffs – methods and quantification of bacteria: a case study of
of analysis for the detection of genetically mod- fire blight and potato brown rot. Anal Bioanal
ified organisms and derived products – general Chem 406:6513–6528
requirements and definitions ISO 24276. 26. Jones M, Williams J, G€artner K, Phillips R,
International Organization for Standardiza- Hurst J, Frater J (2014) Low copy target detec-
tion, Geneva tion by Droplet Digital PCR through applica-
24. Vallone PM, Butler JM (2004) AutoDimer: a tion of a novel open access bioinformatic
screening tool for primer-dimer and hairpin pipeline, ‘definetherain’. J Virol Methods
structures. BioTechniques 37:226–231 202:46–53
Chapter 6
Using Droplet Digital PCR to Detect Coinfection of Human

Herpesviruses 6A and 6B (HHV-6A and HHV-6B) in Clinical
Samples
Ashley Vellucci, Emily C. Leibovitch, and Steven Jacobson
Abstract
Droplet digital™ polymerase chain reaction (ddPCR™) is a unique digital PCR technique that allows for
absolute quantification of nucleic acid samples. This technique operates on the basis of amplification within
water–oil emulsion droplets and can detect very small quantities of target molecules, yielding extremely
precise data. Here, we describe in detail a ddPCR procedure for multiplexed detection of two clinically
relevant herpesviruses, HHV-6A and HHV-6B.
Key words ddPCR, Droplet digital, Polymerase chain reaction, HHV-6A, HHV-6B, Human herpes-
virus 6, Viral detection, Absolute quantification
1 Introduction
Given the widespread use of PCR, there is an ever-growing drive for

innovation and improvement. While early generations of PCR
enabled qualitative and relative quantification, a third generation
technology, ddPCR, enables absolute quantification. Unlike first-
and second-generation PCR methods, ddPCR enables the calcula-
tion of absolute copy numbers independent of a standard curve or
other external calibrator [1–3]. Positive and negative droplet PCR
data are fit to a Poisson algorithm [4], resulting in highly precise
quantitative data of targeted nucleic acids. To provide a brief over-
view of the ddPCR technology, nucleic acid samples are first
digested with a restriction enzyme, and then combined with the
relevant primers, fluorescent probes, and other standard PCR com-
ponents. This mixture is emulsified with oil, resulting in its parti-
tioning over approximately 20,000 nanoliter-sized droplets. Each
droplet then undergoes amplification by thermal cycling and is
assessed for fluorescence level, indicating the internal fluorescence
99
100 Ashley Vellucci et al.
of the probe [2]. This creates positive and negative droplet popula-
tions with a readout of copies/μl.
Through the use of ddPCR, clinically relevant pathogens such
as HHV-6 can be easily and accurately detected in clinical samples
[1]. As HHV-6 is a ubiquitous virus and can be found in healthy
individuals as well as those with HHV-6 associated diseases, a
highly quantitative approach is required for the precise determina-
tion of HHV-6 viral loads in clinical samples. In this report, we
detail two multiplex ddPCR procedures for the detection of
HHV-6. A multiplex technique allows for the detection of multiple
viruses, while saving money, resources, and time. HHV-6 is becom-
ing an increasingly recognized pathogen in the clinical setting [5],
and there are now two recognized species of HHV6, HHV-6A and
HHV-6B. Both are associated with several CNS disorders, includ-
ing epilepsy [6] and encephalitis [7], and HHV-6B is the etiologic
agent of roseola, a childhood febrile illness [8]. HHV-6 reactivation
with or without a symptomatic manifestation is increasingly recog-
nized in the context of transplantation or other severely immuno-
suppressive regimens. HHV-6A and HHV-6B were recently
reclassified as distinct viral species [9] due to differences in disease
associations [8], tropism [10], and other biological and immuno-
logical properties [9]. These viruses may also respond differently to
antiviral therapy, and it is therefore important in a clinical setting to
accurately distinguish between HHV-6A and HHV-6B, particu-
larly as currently available serological assays cannot distinguish
between the two viral species [11].
The observation of elevated viral DNA levels in patients with
HHV-6 associated diseases compared to controls is the basis for
utilizing PCR for HHV-6 detection in clinical samples [8]. In
addition, in approximately 1% of the human population, HHV-6
is chromosomally integrated (ciHHV-6) in subtelomeric regions of
genomic DNA [5], a distinctive aspect of these viruses. It is
unknown whether there are clinical consequences associated with
this integration. Clinically, ciHHV-6 patients are often misdiag-
nosed with reactivated HHV-6, leading to inappropriate and
unnecessary treatment and side effects [12]. ddPCR is a reliable
tool for the rapid and accurate detection of ciHHV-6, particularly
from cell-rich sources such as peripheral blood mononuclear cells
(PBMC) [12, 13]. This protocol will detail the methodology for
using ddPCR to distinguish between HHV-6A and HHV-6B, with
a cellular housekeeping reference gene [2], to detect single or
coinfection of these viruses in clinical samples, including samples
from individuals with ciHHV-6 [8]. These primers and probes were
newly designed for this application, and their initial characteriza-
tion has been previously described [8].
Using Droplet Digital PCR to Detect Coinfection of Human. . . 101
2 Materials
2.1 DNA Extraction 1. Qiagen DNeasy Blood & Tissue Kit (for DNA extraction from
whole blood or isolated PBMC, tissues, CSF cells and saliva).
2. Qiagen QIAamp UltraSens Virus Kit (for DNA extraction from
CSF supernatant and serum).
3. NanoDrop 2000 UV-Vis spectrophotometer (Thermo
Scientific).
2.2 Digest Mix 1. Microcentrifuge tubes (0.5–1.5 ml) (see Note 1).
Components 2. HindIII restriction endonuclease; 20,000 U/ml (New Eng-
land Biolabs).
3. NEB 2.1 Buffer (New England Biolabs).
4. PCR water.
5. 37 C shaker.
6. DNA samples diluted to appropriate concentration (see Note 2).
7. Thermomixer.
2.3 Super Mix 1. Microcentrifuge tubes (0.5–1.5 ml) (see Note 1).
Components 2. 2 ddPCR Super Mix for Probes (Biorad).
3. 10 or 20 primer–probe mixes.
(a) 20 is defined as a final concentration of 900 nm primers
and 250 nm probe.
2.4 Droplet 1. QX200 Droplet Digital PCR (ddPCR) System (Bio-Rad).

Generation and Droplet (a) QX100 Droplet Generator.
Reader Components
(b) QX100 Droplet Reader.
2. PCR thermocycler.
3. 96-well plate (see Note 3).
4. Heat sealer with block for 96-well plate.
5. DG8 Single-Use Disposable Cartridges (Bio-Rad).
6. DG8 Gaskets (Bio-Rad).
7. Droplet Generation Oil for Probes (Bio-Rad).
8. ddPCR Buffer Control for Probes (Bio-Rad).
9. Reagent reservoir, 25 ml capacity.
10. ddPCR Droplet Reader Oil (Bio-Rad).
11. Low Retention Filter Tips, 200 μl (Rainin).
12. Low Retention Filter Tips, 20 μl (Rainin).
13. Single channel manual pipette LTS lite 2–20 μl (Rainin) (see
Note 4).
14. Eight channel manual pipette LTS 5–50 μl (Rainin) (see Note 5).
15. Eight channel manual pipette LTS 2–200 μl (Rainin)
(see Note 6).
16. Pierceable sealing foil sheets (Thermo Scientific).
3 Methods
All procedures are to be carried out at room temperature in a

biosafety cabinet and reagents should thaw on ice. Restriction
enzyme should be removed from 20 C immediately prior to
use, and returned to 20 C immediately after use. Fluorescent
probes are light sensitive, so lights in the hood should be turned off
after the DNA digestion step.
For duplexing, HHV-6A FAM 20 and HHV-6B VIC 20
are run in one well, while RPP30 VIC 20 is run in a separate well
(Fig. 1). For triplexing, all primer–probe mixes are run in the same
well, with different concentrations of primer–probe mixes on a
given channel, i.e., RPP30 VIC 10 and HHV-6B VIC 20
(Fig. 2) (see Notes 7 and 8).
We recommend running samples in duplicate, though triplicate
or quadruplicate wells may be used depending on the desired
sensitivity.
3.1 DNA Extraction 1. Extract DNA from the designated sample using the appropriate
Qiagen kit, in accordance with the manufacturer’s instructions.
2. Obtain DNA concentration using NanoDrop 2000 UV-Vis
spectrophotometer. The DNA concentration will vary with
the sample type. We typically obtain values of less than
10 ng/μl for CSF cells, serum, and saliva.
3.2 Reaction Setup 1. Calculate the amount of reagents the supermix and digest
mixtures will each contain (see Note 9).
(a) For one well the digest mix contains:
l 3.65 μl water.
l 1 μl 10 NEB Buffer.
l 0.1 μl BSA.
l 0.25 μl restriction endonuclease (20 U/μl).
(b) For one well the duplex supermix contains (Fig. 1):
l 12.5 μl 2 ddPCR Supermix for Probes (see Note 10).
l 1.25 μl 20 FAM primer–probe mixture.
l 1.25 μl 20 VIC primer–probe mixture.
(c) For one well the triplex supermix contains (Fig. 2).
Ch1+Ch2+:0 Ch1+Ch2_:22 Ch1_Ch2+:97 Ch1_Ch2_:33945
A
12000 1687
10000
8000
Channel 1 Amplitude
FAM
6000
4000
2468
2000
0
0 1000 2000 3000 4000 5000 6000 7000
Channel 2 Amplitude
VIC
B 6000 G01 H01
5000
4000
Ch2 Amplitude
VIC
3000
1978
2000
1000
0
0 5000 1000015000 20000 25000 30000 35000
Event Number
Fig. 1 HHV-6A and HHV-6B duplexing. (a) ddPCR 2D droplet plot generated using HHV-6A 20 and HHV-6B
20. HHV-6A is on the FAM channel and the positive droplet population is located in the upper left quadrant.
HHV-6B is on the VIC channel and the positive droplet population is located in the lower right quadrant.
(b) ddPCR 1D droplet plot generated using RPP30 20
l 11.25 μl 2 ddPCR Supermix for Probes (see

Note 10).
l 1.25 μl 20 FAM primer–probe mixture.
l 1.25 μl 20 VIC primer–probe mixture.
l 1.25 μl 10 VIC primer–probe mixture (see Note 11).
Ch1+Ch2+:4294 Ch1+Ch2+:1558 Ch1_Ch2+:14548 Ch1_Ch2_:5375

10000 1789
9000
8000
7000
Channel 1 Amplitude
6000
FAM
5000
4000
3000
2451
2000
1000
0
0 1000 2000 3000 4000 5000 6000 7000 8000
Channel 2 Amplitude
VIC
Fig. 2 HHV-6A, HHV-6B, and RPP30 triplexing. By differing the primer/probe concentrations, RPP30 (10) and
HHV-6B (20) can be used simultaneously on the VIC channel, with a clear separation between the VIC
positive droplet populations. Populations double positive for RPP30 and HHV-6A or RPP30 and HHV-6B can
also be appreciated. Single primer–probe sets should be used to confirm the RPP30/HHV-6A/HHV-6B
specificity of the unlabeled populations in the upper right quadrant
Table 1
Primer and probe sequences
Reverse primer Probe sequencea Probe Probe

0 0
Forward primer (5 –3 ) (50 –30 ) (50 –30 ) fluoreb quencher
HHV- CCGTGGGATCGTCT CCACACTAGTC CTGGAACT 6FAM MGBNFQ
6A7 AAAATTATAGATGT CGGACGGATAA GTATAATAGG
HHV- CCGTGGGATCGTCT CCACACTAGTC CTGGAGCT VIC MGBNFQ
6B7 AAAATTATAGATGT CGGACGGCTAA GTACAACAG
RPP302 GATTTGGACCTGCG GCGGCTGTCTC CTGACCT VIC MGBNFQ
AGCG CACAAGT GAAGGCTCT
a
See Note 20
b
See Note 8
2. Calculate the amounts of primers and probe for each 10 or

20 mixture. (Sequences are listed in Table 1).
(a) For 100 μl of a 20 primer–probe mixture (see Note 12).
l 18 μl primer 1 (100 μM).
l 5 μl probe (100 μM).
l QS with PCR water to 100 μl.
(b) For 100 μl of a 10 primer–probe mixture (see Notes 13).

l 2.5 μl probe (100 μM).
l QS with PCR water to 100 μl.
3.3 Digesting Before beginning, turn on plate sealer and PCR thermocycler and
the Samples set the thermomixer to 37 C.
1. Dilute DNA samples as needed (see Note 2).
2. To a microcentrifuge tube, add 5 μl diluted DNA (30 ng/μl is a
recommended DNA concentration for ddPCR) for each sam-
ple replicate. For example, add 10 μl DNA if the sample is to be
run in duplicate.
3. In a separate microcentrifuge tube, prepare the digestion mix-
ture according to the calculations in Subheading 3.1. The
digestion mixture includes PCR water, NEB 2.1 buffer, and
HINDIII restriction enzyme.
4. Add the digest mixture to the DNA in a 1:1 ratio. For example,
add 10 μl digest mixture to 10 μl DNA. Vortex to mix.
5. Incubate at 37 C with shaking (approximately 300 rpm) for
30 min.
3.4 Preparing The ddPCR supermix can be prepared while the DNA samples are
and Aliquoting digesting. Ensure that the lights are off, as the probes are
the ddPCR Supermix fluorescent.
1. Label one microcentrifuge tube for each sample.
2. In a separate tube, prepare the ddPCR supermix according to
the calculations in Subheading 3.1. This includes the
2 ddPCR Supermix for Probes, and the appropriate 10 or
20 primer–probe mixes.
3. Aliquot 30 μl of the ddPCR supermix into each labeled tube.
Place at 4 C in the dark until use.
3.5 Diluting 1. Dilute the digested DNA with PCR water in a 1:2.5 dilution.
the Digested DNA, For example, add 30 μl PCR water to the 20 μl of digest
and Combining reaction (containing 10 μl digest mix and 10 μl DNA added
with ddPCR Supermix in Subheading 3.2) (see Note 14).
2. Add 20 μl of the diluted digested DNA to a tube containing
30 μl supermix. This volume is sufficient to run a given sample
in duplicate wells. Discard any remaining digested,
diluted DNA.
3.6 Droplet 1. Prepare a cartridge holder with a clean DG8 single-use, dispos-
Generation able cartridge.
2. Slowly dispense 20 μl of the DNA/ddPCR supermix mixture
into the sample wells of the droplet generator cartridge.
Add 20 μl 1 Buffer Control for Probes to any wells without

DNA–ddPCR supermix.
3. When all eight sample wells across the cartridge are full, slowly
add 70 μl Droplet Generation Oil for Probes into the cartridge
wells labeled for oil.
4. Secure a DG8 gasket over the cartridge and place into the
QX100 Droplet Generator (see Note 15) and close the lid.
5. When the green light stops flashing, remove the cartridge from
the holder, and discard the gasket. Close the droplet generator
system while not in use.
6. Manually ensure that 200 μl tips are fitted tightly onto each
channel of the 50 μl multichannel pipette. From the center of
the well, very slowly (over approximately ten seconds) draw up
the droplets, avoiding any quick movements, and working to
minimize air bubbles (see Note 16).
7. Dispense the droplets slowly (over approximately 5 s) into one
column of the 96-well plate, with the tips lightly contacting the
inner wall of each well (see Note 17).
8. Repeat steps 1–7 for the remaining DNA–supermix samples
(see Note 18).
9. When all the samples have been converted to droplets, place
the 96-well plate containing the droplets in the heat sealer.
10. Align a pierceable foil seal over the top, with the correct
orientation according to the foil sheet instructions, and firmly
press down for five seconds. Rotate the plate 180 and firmly
press down again for another 5 s.
3.7 PCR PCR should be performed with a 2 C/s ramp rate to allow for a
better heat transfer throughout the droplet oil mixture. Thermal
cycling conditions are as follows (see Note 19) [2].
95 C 10 min (1 cycle).
94 C 30 s, 59 C 60 s (40 cycles).
98 C 10 min (1 cycle).
12 C hold.
Once amplified, the plate can be placed into the QX100 Drop-
let Digital PCR Droplet Reader for analysis. Do not remove the
foil seal.
3.8 Preparing While the plate is in the PCR thermocycler, the template can be set
QuantaSoftTM up using the Bio-Rad QuantaSoft software.
Template
(a) Select “New” plate, and double click in a well to begin the
template set up.
(b) Highlight all designated columns and select ddPCR supermix

for probes in the supermix drop-down menu.
(c) Enter primer/probe name and concentration (10 or 20)
for each channel. The type for each channel is “unknown.”
(d) Click “apply” and ensure that the “U” symbol for each chan-
nel appears in each well.
(e) Save the plate.
(f) Following thermal cycling, place the 96-well plate into the
reader. Select run, set the plate to be read in columns, and
select the appropriate fluor for each channel.
3.9 Data Analysis 1. For calculation of HHV-6 viral load in whole blood, PBMC or
tissue, the copies of HHV-6 are normalized to the copies of the
reference gene RPP30, such that the final data are represented as
copies/106 cells. Before normalization, the RPP30 copies are
divided by two, as there are two copies of RPP30 per diploid cell.
2. For calculation of HHV-6 viral load in saliva, CSF supernatant
or serum, the copies of HHV-6 are not normalized to the copies
of the reference gene RPP30, and the final data are represented
as copies/ml original sample. This calculation must take into
account the volume of sample used for DNA extraction and the
dilution of the DNA throughout the ddPCR procedure.
3. The positive populations for each primer/probe are identified
using positive and negative controls with single (i.e., not multi-
plexed) primer–probe sets.
4 Notes
1. We use Eppendorf DNA LoBind tubes to increase DNA recov-

ery and reduce interplate and intraplate variability.
2. For a sample run in duplicate wells, the optimal DNA concen-
tration is 100 ng per well. DNA input should not exceed
350 ng/assay (n ¼ 2 wells). This translates to a DNA concen-
tration of <35 ng/μl, if run in duplicate (5 μl per well, or 10 μl
total input). We typically dilute DNA extracted from tissues
and cells, but not from relatively acellular compartments such
as saliva, CSF, serum or plasma. We typically run samples in
duplicate wells, though more wells can be run (in multiples of
two) if greater sensitivity is desired.
3. We use the Eppendorf twin.tec PCR 96-well plates.
4. This single channel pipette is used to transfer the DNA–ddPCR
supermix mixture to the sample wells of the cartridge.
5. This eight-channel pipette is used for the transfer of droplets
from the droplet wells of the cartridge to one column of a
96-well plate.
6. This eight-channel pipette is used for the transfer of droplet

generation oil from the reagent reservoir to the oil wells of the
cartridge.
7. In order to use multiple primers and probe mixtures on the
same channel, it is necessary to vary the concentrations, i.e.,
10 and 20. This ensures a separation of the positive droplet
populations on a given channel. It is advisable to test concen-
trations between 10 and 20 for each primer–probe set to
determine optimal separation.
8. HEX can be substituted for VIC. The HEX dye is compatible
with updated versions of the QuantaSoft software (v1.7.4 and
above).
9. When calculating volumes for the digest and PCR supermix
mixes, add 10–20% excess to account for volume loss during
pipetting and tube transfers.
10. The final concentrations in the PCR reaction will be 1 for the
supermix, 1 for the 20 primer–probe mixture, and 0.5 for
the 10 primer–probe mixture.
11. A primer/probe concentration less than 20 should be deter-
mined for each assay. We determined a 10 primer–probe
mixture is optimal for separation of HHV-6B and RPP30.
However, other primer–probe mixes may require different
sequences and it is therefore recommended that the primer–p-
robe mixture concentrations should be tested to determine
concentrations that provide optimal separation.
12. For a 20 primer–probe mix, the final reaction concentrations
are 900 nM for each primer and 250 nM for the probe.
13. For a 10 primer–probe mix, the final reaction concentrations
are 450 nM for each primer and 125 nM for the probe.
14. The purpose of the dilution is to reduce components present in
the digestion buffer that may interfere with PCR amplification.
15. The cartridge holder should lock into a secure position within
the QX100 Droplet Generator. A green light indicates success-
ful operation. An amber light may flash if there is an issue with
the cartridge holder or the samples within the cartridge. Per-
form one or more of the following to address a flashing amber
light: (a) Ensure there is oil in the wells indicated for oil;
(b) Wipe the bottom of the cartridge and the QX100 Droplet
Generator using an ethanol-moistened Kimwipes; (c) Place a
new gasket on the current cartridge; (d) Transfer samples to a
new cartridge.
16. One technique to collect droplets from the cartridge is to lower
the pipette tips into the depression of the wells and draw up the
contents very slowly (droplets will appear cloudy). The goal is
to minimize the uptake of air, as air bubbles may cause droplet

shearing and diminish data quality.
17. We place the pipette tips at a 45 angle and slowly, over
approximately ten seconds, release the droplets, with the goal
of minimizing droplet shearing. If there are too few droplets
per well, a concentration may not appear in Quantasoft due to
insufficient droplets for autoanalysis. The accepted droplets
should be greater than 10,000; however in our experience,
the best data are obtained with a minimum of 15,000 droplets.
18. We use the lid of a pipette tip box to cover the 96-well plate
during droplet generation, to prevent intersample contamina-
tion and the droplet generation oil from evaporating.
19. A thermal cycling gradient should be performed to find an
optimal annealing temperature for new primer–probe sets.
20. One primer set is used to amplify both viruses, while the probes
are specific to each virus. This SNP-like assay design enables
comparable sensitivity and kinetics across amplifications, while
distinguishing between HHV-6A and HHV-6B with a high
degree of specificity.
References
1. Brunetto S, Massoud R, Leibovitch E et al (HHV-6A) and HHV-6B as demonstrated by
(2014) Digital droplet PCR (ddPCR) for the novel digital droplet PCR assay. PLOS One.
precise quantification of human https://doi.org/10.1371/journal.pone.
T-lymphotropic virus 1 proviral loads in 0092328
peripheral blood and cerebrospinal fluid of 8. Ablashi D, Agut H, Alvarez-Lafuente R et al
HAM/TSP patients and identification of viral (2014) Classification of HHV-6A and
mutations. J Neurovirol 20:341–351 HHV-6B as distinct viruses. Arch Virol
2. Hindson B, Ness KD, Masquelier DA et al 5:863–870
(2011) High-throughput droplet digital PCR 9. De Bolle L, Van Loon J, De Clercq E et al
system for absolute quantification of DNA (2005) Quantitative analysis of human herpes-
copy number. Anal Chem 83:8604–8610 virus 6 cell tropism. J Med Virol. 75:76–85
3. Pinheiro L, Coleman V, Hindson C et al 10. Gautheret-Dejean A, Agut H (2014) Practical
(2012) Evaluation of a droplet digital polymer- diagnostic procedures for HHV-6A, HHV-6B,
ase chain reaction format for DNA copy num- and HHV-7. In: Flamand L, Lautenschlager I,
ber quantification. Anal Chem 81:1003–1011 Krueger G, Ablashi D (eds) Human herpes-
4. Agut H, Bonnafous P, Gautheret-Dejean A viruses HHV-6A, HHV-6B & HHV-7, 3rd
(2015) Laboratory and clinical aspects of edn. Elsevier, Amsterdam, Netherlands, pp
human herpesvirus 6 infections. Clin Micrbiol 9–34
Rev 28:313–335 11. Sedlak R, Cook L, Huang M et al (2014) Iden-
5. Theodore W, Epstein L, Gaillard W et al (2008) tification of chromosomally integrated human
Human herpes virus 6B: a possible role in epi- herpesvirus 6 by droplet digital PCR. Clin
lepsy? Epilepsia 49:1828–1837 Chem 60:765–772
6. Yao K, Honarmand S, Espinosa L et al (2009) 12. Sedak H, Hill J, Nguyen T et al (2016) Detec-
Detection of human herpesvirus-6 in cerebro- tion of HHV-6B reactivation in hematopoietic
spinal fluid of patients with encephalitis. Ann cell transplant recipients with inherited chro-
Neurol 65:257–267 mosomally integrated HHV-6A by droplet
7. Leibovitch E, Brunetto G, Caruso B et al digitial PCR. J Clin Microbiol pii:
(2014) Coinfection of human herpes virus 6A JCM.03275-15
Chapter 7
Biomarkers in Cerebrospinal Fluid: Analysis of Cell-Free

Circulating Mitochondrial DNA by Digital PCR
Petar Podlesniy and Ramon Trullas
Abstract
Cerebrospinal fluid (CSF) contains molecules directly linked with brain function because it permeates brain
tissue. The analysis of protein biomarkers in CSF is currently recommended for the diagnosis of neurode-
generative disorders, but the clinical sensitivity and specificity are still being investigated. A major drawback
is that most of the currently used biomarkers of neurodegenerative diseases are proteins that are found at
very low concentrations in CSF and need to be measured by immunoassays that provide relative values,
which sometimes are difficult to reproduce between laboratories. In contrast, the recent availability of
digital PCR platforms allows the absolute quantification of nucleic acids at single-molecule resolution, but
their presence in CSF has not been characterized. CSF contains cell-free mitochondrial DNA (mtDNA) and
changes in the concentration of this nucleic acid are linked to neurodegeneration. Here we describe a
method to measure the concentration of cell-free circulating mtDNA directly in unpurified CSF using
droplet digital PCR with either hydrolysis probes or fluorescent DNA-binding dye methods. This protocol
allows the detection and absolute quantification of mtDNA content in the CSF with high analytical
sensitivity, specificity, and accuracy.
Key words Mitochondrial DNA, Droplet digital PCR (ddPCR), Cerebrospinal fluid (CSF), Inhibitor
tolerance
1 Introduction
Cerebrospinal fluid (CSF) permeates brain tissue and directly

reflects the biochemical processes that occur in the brain, assisted
by the limitation that the blood brain barrier imposes to the free
flow of molecules from brain to other compartments and the
opposite way. Consequently, CSF is considered the best fluid to
measure and identify new biomarkers for brain diseases [1]. At
present, the diagnostic use of CSF biomarkers, although recom-
mended for preclinical diagnoses of neurodegenerative diseases [2],
still requires further research, validation, and standardization
[3, 4]. Currently, the diagnosis of the different neurocognitive
disease subtypes without biomarker evidence is difficult because
many of these diseases exhibit similar clinical symptoms of
111
112 Petar Podlesniy and Ramon Trullas
dementia. Moreover, the majority of these diseases have a long

preclinical phase that may last decades. Preclinical diagnosis relies
on molecular biomarkers that are hypothesized to be linked with
the underlying pathophysiology of the disease. However, the
molecular mechanisms that underlie the majority of neurocognitive
and neuropsychiatric diseases are still not well known, which ham-
pers not only disease diagnosis but also the discovery of drugs
capable of modifying the course of these diseases.
The pathophysiological biomarkers in the CSF currently used
for preclinical diagnosis of neurodegenerative diseases are proteins
which include amyloid beta 1-42 (Ab1-42), total levels of tau (t-tau),
phosphorylated tau (p-tau), alpha-synuclein, and 14-3-3 proteins.
A major difficulty for the diagnostic use of these biomarker proteins
is that in body fluids some of them are found at very low concen-
trations, which fall near the inferior limit of the analytical sensitivity
range of currently available protein detection techniques [5]. In
addition, a major challenge in protein assays of body fluids in
clinical analyses is the inability to establish absolute values that are
reproducible and comparable between laboratories, mainly because
protein quantification is based on antibody immunoassays that
need to be referenced to an external standard and quantification
requires a comparison with a standard calibration curve that is
generally different amongst laboratories. An additional drawback
is that the antibodies used for protein analysis may change signifi-
cantly between different lots; for example, the epitopes recognized
by the antibodies are generally unknown, the target epitopes differ
markedly amongst polyclonal antibodies, and when monoclonal
antibodies are used, the quality between different lots may vary.
Another general source of variability for protein quantification in
bodily fluids is the degree of protein integrity due to sample extrac-
tion, handling and long term storage. All these factors introduce a
significant amount of variability in protein quantification assays in
CSF or other bodily fluids that lessens reproducibility between
laboratories.
To overcome the technical difficulties associated with protein
quantification in CSF samples, we decided to shift the focus from
proteins to deoxyribonucleic acid (DNA) because the latter may be
less susceptible to degradation due to sample manipulation or long
term storage. We opted to detect mitochondrial DNA (mtDNA)
because we hypothesized that the amount of mtDNA in CSF could
be an index of brain injury or metabolism. Based on this hypothesis,
we found that a drop in mtDNA content in CSF is a biomarker for
preclinical Alzheimer’s disease [6].
There are several advantages of measuring mtDNA over pro-
teins as a biomarker in CSF. In theory, the circular nature of intact
mtDNA should make it more resistant to potential degradation by
endonucleases, which are likely present in CSF and other body
fluids. Another advantage of mtDNA over protein in diagnostic
Biomarkers in Cerebrospinal Fluid: Analysis of Cell-Free Circulating. . . 113
assays is that the former is amenable to detection and quantification

with PCR amplification techniques, which allow higher analytical
sensitivity and specificity than immunoassay procedures. Indeed,
the launch of digital PCR has made possible the precise detection
and absolute quantification of nucleic acid content at a single-
molecule resolution of target sequence [7, 8]. This is achieved by
performing an end-point PCR reaction in a large number of sample
partitions and applying Poisson analysis to the fraction of positive
partitions for quantification. The ability to perform absolute quan-
tification offered by digital PCR, avoiding the use of external
reference standards, markedly improves repeatability and reproduc-
ibility of biomarker measurements. In addition, one of the advan-
tages of the digital PCR protocol for quantifying the amount of
mtDNA in CSF described here is that it does not require previous
extraction or processing of the CSF sample, because we found that
the presence of inhibitory molecules in CSF does not significantly
influence end-point digital PCR quantification (Fig. 1c, d).
Another important benefit of digital PCR is analytical sensitivity
because it allows higher precision measurements than qPCR at low
concentrations of the target sequence.
Droplet digital PCR (ddPCR) incorporates all the advantages
of digital PCR by partitioning the reaction in oil emulsion droplets
and analyzing the fraction of droplets that exhibit positive
end-point PCR amplification [9, 10]. Here we describe a protocol
to measure by ddPCR the concentration of cell-free circulating
mtDNA directly in unpurified CSF using either hydrolysis probe
or fluorescent DNA-binding dye methods with primers that pro-
duce an amplicon of 85 base pairs (mtDNA-85). A common prob-
lem of CSF samples is that some of them may be contaminated with
cells that provide an erroneous measurement of cell-free mtDNA.
To circumvent this problem, the hydrolysis probe protocol includes
the simultaneous measurement of cell-free mtDNA and a nuclear
gene in a multiplex ddPCR assay using the mtDNA-85 primers
together with primers that amplify an amplicon of 103 base pairs
(BAX-103) corresponding to the apoptosis regulator BAX isoform
alpha. Alternatively, other single copy nuclear genes can be used for
this purpose. The presence of at least one copy of the nuclear gene
per microliter of CSF indicates that the sample is contaminated with
genomic DNA and is excluded from analysis.
Characterization of the ddPCR reaction amplifying mtDNA in
CSF with our protocol shows that the optimum temperature to
reach the maximum separation between positive and negative dro-
plets is 60 C for both the hydrolysis probe (Fig. 1a) and the
fluorescent dye (Fig. 1b). The optimal temperature for amplifica-
tion of the BAX-103 amplicon is also 60 C (data not shown) which
allows the multiplex assay of this nuclear gene and mtDNA with
hydrolysis probes. The volume of unpurified CSF sample that can
be added to the ddPCR reaction without inhibition of the
A mtDNA-85 Hydrolysis probe B mtDNA-85 Fluorescent dye
Pooled CSF Pooled CSF

NTC
6000 6000 25000
5000 5000 20000
EvaGreen Fluorescence
FAM Fluorescence
4000 4000
15000
3000 3000
10000
2000 2000
5000
1000 1000
0 0 0
70.0 68.5 65.9 62.0 57.5 53.8 51.4 50.0 60.0 70.0 68.5 65.9 62.0 57.5 53.8 51.4 50.0
Annealing Temperature ºC Annealing Temperature ºC
C mtDNA-85 Hydrolysis probe D mtDNA-85 Fluorescent dye
100 100
75 75
% Control
% Control
50 ddPCR 50 ddPCR
qPCR qPCR
25 25
0 0
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9
CSF ml / reaction CSF ml / reaction
E BAX-103 Hydrolysis probe F

4
Log BAX copies detected /ml CSF
5000
1500
base pairs
2 500
300
1 200
75
3
85
0
10
A-
X-
0 1 2 3 4
N
BA
tD
m
Log BAX copies added /ml CSF
Fig. 1 Analysis of cell-free mtDNA directly in unpurified CSF using hydrolysis probes or EvaGreen Fluorescent
DNA-binding dye: Comparison between ddPCR and qPCR. Assay conditions were studied in a human CSF
amplification due to factors present in CSF can be up to 6 μl in a

total reaction volume of 20 μl. In contrast, this volume of CSF
causes significant inhibition (approximately 25%) in qPCR reac-
tions (Fig. 1c, d). In the ddPCR protocol we choose to use 4.5 μl
of CSF because we found that this volume provides the highest
signal with the lowest possibility of amplification inhibition. To
check the possibility that, when individual CSF samples contain a
high concentration of potential PCR inhibitors, this volume of CSF
might inhibit the reaction, we measured CSF samples from subjects
in an advanced stage of a rapid neurodegenerative disease that
contain a very high content of neuronal death biomarkers such as
tau (Fig. 2d, e). Despite a difference in approximately two orders of
magnitude in the concentration of tau and possibly of many other
proteins, we found no significant differences in the ddPCR reaction
profile between these samples and the corresponding control sam-
ples. The absence of inhibition in the ddPCR is exemplified by a
well-defined separation between the populations of positive and
negative droplets in both low and high tau protein samples. In
addition, the mean absolute values of FAM fluorescence of positive
droplets are similar in low and high tau samples, showing that the
PCR amplification reaches a similar endpoint in both reactions and
that there are no inhibitory factors that differentially alter PCR in
the samples (Fig. 2d, e).
The dynamic range, linearity, accuracy, and analytical sensitivity
of our assay to measure the content of mtDNA and BAX in CSF
were assessed by adding different known concentrations of the
purified PCR amplicons to a sample of pooled CSF. Measurement
of mtDNA or BAX in a sample volume of 4.5 μl of CSF is linear up
Fig. 1 (continued) sample pool from several subjects. (a and b) Temperature gradient optimization assays
were performed in a ddPCR reaction amplifying the 85 base pairs mtDNA amplicon (mtDNA-85) using a
hydrolysis probe (a) or the EvaGreen fluorescent DNA binding dye (b). The maximum separation between
positive (upper) and negative (lower) droplets for EvaGreen is reached at 66 C and is constant between
50 and 66 C, whereas for the hydrolysis probe assay, maximum separation is reached at approximately 60 C
and remains constant down to 50 C. We choose 60 C for maximum stringency in both assays, either in
duplex or singleplex forms. In nontemplate controls (NTC) there is a nonsignificant number of positive (upper)
droplets. (c, d) The influence of CSF volume in the PCR reaction was studied in hydrolysis probe (c) and
EvaGreen fluorescent dye (d) assays using either ddPCR (square dots, continuous line) or qPCR (triangular
dots, discontinuous line). In contrast to qPCR, addition of up to 6 μl of CSF to the PCR reaction did not inhibit
ddPCR assay. (e) Accuracy and linearity of ddPCR reaction to directly measure mtDNA in CSF were assayed by
adding different known concentrations of purified BAX-103 amplicons to a pooled CSF sample. Detection of
mtDNA directly in CSF up to 3000 copies/μl fits to a linear model (r2 ¼ 0.999). The detection limit under these
conditions is less than 1 copy of target per microliter of CSF with a signal above the 95% confidence interval of
average noise in nontemplate controls. (f) Representative image of an agarose gel showing a DNA ladder
75–20,000 base pairs (bp) and the Bax-103 (Lane 1) and mtDNA-85 (Lane 2) amplicons obtained in a CSF
sample
A Low mtDNA content sample

B High mtDNA content sample
18 copies/µl of CSF 117 copies/µl of CSF
8000 8000
7000 7000
FAM Fluorescence
FAM Fluorescence
6000 6000
5000 5000
4000 4000
3000 3000
2000 2000
1000 1000
0 0
1 2 3 1 2 3
Sample Replicate Sample Replicate
C A & B Droplet Histogram

2500
Negative
Threshold
Droplets
Number of droplets
2000
1500
1000
500
Positive
Droplets
0
0 1000 2000 3000 4000 5000 6000 7000 8000
FAM Fluorescence
D E
Low Tau protein sample (146 pg/ml) High Tau protein sample (19300 pg/ml)
mtDNA 61 copies/µl of CSF mtDNA 63 copies/µl of CSF
8000 8000
7000 7000
FAM Fluorescence
FAM Fluorescence
6000 6000
5000 5000
4000 4000
3000 3000
2000 2000
1000 1000
0 0
1 2 3 1 2 3
Sample Replicate Sample Replicate
Fig. 2 Representative examples of results from ddPCR analyses of various CSF samples containing different
levels of mtDNA or potential PCR inhibitors. The mtDNA content was measured in 4.5 μl of CSF sample, using
the mtDNA-85 mix and the ddPCR procedure for hydrolysis probes. (a and b) Representative one-dimensional
scatter plots of three replicates from a CSF sample containing a low concentration of mtDNA (a) and a different
to 3000 copies/μl of CSF, indicating that in the conditions of our

ddPCR reaction there is no influence of the inhibitory factors
present in the CSF on the multiplex detection of mtDNA and
BAX as exemplified by the measurement of added copies of BAX
into a pooled CSF sample (Fig. 1e). The limit of detection of our
ddPCR reaction was close to the theoretical limit of three copies of
target per reaction, corresponding to less than one copy of target
per microliter of CSF with a signal above the 95% confidence
interval of average noise in nontemplate controls. Agarose gel
electrophoresis analyses confirmed that the ddPCR reaction condi-
tions used in our experiments produced single amplicons of the
corresponding size, 103 and 85 base pairs for Bax-103 and
mtDNA-85 primer combinations, respectively (Fig. 1f).
2 Materials
2.1 Droplet Digital 1. Forward and reverse primers and hydrolysis probe targeting
PCR Using Hydrolysis mtDNA (see Note 1). Forward: mtDNA-85F, (50 -CTCAC
Probes TCCTTGGCGCCTGCC-30 ); reverse: mtDNA-85R (50 -GGC
GGTTGAGGCGTCTGGTG-30 ); hydrolysis probe: FAM-mt
DNA-85P (6-carboxyfluorescein (FAM)-50 -CCTCCAAATCA
CCACAGGACTATTCCTAGCCATGCA-30 -Black Hole Quen
cher-1(BHQ-1)).
2. Forward and reverse primers and hydrolysis probe targeting a
single copy nuclear gene, e.g., BAX. Forward: BAX-103F,
(50 -TTCATCCAGGATCGAGCAGG-30 ); reverse: BAX-103R,
(50 -TGAGACACTCGCTCAGCTTC-30 ) and hydrolysis probe:
HEX-BAX-103P, (6-carboxy-20 ,4,40 ,50 ,7,70 -hexachlorofluores
cein (HEX)-50 -CCCGAGCTGGCCCTGGACCCGGT-30 -B
HQ1). (see Notes 2 and 3).
ä
Fig. 2 (continued) CSF sample containing a high concentration of mtDNA (b). Good repeatability is observed
between replicates. The amount of mtDNA concentration does not influence the well-defined separation
between positive (upper) and negative (lower) droplets. In addition, the proportion of positive/negative droplets
is equivalent among replicates. The line drawn at 4000 fluorescence units represents the threshold used to
define positive and negative droplets. (c) Histogram of total positive and negative partitions from assays
shown in a and b. Threshold is determined by choosing a point between the two droplet partitions, for
example, adding the mean fluorescence value of negative droplets plus the mean fluorescence value of
positive droplets and dividing the result by 2. (d, e) Representative one-dimensional scatter plots of three
replicates from a CSF sample containing a low concentration of tau protein (d) and a different CSF sample
containing a high concentration of tau protein (e). In addition to different levels of tau, these two samples may
contain also different amounts of other potential PCR inhibitors. Both samples exhibit a similar concentration
of mtDNA and similar fluorescence values of positive and negative droplet populations, showing the resistance
of ddPCR to the presence of potential PCR inhibitors
3. ddPCR Supermix for Probes (No-dUTP) (catalog number

186-3023) (Bio-Rad, Hercules, CA, USA).
4. ddPCR Droplet Generation Oil for probes (catalog number
186-3005) (Bio-Rad).
2.2 Droplet Digital 1. Forward and reverse primers targeting mtDNA (see Note 1).
PCR Using Fluorescent Forward: mtDNA-85F, (50 -CTCACTCCTTGGCGCCTGCC-
DNA-Binding Dyes 30 ); reverse: mtDNA-85R, (50 -GGCGGTTGAGGCGTCTGG
TG-30 ).
2. Forward and reverse primers targeting a single copy nuclear
gene, e.g., BAX. Forward: BAX-103F, (50 -TTCATCCAGGAT
CGAGCAGG-30 ); reverse, BAX-103R (50 -TGAGACACTCG
CTCAGCTTC-30 ).
3. QX200™ ddPCR EvaGreen Supermix (catalog number
186-4033) (Bio-Rad). (see Note 3)
4. QX200™ Droplet Generation Oil for EvaGreen (catalog num-
ber 186-4005) (Bio-Rad).
2.3 Common 1. Cerebrospinal fluid sample (see Note 4).

Materials to Assay 2. Calibrated micropipettes and low retention, presterilized, aero-
Cerebrospinal Fluid sol barrier filter pipette tips, (Biotix, San Diego, CA, USA)
Using ddPCR (see Note 5).
3. Polypropylene PCR tubes, 0.2 ml.
4. Vortex mixer.
5. Microcentrifuge (5000 g).
6. DG8™ Cartridges for QX200™/QX100™ Droplet Genera-
tor (catalog number 186-4008) (Bio-Rad).
7. DG8™ Cartridge Holder (catalog number 186-3051) (Bio-Rad).
8. DG8™ Gaskets for QX200™/QX100™ Droplet Generator
(catalog number 186-3009) (Bio-Rad).
9. Electronic multichannel pipette with wide bore tips (see Note 6).
10. Eppendorf twin.tec® 96-Well PCR plates, semiskirted (catalog
number 0030 128.XXX) (Eppendorf).
11. PX1™ PCR Plate Sealer (catalog number 186-4000) (Bio-Rad).
12. Pierceable Foil Heat Seal (catalog number 181-4040)
(Bio-Rad) (see Note 7).
13. C1000 Touch™ Thermal Cycler with 96-Deep Well Reaction
Module (catalog number 185-1197) (Bio-Rad).
14. ddPCR Droplet Reader Oil (catalog number 186-3004)
(Bio-Rad).
15. QX200™ Droplet Digital™ PCR System (catalog number
186-4001) (Bio-Rad).
16. QuantaSoft™ Software (catalog number 186-4011) (Bio-Rad).
3 Methods
3.1 Sample CSF should be collected and processed following standard

Acquisition operating procedures after informed consent of the patient. In
and Processing our studies, CSF was collected from subjects recruited at the Alz-
heimer’s disease and other cognitive disorders unit of the Hospital
Clinic of Barcelona. The CSF samples were obtained with informed
consent at the Alzheimer’s disease and other cognitive disorders
Unit of the Neurology Service, following the procedure approved
by the ethics committee of the Hospital Clinic of Barcelona. CSF
was obtained by lumbar puncture between 9 a.m. and 12 p.m. A
small sample of the first CSF volume was reserved for later mea-
surement of cell contamination and this was followed by collection
of 10 ml of CSF. The whole volume of CSF was centrifuged
(2000 g/4 C/10 min), aliquoted after centrifugation in
500 μl polypropylene tubes and stored at 80 C within two
hours of the beginning of the collection. The mtDNA assay was
performed in unpurified CSF that underwent only one thaw. It is
convenient to avoid using CSF samples that have undergone
repeated freeze–thaw cycles. However, we find that up to three
freeze–thaw cycles do not significantly modify the measured con-
centration of mtDNA by ddPCR.
3.2 Digital PCR Bring all reagents and samples to room temperature before use.
Reaction Setup Centrifuge briefly, mix thoroughly by vortexing and centrifuge
and Amplification briefly again to properly mix and collect the components to the
bottom of the tubes.
1. Prepare a master mix in a polypropylene microcentrifuge tube
containing ddPCR Supermix, oligonucleotide mix, and
double-distilled nuclease-free water in a total volume sufficient
to obtain 15.5 μl per assay as shown in Table 1 (see Note 8). It is
essential to include nontemplate control assays, with master
mix but adding nuclease-free dH2O instead of CSF sample
(see Note 9). In addition, for assay characterization, it is neces-
sary to include positive control samples containing known
amounts of nuclear DNA or mitochondrial DNA target ampli-
cons cloned in vectors or purified from PCR reactions (see Note
10). The process of assay characterization includes assessment
of accuracy, sensitivity, setting the threshold for positive dro-
plets and the level of false positive droplets for mtDNA and
nuclear DNA respectively.
(a) The ddPCR assay using hydrolysis probe is designed as a
multiplex assay to monitor simultaneously mtDNA and
BAX in the same reaction. The composition of the master
mix for ddPCR using hydrolysis probes is as shown in
Table 1 (see Note 11):
Table 1
Master mix for ddPCR using hydrolysis probes (see Note 8)
Initial Final Volume per reaction

Component concentration concentration (20 μl)
ddPCR Supermix for Probes (No-dUTP) 2 1 10.0 μl
(186-3023)
Oligonucleotide Mix in H2O:
mtDNA-85F 18.0 μM 0.9 μM 1.0 μl
mtDNA-85R 18.0 μM 0.9 μM
FAM-mtDNA-85P 2.0 μM 0.1 μM
BAX-103F 18.0 μM 0.9 μM
BAX-103R 18.0 μM 0.9 μM
HEX-BAX-103P 6.0 μM 0.3 μM
dH2O, nuclease free 4.5 μl
Total volume/assay 15.5 μl
Table 2
Master mix for ddPCR using EvaGreen fluorescent dye (see Note 8)
Initial Final Volume per reaction

Component concentration concentration (20 μl)
QX200™ ddPCR EvaGreen Supermix 2 1 10.0 μl
(186-4033)
Oligonucleotide mix in H2O: 1.0 μl
mtDNA-85F 2.0 μM 0.1 μM
mtDNA-85R 2.0 μM 0.1 μM
or
BAX-103F 2.0 μM 0.1 μM
BAX-103R 2.0 μM 0.1 μM
dH2O, nuclease free 4.5 μl
Total volume/assay 15.5 μl
(b) The ddPCR assay of mtDNA and the nuclear gene BAX
using fluorescent DNA binding dyes is performed in sep-
arate reactions. The composition of the master mix for
ddPCR using EvaGreen fluorescent dye is as shown in
Table 2 Master mix for ddPCR using EvaGreen fluores-
cent dye (see Note 8).
2. Distribute 15.5 μl of the master mix with a 20 μl pipette to

0.2 ml Polypropylene PCR tubes (see Note 8).
3. Add 4.5 μl of unpurified CSF sample to the tube (see Note 8).
4. Mix well by vortexing and centrifuge briefly to collect the mix
to the bottom of the tube and remove bubbles. The Supermix
is viscous and insufficient mixing will lead to less reproducible
results.
5. Place the DG8™ Cartridge for Droplet Generator into the
DG8™ Cartridge Holder.
6. Transfer each 20 μl of reaction mixture to one of the
eight sample wells in the middle of the DG8™ Cartridge (see
Note 8).
7. Add 70 μl of either ddPCR Droplet Generation Oil for probes
(for the probe hydrolysis method) or QX200™ Droplet Gen-
eration Oil for EvaGreen (for the fluorescent DNA-binding
dye method) to the oil wells of the DG8™ Cartridge adjacent
to the samples (see Note 12).
8. Cover the cartridge with a DG8™ Gasket for QX200™/
QX100™ Droplet Generator.
9. Place the covered cartridge into the QX100/QX200™ Digital
Droplet Generator to generate the droplets (see Note 13).
10. Transfer 40 μl of the emulsified sample to an Eppendorf twin.
tec® 96-Well semiskirted PCR plate with an electronic multi-
channel pipette set on the lowest aspirating and dispensing
speed using wide bore tips. Make sure to keep the pipet at a
15 –30 angle when aspirating the emulsified sample to avoid
damaging the droplets.
11. Cover the 96-well PCR plate with pierceable foil heat seal and
place it into PX1™ plate sealer. Seal at 170 C for 4 s.
12. Place the Eppendorf twin.tec® semiskirted 96-Well PCR plate
into a PCR thermal cycler that has the ability to adjust a ramp
temperature and that is compatible with deep wells. Set ramp
rate to 2 C per second for all steps.
13. Use the following PCR profiles (see Note 14):
(a) For hydrolysis probes:
95 C: 10 min.
40 cycles:
94 C: 30 s.
60 C: 1 min.
98 C: 10 min.
4 C: infinite.
(b) For fluorescent dyes (see Note 15):

95 C: 5 min.
40 cycles:
95 C: 30 s.
60 C: 1 min.
4 C: 5 min.
90 C: 5 min.
4 C: infinite.
14. After PCR the plate can be left in the thermal cycler overnight
or stored for a few days at 4 C before droplet reading.
3.3 Digital Data 1. Power on the QX200™ Droplet Digital™ PCR Droplet
Acquisition Reader and let it warm up for at least 30 min, making certain
that there is enough ddPCR Droplet Reader Oil in the reader
compartment. Prime the system if the instrument has been
unused for longer than a week.
2. Place the 96-well PCR plate to the QX200™ Droplet Digital™
PCR Droplet Reader.
3. Introduce the appropriate settings in the PCR Droplet reader
software indicating the PCR supermix used and the fluores-
cence acquisition channels (see Note 15).
3.4 Data Analysis 1. After reading all samples, open the file with the appropriate
QuantaSoft software and in the Experiments window select
Absolute Quantification method (ABS) and then select all
samples for data analysis.
2. Click on the Analyze tab and then the multiwell thresholding
icon. Choose a point between the two droplet populations
shown in the fluorescence histogram (Fig. 2c) to identify a
threshold value that reliably separates positive and negative
droplets. For example, add the mean fluorescence value of
negative droplets plus the mean fluorescence value of positive
droplets and divide the result by 2. Enter the resulting value
into the Set Threshold window (see Note 16)
3. Determine target concentrations by analyzing the proportion
of negative versus positive droplets using the Absolute Quanti-
fication method (see Note 17). Replicates that do not exhibit
equivalent mean fluorescence values of positive and negative
droplets between them indicate a droplet generation failure and
should be discarded. When all replicates of one sample exhibit
equivalent mean fluorescence that is significantly lower than
the exhibited by the replicates of another sample, this indicates
inhibition of the PCR and values cannot be compared.
4. Obtain the total number of copies per 20 μl of reaction for each

target and divide by the volume (e.g., 4.5 μl) of CSF intro-
duced into the reaction to obtain copies per microliter of CSF.
4 Notes
1. These primers target an mtDNA sequence that is not present in

the nuclear genome as a nuclear mtDNA (Numt). When
designing primers targeting other mtDNA sequences it is
important to check that these sequences do not have a
corresponding Numt.
2. The reporter fluorophores used to label the hydrolysis probes
can be FAM combined interchangeably with either HEX or
VIC in any of the probes. We chose to use FAM for mtDNA
and HEX for BAX, but it could also be the opposite. It is likely
that in the near future there will be multichannel digital PCR
platforms that will allow more fluorophore combinations for
multitarget detection. We use the dark quencher BHQ-1
because it is capable of quenching both FAM and HEX with
high efficiency. We found that quenchers such as TAMRA do
not completely quench probe fluorescence and do not permit a
clear separation between positive and negative droplets.
3. The oligomix with fluorophores or EvaGreen supermix should
not be exposed to light for prolonged periods of time. We cover
with aluminum foil all tubes that contain fluorophores.
4. After centrifugation of the whole volume of CSF obtained by
lumbar puncture, the unpurified CSF should be placed in
sterile siliconized polypropylene microcentrifuge tubes to
avoid mtDNA absorption to the tube, and stored at 80 C
in small aliquots (e.g., 500 μl) to avoid unnecessary transfer to
different tubes or freezing–thawing cycles.
5. The use of low retention pipette tips is essential to obtain
appropriate accuracy and repeatability of mtDNA measure-
ments in CSF. Low retention tips are designed to sustain
minimal sample loss from viscous liquids such as CSF. Regular
calibration of pipettes and correct pipetting techniques with
slow aspiration/delivery rates is crucial to get low coefficients
of variation for repeatability and reproducibility.
6. Wide bore tips with an orifice size wider that 1.5 mm are
convenient to prevent mechanical damage to the droplets
when transferring from the DG8 cartridge to the PCR plate.
7. This foil seal cannot be replaced by another material because it
may damage the ddPCR droplet reader.
8. We find that it is best to prepare the master mix for all the
planned reaction tubes in an excess volume of 10% and
distribute to each tube 17.05 μl of master mix. To this volume,

we also add a 10% excess of CSF sample (4.95 μl). This is
because in Subheading 3.2.6 of the protocol it is necessary to
transfer without bubbles exactly 20 μl of this mixture to the
cartridge well, and we found that if the initial volume is exactly
20 μl it is very difficult not to draw bubbles in the pipette
during the transfer.
9. We find it is important that each ddPCR plate contains several
nontemplate control assays (NTCs) that have master mix and
nuclease-free dH2O instead of CSF sample. NTCs allow iden-
tification of possible contamination of analysis materials and
reagents. It is essential to perform all procedures with sterile
gloves using material specially reserved for PCR in laboratory
areas only dedicated to the setup of PCR reactions.
10. Care should be taken to avoid contamination of equipment and
samples with DNA standards that may contain significantly
higher concentrations than the samples.
11. The reaction mixture contains a different concentration of
FAM versus HEX labelled probe. The reason is that we found
that there is a marked bleed-through cross-emission of FAM
into HEX channel. One way to reduce such fluorophore cross-
emission is by lowering the concentration of FAM labelled
probe. The concentration ratio of FAM/HEX in our reaction
conditions is optimized for our particular probes and if using
other probes or quenchers we recommend finding again the
optimal conditions.
12. It is important to add the oil after the reaction mixture because
if added before, it will block the microfluidic channels.
13. We find that it is important to inspect the cartridge after
droplet generation to ensure that droplets have been correctly
formed. A simple way when using hydrolysis probes is to slowly
rotate the cartridge in front of a light source and if droplets are
formed there is a change in reflection of the light source every
60 rotation. If there are no droplets in one of the wells of the
cartridge (either because of air bubbles or obstruction in the
microfluidic channel) it is advisable to discard the whole car-
tridge. When using EvaGreen fluorescent dye it is more diffi-
cult to identify whether droplets have been correctly formed
prior to the analysis.
14. The PCR temperature profiles described here are the optimal
for the particular primer combinations of our reactions and our
laboratory conditions. We recommend to follow the digital
MIQE guidelines [11] and perform annealing temperature
gradients to optimize the separation between positive and
negative droplets for each primer combination and laboratory
setting.
15. The total time to analyze a 96-well plate is approximately 7 h.

This includes: 2.5 h for sample preparation and droplet gener-
ation: 2 h for thermocycling and 2.5 h for droplet reading.
16. We recommend setting the same threshold value for all the
wells in the plate (Fig. 2). Some samples may show only one of
the droplet partitions and the fluorescence histogram of those
samples will have only one Gaussian curve, hampering the
detection of a reliable threshold value to separate positive and
negative partitions. Therefore, we find convenient to select all
wells for analysis because then the histogram shows the ampli-
tude of the total events in all samples, making unlikely the
possibility that there is only one droplet population in all of
them. We find that when the ddPCR assay is performed in the
most optimal conditions, the threshold value does not signifi-
cantly influence the result.
17. Only samples that exhibit a total of more than 12,000 droplets
or events should be included in the final analysis. However, for
most accurate measurements, setting the limit for sample inclu-
sion to more than 15,000 droplets provides results with signif-
icantly lower coefficient of variation among replicates. Using
this ddPCR method to measure mtDNA concentration in
4.5 μl of CSF we find that three sample replicates is sufficient
to obtain precise values with repeatability levels equivalent to
the sampling error expected for volumes of less than 10 μl.
Acknowledgment
This work is supported by the Ministerio de Economia y Compe-

titividad of Spain (Grants: SAF2011-23550, SAF2014-56644-R)
and by the Instituto Carlos III (Grant: PI2013/08-3) from Centro
de Investigacion Biomedica en Red de Enfermedades Neurodegen-
erativas to R.T.
References
1. Lleo A, Cavedo E, Parnetti L, for Alzheimer’s disease. Alzheimers Dement
Vanderstichele H, Herukka SK, Andreasen N, 7:263–269
Ghidoni R, Lewczuk P, Jeromin A, Winblad B 3. Jack CR Jr, Albert MS, Knopman DS,
et al (2015) Cerebrospinal fluid biomarkers in McKhann GM, Sperling RA, Carrillo MC,
trials for Alzheimer and Parkinson diseases. Nat Thies B, Phelps CH (2011) Introduction to
Rev Neurol 11:41–55 the recommendations from the National Insti-
2. McKhann GM, Knopman DS, Chertkow H, tute on Aging-Alzheimer’s Association work-
Hyman BT, Jack CR Jr, Kawas CH, Klunk groups on diagnostic guidelines for
WE, Koroshetz WJ, Manly JJ, Mayeux R et al Alzheimer’s disease. Alzheimers Dement
(2011) The diagnosis of dementia due to Alz- 7:257–262
heimer’s disease: recommendations from the 4. Dubois B, Feldman HH, Jacova C, Hampel H,
National Institute on Aging-Alzheimer’s Asso- Molinuevo JL, Blennow K, DeKosky ST,
ciation workgroups on diagnostic guidelines Gauthier S, Selkoe D, Bateman R et al (2014)
Advancing research diagnostic criteria for Alz- 8. Vogelstein B, Kinzler KW (1999) Digital PCR.
heimer’s disease: the IWG-2 criteria. Lancet Proc Natl Acad Sci U S A 96:9236–9241
Neurol 13:614–629 9. Hindson BJ, Ness KD, Masquelier DA,
5. Blennow K, Zetterberg H (2015) Understand- Belgrader P, Heredia NJ, Makarewicz AJ,
ing biomarkers of neurodegeneration: Ultra- Bright IJ, Lucero MY, Hiddessen AL, Legler
sensitive detection techniques pave the way TC et al (2011) High-throughput droplet digi-
for mechanistic understanding. Nat Med tal PCR system for absolute quantitation of
21:217–219 DNA copy number. Anal Chem 83:8604–8610
6. Podlesniy P, Figueiro-Silva J, Llado A, 10. Pinheiro LB, Coleman VA, Hindson CM,
Antonell A, Sanchez-Valle R, Alcolea D, Herrmann J, Hindson BJ, Bhat S, Emslie KR
Lleo A, Molinuevo JL, Serra N, Trullas R (2012) Evaluation of a droplet digital polymer-
(2013) Low CSF concentration of mitochon- ase chain reaction format for DNA copy num-
drial DNA in preclinical Alzheimer’s disease. ber quantification. Anal Chem 84:1003–1011
Ann Neurol 10:655–668 11. Huggett JF, Foy CA, Benes V, Emslie K, Gar-
7. Sykes PJ, Neoh SH, Brisco MJ, Hughes E, son JA, Haynes R, Hellemans J, Kubista M,
Condon J, Morley AA (1992) Quantitation of Mueller RD, Nolan T et al (2013) The digital
targets for PCR by use of limiting dilution. MIQE guidelines: Minimum Information for
BioTechniques 13:444–449 Publication of Quantitative Digital PCR
Experiments. Clin Chem 59:892–902
Chapter 8
Testing of General and Human-Associated Fecal

Contamination in Waters
Yiping Cao, Meredith R. Raith, and John F. Griffith
Abstract
qPCR has become increasingly popular for microbial water quality testing because it is faster, more specific,
and more flexible than culture-based methods. However, qPCR method limitations such as quantification
bias introduced by reliance on standards and susceptibility to PCR inhibitors are major obstacles for
implementation in water testing. This is because water testing requires accurate quantification of rare
targets and because environmental waters often contain PCR inhibitors. Digital PCR offers the opportunity
to maintain qPCR’s advantages over culture-based methods while ameliorating two of qPCR’s major
limitations: the necessity to run standard curves and high susceptibility to inhibition. Here we describe a
complete method for simultaneous testing for a general microbial water quality indicator
(Enterococcus spp.) and a human-associated fecal marker in environmental waters. The complete method
includes water sampling and filtration to capture bacteria, DNA extraction from bacteria captured on the
filter, and droplet digital PCR to quantify the genetic markers from bacteria indicative of general and
human-associated fecal contamination.
Key words Fecal contamination, Water quality, Enterococcus, HF183, Microbial source tracking,
Duplex digital PCR
1 Introduction
Fecal contamination in water is a public health concern worldwide.

In most countries, fecal indicator bacteria (FIB) such as Enterococ-
cus, E. coli, and total and fecal coliforms are routinely tested in
surface water, treated wastewater, and source and finished drinking
water for public health protection. However, while FIB offer a
conservative indication of general fecal contamination, they do
not provide information on host sources of fecal contamination
because all warm-blooded animals shed FIB in their fecal material
[1]. FIB are therefore limited in their usefulness for pollution
remediation because effective remediation requires knowledge of
the origin of the contamination, and for accurate risk assessment
because human fecal contamination generally presents a much
127
128 Yiping Cao et al.
higher public health risk than nonhuman contamination [2]. Addi-

tionally, FIB may also originate from nonfecal sources [3], further-
ing the need to track fecal contamination to a specific host. Genetic
markers specific to fecal contamination from different hosts are now
available [4]. Known as microbial source tracking markers, the most
widely used human-associated marker is HF183, which can detect
human fecal contamination with high diagnostic sensitivity and
specificity [5].
While FIB in water have historically been measured by growth-
based methods such as membrane filtration and substrate utiliza-
tion, qPCR has become a popular alternative to these methods
because qPCR provides faster sample-to-result time, higher speci-
ficity, and greater flexibility [1, 6]. In fact, the revised U.S. EPA
recreational water quality criteria published in 2012 allows use of
qPCR for Enterococcus spp. in routine recreational water monitor-
ing [7]. For the HF183 marker specific to human fecal contamina-
tion, the only available measurement methods use (q)PCR [8]. In
fact, qPCR is the dominant technology for measuring host-specific
fecal contamination in the field of microbial source tracking
because of the same advantages qPCR affords for measuring
FIB [1].
Despite wide acceptance of qPCR in the medical and food
safety fields, method limitations such as quantification bias intro-
duced by reliance on standards and susceptibility to PCR inhibitors
are still major obstacles for implementing qPCR in water testing
[9–11]. It is also difficult to multiplex in qPCR, making it infeasible
to measure both general and human-associated fecal contamination
in one analysis [12]. However, digital PCR provides direct quanti-
fication without standard curves, is more resistance to inhibition,
and much easier to multiplex [13]. A recently developed duplex
digital PCR assay (EntHF183 duplex ddPCR) enabled unbiased,
precise, sensitive, specific, highly inhibition resistant, and highly
reproducible, simultaneous quantification of Enterococcus spp. and
the HF183 marker. This is the first assay to simultaneously measure
concentrations of a general microbial water quality indicator used
for regulatory purposes and provide information regarding human-
associated fecal contamination [12].
Here we describe the complete protocol for simultaneous test-
ing of general and human-associated fecal contamination in water.
The protocol consists of three major procedures: (1) water sam-
pling and filtration to capture bacteria, (2) DNA extraction from
bacteria captured on the filter, and (3) droplet digital PCR to
quantify DNA representing the general microbial water quality
indicator Enterococcus spp. and the HF183 human-associated fecal
marker.
EntHF183 Duplex ddPCR 129
2 Materials
Carry out all procedures at room temperature unless otherwise

specified. Aseptic technique must be used with all portions of the
following procedures and gloves, a clean laboratory coat and other
personal protective equipment should be worn for safety and to
avoid contamination. Workspaces and equipment must be cleaned
(using DNAway or a 10% solution of chlorine bleach) prior to
starting all laboratory procedures. The water sampling procedure
described is specific to recreational water quality sampling at marine
or freshwater beaches. Sampling of other water bodies may require
different materials and follow different protocols. Unless bottles are
new and certified to be nucleic acid free, it is recommended that
sampling bottles be treated by adding a small volume (ca. 10 mL) of
10% HCl to conveniently destroy potential residual DNA/RNA
and stored with the acid inside until use.
2.1 Water Sampling 1. 10% hydrochloric acid.

and Filtration 2. Deionized water.
2.1.1 Sample Bottle
Preparation
2.1.2 Water Sampling 1. New or acid treated polypropylene or polyethylene sample

collection bottle.
2. Cooler with ice packs or ice.
3. Sampling field sheet.
2.1.3 Water Filtration 1. Polycarbonate membrane (47 mm Isopore™ or Nuclepore™,

0.4 μm).
2. Disposable filtration funnel (such as Nalgene® Analytical Test
Filter Funnels).
3. 2 mL screw-cap bead beating tube preloaded with glass beads
(GeneRite S0205-50 or equivalent).
4. Sterile PBS rinsing buffer: pH 7.4, sodium dihydrogen phos-
phate at 4 mM, sodium monohydrogen phosphate at 21 mM,
sodium chloride at 145 mM in 1 L of molecular grade water.
5. Anhydrous ethanol (for alcohol lamp).
6. Liquid nitrogen.
2.2 DNA Extraction 1. GeneRite Extraction kit (K200-02C-50).

2. Bead beater (Biospec 607 or equivalent).
3. Low binding 1.7 mL microcentrifuge tubes.
Table 1
Target Primer and probe sequences Reference

Enterococcus spp. EnteroF1A: GAGAAATTCCAAACGAACTTG [15]
EnteroR1: CAGTGCTCTACCTCCATCATT
GPLQ813TQ: [6-FAM]-
TGGTTCTCTCCGAAATAGCTTTAGGGCTA-[BHQ1]
Human fecal-associated HF183ND: ATCATGAGTTCACATGTCCG [5]
Bacteroidales BthetR1: CGTAGGAGTTTGGACCGTGT
BthetP1: [HEX]-CTGAGAGGAAGGTCCCCCACATTGGA-
[BHQ1]
2.3 Droplet DigitalTM 1. Enterococcus and HF183 primers and probe (Table 1).
PCR 2. TE pH 8 buffer.
2.3.1 Making Assay 3. ddPCR™ Supermix for Probes (no dUTP) (Bio-Rad,
Master Mixes 186-3024).
4. Nuclease-free PCR grade water.
5. Low binding 1.7 mL microcentrifuge tubes.
6. Aluminum Sealing Film.
7. Hard-shell 96-well Plate.
2.3.2 Making Droplets 1. Droplet Generation Oil for Probes (Bio-Rad, 186-3005).
2. DG8TM Cartridges for Droplet Generator (Bio-Rad,
186-4008).
3. DG8 Gaskets for Droplet Generator (Bio-Rad, 186-3009).
4. DG8 Cartridge Holders (Bio-Rad, 186-3051).
5. Droplet Generator (Bio-Rad).
6. Rainin 20 μL multichannel pipet/tips (Rainin 20 μL XLT and
GPL20F).
7. Rainin 50 μL multichannel pipet/tips (Rainin 50 μL XLT and
GPL200F).
8. Rainin 200 μL multichannel pipet/tips (Rainin 200 μL XLT
and GPL200F).
2.3.3 ddPCR Thermal 1. Hard-shell 96-well PCR plate (Such as the Eppendorf Twin
Cycling, Detection, Tec PCR plate).
and Data Analysis 2. Pierceable Heat Seal Foil (Bio-Rad).
3. PX1TM PCR Plate Sealer (Bio-Rad).
4. CFX96TM Thermalcycler (Bio-Rad).
5. QX100TM Droplet Reader (Bio-Rad).
6. Droplet Reader Oil (Bio-Rad, 186-3004).
7. QuantaSoftTM Software.
3 Methods
3.1 Water Sampling 1. Rinse the bottles three times with deionized water (DI) in
and Filtration order to remove any large particles.
3.1.1 Preparing Sampling 2. Pour all the water out and add about 10 mL of 10% hydro-
Bottles chloric acid (10% HCl). Securely cap the bottle and swirl acid to
coat inner bottle surface. Store bottle capped with acid inside
until use.
3.1.2 Beach Water 1. Dispose of acid in sampling bottle by making a small divot in
Sampling the sand with your toe and pouring acid into divot. Avoid
getting acid on your skin or clothes.
2. Fill the sampling bottle <¼ full with sample water. Close
bottle, shake vigorously, and discard rinse water away from
collection area. Repeat this process twice more.
3. Collect full volume water sample at ankle depth on an
incoming wave.
4. Place water sample in cooler, transport sample on ice pack/ice
to lab for filtration within 6 h.
3.1.3 Water Filtration 1. Setup vacuum filtration system (see Note 1). Light the alcohol
lamp or Bunsen burner (for flame sterilization). Soak the tips of
two pairs of stainless steel forceps in a beaker with 95% ethanol.
2. Prepare the disposable filtration funnel for water filtration by
replacing the manufacturer’s preloaded filter (generally not the
one specified in this method) with the specified polycarbonate
membrane (see Note 2).
3. Filter 100 mL water sample (see Note 3), rinse funnel wall with
approximately 10 mL of PBS so all materials are captured on
the membrane filter. After all liquid passes through the mem-
brane filter, close vacuum valve and remove the funnel.
4. Using flame sterilized (and cooled) forceps (see Note 4), care-
fully roll up the polycarbonate membrane filter on the backing
plate. First, fold one edge of the membrane onto itself (about
¼ of the diameter of the membrane) and hold in place with
second forceps. Then, using each pair of forceps alternately, roll
the membrane to create a tube. Place into the corresponding
collection tube (i.e., 2 mL screw-cap microtube suitable for
80 C storage and the bead beating step during DNA extrac-
tion), screw the cap securely (see Note 5).
5. Flash-freeze the tube containing the filter in liquid nitrogen
and transfer to 80C freezer for storage until DNA extraction
(see Note 6).
3.2 DNA Extraction This protocol employs the published bead beating DNA extraction
procedure using the GeneRite DNA EZ kit (commonly employed
for molecular testing of recreational waters [1, 9, 11]) (see Note 7).
3.2.1 Bead Beating 1. Remove filter tubes to be analyzed from 80 C freezer and
pipet 600 μL lysis buffer into each. Bead beat the tubes for
1 min (see Note 8).
2. Centrifuge the bead beaten tubes for 3 min at 12,000 g. Pull
out all liquid from tubes and place into a new 1.7 mL low
binding microtube. Centrifuge the supernatants in the micro-
tubes for 1 min at 12,000 g.
3.2.2 Binding DNA 1. Pull out 380 μL of supernatant from the centrifuged tube and
to Column place liquid into a new 1.7 mL low binding tube containing
760 μL of binding buffer. Vortex and spin the mixture briefly.
2. Add ~650 μL of mixture to the provided column (the column is
already sitting in the provided collection tube) and spin for
1 min at 12,000 g. Discard the collection tube containing the
flow through and place the column in a new collection tube.
Repeat step 4 until all of the mixture from step 3 is put
through the column.
3.2.3 Column Washing 1. After placing the column in a new collection tube add 500 μL
of wash buffer to the column and spin for 1 min at 12,000 g.
Discard the collection tube containing the flow through and
place column in new collection tube. Repeat this step.
3.2.4 Column Elution 1. Place column in new 1.7 mL low binding microtube. Add
50 μL elution buffer to the column and let sit 1 min. Spin the
column for 1 min at 12,000 g. Repeat this step.
2. After 100 μL of DNA extract has been collected, dispose of the
column and aliquot the extract if needed.
3. Quantify total DNA in each sample by spectrophotometer (see
Note 9)
3.3 Droplet Two hard-shell 96-well plates are used in this procedure: one for
Digital PCR preparing the assay mixture and the other as the final PCR plate.
The former can be any plate that is easy to work with, while the
latter should be a PCR plate with thermal conductivity, plate height
compatible with the PCR thermal cycling parameters, the thermal
cycler, and the Droplet Reader. However, currently, it is recom-
mended to use an Eppendorf Twin Tec 96-well plate (Bio-Rad,
personal communication). Figure 1 depicts the transfer of liquid
(reagents, sampled DNA, oil, and droplets) in this section. The
steps described in this section are also demonstrated in a video
article [14].
Fig. 1 Liquid transfer flow diagram for Subheading 3.3
1. Make 100 μmol/L stock concentrations for all Enterococcus

and HF183 primers and probes. Use molecular grade water
to hydrate and dilute primers and TE Buffer pH 8 to hydrate
probes (see Table 1).
2. Use the primers and probes to prepare the master mix by
adding 12 μL Droplet PCR Supermix (2 stock), 0.216 μL
each forward and reverse primer, 0.06 μL each probe, and
5.016 μL nuclease free PCR grade water, per reaction well.
Prepare master mix in bulk for all reactions. Ensure that the
master mix is well vortexed and spun down before pipetting
into the hard-shell 96-well plate (no specific requirement on
plate height, thermal conductivity, or supplier).
3. Pipet 18 μL master mix into each well on the 96-well reaction
mix plate and add 6 μL DNA template. If running samples in
duplicate, pipet 36 μL of master mix per well and 12 μL DNA
template into each well and leave the corresponding replicate
wells (e.g., in the adjacent plate column) empty on the plate (see
Note 10).
4. Pipet the assay mixture up and down with the 20 μL multi-
channel pipet a minimum of 12 times to ensure adequate
mixing. Avoid making excess bubbles within the mixture (see
Note 11).
5. Place a cartridge in the cartridge holder, click shut, and pipet

20 μL of the assay mixture using the multichannel pipet into
the section of the cartridge marked “Sample” (see Note 12).
Use the 200 μL pipet to transfer 70 μL of droplet generation oil
to the section of the cartridge marked “Oil.” Cover cartridge
with a gasket and gently place on droplet generator.
6. While droplet generation is occurring prepare the next car-
tridge in the same manner as described in step 5 to save total
setup time (see Note 13).
7. When the droplet generator is done, take off the first cartridge,
set aside, and gently place the second cartridge onto the drop-
let generator.
8. Remove the gasket from the first cartridge and discard. Do not
unclick the cartridge from the cartridge holder as this may
break the newly generated droplets. Gently transfer the total
volume of generated droplets into the final PCR plate for
thermal cycling (see Note 14). Repeat droplet generation
with the remaining samples.
9. When all the droplets are in the final PCR plate, place plate on
plate sealer with pierceable foil on top and heat seal at 180 C
for 10 s.
10. Remove the plate from sealer and place on a thermal cycler
compatible with the final PCR plate and a temperature ramping
speed of 2.0 C/s. Run thermal program as follows: 10 min at
95 C, followed by 40 cycles of 30 s at 94 C and 60 s at 60 C,
followed by a 10 min hold at 98 C (optional: final hold at
4 C, or 12 C if left overnight).
11. After cycling, transfer the plate to the Droplet Reader (see Note
15). The PCR plate may also be stored at 4 C for up to 3 days
before droplet reading.
12. Once the plate is on the droplet reader, configure the software
to read RED (rare event detection) for each well. Define all
sample names and select Channel 1 (FAM) and Channel
2 (HEX or VIC) for data collection. Select Run and choose
FAM-HEX (or FAM-VIC) to start the run (see Note 16).
3.4 Data Analysis 1. When the run is complete, open the data file and check the
events number (i.e., number of acceptable droplets in each
well). All wells containing less than 10,000 droplets should
be excluded.
2. The fluorescence threshold should then be set approximately
one standard deviation (500–700 fluorescence units, Fig. 2)
above the negative droplets in NTC wells (see Note 17). Fluo-
rescence values of unknown samples should also be compared
to that of positive controls to check for signs of PCR inhibition

(see Note 18).
3. Results may be exported in.csv file for further data analysis (see
Note 19).
4 Notes
1. Test whether the vacuum assembly is working properly before

moving forward. Turn on the pump and close all valves. Then
adjust the vacuum pressure using the knurled knob on the
vacuum inlet of the pump (<20 in.Hg, or <0.6 atm). Ensure
that suction is carried from the pump to the filtration manifold.
2. If the commercially available prepacked disposal filtration fun-
nels are preloaded with a filter other than the polycarbonate
membrane filter, the following steps can be followed to asepti-
cally replace the preloaded filter:
(a) Mount the filtration funnel (i.e., housing) on to the adap-
tor. Remove the funnel from the base and place it upside
down on the bench on top of its lid.
(b) Take one pair of forceps that have been soaking in 100%
ethanol, flame to burn off the ethanol, allow forceps to
cool, then pick up the preloaded filter from the filtration
housing being careful not to damage the filter support
underneath the grid filter. Discard the gridded filter.
(c) Use the same forceps to carefully pick up one isopore
polycarbonate membrane (clear grayish membrane
in-between paper separators, do not mistake the paper sepa-
rator for the membrane filter), and carefully place it onto
the center of the filter support in place of the gridded
filter.
(d) Snap the filtration funnel back onto the housing to secure
the polycarbonate membrane (make sure that there is no
gap between the edge of polycarbonate membrane and the
bottom of the filtration funnel and that the edge of the
membrane is not folded, i.e., no liquid should go through
the housing/support without passing through the polycarbon-
ate membrane first).
ä
Fig. 2 Example 1-D plots for the two targets (A: Enterococcus, B: HF183) of the duplex ddPCR assays. Positive
control, no-template controls, and samples with no, low, and high levels of inhibitors are labeled on each well
on top of each plot as POS, NTC, and numeric numbers, respectively. Numeric numbers denote humic acid
concentration (spiked into the same sample) in ng per μL reaction. Green, orange, red, and black colors of the
label indicate no effect on quantification, underestimation, failure to detect, and NTC, respectively
3. Depending on turbidity of the water, smaller volumes may

need to be added incrementally (e.g., 20 mL) to the funnel
until the filter clogs or 100 mL successfully filters through the
membrane. The filtered volume must be recorded for data
analysis downstream.
4. Forceps must be used to move the filter to the final microtube.
The forceps must be flame sterilized but cannot be used right
away. If the forceps are too hot then portions of the sample
and/or filter will be destroyed. After flame sterilization, let
forceps cool for ~15 s.
5. The membrane can be folded twice before placing into the
2 mL screw-cap microtube. Keep in mind that the material
collected on the filter should not be touched or damaged
during folding as this compromises the integrity of the sample.
Other tubes compatible with 80 C storage and downstream
DNA extraction protocol can also be used.
6. If liquid nitrogen is not available, alternatives (such as etha-
nol–dry ice bath) suitable to flash-freezing a sample can also be
used. It is important to flash freeze samples to minimize enzy-
matic activity causing sample degradation on the membrane
filter.
7. DNA extraction using other procedures/kits can be used but
should be evaluated for recovery, purity, and the type of elution
buffer used. Recovery and purity of the DNA are necessary to
ensure the integrity of the sample. The elution buffer should
not contain chemicals that might interfere with downstream
analysis (e.g., ddPCR), either with enzyme function nor with
droplet formation and stability.
8. Bead beating is necessary to break open cells and free DNA.
However, too much bead beating may lead to DNA fragmen-
tation and sample degradation, which will alter the concentra-
tions of measurable target DNA downstream. One minute of
bead beating is sufficient for environmental samples using the
filters described in this protocol.
9. Total DNA concentration should be measured to ensure that
the ddPCR reaction is not overloaded. If DNA concentration is
greater than 13 ng/μL (i.e., 66 ng per ddPCR reaction), a
dilution should be considered when prepping ddPCR.
10. When setting up ddPCR reaction mix plates, it is convenient to
set up samples for each DG8 cartridge vertically in a plate
column.. All wells in a cartridge, and thus in a column if
using an eight-channel pipettor to load the samples, must be
filled in order for successful droplet generation. If all eight
unique samples on a chip are run with two technical replicates,
then both chips can be loaded from a single plate column
containing a double volume (2 24 μL) of reaction mix.
11. New pipet tips must be used for each column when mixing the
samples. During mixing, pipet less than 20 μL and keep the tips
in the liquid, as this avoids the formation of bubbles in the
sample. Ensure that all the sample mixture is pushed out of the
pipet tip before disposing.
12. When transferring the sample assay mixture from the hard-shell
96-well plate to the cartridge, ensure that all 20 μL of volume
has been taken up in each of the tips. Then placing the pipet tip
into the cartridge, hold the pipet perpendicular to the bench
top and ensure that each pipet tip touches the side near but not
touching the bottom of the cartridge. Once touching, slowly
depress the plunger to release ~20% of the volume. Continue
releasing volume slowly but work the pipet tip up the side of
the cartridge walls. This action ensures minimal bubbles are
introduced into cartridge. If excess bubbles are present droplet
generation will end prematurely.
13. The second cartridge may be prepped when the first cartridge
has been placed on the droplet generator. If the second car-
tridge corresponds to the replicate column of the first cartridge
then sample mixture is taken from the same column in the
hard-shell 96-well plate as when prepping cartridge one.
Remember that less volume is now present in the once-used
reaction mix column and pipetting must be done slowly to
ensure that a full 20 μL is put onto the second cartridge.
14. Transferring the droplets from the cartridge to the final PCR
96-well plate (such as the Eppendorf Twin Tec PCR plate)
must be done slowly and precisely. Using the 50 μL multichan-
nel pipet place the tips into the droplets at a 45 angle. The tips
should not touch any surface but the droplets. Slowly pipet the
droplets into the tips all the while following the liquid level as it
decreases. At this point, the tips will touch the hard surface of
the cartridge, continue to slowly pipet the droplets. Once the
maximum amount of volume has been pipetted place the tips
containing the droplets to the side of the final PCR plate at a
45 angle, close to the bottom of the well. Slowly depress the
pipet while transferring the droplets to the well of the droplet
PCR plate. Check that all the droplets have been recovered
from the cartridge. If not, pipet the remaining droplets out and
place them into the final plate in the same manner as just
described. Note that pipetting too quickly at this step will
shear the droplets and reduce the accuracy of the final
concentration.
15. The plate can only be placed on the droplet reader with well A1
in the top left corner. If the plate is placed rotated 180 , the
components will not sit correctly and reading cannot com-
mence. When the plate is placed on the reader, it is recom-
mended that the samples are at room temperature.
16. The QX100 droplet reader reads two fluorophores. FAM is

read on channel 1 and HEX or VIC are read on channel 2. In
this protocol FAM is used for the Enterococcus target and HEX
is used for human HF183 target.
17. A single manual fluorescence threshold applied to all samples is
recommended for the following reasons. First, quantification
of target concentrations have been shown to be relatively
insensitive to threshold position. Second, a single threshold
facilitates method implementation among water testing
practitioners.
18. This duplex ddPCR assay is much more resistant to PCR
inhibition than its qPCR counterparts [12]. The ddPCR
continued to provide quantification for both Enterococcus and
HF183 even at humic acid (HA) concentration of 15 ng per μL
reaction. Nevertheless, extremely high inhibitor concentration
did lead to underestimation (20–25 ng HA per μL reaction) or
false negatives (30 ng HA per μL reaction) (Fig. 2). Typically, a
lower fluorescence value for the positive droplets is observed
before underestimation occurs (Fig. 2).
19. Once the data is exported, multiply the exported target con-
centration by 4 to convert it from copies of target (23S gene of
Enterococcus spp. or the HF183 marker) per μL reaction to
copies of target per μL DNA template. Additional calculations
may be needed in order to account for filtration volume and to
compute the target copies per unit volume of sample water.
References
1. Boehm AB, Van De Werfhorst LC, Griffith JF 6. Griffith JF, Weisberg SB (2011) Challenges in
et al (2013) Performance of forty-one micro- implementing new technology for beach water
bial source tracking methods: a twenty-seven quality monitoring: lessons from a California
lab evaluation study. Water Res 47:6812–6828 demonstration project. Mar Techol Soc J
2. Soller JA, Schoen ME, Bartrand T et al (2010) 45:65–73
Estimated human health risks from exposure to 7. U.S. EPA (2012) Recreational water quality
recreational waters impacted by human and criteria. 820-F-12-058. Office of Water,
non-human sources of faecal contamination. Washington, DC
Water Res 44:4674–4691 8. Bernhard AE, Field KG (2000) A PCR assay to
3. Whitman RL, Nevers MB (2003) Foreshore discriminate human and ruminant feces on the
sand as a source of Escherichia coli in nearshore basis of host differences in Bacteroides-Prevo-
water of a lake Michigan beach. Appl Environ tella genes encoding 16S rRNA. Appl Environ
Microbiol 69:5555–5562 Microbiol 66:4571–4574
4. Field KG, Samadpour M (2007) Fecal source 9. Cao Y, Griffith JF, Dorevitch S et al (2012)
tracking, the indicator paradigm, and manag- Effectiveness of qPCR permutations, internal
ing water quality. Water Res 41:3517–3538 controls and dilution as means for minimizing
5. Layton BA, Cao Y, Ebentier DL et al (2013) the impact of inhibition while measuring
Performance of human fecal anaerobe- Enterococcus in environmental waters. J Appl
associated PCR-based assays in a multi- Microbiol 113:66–75
laboratory method evaluation study. Water 10. Cao Y, Sivaganesan M, Kinzelman J et al
Res 47:6897–6908 (2013) Effect of platform, reference material,
and quantification model on enumeration of
Enterococcus by quantitative PCR methods. system for absolute quantitation of DNA copy
Water Res 47:233–241 number. Anal Chem 83:8604–8610
11. Shanks OC, Sivaganesan M, Peed L et al 14. Cao Y, Raith MR, Griffith JF (2016) A duplex
(2012) Inter-laboratory general fecal indicator digital PCR assay for simultaneous quantifica-
quantitative real-time PCR methods compari- tion of the Enterococcus spp. and the human
son study. Environ Sci Technol 46:945–953 fecal-associated HF183 marker in waters. J Vis
12. Cao Y, Raith MR, Griffith JF (2015) Droplet Exp:e53611. https://doi.org/10.3791/
digital PCR for simultaneous quantification of 53611
general and human-associated fecal indicators 15. U.S. EPA (2012) Method 1611: enterococci in
for water quality assessment. Water Res water by TaqMan® quantitative polymerase
70:337–349 chain reaction (qPCR) assay. EPA-821-R-12-
13. Hindson BJ, Ness KD, Masquelier DA et al 008. Office of Water, Washington, DC
(2011) High-throughput droplet digital PCR
Part III
Copy Number Variation

Chapter 9
Analyzing Copy Number Variation with Droplet Digital PCR

Avery Davis Bell, Christina L. Usher, and Steven A. McCarroll
Abstract
Many genomic segments vary in copy number among individuals of the same species, or between cancer
and normal cells within the same person. Correctly measuring this copy number variation is critical for
studying its genetic properties, its distribution in populations and its relationship to phenotypes. Droplet
digital PCR (ddPCR) enables accurate measurement of copy number by partitioning a PCR reaction into
thousands of nanoliter-scale droplets, so that a genomic sequence of interest—whose presence or absence in
a droplet is determined by end-point fluorescence—can be digitally counted. Here, we describe how we
analyze copy number variants using ddPCR and review the design of effective assays, the performance of
ddPCR with those assays, the optimization of reactions, and the interpretation of data.
Key words Copy number variants, Genomic structural variation, Droplet digital PCR, Digital PCR,
Genotyping, Genotyping assay design
1 Introduction
Even within a single species, such as humans, thousands of genomic

segments vary in copy number from individual to individual. In the
context of cancer and other proliferative disorders, substantial parts
of the genome can also differ in copy number between disease and
healthy cells from the same person. Precisely measuring such differ-
ences is key to ascertaining their biological import.
Though precise and accurate measurement is critical in all
research, many research contexts present particular challenges for
accurate copy number determination. In cancer cells, many onco-
genes become amplified to high copy numbers. Many inherited
copy number variants (CNVs) are also present in a wide range of
copy numbers (e.g., from two to ten) within different individuals’
diploid genomes, due to multiallelism. In humans, such CNVs
appear to generate most inherited gene-dosage variation and
make a substantial contribution to gene-expression variation [1],
143
144 Avery Davis Bell et al.
suggesting that they may contribute to variation in phenotypes. To

understand how copy number variation contributes to phenotypes,
how alleles are distributed within and across populations, and how
CNVs relate to SNPs and haplotypes, it is crucial to accurately
measure (or “genotype”) such CNVs.
Droplet digital PCR (ddPCR) obtains precise and accurate
measurements of copy number by partitioning the reagents for
two fluorescence assays (one detecting the CNV of interest and
one detecting a control reference locus of known copy number)
into thousands of droplets—creating thousands of individual reac-
tions—and determining whether each droplet contained either
DNA molecule by measuring the fluorescence of each droplet
after PCR [2, 3]. Copy number is calculated by comparing the
number of molecules arising from the CNV segment of interest
(calculated from the number of positive droplets) to the number of
molecules arising from the reference genomic locus. Because the
fluorescence measurement is taken after (rather than during) PCR,
its accuracy relies only on distinguishing the fluorescence-positive
from the fluorescence-negative droplets, not on the quantitative
PCR kinetics that classical real-time PCR attempts to measure. This
yields a powerful improvement in the precision of analysis, allowing
a precise determination of integer copy number at loci where rtPCR
has been unable to do so [2–4]. For example, at the highly copy-
number variable sperm gene SPANXB [5], studies using qPCR
have only estimated the copy numbers that are present in each
genome [6, 7], whereas ddPCR can measure the precise, integer
level in each individual’s genome (Fig. 1b).
Here, we share a detailed protocol for analyzing copy number
variation with ddPCR including (1) designing successful assays
targeting genomic segments of interest, (2) using those assays in
ddPCR and optimizing reaction conditions, and (3) improving the
ddPCR analysis results after data generation. We pay particular
attention to assay design and optimization, which can greatly affect
data quality (Fig. 1). We have used this method for deep interroga-
tion of particular genomic regions, including characterization and
phenotype association analyses [4, 8], as well as for confirmation,
validation, and population-based analysis of copy number
variants [1].
While this protocol includes many details that are most helpful
for typing germline copy number variants in stable euploid gen-
omes, the protocol is readily adapted for analyzing copy-number-
variable segments in cancer genomes. A key difference is that
because cancer samples are often mosaic (a mixture of clones with
different genomes), analysis results for cancer samples may involve
noninteger copy-number levels that represent an average across the
cells in a sample. Another useful application of ddPCR involves
quantifying the copy number of transgenes.
ddPCR to Measure copy Number Variation 145
a. Un-optimized ddPCR
15
Counts
10
1 2 3 4 5 6 7 8 9 10 11
SPANXB copy number
b. Optimized ddPCR
25
20
Counts
15
10
5
0
1 2 3 4 5 6 7 8 9 10 11
SPANXB copy number
Fig. 1 ddPCR-generated copy numbers for 179 individuals at the SPANXB locus before (a) and after (b) the
assay and reaction optimization techniques outlined in this protocol. The optimized copy numbers were
generated by combining data from replicates run with two separate X chromosome-located replication-timing
matched control assays (see Notes 26–28)
2 Materials
Prepare all solutions with ultrapure, molecular biology-grade water.

Protect all solutions containing fluorescently labeled probes from
light. Mix all reagents by briefly vortexing and centrifuging them
before use.
2.1 Locus-Specific 1. Assay targeting CNV region of interest (20 target mix): 18 μM
Reagents forward primer, 18 μM reverse primer, and 5 μM 50 FAM-labeled,
30 ZEN or Black Hole-quenched probe designed to genomic
region of interest. To make, combine 25.2 μL of 100 μM forward
primer, 25.2 μL of 100 μM reverse primer, and 7 μL of 100 μM
probe with 82.6 μL water. Store at 20 C (see Note 1).
2. Assay targeting control region (20 control mix): 18 μM for-
ward primer, 18 μM reverse primer, and 5 μM 50 HEX-labeled,
30 ZEN or Black Hole-quenched probe designed to non-copy
number variable genomic region (see Notes 2 and 3). To make,
combine 25.2 μL of 100 μM forward primer, 25.2 μL of
100 μM reverse primer, and 7 μL of 100 μM probe with
82.6 μL water. Store at 20 C (see Note 1).
2.2 ddPCR 1. Genomic DNA at a concentration of 5 ng/μL or higher, total-

Components ing at least 50 ng (see Note 4).
and Equipment 2. Restriction enzyme and associated buffer for digesting geno-
mic DNA, potentially AluI with 10 CutSmart® buffer (New
England Biolabs) (see Note 5).
3. 2 ddPCR™ Supermix for Probes, with or without dUTP
(Bio-Rad).
4. DG8™ Cartridges for droplet generation (Bio-Rad).
5. DG8™ Gaskets for droplet generation (Bio-Rad).
7. Droplet Reader Oil (Bio-Rad).
8. QX200™ Droplet Digital PCR System: droplet generator and
cartridge holders, droplet reader, and QuantaSoft reader soft-
ware (Bio-Rad).
9. Rainin multichannel pipettors and corresponding tips for
pipetting 20 μL and 40 μL volumes (see Note 6).
10. Half-skirted Eppendorf twin.tec 96-well plates for droplet
thermal cycling and reading.
11. Pierceable, heat-sealable foil seals (Bio-Rad Pierceable Foil
Heat Seal).
12. Plate sealer capable of sealing for 5 s at 180 C (e.g., Bio-Rad
PX1™ Plate Sealer).
13. Thermal cycler.
2.3 Web Resources 1. UCSC genome browser (hg19): http://genome.ucsc.edu/cgi-

bin/hgGateway.
2. Primer3 primer design tool: http://bioinfo.ut.ee/primer3/
[9, 10].
3. SNP masking tool: http://bioinfo.ut.ee/snpmasker/ [11].
4. NEB cutter: http://nc2.neb.com/NEBcutter2/ [12].
5. IDT oligoanalyzer: http://www.idtdna.com/calc/analyzer.
6. Multiple primer heterodimer analyzer: http://www.
thermoscientificbio.com/webtools/multipleprimer/.
3 Methods
3.1 Assay Design 1. The first step of assay design is to determine the best region for
assay placement in order to optimize detection of the genomic
segment of interest and ddPCR performance. To begin, obtain
the DNA sequence for the copy-number-variable region of
interest by entering the coordinates spanning the region into
the UCSC genome browser. Determine whether the region is
present once or more than once in the reference genome by

displaying segmental duplications. Select “dense” from the
“Segmental Dups” pull-down menu under the “Repeats” sec-
tion at the bottom of the page (see Note 7). Many copy number
variants are found more than once in the reference genome,
raising special considerations; if this is the case for the region of
interest, see Note 8.
2. When the specific region of interest is identified and displayed
in the genome browser, set the “RepeatMasker” track (under
“Repeats”) pull-down menu to “dense” and reload the page.
Get the sequence for the visualized region by selecting “DNA”
under the “View” menu at the top of the page. In order to
prevent the assay from being designed to target repeat regions,
check the box next to “Mask repeats” and select “to N,” then
click the “get DNA” button (see Note 9).
3. Design the primers and probe to assay this region using the
Primer3 primer design tool. Enter the DNA sequence obtained
in step 1 into the box at the top of the webpage. Check
“Pick hybridization probe (internal oligo)” under the input
sequence.
(a) From the “Mispriming library (repeat library)” pull-down
menu above the sequence box, choose “HUMAN.”
(b) Under “General Primer Picking Conditions,” set the opti-
mal primer length (“Primer size”) to 22 bp, “Primer Tm”
Min to 59, Opt to 60, and Max to 61. Set the product size
range to 60–90 bp (which can be relaxed to 60–150 bp if
no assays are found) (see Note 10).
(c) Under “Internal Oligo (Hyb Oligo) General Condi-
tions,” set “Internal Oligo Tm” Min to 68, Opt to
69, and Max to 70, and choose “HUMAN” from the
“Internal Oligo Mishyb Library” pull-down menu.
Leave the remaining options unchanged and click “Pick
primers.” The temperatures can be adjusted if no suitable assays
are found, as long as the internal oligo (probe) melting tem-
perature is still higher than the primer temperature.
4. Choose an assay from the results of step 3 that contains any
necessary sequences and is likely to perform well. Avoid probe
sequences that start with G (see Note 11). Use the UCSC
BLAT tool (under the “Tools” menu at the top of the page)
to check that the forward and reverse primers match the region
of interest uniquely and perfectly (see Note 12). View the
region with the “Common SNPs” track displayed to ensure
the primers and probe do not bind over a SNP. In addition,
check whether the primers or probes are likely to bind each
other or the control assay by using the “Hetero-Dimer” option
in the right menu bar of IDT’s oligoanalyzer; delta-Gs lower
than 7 should be avoided because their heterodimerization

may interfere with the PCR (see Note 13).
5. Ensure that the amplicon generated by the primers does not
contain a cut site for the restriction enzyme that will be used to
digest the DNA prior to ddPCR. Obtain the amplicon
sequence from UCSC genome browser. Do not mask repeats
this time. Copy and paste this sequence into the NEB cutter
webpage. Select “All commercially available specificities” to the
right of “Enzymes to use,” then click the “Submit” button to
the right of the box containing the DNA sequence. Under the
resulting graphic in the “List” box, click “0 Cutters” and make
sure the enzyme of interest is included (see Note 14).
6. Order the primers and probe from your usual oligo supplier.
We prefer the FAM and HEX probe fluorophores, both with
the ZEN quencher (Integrated DNA Technologies), though
other combinations of fluorophores, quenchers, and suppliers
also perform well. When making or ordering the 20 assay
mix, please note that the proportion of primers to probes is
different than in qPCR.
3.2 ddPCR for Copy 1. Digest the genomic DNA with a restriction enzyme to separate
Number Determination the copies of the CNV. For each sample, make an enzyme
master mix consisting of 0.2 units/μL AluI and 2 CutSmart
buffer (New England Biolabs). Add 10 μL of this master mix to
50 ng DNA in 10 μL, for a total reaction volume of 20 μL. Mix
by pipetting up and down. Do not vortex the enzyme or
enzyme solution.
2. Incubate the enzyme-DNA mixture at 37 C for 1 h.
3. Dilute the digested DNA twofold by adding 20 μL of water to
each sample, yielding a DNA concentration of 1.25 ng/μL.
Keep digested DNA at 4 C or on ice for immediate use, or
at 20 C for long-term storage. (See Note 15 for an alterna-
tive restriction digestion strategy.)
4. For each sample add:
(a) 12.5 μL of 2 ddPCR Supermix for Probes (Bio-Rad).
(b) 1.25 μL of 20 assay targeting the CNV region.
(c) 1.25 μL of 20 assay targeting control region (these first
three reagents can be combined to form a master mix.).
(d) 10.0 μL of the digested, diluted DNA (see Notes 16 and 17).
5. Mix well by pipetting up and down ten times. Proper mixing is
critical. Spin the plate to collect the liquid at the bottom of
wells. Keep the plate protected from light until droplet genera-
tion, and allow the reactions to equilibrate to room tempera-
ture for 3 min prior to droplet generation.
6. Place a DG8™ cartridge into the QX200 droplet generation

cartridge holder and snap the holder closed. Pour Droplet
Generation Oil for Probes into a reservoir for ease of multi-
channel pipetting.
(a) Pipette 20 μL of the PCR mix into the middle row of the
cartridge (the smallest wells) (see Note 18). Only push
down to the first stop when ejecting liquid, and ensure
there are no air bubbles in the sample (see Note 19).
Using a Rainin multichannel pipettor with Rainin tips is
preferred at this stage (see Note 6).
(b) Pipette 70 μL of oil into the bottom row of wells in the
cartridge. Always be sure to pipette the oil after the sam-
ples. The top row is left empty.
(c) Place a DG8™ rubber gasket over the cartridge by hook-
ing the prongs of the cartridge holder through the gas-
ket’s four holes.
7. Place the cartridge holder with cartridge and gasket into the
QX200 droplet generator. Close the generator; droplets will be
formed. Prepare the next cartridge while the first set of droplets
is being generated.
8. When the triangles on the button on the lid of the droplet
generator return to being lit solid green and the generator
stops making noise, remove the cartridge. Carefully discard its
gasket and transfer the droplets in the top row to a clean, half-
skirted Eppendorf plate. The output sample has greater volume
than the input, so set the Rainin pipette to 40 μL. It is impor-
tant that the pipetting at this stage is slow and careful, with the
pipette oriented at 45 , otherwise the droplets may shear.
Afterward, discard the gasket and cartridge (see Note 20).
9. After all droplets are made, seal the droplet plate with a foil seal
by heating the seal on the plate to 180 C for 5 s.
10. Thermal cycle the plate as follows:
(a) 95 C for 10 min
(b) 40 cycles of 94 C for 30 s followed by 60 C for 1 min
(see Notes 21 and 22)
(c) 98 C for 10 min
(d) 8 C hold
Use a 2.5 C per cycle ramp rate for all steps. Droplets
can be stored protected from light at 4 C after cycling for up
to 24 h before reading.
11. Set up a template on the QX200 droplet reader computer.
Open QuantaSoft. Under “Template” in the top left corner,
select “New” in order to fill in a new plate map. To fill in the
information for each sample, double-click on the first
non-empty well. In the “Sample” box, under “Experiment,”

select any of the “CNV” experiments, and then, under “Super-
mix,” select “ddPCR Supermix for Probes” (see Note 23). In
the “Target 1” box, enter the name of the FAM assay in the
“Name” field. From the “Type” menu, select “Ch1
Unknown” if this is the target assay or “Ch1 Reference” if
this is the control assay. In the “Target 2” box, enter the
name of the HEX or VIC assay in the “Name” field. From
the “Type” menu, select “Ch2 Reference” if this is the control
assay or “Ch2 Unknown” if this is the target assay. Without
closing this menu or double clicking, select all wells of the plate
that will contain samples that are using the same assays. Click
the blue “Apply” button in the top window to set the assays
and experiment for all these wells. Once finished, click “OK”
and save the template.
12. Read the droplets on the QX200 droplet reader. Put the plate
into the plate holder in the QX200 compartment under the
door, then place the black plate holder on top and click the
silver tabs on either side down into place, making sure the A1
well is in the top left corner. Close the lid of the QX200. In
QuantaSoft on the QX200 computer, make sure the template
created in step 10 is loaded, then click “Run” in the column of
options to the left of the plate map. On the popup menu that
appears, select “FAM/HEX” or “FAM/VIC,” depending on
the pair of fluorophores used, then click “OK.”
3.3 Data Finalization 1. While the initial output from QuantaSoft can be sufficient for
and Quality Control downstream data analysis, careful quality control and optimi-
zation of this data often yields more accurate, more reliable
copy number calls. So when all the wells containing samples
have been run, perform a well-by-well visual inspection of
droplet clusters in QuantaSoft by clicking “Analyze” on the
leftmost menu followed by “2D Amplitude.” Ensure that there
is clear separation between the positive and negative clusters for
both the target and reference assay channels. Some bleeding of
droplets between the positive and negative channels (some-
times referred to as “rain”) is acceptable, but a substantial
amount can cause inaccuracy (Fig. 2). If only a few samples
show poor cluster separation, exclude these from analysis. If all
wells have bad cluster separation, see Notes 17, 21, and 22 or
redesign the assay according to Subheading 3.1.
2. Determine that the software has made the correct call for each
droplet cluster in each well. Make sure that all droplets are
correctly labeled by the software: droplets in the top left corner
of the 2D amplitude plot are FAM positive only, droplets in the
bottom right corner are HEX/VIC positive only, droplets in
the top right corner are positive for both fluorophores, and
Fig. 2 Common assay issues and solutions. (a) Examples of common issues. A poor separation of clusters (left)
can be resolved by optimizing the thermocycling conditions or assay design. Two positive clusters (center)
likely result from a SNP being in the assay-binding region or the amplification of a secondary genomic region.
Droplet shearing or excess rain (right) can result from not handling the droplets properly. Assays displaying
these characteristics should be redesigned or optimized following the suggestions in the protocol. (b)
Examples of PCR reaction optimization. A temperature gradient (left) can be used to determine the optimal
annealing temperature for the PCR, as shown by greatest cluster separation (here, 56.4 C yields the cleanest
clusters). Increasing the number of PCR cycles from 40 (center) to 50 (right) can increase cluster separation to
an acceptable amount
droplets in the bottom left corner are negative for both fluor-
ophores. If a well has some droplets called incorrectly, manually
assign them to clusters. (Using QuantaSoft version 1.6.6, this
is accomplished by designating the groupings with the
“Threshold” or “Lasso” tools). For wells where droplets have
been correctly assigned to clusters, make sure the “Status”
column is set to “OK”—if it says “Check,” click anywhere in
the amplitude plot to get the software to recognize the data (see
Note 24).
3. Export data for all wells that passed visual inspection. Select
these wells and click “Export CSV.”
4. Perform further sample-level quality control (see Note 25).
Exclude from analysis samples with data drawn from fewer
than 5000 droplets (“AcceptedDroplets” column of the
exported CSV). Mark samples with mid-integer CNV calls
(those 0.35–0.65 away from an integer number). Samples
with CNV confidence intervals wider than 1 and samples with
fewer than 10% double negative droplets as unreliable; opti-
mize and rerun them (see Note 26).
5. Repeat the ddPCR for the rare individual samples that failed
visual inspection (step 1) or quality control (step 4) (see Note
27). If many samples failed a given test, there may be a systemic
issue that needs to be remedied before repeating (see Notes
28–30) (Figs. 2 and 3).
Fig. 3 An example of assay optimization for an exceptionally difficult CNV locus (AMY1 in the amylase locus).
The cluster plots (top and middle) and final copy number calls (bottom) both improve after assay and
PCR-reaction optimization. The reaction was optimized by designing an assay that conformed to the assay
design guidelines in the Methods, running a melting temperature gradient to determine the optimal melting
temperature (see Note 20), adding ten extra cycles to the PCR (see Note 19), using a replication-matched
control assay (see Note 26), and using the optimal amount of DNA input (see Note 24). It can be improved
further by averaging replicates
6. For germline CNV studies where integer copy numbers are

expected, round the copy number calls to the nearest integer
for all wells that pass visual inspection and quality control if
copy numbers generally cluster around integers. If not, a sys-
temic issue may need to be remedied (see Notes 28 and 30)
(Figs. 2 and 3).
4 Notes
1. 20 assay mixes should be kept at 20 C for long-term

storage, but avoid repeated freezing and thawing. We have
found 20 mixes to be stable at 4 C for at least 1 month.
2. Any region of the genome that is known or strongly expected
to be invariant in copy number can be used as a control. For
human genomes, a particularly well-validated assay targets the
RPP30 gene and is a useful control to use as a starting point for
target assay testing. The sequences for this assay are: forward
primer, 50 -GATTTGGACCTGCGAGCG-30 ; reverse primer,
50 -GCGGCTGTCTCCACAAGT-30 ; probe, 50 -CTGACCT
GAAGGCTCT-30 . When working with aneuploid samples
(e.g., those from cancers), it may be necessary to either use
multiple control assays or empirically determine the copy num-
ber and copy number stability of the control locus.
3. FAM and HEX probes can be ordered through Integrated
DNA Technologies. VIC-labeled probes can be ordered
through Life Technologies. HEX and VIC probes are read in
the same channel in ddPCR, so either a HEX or a VIC probe
can be paired with a FAM probe. To minimize costs, we use the
more-expensive HEX or VIC fluorophores for control assays,
and FAM for the (more numerous and diverse) target locus
assays.
4. This protocol is suitable for high-quality, nondegraded DNA
from any source (e.g., from cell lines, PBMCs, and fresh tis-
sues). We have had success using ddPCR to type copy number
variation in DNA extracted using Qiagen’s DNeasy DNA
extraction kits. When using DNA from sources, such as FFPE
tissue and urine, in which the DNA may be degraded, special
assay design considerations should be taken into account (see
Note 10).
5. To obtain accurate copy number calls, it is crucial that the DNA
is digested. In particular, it is important that the restriction
enzyme cuts between the assay sites (and not within them).
This ensures that intact, individual copies of the region of
interest segregate independently into droplets. Any restriction
enzyme that accomplishes this goal can be used; AluI is pre-
ferred because its recognition site occurs frequently in DNA.
6. Using low-quality pipette tips can cause droplet shredding
during droplet generation, likely because they shed tiny pieces
of plastic into the reaction. Rainin pipettors and tips perform
extremely well when working with droplets but are not strictly
necessary in all applications. The use of Rainin pipettors and
tips at all stages of DNA extraction, preparation, and droplet
generation ensures that plastic particles are not present in the

ddPCR reaction and is considered as the best practice.
7. The segmental duplication track on the UCSC browser
includes only regions larger than 1 kilobase; shorter regions
of high identity will be missed. For advanced assay design, to
determine whether the region of interest is within one of these
shorter regions, display the “Mapping” track under the
“Mapping and Sequencing” menu. Regions with high (dark)
uniqueness values are present only once on the reference, while
regions with lower uniqueness values are present more
than once.
8. If the region of interest is within a segmental duplication (i.e.,
present more than once in the reference genome), there are
likely differences between the two (or more) copies of the
region present (“paralogs”). In this situation, an assay can be
designed to target a specific paralog or all copies of the region,
depending on the goal of the experiment. For example, a
specific paralog might be targeted if its particular function is
of interest, while total copy number might be desired if differ-
ences between the paralogs are not expected to impact the
biological question of interest.
If one paralog is to be targeted, design the assay to exploit
differences between the two copies. To find these differences,
use the segmental duplication track within the UCSC genome
browser to identify the locations of the duplicated regions,
then align their sequences and search for sites with several
nucleotide differences (termed paralogous sequence variants
or PSVs). Design the assays to include these PSVs, particularly
by placing the PSVs in the probe-binding region or in the 30
end of primers.
If total copy number is to be targeted, the assay should be
designed to avoid differences between the copies. Perform the
alignment of the paralogs (segmental duplications) as above,
but find regions that do not contain PSVs. Restrict the region
used for assay targeting to these PSV-free regions.
9. SNPs in the primer or probe binding sites can prevent or hinder
assay binding, so the sites of common SNPs must be excluded
from the sequence used to design assays. If the region of
interest is duplicated on the reference, make sure to avoid
SNPs that occur in any copies of the sequence. One way to
avoid SNPs is to turn on the “CommonSNPs(138)” or “Com-
monSNPs(141)” track under the “Variation” section at the
bottom of the USCS genome browser page and subsequently
narrow the region of sequence to avoid any common SNPs.
This is workable for small regions, but can be tedious for larger
ones. Another option is to use the SNP masking tool website:
after masking the repeat regions using the UCSC genome
browser, feed that sequence into the SNP masking tool to

create a sequence where all SNPs and repeat regions are masked
to “N.”
10. When working with high-quality DNA, it is generally unneces-
sary to match or restrict amplicon sizes beyond the guidelines
presented in the Methods. However, if the DNA is composed
of short fragments due to degradation or shearing, having PCR
amplicons of different lengths can result in a bias, as longer
stretches of DNA are less likely to be intact than shorter
stretches. Designing target and reference assays that have simi-
lar, preferably short, amplicon lengths can maximize and match
amplification efficiency between the target and reference
regions.
11. G nucleotides at the beginning of probes can quench nearby
fluorophores. If Primer3 suggests a probe beginning in G, use
the probe’s reverse complement sequence as the assay probe.
(Redesign assays if the reverse complement also begins with a G.)
12. If the primers are too short to BLAT, use the in silico PCR
function (“In-Silico PCR” under the UCSC genome browser
“Tools menu”) instead, though BLAT is preferable. If the
region of interest is in a segmental duplication and the assay
is designed to target one copy specifically, one unique and
perfect match may not be possible; make sure that the best
match is the duplication of interest. If the assay is designed to
capture all copies of a region, make sure all of these regions are
present in BLAT’s output.
13. Multiple pairs of oligos can be checked for heterodimerization
using ThermoScientific’s multiple primer tool. This tool is
quite sensitive, so use it for preliminary screening and then
check any proposed heterodimers with the IDT oligoanalyzer.
Paste the named oligo sequences (tab delimited) into the box
at the top of the ThermoScientific multiple primer web page to
get results.
14. If the amplicon does contain a restriction site for the selected
enzyme, change either the enzyme or the assay, making sure
that any new enzyme is compatible with the control assay.
Generally, it is simpler to redesign the assay, unless the genomic
context restricts the assay to a very specific sequence.
15. It is possible to digest the DNA in the ddPCR reaction mix-
ture, rather than predigesting the DNA as explained in Sub-
heading 3.2, steps 1–3. This in-Supermix digestion is useful
when the sample is limited, as a lower total amount of sample
can be used. To perform this in-Supermix digestion, include
2–5 units of restriction enzyme diluted to a volume of 1 μL in
the enzyme’s buffer in the ddPCR reaction described in Sub-
heading 3.2, step 4, and decrease the volume of the
DNA–water mixture commensurately. Undigested DNA of

high concentration, totaling 10 ng, should be substituted for
the digested, diluted DNA.
16. ddPCR droplets are made in sets of eight samples at a time and
read in 96-well plate format, so it is easiest to set up the PCR in
a 96-well plate.
17. Adding 10 μL of DNA equates to using 10 ng of DNA in the
assay, since only 20 μL of this PCR mix is used for droplet
generation. Ten nanograms of DNA is generally a good start-
ing amount for ddPCR, but often individuals with high copy
numbers need to be regenotyped using half this much DNA to
avoid overwhelming the droplets with CNV-containing mole-
cules. In general, if double-negative droplets constitute less
than 10% of the droplets generated, decrease the input concen-
tration of DNA; if the error bars on the CNV estimate are too
large, increase the input concentration of DNA. In all cases,
keep the volume of DNA and water added constant at 10 μL.
We often genotype all samples using 10 μL digested DNA
input, then regenotype individuals with high copy numbers
or low numbers of double negative droplets using 5 μL
digested DNA and 5 μL water.
18. Though only 20 μL of the 25 μL PCR mix is used for droplet
generation, the 5 μL excess prevents air bubbles from being
pipetted into the droplet generation reaction, ensuring that the
full 20 μL is converted into droplets. If the sample is limited,
however, a PCR mix with final volume of 22 μL can be
substituted.
19. Pushing the pipette down to the final stop introduces air
bubbles, which compromises the number and quality of dro-
plets. Pipette only until the first stop. If air bubbles are intro-
duced into the sample chamber, manually pop them with a
clean pipette tip to improve droplet generation.
20. An automatic droplet generator (Bio-Rad QX200 AutoDG)
and associated consumables can be used instead of the manual
droplet generation described in Subheading 3.2, steps 6–8.
The AutoDG can be run using increments of eight samples,
though some reagents are partitioned into 32-sample sets. The
AutoDG is best suited for use with full 96-well plates.
21. If clusters are too close together on the scatterplot, the intensity,
and thus the separation of the fluorescent signals, may be
increased by adding ten extra cycles to the PCR (Figs. 2b and 3b).
22. Although assays are designed to work best at 60 C, certain
assays—or combinations of assays—may give cleaner data at
another temperature. We find it best to run a temperature
gradient (55–65 C) on one sample to determine which tem-
perature yields the cleanest, most clearly separated clusters
(Fig. 2b).
23. We like to leave the “Name” field blank and merge the final
data with a plate map using Excel or another statistical pro-
gram. Otherwise the name of each and every sample must be
entered by hand.
24. The software only calls wells with 10,000 or more droplets and
sets the “Status” column to “Check” for wells with fewer
droplets, but we have found that CNV calls are reliable down
to 5000 droplets, at least for individuals carrying 0–3 copies.
25. This quality control step can be performed in any software for
quantitative or statistical analysis. Doing it in R is convenient
for automation and repetition, but it can be done in Excel
either manually or with formulas.
26. Wells with fewer than 5000 accepted droplets may not contain
enough droplets to accurately determine copy number, espe-
cially when copy number is above four. Wells with mid-integer
CNV calls are not informative (e.g., it is not clear whether a
called copy number of 3.5 corresponds to an actual copy
number of 3 or 4). Wide confidence intervals suggest the
DNA concentration was too low to make a definitive call.
Reactions in which fewer than 10% of droplets are negative
typically involve situations in which the reaction is too close to
saturation with DNA template. In these situations, the Poisson
statistics used to estimate the number of droplets with more
than one locus copy may be inaccurate, and it may be preferable
to rerun the reaction with a lower concentration of genomic
DNA. If using a lower concentration of input DNA results in a
reference concentration too low to be reliable, multiple wells
can be run for each sample, with the resulting data merged
during analysis to increase precision.
27. Increasing input DNA concentration typically decreases confi-
dence interval size; samples that failed quality control because
of CNV confidence interval size should be repeated with
higher DNA input. Mid-integer CNV calls for high copy num-
bers (above six or so) can often be resolved by repeating the
assay using a lower amount of input DNA. For mid-integer
calls with lower copy numbers, see Note 25.
28. An over-abundance of mid-integer copy number calls can be
caused by degraded DNA, undigested DNA, or an incompati-
bility between the target and control assays. (For cancer sam-
ples, it can also reflect clonal mosaicism or mixtures of tumor
and stromal cells, and therefore will not benefit from the
corrections proposed here.) When DNA is derived from repli-
cating cells (such as a cell line), another cause of mid-integer
calls is a difference in the replication timing of the control and
target loci. DNA replication occurs in different stages across
the genome; this timing is heritable, visible in sequencing data,
and largely the same across individuals [13–15]. DNA from

genomic regions that replicate early in the cell cycle is more
abundant in asynchronous cell culture, because these regions
exist in a duplicated state for much of the cells’ lives. Most
genes (and as a consequence, most popular control assays) are
in these early-replicating regions.
Mid-integer copy number calls are often observed in
regions of the genome that replicate late but were paired with
an early-replicating control region for ddPCR. In our experi-
ence, this discrepancy results in copy number calls that are
about 10% under the true call, though this amount varies
from sample to sample depending on the proportion of repli-
cating cells at the time of DNA extraction. This can have a large
impact, especially on samples with high copy number. Design-
ing a control assay to a region of the genome that replicates at
the same time as the region of interest improves copy number
analysis. Ideally this control assay could be located very close to
(but still genomically outside) the CNV. Replication profiles
for lymphoblastoid cell lines can be found using the data from a
recent study of replication timing in humans [15]. (Figures 1
and 3b demonstrate using replication-matched controls as well
as other optimizations.)
29. In germline-CNV analyses that use DNA derived from prolif-
erating cells (such as the HapMap and 1000 Genomes Project
DNAs, widely used as controls), we have found that copy
number calls for CNVs on the X chromosome can be improved
by using a control assay targeted to nearby X chromosome
sequence. The late and unstructured replication of the inactive
X chromosome in females and the resulting varying number of
X chromosomal regions in asynchronous cell culture may
explain this phenomenon [16]. (Figure 1 demonstrates using
an X chromosome control as well as other optimizations.)
When using an X chromosomal control assay, make sure to
divide the CNV estimates for males given by QuantaSoft by
two, as QuantaSoft assumes a diploid control is used and males
are haploid for the X chromosome.
30. For germline CNVs, for which integer copy numbers are
expected, we have also found that using two different control
assays in separate reactions (or two slightly different target
assays) and pooling the data to obtain a final copy number
increases the proportion of samples with clear integer copy
number calls, especially for samples with high copy number.
If the same input DNA concentration is used for both repeti-
tions, data can be pooled at the droplet level and then reana-
lyzed with Poisson statistics. However, if the DNA input was
changed between the replicates, data should be pooled by
averaging the copy number calls. (Figure 1 demonstrates
pooling the data from two control assays as well as other

optimizations.) Alternatively, two control assays with the
same fluorophore can be used in the same reaction with the
target assay, creating a synthetic four-copy reference. This
method may increase the precision of calls made from a single
reaction for genomic segments that are present at high copy
numbers.
For a few loci, ddPCR may tend to slightly undercount or
overcount a genomic locus for an unknown reason; this effect is
usually very small at low copy numbers but can become more
visible at high copy numbers. If all copy number measurements
trend away from integers in the same direction (e.g., if all copy
numbers tend to be below integer values), applying a plate-wide
multiplicative correction factor that moves all measurements
closer to the corresponding integer value appears to be a legiti-
mate correction (as validated by correspondence to sequencing-
based measurements of copy number). If attempting this, opti-
mize this correction factor by multiplying the copy numbers by a
series of factors between 0.9 and 1.1 (in increments of 0.001)
and choose the factor that gives the lowest overall deviation
from the closest integer (summing the absolute values of the
deviations). Generally, a correction factor within 3% is optimal.
Acknowledgment
Our understanding of CNVs and assays has benefited greatly from

interactions with our colleagues Robert Handsaker, Aswin Sekar,
and Linda Boettger. We also thank Katherine Tooley for helpful
discussions of this protocol. This work was supported by a grant
from the National Human Genome Research Institute (R01
HG006855, to S.A.M.).
References
1. Handsaker RE, Van Doren V, Berman JR, White HE, Regan JF, Karlin-Neumann GA,
Genovese G, Kashin S, Boettger LM, McCar- Hindson CM, Saxonov S, Colston BW (2011)
roll SA (2015) Large multiallelic copy number High-throughput droplet digital PCR system
variations in humans. Nat Genet 47 for absolute quantitation of DNA copy num-
(3):296–303. https://doi.org/10.1038/ng. ber. Anal Chem 83(22):8604–8610. https://
3200 doi.org/10.1021/ac202028g
2. Hindson BJ, Ness KD, Masquelier DA, 3. Pinheiro LB, Coleman VA, Hindson CM,
Belgrader P, Heredia NJ, Makarewicz AJ, Herrmann J, Hindson BJ, Bhat S, Emslie KR
Bright IJ, Lucero MY, Hiddessen AL, Legler (2012) Evaluation of a droplet digital polymer-
TC, Kitano TK, Hodel MR, Petersen JF, Wyatt ase chain reaction format for DNA copy num-
PW, Steenblock ER, Shah PH, Bousse LJ, ber quantification. Anal Chem 84
Troup CB, Mellen JC, Wittmann DK, Erndt (2):1003–1011. https://doi.org/10.1021/
NG, Cauley TH, Koehler RT, So AP, Dube S, ac202578x
Rose KA, Montesclaros L, Wang S, Stumbo 4. Usher CL, Handsaker RE, Esko T, Tuke MA,
DP, Hodges SP, Romine S, Milanovich FP, Weedon MN, Hastie AR, Cao H, Moon JE,
Kashin S, Fuchsberger C, Metspalu A, Pato https://doi.org/10.1093/bioinformatics/

CN, Pato MT, McCarthy MI, Boehnke M, Alt- btm091
shuler DM, Frayling TM, Hirschhorn JN, 10. Untergasser A, Cutcutache I, Koressaar T, Ye J,
McCarroll SA (2015) Structural forms of the Faircloth BC, Remm M, Rozen SG (2012)
human amylase locus and their relationships to Primer3—new capabilities and interfaces.
SNPs, haplotypes and obesity. Nat Genet 47 Nucleic Acids Res 40(15):e115. https://doi.
(8):921–925. https://doi.org/10.1038/ng. org/10.1093/nar/gks596
3340 11. Andreson R, Puurand T, Remm M (2006)
5. Kouprina N, Pavlicek A, Noskov VN, SNPmasker: automatic masking of SNPs and
Solomon G, Otstot J, Isaacs W, Carpten JD, repeats across eukaryotic genomes. Nucleic
Trent JM, Schleutker J, Barrett JC, Jurka J, Acids Res 34(Web Server):W651–W655.
Larionov V (2005) Dynamic structure of the https://doi.org/10.1093/nar/gkl125
SPANX gene cluster mapped to the prostate 12. Vincze T, Posfai J, Roberts RJ (2003) NEB-
cancer susceptibility locus HPCX at Xq27. cutter: a program to cleave DNA with restric-
Genome Res 15(11):1477–1486. https://doi. tion enzymes. Nucleic Acids Res 31
org/10.1101/gr.4212705 (13):3688–3691
6. Salemi M, Bosco P, Cali F, Calogero AE, Soma 13. Hiratani I, Takebayashi S, Lu J, Gilbert DM
PF, Galia A, Lanzafame M, Romano C, (2009) Replication timing and transcriptional
Vicari E, Grasso G, Sirago P, Rappazzo G control: beyond cause and effect—part II. Curr
(2008) SPANX-B and SPANX-C (Xq27 Opin Genet Dev 19(2):142–149. https://doi.
region) gene dosage analysis in Sicilian patients org/10.1016/j.gde.2009.02.002
with melanoma. Melanoma Res 18
(4):295–299. https://doi.org/10.1097/ 14. Koren A, Handsaker RE, Kamitaki N, Karlic R,
CMR.0b013e32830aaa90 Ghosh S, Polak P, Eggan K, McCarroll SA
(2014) Genetic variation in human DNA repli-
7. Hansen S, Eichler EE, Fullerton SM, Carrell D cation timing. Cell 159(5):1015–1026.
(2010) SPANX gene variation in fertile and https://doi.org/10.1016/j.cell.2014.10.025
infertile males. Syst Biol Reprod Med
55:18–26. https://doi.org/10.3109/ 15. Koren A, Polak P, Nemesh J, Michaelson JJ,
19396360903312015 Sebat J, Sunyaev SR, McCarroll SA (2012)
Differential relationship of DNA replication
8. Boettger LM, Handsaker RE, Zody MC, timing to different forms of human mutation
McCarroll SA (2012) Structural haplotypes and variation. Am J Hum Genet 91
and recent evolution of the human 17q21.31 (6):1033–1040. https://doi.org/10.1016/j.
region. Nat Genet 44(8):881–885. https:// ajhg.2012.10.018
doi.org/10.1038/ng.2334
16. Koren A, McCarroll SA (2014) Random repli-
9. Koressaar T, Remm M (2007) Enhancements cation of the inactive X chromosome. Genome
and modifications of primer design program Res 24(1):64–69. https://doi.org/10.1101/
Primer3. Bioinformatics 23(10):1289–1291. gr.161828.113
Chapter 10
Assessing HER2 Amplification in Plasma cfDNA

Isaac Garcia-Murillas and Nicholas C. Turner
Abstract
Digital PCR (dPCR) is a highly accurate method to determine DNA concentration. In dPCR, DNA is
portioned into many discrete single entities, and these are analyzed individually for the presence or absence
of a target molecule of interest. Here we describe how digital PCR can be employed to determine the
presence of oncogenic amplification through noninvasive analysis of circulating free DNA (cfDNA), and
exemplify this approach by developing a plasma circulating free DNA dPCR assay for HER2 copy number.
Key words Breast cancer, HER2, Circulating free DNA, Plasma, Digital PCR
1 Introduction
Genomic amplifications are important therapeutic targets. In routine

clinical practice, the presence of amplifications is determined by
analysis of a tumor biopsy at initial diagnosis. To optimally deliver
targeted therapy, repeated sampling of a tumor is required to deter-
mine whether the genetic profile of a cancer has changed following
prior therapy. This would require repeated biopsies of recurrent and
metastatic cancers, yet this approach has limitations. Biopsy has asso-
ciated risks, may be technically challenging depending on the site
(s) of relapsed disease, and is often costly. A biopsy usually samples
only a single area of tumor and in heterogeneous tumors may under-
estimate the array of genetic aberrations present [1]. Ideally, to
overcome these limitations, and to allow repeated sampling, the
presence of genomic amplifications could be diagnosed noninvasively.
DNA arising from tumor cells is found in the plasma of patients
with cancer and this circulating DNA represents a potential source
to noninvasively analyze tumor DNA [2]. High sensitivity assays of
coding mutations on circulating free DNA (cfDNA) have reported
high concordance with cancer mutational status [3, 4]. Analysis of
cfDNA is noninvasive, can be repeated at multiple occasions
throughout the disease course, and potentially may assess the full
heterogeneity of genomic aberrations present. Analysis of cfDNA
161
162 Isaac Garcia-Murillas and Nicholas C. Turner
requires an assay of high sensitivity as DNA is frequently present at

only low concentration in plasma and tumor cell derived DNA may
be only a small fraction of the total plasma DNA [2, 5] with the
remainder being derived from non-tumorous cells.
Digital PCR (dPCR) has the potential to accurately quantify the
concentration of nucleic acids in a sample, to a much greater degree
than traditional quantitative PCR (qPCR), by counting individual
DNA molecules [6]. To examine the potential of dPCR for amplifi-
cation detection we developed an assay to test for HER2 amplifica-
tion on cfDNA extracted from plasma [7]. This assay can also be used
to test for HER2 amplification on formalin fixed paraffin embedded
(FFPE) tissue and Fresh Frozen Tissue [8]. Chromosomal aneu-
ploidy complicates copy number assessment in plasma, and the
appropriate selection of a control probe. Here, we employ a copy
number reference gene on the same chromosomal arm as HER2 that
is very rarely coamplified with HER2, but is of stable neutral relative
copy number in nonamplified cancers.
2 Materials
Ensure that the work is divided into clearly separated plasma pro-
cessing, pre-dPCR, and post-dPCR areas and that no reagents or
equipment is shared between spaces. This includes lab coats,
reagents, plasticware, lab books, notebooks, computers, and any
other material.
2.1 Blood Collection, Blood should be collected into Vacutainer Plastic K2 EDTA
Plasma Separation, (BD Biosciences) blood collection tubes as per local guidelines.
and Storage Blood should be separated using a horizontal rotor swing out
head centrifuge ideally within 2 h of collection, aliquoted into
cryogenic vials and stored at 80 C. Avoid repeated freeze-thaw
cycles of plasma aliquots.
2.2 Circulating Free Circulating free DNA (cfDNA) should be extracted from plasma
DNA (cfDNA) using Qiagen’s QIAamp circulating nucleic acid kit on a dedicated
Extraction from plasma processing area using a QIAVac 24 Plus vacuum as instructed
Plasma by the manufacturer. cfDNA should be eluted into nuclease-free
nonstick 1.5-ml microcentrifuge tubes. Extracted cfDNA can be
stored short-term at 20 C. For long-term storage, consider stor-
ing cfDNA at 80 C. Avoid repeated thaw–freeze cycles of
extracted cfDNA and consider aliquoting it into smaller volumes.
2.3 Quantification cfDNA should be quantified using a fluorescence-based method like

and Quality Control Quant-iT PicoGreen dsDNA assay (Thermo Fisher Scientific) or
of cfDNA Extracted Qubit dsDNA assays (Thermo Fisher Scientific) on 1 μl of eluate as
from Plasma per manufacturer’s instructions. Do not quantify cfDNA using a
spectrophotometric method (260/280) as it will not be accurate
enough. Alternatively cfDNA can be quantified using Agilent’s
Assessing HER2 Amplification in Plasma cfDNA 163
Table 1
Sequences for the primers and probes used for the HER2 amplification assay
Sequence 50 –30
HER2 assay
Forward primer sequence ACAACCAAGTGAGGCAGGTC
Reverse primer sequence GTATTGTTCAGCGGGTCTCC
Probe sequence 6-FAM/CCCAGCTCTTTGAGGACAAC/MGBNFQ
EFTUD2 assay
Forward primer sequence GGTCTTGCCAGACACCAAAG
Reverse primer sequence TGAGAGGACACACGCAAAAC
Probe sequence VIC/GGACATCCTTTGGCTTTTGA/MGBNFQ
Table 2
Constituents for the HER2 Primer-Probe cocktails
20 HER2 cocktail 20 EFTUD2 cocktail

90 μl 100 μM primer forward 90 μl 100 μM primer forward
90 μl 100 μM primer reverse 90 μl 100 μM primer reverse
25 μl 100 μM HER2 FAM/MGBNFQ probe 25 μl 100 μM EFTUD2 VIC/MGBNFQ probe
295 μl nuclease-free H2O 295 μl nuclease-free H2O
Bioanalyzer/Tapestation (or similar) when checking its quality.

cfDNA quality should be checked using Agilent’s Bioanalyzer/
Tapestation (or similar). These procedures should be carried out on
a dedicated pre-PCR area and with dedicated pre-PCR equipment.
2.4 HER2 Droplet Droplet digitalTM PCR (ddPCRTM) should be performed on

Digital PCR Bio-Rad’s QX100/QX200, as described in Subheading 3, in a
pre-PCR area. Reactions should be performed using 2 ddPCR
Supermix for probes (Bio-Rad). Primers and probes (Table 1)
should be used at a final concentration of 900 nM and 250 nM
respectively on a final volume of 20 μl PCR mix. Prepare Primer-
Probe cocktails as shown in Table 2.
3 Methods
3.1 Venous Blood 1. Collect blood via vein puncture directly into an EDTA blood
Collection, Isolation tube.a
of Plasma from Blood 2. Centrifuge at 1,600 g for 20 min at room temperature on a
and Storage horizontal rotor, swing out head centrifuge.b
Pipette plasma into

Plasma cryovial leaving a small
ammount (~5mm) above
Buffy Coat
Buffy Coat
Erythocytes
Fig. 1 When separating plasma for extraction of cfDNA, only the top layer should
be taken. The buffy coat layer containing white blood cells should be left
undisturbed
3. Remove plasma carefully without disturbing the buffy coat

layer (white blood cells and platelets) and transfer into a clean
conical centrifuge tube (Fig. 1).c
4. Centrifuge plasma at 16,000 rpm for 10 min to pellet cellular
debris.
5. Pipette off all the plasma into labeled Cryogenic vials.d
6. Place the cryogenic vials containing plasma upright into a
freezer storage box in a 80 C freezer.e
3.2 Circulating Free Extract cfDNA from 2 ml of plasma using the QIAamp circulating
DNA Extraction from nucleic acid kit from Qiagen. Ensure that the plasma has been
Plasma Using a QIAvac thawed on ice before starting the procedure.
1. Pipet 200 μl QIAGEN Proteinase K into a 50 ml
centrifuge tube.
2. Add 2 ml of plasma to the 50 ml tube.
3. Add 1.6 ml Buffer ACL (containing 1.0 μg carrier RNA). Close
the cap and mix by pulse-vortexing for 30 s.a
4. Incubate at 60 C for 30 min.
5. Place the tube back on the lab bench and unscrew the cap.
6. Add 3.6 ml Buffer ACB to the lysate in the tube. Close the cap
and mix thoroughly by pulse-vortexing for 15–30 s.
7. Incubate the lysate–Buffer ACB mixture in the tube for 5 min
on ice.
8. Insert the QIAamp Mini column into the VacConnector on the
QIAvac 24 Plus. Insert a 20 ml tube extender into the open
QIAamp Mini column.b
9. Carefully apply the lysate–Buffer ACB mixture from step 7 into
the tube extender of the QIAamp Mini column. Switch on the
vacuum pump. When all lysate have been drawn through the
column completely, switch off the vacuum pump and release
the pressure. Carefully remove and discard the tube extender.c
10. Apply 600 μl Buffer ACW1 to the QIAamp Mini column.
Leave the lid of the column open, and switch on the vacuum
pump. After all of Buffer ACW1 has been drawn through the
QIAamp Mini column, switch off the vacuum pump and
release the pressure.
11. Apply 750 μl Buffer ACW2 to the QIAamp Mini column.
release the pressure.
12. Apply 750 μl of ethanol (96–100%) to the QIAamp Mini
column. Leave the lid of the column open, and switch on the
vacuum pump. After all of ethanol has been drawn through the
spin column, switch off the vacuum pump and release the
pressure.
13. Close the lid of the QIAamp Mini column. Remove it from the
vacuum manifold, and discard the VacConnector. Place the
QIAamp Mini column in a clean 2 ml collection tube, and
centrifuge at full speed (14,000 rpm) for 3 min.
14. Place the QIAamp Mini column into a new 2 ml collection
tube. Open the lid, and incubate the assembly at 56 C for
10 min to dry the membrane completely.
15. Place the QIAamp Mini column in a clean 1.5 ml elution tube
and discard the 2 ml collection tube from step 14. Carefully
apply 25 μl of Buffer AVE to the center of the QIAamp Mini
column membrane. Close the lid and incubate at room tem-
perature for 3 min.d
16. Centrifuge in a microcentrifuge at full speed (14,000 rpm) for
1 min to elute the nucleic acids.
17. Repeat steps 15 and 16 again eluting into the same tube. Final
volume ~50 μl, labeled as eluate #1.
18. Repeat steps 15 and 16 (2 25 μl) eluting into a fresh 1.5 ml
elution tube. Label as eluate #2.
19. Store extracted cfDNA at 20 C.
3.3 Quantification 1. Take the Qubit dsDNA Assay Kit out of the freezer and equili-
of Extracted cfDNA brate at room temperature for 30 min.
2. Make the Qubit working solution by diluting the Qubit
dsDNA BR reagent 1:200 in Qubit dsDNA BR buffer. Do
not mix the working solution in a glass container. Label two
tubes for the standards.
ue
d
[FU]
Re
Bl
cfDNA
Higher Marker
gDNA
Lower Marker
Higher Marker
150 bp
Lower Marker
35 50 150 200 300 500 700 2000 10380 [bp]
Fig. 2 cfDNA extracted from plasma free from contaminant gDNA should exhibit a peak at around 150–300 bp
when analyzed on the Bioanalyzer/Tapestation (blue trace), while cfDNA contaminated with gDNA will show an
extra peak/s above 1000 bp (red trace)
3. Load 190 μl of Qubit working solution into each of the labeled

standard tubes.
4. Add 10 μl of each Qubit standard to the appropriate tube and
mix by vortexing 2–3 s. The final volume in each tube should
be 200 μl.
5. Load 199 μl of Qubit working solution into individual assay
tubes.
6. Add 1 μl of plasma eluate to the tubes containing Qubit work-
ing solution. The final volume in each tube should be 200 μl.
7. Incubate tubes at room temperature for 2 min.
8. Read the standards and plasma samples on a Qubit. 2.0 Fluo-
rometer as instructed by the apparatus.
9. Calculate the concentration of the plasma samples taking into
account the dilution step on [6].
3.4 Checking cfDNA Run 1 μl of eluate on a High Sensitivity DNA chip as per manu-
Quality Using Agilent’s facturer’s instructions. Assess the quality of the extracted cfDNA
Bioanalyzer based on the presence of a peak at around 150–180 bp and the
absence of a high molecular weight peaks of more than 1000 bp
(Fig. 2, blue trace). Genomic DNA (gDNA) contaminated cfDNA
will exhibit high molecular weight peaks of more than 1000 bp
(Fig. 2, red trace).
3.5 Droplet Digital 1. For each sample prepare a reaction mix, minus the DNA, into
PCR (ddPCR) for HER2 nuclease-free nonstick 1.5-ml microcentrifuge tubes as shown
Amplification in Table 3.
2. Aliquot reaction mix into 0.2 ml PCR plate wells (or tubes),
add DNA to be analyzed to tube, vortex and spin down.
3. Assemble droplet generator cartridge into cartridge holder.
4. Aliquot 20 μl of PCR reaction mix onto central row reservoirs
of droplet generator cartridge.a
Table 3
PCR master mix components and volumes used for HER2 amplification
testing on cfDNA
1 (μl)
DNA x
20 HER2 cocktail 1
20 EFTUD2 cocktail 1
Supermix for probes 10
Nuclease-free H2O to 20
Table 4
PCR conditions for HER2 amplification testing on cfDNA extracted from
plasma
Ramp temp
Step Temperature ( C) Time (min) # Cycles increment
Heated lid 105
Denaturing 95 10:00 1 2.5 C/s
Denaturing 95 00:15 40 2.5 C/s
Anneling/ 60 01:00 1 2.5 C/s
extension
Completion 98 10:00 1 2.5 C/s
Hold 10 Indefinite 1
5. Aliquot 70 μl of droplet generator oil (using same tip for all

wells) onto bottom row reservoirs of droplet generator
cartridge.b
6. Cover cartridge with gasket making sure ends fit over plastic
overhangs. Place into droplet generator.c When droplets have
been generated, remove cartridge from generator, remove gas-
ket, dispose and collect 40 μl of droplets from top row to
dispense into a PCR plate (Fisher Scientific). Collect droplets
using an eight channel multichannel air displacement pipette
by tipping pipette at 20–25 angle and slowly (up to 15 s)
collect the droplets.d Dispense droplets on the PCR plate by
tipping the pipette 20–25 angle and touching the wall of the
plate well, slowly (up to 15 s) release droplets by letting them
slide along the wall of the well. Dispose of cartridge.e
7. Repeat as many times as needed, covering wells between drop-
let generation steps.
8. Seal Plate and run samples on thermal cycler as shown in
Table 4.
Fig. 3 Representative droplet digital plots from a sample with high level amplification (a), and a non-amplified
sample (b). The four quadrants represent top left—droplets with HER2 DNA only (blue population), top right—
droplets with both HER2 and EFTUD2 DNA (brown population), bottom right—droplets with EFTUD2 DNA
only (green population), and bottom left—droplets with no DNA (black population)
9. When PCR cycling is complete, read droplets on a QX100/

QX200 Droplet Reader as per manufacturer’s instructions.
Read the plate using Absolute Quantification.
10. Samples that do not have at least 400 WT droplets should be
enhanced by running more sample until at least 400 WT dro-
plets are analyzed.
3.6 Droplet DigitalTM Data produced on the QX100/QX200 reader may be analyzed
PCR (ddPCR) Analysis using Bio-Rad QuantaSoftTM software. Gate the four distinct
populations produced by the assay as shown in Fig. 3 and use the
built-in analysis tool to calculate the ratio of HER2 to EFTUD2
reference assay. Values above a ratio of 1.25 should be considered
amplified, while ratios below that should be considered
non-amplified.
For optimal results in samples, and in particular in samples with
a copy number ratio near to the cutoff, the sequential probability
ratio test (SPRT) (Fig. 4a) should be used to determine whether
the copy number is elevated, not-elevated, or indeterminate. The
SPRT uses the maximum likelihood method to determine whether
the observed ratio is above or below the threshold within accept-
able bounds of error, or whether further digital PCR runs are
required to determine the copy number ratio.
For data analysis using the SPRT test, a threshold likelihood
ratio of 8, as previously reported with modifications [9, 10], should
be used. Only informative droplets should be analyzed, i.e., those
droplets positive for either HER2 alone or EFTUD2 alone. The
proportion of informative droplets positive for HER2 is calcu-
lated as PHER2 ¼ NHER2/N where N ¼ (NHER2 + NEFTUD2) the
number of informative droplets, NHER2, is the number of droplets
positive for HER2 alone, and NEFTUD2, is the number of droplets
positive for EFTUD2 alone.
The boundaries of the SPRT test are calculated as follows:
Upper ¼ ððln 8Þ=N ln d Þ= ln g
Lower ¼ ððln 1=8Þ=N ln d Þ= ln g
Fig. 4 (a) Droplet digital PCR with a FAM labeled HER2 probe and VIC labeled EFTUD2 probe. DNA is partitioned
into droplets, and after PCR, droplets are assessed by a fluorescent reader. The concentration of DNA in each
sample can be quantified from the number of wells positive using the Poisson distribution. This is further
analyzed with the SPRT using informative droplets, those droplets positive for HER2 or EFTUD2 alone, and not
those positive for both or neither. The SPRT assesses whether the proportion of informative wells positive for
HER2, informative wells ratio, is elevated as data accumulates. SPRT defines two boundaries, with a ratio
above the upper boundary being considered HER2-positive and below the lower boundary considered HER2-
negative. A ratio between the two boundaries is considered as unassigned, and the sample is subjected to
further rounds of digital PCR until the result is above or below the boundaries. (b) Analysis of digital PCR with
Sequential Probability Ratio Test (SPRT) using informative droplets on a cohort of 58 metastatic breast cancer
patients, 11 patients with HER2-amplified cancers and 47 patients with HER2-nonamplified cancers. Red
triangles indicate patients with HER2-amplified tumors and black triangle HER2-nonamplified tumors. Adapted
from H. Gevensleben et al., Noninvasive detection of HER2 amplification with plasma DNA digital PCR. Clinical
cancer research: an official journal of the American Association for Cancer Research 19, 3276–3284 (2013)
with permission from The American Association of Cancer Research (AACR)
where

d ¼ ð1 q 1 Þ= 1 q 0

g ¼ q 1 1 q 0 =q 0 ð1 q 1 Þ
and
q1 ¼ the proportion of informative droplets positive for HER2
if alternative hypothesis is accepted (plasma sample is from a patient
with HER2 amplification).
q0 ¼ the proportion of informative droplets positive for HER2
if null hypothesis is accepted (plasma sample is from a patient
without HER2 amplification).
q1 is calculated from HER2/EFTUD2 copy number ratio
(TAMP ¼ 1.3) for assigning a sample as HER2 positive sample,
and varies according to MEFTUD2.
q 1 ¼ ðXAMP XAMP nEFTUD2 =nÞ=
ðXAMP þ nEFTUD2 =n 2XAMP nEFTUD2 =nÞ
where XAMP ¼ 1 exp (TAMPMEFTUD2) the expected proportion
of informative droplets positive for HER2 at the threshold ratio
TAMP.
q0 is similarly calculated from HER2/EFTUD2 copy number
ratio (TNONAMP ¼ 1.2) for assigning a sample as HER2 positive or
negative, and also varies according to MEFTUD2.
q 0 ¼ ðXNONAMP XNONAMP nEFTUD2 =nÞ=
ðXNONAMP þ nEFTUD2 =n 2XNONAM PnEFTUD2 =nÞ,
where XNONAMP ¼ 1 exp (TNONAMPMEFTUD2).
Using these equations for any given N, the upper and lower
boundaries of the SPRT curve are calculated. If PHER2 is greater
than the upper boundary then the test result is HER2 positive. If
PHER2 is less than the lower boundary then the test result is HER2
negative. If PHER2 lies between the two boundaries, then further
round(s) of digital PCR are required until a sample is above, or
below, the boundaries.
The SPRT uses a likelihood ratio of 8, which corresponds
roughly to two sided 95% confidence for differentiating between
ratios of 1.30 and 1.20 [9]. With these parameters a sample with an
actual HER2:EFTUD2 ratio of, for example, 1.15 would have a
very high probability (>99.9%) of being correctly called as nega-
tive, and likewise a sample with an actual ratio of 1.35 would have a
very high probability of being correctly called as positive.
Figure 4b illustrates the use of the SPRT test on a cohort of
58 metastatic breast cancer patients (11 with HER2-amplified can-
cers and 47 with HER2-nonamplified cancers). Red triangles indi-
cate patients with HER2-amplified tumors and black triangle
HER2-non-amplified tumors. The displayed SPRT decision
boundaries are for illustrative purposes only, as the exact level varies
according to the EFTUD2 control probe concentration (MEF-
TUD2), with the displayed boundaries calculated with MEF-
TUD2 ¼ 0.025. Cases with a number of informative droplets
>5000 are not displayed [7].
4 Notes
1. Venous blood collection, isolation of plasma from blood and

storage (Section 3.1)
(a) When collecting blood, hemolysis must be avoided.
(b) EDTA BCTs should be stored in an upright position at
room temperature and centrifuged within 2 h of sample
collection.
(c) Invert the ETDA tube gently 8–10 times immediately
prior to centrifugation.
(d) Use a piston-driven air displacement pipette (avoid using a
Pasteur pipette) to remove the plasma.
(e) When freezing plasma do not use polystyrene boxes.
2. Circulating free DNA extraction from plasma using a
QIAvac (Section 3.2)
(a) Make sure that a visible vortex forms in the tube. In order
to ensure efficient lysis, it is essential that the sample and
Buffer ACL are mixed thoroughly to yield a homoge-
neous solution Do not interrupt the procedure at this
time. Proceed immediately to the lysis incubation step.
(b) Make sure that the tube extender is firmly inserted into the
QIAamp Mini column in order to avoid leakage of sample.
Keep the collection tube for the dry spin in step 13.
(c) Note that large sample lysate volumes may need up to
10 min to pass through the QIAamp Mini column mem-
brane by vacuum force. To avoid cross-contamination, be
careful not to move the tube extenders over neighboring
QIAamp Mini columns.
(d) Ensure that the elution buffer AVE has been warmed to
42 C before use. Elution volume is flexible and can be
adapted according to the requirements of downstream
applications. The recovered eluate volume will be up to
5 μl less than the elution volume applied to the QIAamp
Mini column.
3. Droplet digital PCR (ddPCR) for HER2 amplification
(Section 3.5)
(a) Carefully load the cartridge and do not introduce bubbles
onto the bottom of the reservoirs. Always fill all wells on a
generator cartridge. If less than eight samples then fill

empty wells with 20 μl of 2 Buffer Control kit.
(b) Droplet generator oil is volatile, so avoid leaving the
tube or well open for very long time to prevent evapora-
tion. Always add the oil after having added the sample to
the central row of the reservoir to avert oil filling up the
microchannels connecting the reservoirs.
(c) When loading the cartridge onto the droplet generator, be
careful as to not spill oil (hold by central part). Droplets
will be generated in around 2 min.
(d) Collecting droplets slowly will prevent droplets from
breaking up and sticking together.
(e) Once droplets are generated and deposited onto the PCR
plate cover filled wells with tape to prevent evaporation.
Acknowledgments
Prof. Nicholas C. Turner is a CRUK Clinician Scientist. We

acknowledge NHS funding to the NIHR Biomedical Research
Centre. Figure 4 is adapted from H. Gevensleben et al., Noninva-
sive detection of HER2 amplification with plasma DNA digital
PCR. Clinical cancer research: an official journal of the American
Association for Cancer Research 19, 3276-3284 (2013) with per-
mission from The American Association of Cancer Research
(AACR).
References
1. Ding L et al (2010) Genome remodelling in a 6. Vogelstein B, Kinzler KW (1999) Digital PCR.

basal-like breast cancer metastasis and xeno- Proc Natl Acad Sci U S A 96:9236–9241
graft. Nature 464:999–1005 7. Gevensleben H et al (2013) Noninvasive detec-
2. Johnson PJ, Lo YM (2002) Plasma nucleic tion of HER2 amplification with plasma DNA
acids in the diagnosis and management of digital PCR. Clin Cancer Res 19:3276–3284
malignant disease. Clin Chem 48:1186–1193 8. Garcia-Murillas I, Lambros M, Turner NC
3. Li M, Diehl F, Dressman D, Vogelstein B, Kin- (2013) Determination of HER2 amplification
zler KW (2006) BEAMing up for detection and status on tumour DNA by digital PCR. PLoS
quantification of rare sequence variants. Nat One 8:e83409
Methods 3:95–97 9. Wei Zhou GG, Goodman SN, Romans KE,
4. Board RE et al (2010) Detection of PIK3CA Kinzler KW, Vogelstein B, Choti MA, Mon-
mutations in circulating free DNA in patients tgomery EA (2001) Counting alleles reveals a
with breast cancer. Breast Cancer Res Treat connection between chromosome 18q loss and
120:461–467 vascular invasion. Nat Biotechnol 19:78–81
5. Diehl F et al (2008) Circulating mutant DNA 10. Lo YM et al (2007) Digital PCR for the molec-
to assess tumor dynamics. Nat Med ular detection of fetal chromosomal aneu-
14:985–990 ploidy. Proc Natl Acad Sci U S A
104:13116–13121
Chapter 11
Detection and Quantification of Mosaic Genomic DNA

Variation in Primary Somatic Tissues Using ddPCR: Analysis
of Mosaic Transposable-Element Insertions, Copy-Number
Variants, and Single-Nucleotide Variants
Bo Zhou, Michael S. Haney, Xiaowei Zhu, Reenal Pattni, Alexej Abyzov,
and Alexander E. Urban
Abstract
Here, we describe approaches using droplet digital polymerase chain reaction (ddPCR) to validate and
quantify somatic mosaic events contributed by transposable-element insertions, copy-number variants, and
single-nucleotide variants. In the ddPCR assay, sample or template DNA is partitioned into tens of
thousands of individual droplets such that when DNA input is low, the vast majority of droplets contains
no more than one copy of template DNA. PCR takes place in each individual droplet and produces a
fluorescent readout to indicate the presence or absence of the target of interest allowing for the accurate
“counting” of the number of copies present in the sample. The number of partitions is large enough to
assay somatic mosaic events with frequencies down to less than 1%.
Key words Droplet digital PCR (ddPCR), Somatic mosaicism, Mobile elements, Copy number
variations (CNVs), Single nucleotide variations (SNVs)
1 Introduction
Droplet digital polymerase chain reaction (ddPCR) is a powerful

approach to validate and to quantify somatic genomic mosaicism in
primary tissues [1–3]. Somatic genomic mosaicism is the phenom-
enon where variations in the genomic DNA sequence are present in
only a subset of cells comprising a given tissue. This type of geno-
mic variation is presumed to arise within an individual due to
postzygotic mutation [4]. Somatic genomic variants will typically
have been detected first using a variety of “next-generation” high-
throughput DNA sequencing-based approaches [5]. In addition to
Bo Zhou and Michael S. Haney contributed equally to this work.
173
174 Bo Zhou et al.
validating such sequencing-based variant-detections, ddPCR can

also more accurately quantify the frequency at which the variant
occurs in a given tissue [1, 3]. Furthermore, ddPCR is one of only a
few approaches that can be used for such a purpose (validation and
quantification of mosaic genomic variation). Since it is a nonse-
quencing based method, it serves as an orthogonal approach to the
typical method of discovery in somatic mosaicism studies [5]. This
makes validation and quantification of somatic variants with ddPCR
particularly useful, especially in the context of a topic where exten-
sive and specific experimental validation of initial discoveries is
essential in order to avoid the reporting of sequencing artifacts.
The approaches described here have been tested in DNA samples
extracted either directly from noncancerous primary tissue or non-
cancer cell lines. We do however expect these same approaches to be
equally applicable to cancer tissues or cell lines, respectively.
In this context of somatic genomic mosaicism, ddPCR can be
used in the analysis of genomic variants spanning the entire spec-
trum of sizes, from the smallest (single-nucleotide variants, SNVs)
over medium-sized (e.g., insertions of retrotransposons and mobile
genomic elements in general—i.e., mobile element insertions,
MEIs) to large copy number variants (CNVs) [1, 3, 6]. There are
only minor adjustments needed to the general ddPCR protocol and
assay design to allow for the analysis of each of these types of
variants. Here, we describe how to employ ddPCR, specifically
the Bio-Rad QX200 Droplet Digital PCR system [7], in the analy-
sis of three types of somatic mosaic genome variants: SNVs, MEIs
and CNVs. By combining high throughput microfluidics with
quantitative (digital) fluorescent readouts, using either TaqMan
style probes or an intercalating fluorescent dye (EvaGreen),
ddPCR assays achieve unprecedented levels of sensitivity and accu-
racy in PCR-based detection [7, 8].
1.1 Principle Briefly, a standard ddPCR reaction mixture, containing template

of Approach DNA, target specific primers and fluorescent probes
(or intercalating fluorescent dye, see below), is partitioned into
approximately 20,000 oil-in-water droplets [7, 8]. The most crucial
aspect of understanding the experimental approach is to keep in
mind that while all other reagents are in excess, the template DNA
is partitioned into the droplets following a Poisson manner [7], and
as a result, target occupancy in droplets or copies per droplet (cpd)
increases with input DNA concentration. At <10% target occu-
pancy or <0.1 cpd the vast majority of droplets contain either 0 of
1 template copy of your PCR target of interest. While maximal
precision for target copy number quantification is achieved at
1.6 cpd or 32,000 human genome equivalents per 20 μL reaction,
accurate quantification can be achieved as long as negative droplet
partitions are present in the reaction (<5 cpd) [8]. Overloading of
template DNA (>5 cpd) results in the absence of null droplets
Detection and Quantification of Mosaic Genomic DNA Variation in Primary. . . 175
(droplets that contain zero copies of template target), thus Poisson

correction cannot be applied rendering accurate quantification
impossible [9, 10]. It is also important to note that the high degree
of sample DNA partition in ddPCR allows for the detection and
quantification of low-frequency sequence variants by giving a
binary-like readout for the presence or absence of the genetic
variant in each droplet [8]. The partitioning effect is perhaps
most beneficial (compared to unpartitioned qPCR) for the detec-
tion and quantification of SNVs where the effect of imperfect probe
specificity is “exploited” by partitioning allowing for the discern-
ment between reference and variant signals (see Subheading 3.5).
For each individual DNA sequence variant, a set of PCR primers
is designed, as well as a specific fluorophore-conjugated, TaqMan-
style probe that will bind to the template DNA during the primer
extension step. Alternatively, when using the EvaGreen assay (i.e., an
intercalating fluorescent dye), only PCR primers are required. There
are two basic principles of probe/primer design. Primers can be
designed to “bracket” a predicted mosaic SNV, CNV-deletion junc-
tion, CNV-duplication junction or MEI-insertion point or designed
to bind within the boundaries of a predicted mosaic deletion or
duplication CNV. These principles are illustrated in Fig. 1.
Fig. 1 The basic principles of primer and probe design for ddPCR in the analysis of somatic genomic mosaics.
Panel a: if the breakpoint junction of a suspected mosaic CNV, the insertion point of a mosaic MEI event, the
coordinate of a mosaic SNV, or the location of a small indel is exactly known, then a primer-and-probe set can
be designed to cover these exact coordinates (green arrows: forward and reverse primers, orange bracket:
specific TaqMan style fluorophore conjugated oligomer or EvaGreen dye intercalating into amplicon). Panel b:
if the exact coordinates of a suspected mosaic CNV are not known then primer-and-probe sets can be
designed to produce an amplicon from a locus somewhere between the suspected approximate breakpoint
coordinates (green arrows: forward and reverse primers, orange bracket: specific TaqMan style fluorophore
conjugated oligomer or EvaGreen dye intercalating into amplicon). In this latter scenario somatic mosaic MEIs
and SNVs/small indels are not accessible to ddPCR analysis by approaches covered in this chapter. However,
these may be accessible with alternative EvaGreen approaches [6]
176 Bo Zhou et al.
Under the first principle of probe-and-primer design (Fig. 1a),

mosaic CNVs and SNVs can be detected because the custom fluo-
rescent probes are designed such that it spans the breakpoint junc-
tion of the mosaic deletion-CNV or the insertion-point sequence of
the mosaic duplication-CNV or MEI, or sits right on top of the
mosaic SNV. If a given droplet with the partitioned sample DNA
contains a DNA fragment with this specific mosaic breakpoint or
insertion point junction or SNV, there will be a discrete digital
readout. In droplets that do not contain such a fragment (because
the frequency of its occurrence in the sample DNA is low or it may
even be completely absent) there will simply be no signal at all.
With this binary, digital readout of the proportion of cells harbor-
ing the mosaic variant in the sample tissue can be calculated with
sensitivity as low as 1% mosaic frequency for CNVs and less than
0.1% for SNVs and MEIs. For example, consider a deletion-CNV
that had been detected in a clonally expanded iPSC line by shallow
whole-genome sequencing, but not in the fibroblast culture from
which the iPSC line hiPSC 6 was derived (Fig. 2, left panel)
[1]. Using ddPCR, it was established that this deletion-CNV was
present as a heterozygous mosaic variant in ~0.8% of the cells from
the original fibroblast culture and present in ~100% of the cells in
the iPSC culture (Fig. 2, right panel), suggesting that the iPSC line
was derived and clonally expanded from a fibroblast cell harboring
Fig. 2 Detection and quantification of a lineage-manifested CNV (LM-CNV). Left panel: low-coverage whole-
genome sequencing (using CNVnator [11]) could not detect a CNV that was present in mosaic fashion in a
fibroblast sample. After clonal expansion of a single cell from the fibroblast sample in the process of producing
human induced pluripotent stem cells (hiPSCs) the CNV became unmasked in low-coverage whole-genome
sequencing (dashed orange lines). Right panel: ddPCR analysis could confirm the presence of the LM-CNV
detected in hiPSC line 6 in the fibroblast sample from which hiPSC 6 was created as well as quantify its
mosaic allele frequency. y-axis: number of ddPCR events per well. Blue bars: reference probe. Green bars:
LM-CNV specific probe. The frequency of CNV events detected in hiPSC 6 indicates that this is a heterozygous
deletion present in almost all cells while the frequency of CNV events detected in fibroblast samples indicate
this heterozygous deletion is present in ~0.8% of cells. ddPCR was carried out in three replicates in each of
the hiPSC and fibroblast genomic DNA samples (Figure adapted with permissions from [1])
this specific deletion-CNV in the mosaic cell population [1]. The

second principle of probe-and-primer design (Fig. 1b) can be
applied when the genomic coordinates of the breakpoint junction
of a CNV is ambiguous or unknown. Primers and probes can be
designed to target the predicted region of duplication or deletion
rather than spanning the breakpoint junction. The sensitivity of this
approach is lower than that of bracketing a given variant, and it
cannot be applied to analyze a specific MEI or SNV. On the other
hand, however, this second approach can be used to still achieve
validation and at least limited quantification of such CNVs that
have been detected by sequencing but for which the exact break-
point sequences have not been resolved. In all scenarios, it is
recommended to pair the mosaic variant assay of interest with an
assay for a reference gene that simultaneously quantifies the num-
ber of genome equivalents in the input sample (see Note 1).
2 Materials
In order to prevent contamination, carefully follow general PCR

preparation guidelines and store ddPCR reagents and consumables
away (ideally in separate rooms) from where ddPCR assays are
prepared. All oils, cartridges, and gaskets should be stored at
room temperature unless otherwise noted. When preparing assays,
primers and probes should be thawed and kept on ice (or stored at
4 C for up to 1 year); probes should be kept in the dark or wrapped
in aluminum foil; all other reagents should be prepared at room
temperature. Follow all waste disposal regulations when disposing
of waste materials.
2.1 Instruments 1. QX200 Droplet Generator (Bio-Rad).

2. PX1 PCR Plate Sealer (Bio-Rad).
3. 96-well thermal cycler with gradient temperature feature (e.g.,
Bio-Rad C1000 or Applied BiosystemsVeriti).
4. QX200 Droplet Reader (Bio-Rad).
5. PC with QuantasSofttm Software (Bio-Rad).
6. Eight-channel P100 or P200 Pipette.
7. Single channel P2, P20, and P200 pipettes for single reaction
preparations.
2.2 ddPCR 1. DG8 Cartridges for QX200 Droplet Generator (Cat#

Consumables (All 186-4008).
Bio-Rad) 2. DG8 Gaskets for QX200 Droplet Generator (Cat# 186-3009).
3. Pierceable Foil Heat Seal (Cat# 1814040).
4. QX200 Droplet Reader Oil (Cat# 186-3004).
178 Bo Zhou et al.
2.3 ddPCR Reagents 1. ddPCR Supermix for Probes (no dUTP) (Bio-Rad Cat#
for Probe Assay 186-3024).
2. ddPCR 2 Buffer Control Kit for Probe (Bio-Rad Cat#
186-3052).
3. Droplet Generation Oil for Probes (Bio-Rad Cat#186-3005).
4. Primers for target and control amplicons (e.g., IDT).
5. FAM-VIC or FAM-HEX probes for target and reference
amplicons (e.g., IDT 60 Fluro FAM/HEX, 30 Quencher ZEN).
2.4 ddPCR Reagents 1. QX200 ddPCR EvaGreen Supermix (Bio-Rad Cat#

for EvaGreen Assay 186-4034).
2. QX200 2 Buffer Control for EvaGreen (Bio-Rad Cat#
186-4052).
3. Droplet Generation Oil for EvaGreen store at room tempera-
ture (Bio-Rad Cat# 186-4005).
4. Primers for target and reference control amplicons (e.g., IDT).
2.5 Other 1. Twin Tec semiskirted 96-well plate (Eppendorf Cat#

Consumables 951020362).
2. 25 mL Reagent trough (Thermo Fisher).
3. P20 barrier pipette tips (Rainin Only, gentler for pipetting
droplets Cat# GP-20F).
4. P200 barrier pipette tips (Rainin Only, gentler for pipetting
droplets Cat# GP-200F).
5. DNA restriction enzyme compatible with assay (e.g., New
England Biolabs).
6. Molecular biology grade water.
2.6 Genomic DNA 1. DNeasy Blood & Tissue kit (Qiagen Cat# 69504).
Isolation
3 Methods
3.1 Primer and Probe The design of the ddPCR primers and probes should follow these
Design general guidelines. For primers, Tm should be 50–65 C with
5 C difference between both primers. The primer length should
3.1.1 General
be 9–40 bp (ideally ~ 20 bp) and the amplicon size should be
Considerations
60–200 bp. The primer GC percentage should be between 50%
and 60%. In addition, these features should be avoided in primer
design:
1. Stretches of 4 or more C or G;
2. Hairpin structures (Tm of the hairpin should be lower than the
annealing temperature);
3. Primer dimers (ΔG for dimer analysis should be between 0 and

9 kcal/mol);
4. Nonspecific priming (check with in silico PCR available on
UCSC Genome Browser).
To design TaqMan probes for ddPCR probe assays, the probe
should be 15–30 bp in length with GC content between 30% and
80%. Probe Tm (noninclusive of quencher effect) should be ~10 C
higher than the primer Tm. If using a 30 -minor groove binder
(MGB) the MGB increases probe Tm by 15–30 C, which allows
for the use of shorter probes with high specificity. MGB Tm
enhancer can be obtained from ABI or Thermo Fisher. Target
and reference probe 50 fluorophores should be either FAM and
VIC (“ABI”) respectively or FAM and HEX respectively (IDT).
Probes should also have a quencher group such as ZEN (IDT),
NFQ or black hole (on the 30 -end). The probe sequence can be
based off of either the Watson or Crick strand; it is preferable to use
the strand that places more Cs than Gs in the probe sequence, as
long as the sequence does not start with a G at the 50 -end.
3.1.2 Reference Primer Primers that target an amplicon region within the human RPP30
and Probe Design gene, for which two copies should be present in each cell, can serve
as reference primers. VIC or HEX fluorophore-conjugated Taq-
Man probe hybridizing this amplicon can serve as the reference
probe (reference RPP30 probes and primers can be also be supplied
by Bio-Rad (https://www.bio-rad.com/digital-assays/#/).
3.1.3 Mosaic Variant Primers should be designed to target a variant such that the ampli-
Primer and Probe Design con contains the SNV or the breakpoint-junction or insertion-point
sequence of the CNV or MEI, respectively. FAM probes should be
designed to hybridize directly onto the predicted SNV (variant
sequence) or the breakpoint-junction or insertion-point sequence
of the predicted CNV/MEI (Fig. 1) and assayed with RPP30 as the
reference target. When using single-color EvaGreen intercalating
dye for the analysis of somatic variants, it is important to consider
that the signal amplitude differentiation between different targets
comes from the differences in the lengths of the amplicons (i.e.,
reference and target amplicon); therefore, target amplicons and
reference amplicons must be sufficiently different in length to
yield clear signal differentiation [6, 12]. For instance, mixing pri-
mers that target a mosaic CNV yielding amplicons 200 bp in size
and primers that target a reference region (60 bp) within RPP30 in
the same reaction may result in clear signal separation, which is not
expected if both target and reference amplicons are similar in
lengths. In certain instances, clusters may be sufficiently discrimi-
nated by as little as 6 bp [6]. The frequency of the target CNV can
180 Bo Zhou et al.
be reliably quantified by comparing the groups of droplets (target,

reference, and null) with distinct EvaGreen intensities. The optimal
annealing and extension temperature should be examined for each
target and reference primer set by testing the separation of Eva-
Green signal intensities from the target-variant and reference-
RPP30 amplicons. If the mosaic event is novel and rare, one can
perform positive and negative control experiments to ensure accu-
rate interpretation of ddPCR results. DNA (synthesized, cloned, or
isolated) containing the variant of interest may be diluted to desired
frequency to mimic mosaic events and mixed with template DNA
to serve as positive control. Synthesized DNA fragments may for
example be purchased as gBlocks from Integrated DNA Technolo-
gies (IDT).
3.2 Genomic DNA Use nondegraded genomic DNA for all ddPCR assays. Perform
Isolation genomic DNA extraction using QIAGEN DNeasy Blood & Tissue
kit (Cat. 69504) following manufacturer guidelines or any other
genomic DNA isolation method of choice as long as genomic DNA
of high quality (nondegraded) and high purity can be obtained.
3.3 DNA Digestion It is essential to load restriction enzyme in the reaction mix if more
than 66 ng of genomic DNA per reaction is used. However, it can
also be very important for assaying a genomic target region that is
poorly accessible, thus restriction enzyme at 10–20 U/1 μg can be
loaded for such cases. A four-cutter enzyme such as MseI will
fragment the DNA every 250 base pairs, on average. Choose a
digestion enzyme that does not cut in either reference or target
amplicon by entering the amplicon sequence into NEBCutter
(http://nc2.neb.com/NEBcutter2/). Restriction enzyme may be
directly loaded into the ddPCR reaction. One may also choose to
digest template genomic DNA prior to loading into ddPCR reac-
tion. However, the user should be aware that the predigested
sample must be diluted at least tenfold in the final ddPCR reaction
to avoid any inhibition of the reaction. This approach requires
significantly more DNA in situ restriction digest.
3.4 ddPCR 1. For each targeted SNV, CNV, or MEI ddPCR reaction the
Experimental Setup following should be prepared at room temperature.
Probe assay:
(a) 10 μL 2 ddPCR Probe Master Mix.
(b) 1.8 μL 10 μM Forward Target Primer.
(c) 1.8 μL 10 μM Reverse Target Primer.
(d) 0.5 μL 10 μM Target Probe.
(e) 1.8 μL 10 μM Forward Reference Primer.
(f) 1.8 μL 10 μM Reverse Reference Primer.

(g) 0.5 μL 10 μM Reference Probe.
(h) 20 ng (5–66 ng) of template DNA (volume variable).
(i) 0.5 μL DNA restriction enzyme (if more than 66 ng of
template DNA is used).
(j) Nuclease-free water to make total volume 20 μL).
EvaGreen assay (see Note 4):
(a) 10 μL 2 ddPCR EvaGreen Master Mix.
(b) 0.2 μL 10 μM Forward Target Primer.
(c) 0.2 μL 10 μM Reverse Target Primer.
(d) 0.2 μL 10 μM Forward Reference Primer.
(e) 0.2 μL 10 μM Reverse Reference Primer.
(f) 20 ng (5–66 ng) of template DNA (volume variable).
(g) 0.5 μL DNA restriction enzyme (if more than 66 ng of
template DNA is used).
(h) Nuclease-free water to make total volume 20 μL).
2. Mix ddPCR solution by slowly pipetting up and down
10 times.
3. Place 8-well DG8 cartridge into cartridge holder.
4. Slowly load 20 μL of PCR solution into the middle row of wells
in the DG8 cartridge using eight-channel multichannel pipette.
Ensure that there are no air bubbles in the bottom of the wells.
If air bubbles are introduced, remove with an extra clean
pipette tip.
5. Load 70 μL of droplet generation oil in the indicated wells in
the DG8 cartridge, again using an eight-channel multipipette.
Droplet generation should begin within 20 after oil addition.
6. Attach gasket to the cartridge holder.
7. Place cartridge in droplet generator and press button to close
the droplet generator lid and begin droplet generation.
8. Wait until droplet generation is complete (~20 ).
9. After droplet generation is complete, the droplet mixture
should appear cloudy.
10. Carefully transfer each ddPCR droplet mixture (~38 μL) to
96-well plate using an eight-channel multipipette.
11. Seal plate with foil using plate sealer.
12. Place 96-well plate in thermocycler and run the following
protocols. (Note: annealing temperatures may vary depending
on primer–probe sets used).
182 Bo Zhou et al.
Probe assay:
95 C 10 min.
94 C 30 s, 60 C 60 s (40 cycles).
98 C 10 min.
4 C hold.
EvaGreen assay:
95 C 5 min.
95 C 30 s, 60 C 60 s (40 cycles).
4 C 5 min.
90 C 5 min.
4 C hold.
13. Once thermocycler protocol is completed, place 96-well plate
in droplet reader.
14. Read assay and analyze results using the QuantaSoft Software,
choosing appropriate droplet reading parameters for probes or
EvaGreen chemistries.
3.5 Mosaic SNV The equations (Table 1) used for estimating the SNV allele fre-
Analysis quency (b r ) and its 95% confidence interval (b r Low and b r High ) are
adopted from [9] where the principles are discussed in more detail.
In theory, these principles can be applied for both probe-based and
EvaGreen assays. Three input values are needed: the total number
of droplets generated (C), the number of droplets that contain the
SNV of interest (H1), and the number of droplets that contain the
reference allele (H2). All three input values can be obtained from
the Bio-Rad QuantaSoft software readout of the ddPCR reaction:
C ¼ “Accepted Droplets,” H1 ¼ “Positives” under Chanel
1, H2 ¼ “Positives” under Chanel 2. If replicates are performed,
then droplet counts should be combined. It is generally encouraged
to perform more ddPCR replicates for low-frequency SNVs to
achieve more precise quantification (see Note 2). In the example
shown in Fig. 3, C is the sum of all droplets which is 12,854; H1 is
be the sum of the droplets contained in clusters E and F which is
367. H2 is be the sum of the droplets contained in the green cluster
and clusters B, D, and F which is 2516. These three values will
subsequently be used as input values for equations listed in Table 1
to calculate allele frequency. Since SNVs only contain a single
nucleotide difference from the nonvariant allele, the TaqMan
probe designed against the variant allele of interest will also hybrid-
ize with the nonvariant allele albeit at a lower amplitude (Fig. 3
bottom, cluster C). It is strongly encouraged to perform the SNV
ddPCR assay on positive and negative control templates to ascer-
tain the optimal ddPCR conditions, primer/probe concentrations,
Table 1
Variant frequency calculation
Input values
Total number of droplets: C
Total number of droplets containing the variant: H1
Total number of droplets containing the reference: H2
Output values
Variant frequency: b
r
Lower bound of the 95% confidence interval of b
r :b
r Low
Upper bound of the 95% confidence interval of b
r :b
r High
Positive fractions for variant b
p 1 and reference b
p 2:
p 1 ¼ HC1 b
b p 2 ¼ HC2
Standard deviations for variant S1 and reference S2:

rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

bp 1 1bp 1 bp 2 1bp 2
S1 ¼ C S 2 ¼ C
p 1, Low ¼ b
b p 1 1:96S 1 p 1, High ¼ b
b p 1 þ 1:96S 1
b
p 2, Low ¼ b
p 2 1:96S 2 b
p 2, High ¼ b
p 2 þ 1:96S 2
Average number of positive events per droplet for variant b λ1

and reference b
λ2:

b p1 b
λ 1 ¼ ln 1 b λ 1, Low ¼ ln 1 pb1, Low bλ 1, High ¼ ln 1 b
p 1, High

b p2 b
λ 2 ¼ ln 1 b p 2, Low b
λ 2, Low ¼ ln 1 b λ 1, High ¼ ln 1 b
p 2, High
H Top ¼ b
λ 1, High b
λ 1 H Bottom ¼ b
λ1 b
λ 1, Low
W Right ¼ b
λ 2, High b
λ 2 W Left ¼ b
λ2 b
λ 2, Low
b
r ¼ λ1
b
bλ 2
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
bλ 1bλ 2 bλ 21bλ 22 H 2Bottom bλ 21 W 2Right bλ 22
b
r Low ¼
bλ 22 W 2Right
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

bλ 1bλ 2 bλ 21bλ 22 H 2Top bλ 21 W 2Left bλ 22
b
r High ¼
bλ 22 W 2Left
and number of ddPCR cycles before performing the assay on the

sample. As positive control, template DNA that contains the variant
of interest may be diluted to desired allele frequency to mimic
mosaic events (see Note 3). The optimal ddPCR conditions for
SNV assays are generally those in which the droplets that contain
184 Bo Zhou et al.
the SNV allele has the widest signal (amplitude) separation from
those that contain the nonvariant allele only (cluster C, in the
example shown in Fig. 3).
3.6 Mosaic MEI The mosaic allele quantification principles as outlined for mosaic
Analysis SNV analysis under Subheading 3.4 may be applied for mosaic MEI
analysis, where H1 is the total number of droplets that contain the
MEI allele of interest in the sample and, as above, H2 is the total
number of droplets that contain the reference allele. An example of
the sensitivity of ddPCR in the detection and quantification of
Line1 mobile genomic elements using different fluorophore
options is illustrated in Fig. 4.
3.7 Mosaic CNV In general, assuming that the ddPCR primers and/or TaqMan
Analysis probe span the breakpoint of the CNV of interest, the mosaic allele
quantification principles as outlined for mosaic SNV analysis under
Subheading 3.4 may be applied for mosaic CNV analysis (Fig. 5),
where H1 is the total number of droplets that contain the target
CNV allele in the sample, and H2 is the total number of droplets
with the reference allele.
4 Notes
1. The accuracy of mosaic frequency quantification depends

mostly on the sample size (or abundance) of the target of
interest as well as the total number of droplets. As long as the
ddPCR assay readout gives clear signal separation, EvaGreen
and probe assays should be equally reliable. EvaGreen assays are
significantly cheaper per reaction as no specific fluorophore-
conjugated TaqMan probe has to be synthesized, but they may
require longer time to optimize for each primer pair as three or
more clusters of droplets need to clearly separate in a single
channel amplitude in order for quantification to proceed.
ä
Fig. 3 Quantifying the frequency of SNV in human genomic DNA using TaqMan Probe Assay. In order to mimic
template DNA that contains a mosaic SNV on chromosome 10, copies of synthesized double-stranded DNA
(500 bp) that contain the SNV at the locus of interest were added to human genomic DNA at ~16% allele
frequency. Data shown are from a single ddPCR well. The target TaqMan probe (Channel 1) designed for this
assay has a perfect sequence match to the variant allele and contains one base pair mismatch to the
nonvariant allele. A probe targeting the RNaseP gene RPP30 was used as the reference allele (Channel 2).
(Top) Negative control assay with NA12878 genomic DNA as template DNA in which only RPP30 is present.
(Middle) Positive control assay with only synthesized double-stranded DNA as template, and Chanel 1 positive
droplets appear at a higher amplitude cluster compared to the negative control assay. (Bottom) Six separate
characteristic clusters of droplets were observed in this assay: no template (A), RPP30 only (B), nonvariant
allele only (C), RPP30 þ nonvariant allele (D), SNV only (E), and SNV þ RPP30 (F)
186 Bo Zhou et al.
Fig. 4 Detection of mosaic Line1 insertions with different ddPCR approaches. To test the power of ddPCR to
detect highly mosaic Line1 insertions we mixed different levels of GM12878 genomic DNA, containing a rare
homozygous insertion of Line1 with precisely known boundaries, and genomic DNA from a cell line that does
not contain the insertion (RUID#04C3729). With a total amount of 12 ng genomic DNA, which contains 100%,
20%, 4%, 0.8%, and 0.16% GM12878 DNA, respectively, we tested the performance of three ddPCR
protocols: EvaGreen (top panel), TaqMan with FAM/HEX probes (middle panel) and TaqMan with FAM/VIC
probes (bottom panel). Overall all three protocols, showed similar very high performance in quantifying MEIs
that are as rare as being present in only ~0.1% of the input DNA
Probe-based assays have the advantage of dual channel readout

and the additional target specificity introduced by the TaqMan
probe. In certain cases, TaqMan probes will be difficult to
design; in these cases, only the EvaGreen assay is the only
available option. It is also possible that for certain variants,
especially small indels and SNVs, the additional sequence spec-
ificity conferred by the TaqMan probe is needed.
2. Generally, the more replicate reactions performed in the valida-
tion of a variant, the greater the precision of estimation of
mosaic level. It is recommended that for low-frequency events
additional replicates be performed, instead of overloading a
Fig. 4 (continued)
reaction with too much template DNA. Always load <5 cpd or
<1.6 cpd for maximal precision. Too much template DNA
(>5 cpd) could adversely affect droplet formation and also
eliminates the number of negative droplets such that Poisson
correction of copy numbers cannot be performed leading to
inaccurate quantification.
3. It is strongly encouraged to perform the SNV ddPCR assay on
positive and negative control templates to first ascertain the
optimal ddPCR conditions, primer/probe concentrations, and
number of ddPCR cycles before performing the assay on the
sample. As a positive control, template DNA that contains the
variant of interest may be diluted to desired allele frequency to
mimic mosaic events. The optimal ddPCR conditions for SNV
assays are generally those in which the droplets that contain the
SNV allele has the widest signal (amplitude) separation with
those that contain the nonvariant allele only or clusters A and C
in the example shown in Fig. 3.
188 Bo Zhou et al.
Fig. 4 (continued)
4. Do not use any intercalating fluorophores other than EvaGreen

(e.g., do not use SYBR Green). Intercalating dyes that have not
been tested on and approved for the Bio-Rad ddPCR Droplet
Reader may potentially leak out of the droplet and damaged the
optics or other parts of the machine.
Acknowledgments
Research reported in this chapter was supported by the National

Institute of Mental Health of the National Institutes of Health
under award numbers R01MH094740 and R01MH100914. We
also acknowledge additional funds from Stanford University
(Department of Psychiatry and Behavioral Sciences, and Depart-
ment of Genetics).
Fig. 5 Detection of mosaic CNV in fibroblast line (A) that is clonally expanded in hiPSC line (B). VIC florescent
event counts measure copy number of RPP30 reference gene while FAM florescent event counts measure
copy number of mosaic CNV by designing the FAM probe such that it spans the CNV breakpoint junction. In this
sample, the heterozygous CNV deletion was present in ~10% of fibroblast cells, while the deletion was
present in ~100% of the hiPSC 7 cells
190 Bo Zhou et al.
References
1. Abyzov A, Mariani J, Palejev D, Zhang Y, 7. Hindson CM, Chevillet JR, Briggs HA, Galli-
Haney MS, Tomasini L, Ferrandino AF, chotte EN, Ruf IK, Hindson BJ, Vessella RL,
Rosenberg Belmaker LA, Szekely A, Tewari M (2013) Absolute quantification by
Wilson M, Kocabas A, Calixto NE, Grigorenko droplet digital PCR versus analog real-time
EL, Huttner A, Chawarska K, Weissman S, PCR. Nat Methods 10:1003–1005. https://
Urban AE, Gerstein M, Vaccarino FM (2012) doi.org/10.1038/nmeth.2633
Somatic copy number mosaicism in human 8. Pinheiro LB, Coleman VA, Hindson CM,
skin revealed by induced pluripotent stem Herrmann J, Hindson BJ, Bhat S, Emslie KR
cells. Nature 492:438–442. https://doi.org/ (2012) Evaluation of a droplet digital polymer-
10.1038/nature11629 ase chain reaction format for DNA copy num-
2. Campbell IM, Yuan B, Robberecht C, ber quantification. Anal Chem 84:1003–1011.
Pfundt R, Szafranski P, McEntagart ME, Naga- https://doi.org/10.1021/ac202578x
mani SC, Erez A, Bartnik M, Wisniowiecka- 9. Dube S, Qin J, Ramakrishnan R (2008) Math-
Kowalnik B, Plunkett KS, Pursley AN, Kang ematical analysis of copy number variation in a
SH, Bi W, Lalani SR, Bacino CA, Vast M, DNA sample using digital PCR on a nanoflui-
Marks K, Patton M, Olofsson P, Patel A, Velt- dic device. PLoS One 3:e2876. https://doi.
man JA, Cheung SW, Shaw CA, Vissers LE, org/10.1371/journal.pone.0002876
Vermeesch JR, Lupski JR, Stankiewicz P 10. Strain MC, Lada SM, Luong T, Rought SE,
(2014) Parental somatic mosaicism is under- Gianella S, Terry VH, Spina CA, Woelk CH,
recognized and influences recurrence risk of Richman DD (2013) Highly precise measure-
genomic disorders. Am J Hum Genet ment of HIV DNA by droplet digital PCR.
95:173–182. https://doi.org/10.1016/j. PLoS One 8:e55943. https://doi.org/10.
ajhg.2014.07.003 1371/journal.pone.0055943
3. Evrony GD, Lee E, Mehta BK, Benjamini Y, 11. Abyzov A, Urban AE, Snyder M, Gerstein M
Johnson RM, Cai X, Yang L, Haseley P, Leh- (2011) CNVnator: An approach to discover,
mann HS, Park PJ, Walsh CA (2015) Cell line- genotype, and characterize typical and atypical
age analysis in human brain using endogenous CNVs from family and population genome
retroelements. Neuron 85:49–59. https://doi. sequencing. Genome Res 21(6):974–984
org/10.1016/j.neuron.2014.12.028
12. McDermott GP, Do D, Litterst CM, Maar D,
4. Edwards JH (1989) Familiarity, recessivity and Hindson CM, Steenblock ER, Legler TC,
germline mosaicism. Ann Hum Genet Jouvenot Y, Marrs SH, Bemis A, Shah P,
53:33–47 Wong J, Wang S, Sally D, Javier L, Dinio T,
5. Freed D, Stevens EL, Pevsner J (2014) Somatic Han C, Brackbill TP, Hodges SP, Ling Y,
mosaicism in the human genome. Genes Klitgord N, Carman GJ, Berman JR, Koehler
(Basel) 5:1064–1094. https://doi.org/10. RT, Hiddessen AL, Walse P, Bousse L,
3390/genes5041064 Tzonev S, Hefner E, Hindson BJ, Cauly TH
6. Miotke L, Lau BT, Rumma RT, Ji HP (2014) 3rd, Hamby K, Patel VP, Regan JF, Wyatt PW,
High sensitivity detection and quantitation of Karlin-Neumann GA, Stumbo DP, Lowe AJ
DNA copy number and single nucleotide var- (2013) Multiplexed target detection using
iants with single color droplet digital PCR. DNA-binding dye chemistry in droplet digital
Anal Chem 86:2618–2624. https://doi.org/ PCR. Anal Chem 85:11619–11627. https://
10.1021/ac403843j doi.org/10.1021/ac403061n
Part IV
Rare Mutation and Rare Allele Detection

Chapter 12
Monitoring of Response and Resistance in Plasma

of EGFR-Mutant Lung Cancer Using Droplet Digital PCR
Yanan Kuang, Allison O’Connell, Adrian G. Sacher, Nora Feeney,
Ryan S. Alden, Geoffrey R. Oxnard, and Cloud P. Paweletz
Abstract
The identification of oncogenic driver mutations has led to the rapid rise of genotype-directed treatments.
However, genetic analysis of tumors remains cumbersome and a morbid experience for patients. Noninva-
sive assessment of tumor genotype, so-called “liquid biopsy,” such as plasma genotyping represents a
potentially transformative tool. Here we describe a genotyping protocol of cell-free plasma DNA
(cfDNA) using Droplet Digital™ PCR (ddPCR™). ddPCR emulsifies DNA into ~20,000 droplets in
which PCR is performed to endpoint in each droplet for both mutant and wild-type DNA. Droplets are run
through a modified flow cytometer where mutant and wild-type DNA emit different colored signals. The
count of these signals upon Poisson distribution analysis allows sensitive quantification of allelic prevalence.
Key words Droplet Digital™ PCR, Cell-free DNA, Liquid biopsy, Genotyping, Non-small-cell lung
cancer, Resistance mutations, EGFR
1 Introduction
Genotype-directed therapy is now the standard of care for advanced

non-small-cell lung cancer (NSCLC) patients. Somatic mutations
in EGFR, ALK, ROS1, and BRAF had been found to be predictors
of the efficacy of kinase inhibitors and are commonly used in either
clinical practice or in clinical trials [1–4]. However, despite the
therapeutic success of this approach, not all patients have sufficient
tumor tissue available for genotyping. In addition, acquired drug
resistance ultimately develops in all patients successfully treated
with kinase inhibitors [5]. The identification of treatment-related
changes in such cancers is of increasing importance given the
development of new targeted therapies specifically for patients
with treatment resistance [6, 7].
Repeated tumor biopsy is the gold standard for evaluating
genetic changes in lung cancers following treatment with targeted
therapies. However, this is not always feasible and can seldom be
193
194 Yanan Kuang et al.
performed more than once. Noninvasive techniques for blood-

based tumor genotyping (“liquid biopsies”) may realize the poten-
tial of genotype-directed lung cancer therapy. Recent studies have
suggested that highly sensitive genotyping assays can detect muta-
tions in circulating cell-free plasma DNA (cfDNA) from cancer
patients, potentially representing the biology of a patient’s cancer
[8–10].
We have now developed [11] a new assay for noninvasive
quantitative genotyping of oncogenic mutations in cfDNA of
NSCLC patients using Droplet Digital PCR™ (ddPCR™). This
technology emulsifies input DNA into ~20,000 droplets so that the
proportion carrying mutant versus wild-type alleles can be counted
using flow cytometry. We have now used this method to serially
analyze cfDNA from EGFR-mutant NSCLC patients treated with
the EGFR kinase inhibitor, erlotinib. We can identify the presence
of the EGFR activating mutation prior to therapy, its reduction or
disappearance during therapy and reemergence along with the drug
resistance EGFR T790M mutation during treatment [11]. The
present chapter guides the reader through several types of
ddPCR™ assay designs for plasma derived cfDNA and its clinical
application.
2 Materials
2.1 Cell-Free DNA 1. EDTA lavender-capped vacutainer tubes (Becton and Dickin-
Extraction son #366643).
2. QIAmp circulating nucleic acid kit (Qiagen #551140).
3. Vacuum pump (Qiagen #84010).
4. QIAvac connecting system (Qiagen #19419).
5. QIAvac 24 Plus (Qiagen #19413).
6. 15 mL polypropylene conical tube (Thermo Scientific
#339650).
7. 50 mL polypropylene conical tube (Thermo Scientific
#339652).
8. 2 mL cryo vial (Thermo Scientific #4000200).
9. Phosphate buffered saline (Thermo Scientific #10010023).
10. Benchtop Vortex mixer.
2.2 Droplet See Note 1.

Digital PCR
tor (Bio-Rad #1864008).
(Bio-Rad #1863009).
Monitoring of Response and Resistance in Plasma of EGFR-Mutant Lung. . . 195
Table 1
Cell lines used as positve controls
Mutation Cell line

EGFR del 19 PC9
EGFR L858R H1975
EGFR T790M PC9 GR
EGFR wild type A549
3. DG8™ Cartridge Holder (Bio-Rad #1863051).

4. Droplet Generation Oil for Probes (Bio-Rad #1863005).
5. ddPCR™ Supermix for probes (No dUTP) (Bio-Rad #186-
3025).
6. ddPCR™ Droplet Reader Oil (Bio-Rad #1863004).
7. Semiskirted 96-well PCR plate (Thermo Scientific
#E951020346).
8. Pierceable Foil Heat Seal (Bio-Rad #1814040).
9. Rainin single and multichannel pipettors (P2, P10, P20,
P1000; multi-P50 and multi-P100).
10. Sharp Precision DNAse- and RNase-free filter pipette tips
(Denville Scientific: #P1096-FR, #P1121, #P1122, and
#P1126).
11. Positive DNA controls: High, medium, or low levels of geno-
mic DNA from EGFR mutant cell lines (see Table 1).
(a) High QC: 1000 mutant copies in background of 5000 wt.
(b) Medium QC: 100 mutant copies in background of
5000wt.
(c) Low QC: 10 mutant copies in background of 5000 wt.
12. Primer and probes are custom ordered from Life Technologies.
For good assay design principles we refer the reader to
Chapter 2 of Bio-Rad’s ddPCR™ application guide [13].
(a) EGFR L858R primer sequences:
Forward: 50 -GCAGCATGTCAAGATCACAGATT-30
Reverse: 50 -CCTCCTTCTGCATGGTATTCTTTCT-30
EGFR L858R probe sequences:
50 -VIC-AGTTTGGCCAGCCCAA-MGB-NFQ-30
50 -FAM-AGTTTGGCCCGCCCAA-MGB-NFQ-3’
(b) EGFR exon19 deletion primer sequences:
Forward: 50 -GTGAGAAAGTTAAAATTCCCGTC-30
Reverse: 50 -CACACAGCAAAGCAGAAAC-30
EGFR exon19 deletion probe sequences are:
50 -VIC-ATCGAGGATTTCCTTGTTG-MGB-NFQ-30
50 -FAM-AGGAATTAAGAGAAGCAACATC-MGB-
NFQ-30
(c) EGFR T790M primer sequences:
Forward: 50 -GCCTGCTGGGCATCTG-30
Reverse: 50 -TCTTTGTGTTCCCGGACATAGTC-30
EGFR T790M probe sequences:
50 -VIC-ATGAGCTGCGTGATGAG-MGB-NFQ-30
50 -FAM-ATGAGCTGCATGATGAG-MGB-NFQ-30
3 Methods
3.1 Plasma Isolation See Notes 2–5.

1. Draw venous blood into one [1] 10 mL EDTA lavender-
capped vacutainer tube and immediately invert the tube eight
to ten times.
2. Immediately centrifuge for 10 min at 1500 (150) g in a
swinging bucket centrifuge.
3. Pipette the ~2–3 mL plasma layer (upper layer) into a sterile
15 mL polypropylene conical tube (Fig. 1).
4. Centrifuge the 15 mL tube containing the plasma for an addi-
tional 10 min at 3000 (150) g in a swinging bucket
centrifuge. This step removes any remaining white blood cells.
5. Transfer, using a fresh pipette, the supernatant into a second
15 mL polypropylene tube. Leave about 0.3 mL of supernatant
in the tube. This leftover contains cellular debris.
6. Using a fresh pipette aliquot plasma into 2 mL cryovials. One
can expect ~2–3 mL of pale yellow plasma from a full EDTA
vacutainer.
7. Freeze immediately upright at 70 C or colder until use.
cfDNA
330 bp
165 bp
Fig. 1 Plasma is isolated by whole blood centrifugation. cfDNA fragments range

between 135 and 480 base pairs
3.2 cfDNA Extraction See Notes 6–9.

This procedure is adapted from the QIAmp circulating nucleic
acid handbook. All buffers and enzymes are provided in the
QIAamp Circulating Nucleic Acid Kit, but require the addition of
100% ethanol and isopropanol as specified by the manufacturer.
Before you start:
(a) Prepare all buffers.
(b) Wipe down the lab bench, hood and pipets with 10% bleach
followed by 70% ethanol.
(c) Heat a water bath or heating block for 15 mL conical tubes
to 60 C and a heating block for 1.5 mL microcentrifuge
tubes to 56 C.
(d) Set up the QIAvac24 Plus manifold.
(e) Thaw 2 mL of plasma to room temperature.
1. Pipet 200 μL QIAGEN Proteinase K into a 15 mL

conical tube.
2. Add 2 mL of plasma into the tube.
3. Add 1.6 mL Buffer ACL (1.0 μg carrier RNA per plasma
sample). Close the cap and mix by pulse-vortexing for 30 s.
4. Incubate in a 60 C water bath or heating block for 30 min.
5. Place the tube back on the lab bench and unscrew the cap.
6. Add 3.6 mL Buffer ACB to the lysate in the tube. Close the cap
and mix thoroughly by pulse-vortexing for 15–30 s.
7. Incubate the lysate–Buffer ACB mixture in the tube for 5 min
on ice.
8. Insert the QIAamp Mini column into the VacConnector on the
QIAvac 24 Plus. Insert a 20 mL tube extender onto the open
QIAamp Mini column.
9. Carefully apply the lysate–Buffer ACB mixture from step 7 into
the tube extender of the QIAamp Mini column. Repeat with all
samples. Switch on the vacuum pump. When all the contents
have been drawn through the columns completely, switch off
the vacuum pump. Carefully remove and discard the tube
extender when the pressure has reached 0 mbar.
10. Apply 600 μL Buffer ACW1 to the QIAamp Mini column.
pump until all of ACW1 has been drawn through the QIAamp
Mini column.
11. Switch off the vacuum pump and release the pressure to
0 mbar.
12. Apply 750 μL Buffer ACW2 to the QIAamp Mini column.

release the pressure to 0 mbar.
13. Apply 750 μL of ethanol (>95%) to the QIAamp Mini column.
pump. After all of ethanol has been drawn through the spin
column, switch off the vacuum pump and release the pressure
to 0 mbar.
14. Close the lid of the QIAamp Mini column. Remove it from the
vacuum manifold and place the QIAamp mini column into a
clean 2 mL collection tube. Centrifuge at full speed
(20,000 g) for 3 min.
15. Place the QIAamp Mini column into a new 2 mL collection
tube. Open the lid, and incubate the assembly at 56 C for
10 min to dry the membrane completely.
16. Place the QIAamp Mini column (from step 15) into a clean
1.5 mL elution tube and discard the 2 mL collection tube from
step 15. Carefully apply 100 μL of Buffer AVE to the center of
the QIAamp Mini membrane. Close the lid and incubate at
room temperature for 3 min.
17. Centrifuge in a microcentrifuge at full speed (20,000 g) for
1 min. The eluate contains cfDNA.
18. Store upright at 80 C.
3.3 Droplet Digital See Notes 10–12.

PCR Sample Follow generally accepted PCR rules to avoid contamination:
Preparation,
(a) Wipe down work surfaces and equipment using 70% ethanol:
Thermocycling, hood, bench, racks, pipettes, cartridge holders, waste beaker,
and Reading droplet generator and heat sealer before you start and after
you finish. Clean your working area on a weekly basis using
DNA Zap.
(b) Change gloves frequently: always use CLEAN gloves when
preparing a master mix, especially when opening a Taqman
probe tube. Change gloves between handling positive controls
and patient samples.
(c) Use aerosol resistant pipette tips and calibrated pipettes.
Check liquid level in the tip before/after pipetting.
1. In a micro centrifuge tube, prepare the reaction mixture in the

exact order indicated below, mix five to ten times by gently
pipetting up and down.
(a) 6.875 μL deionized Water
(b) 12.5 μL 2 ddPCR Supermix
(c) 0.625 μL 40 Taqman primer–probe mix
(d) When preparing master mix for multiple samples, allow
10% extra volume for pipetting.
2. Transfer master mix into a DNase- and RNase-free pipetting

reservoir.
3. Use an eight-channel pipettor, transfer 20 μL reaction mixture
from step 1 into wells of a PCR plate.
4. Add 5 μL of isolated cfDNA to each well according to prede-
termined sample layout. Each assay from one specimen is run in
triplicates of 5 μL of isolated cfDNA for a total of 15 μL
representing approximately 300 μL of isolated plasma (see sam-
ple plate layout in step 14). Pipette up and down slowly three
times to mix.
5. Secure the DG8 cartridges in cartridge holders. Load 20 μL of
reaction mix to the middle row.
6. Fill oil wells, in the lower row of the cartridge, with 70 μL of
Droplet Generation Oil.
7. Cover the cartridge with the DG8 gasket and load the cartridge
holder with the DG8 cartridge into Droplet Generator.
8. When the light on Droplet Generator turns green, take out the
cartridge and holder.
9. Use a manual 50 μL eight-channel pipettor, gently and slowly
pipette up 40uL droplets from the top row of the cartridge over
a 5 s period. Gently release the droplets into a 96-well PCR
plate over a 5 s period with each pipette tip touching the side of
a well near—but not at—the well bottom to avoid shearing the
droplets (Fig. 2).
10. Repeat droplet generation until all the cartridges are processed
and the 96-well plate is filled to the extent desired.
11. Cover the PCR plate with a sheet of Easy Pierce Foil PCR Plate
Seal. Mark A1 at the right corner. Seal the PCR plate using a
preheated Eppendorf PCR Plate Sealer by pressing to the sec-
ond tier and counting to six. Rotate the plate 180˚ and repeat.
Fig. 2 Example of good emulsification of droplet generation. A successful

emulsification should yield a clear separation of two phases (a) that consist of
individual droplets (b)
12. Place the sealed plate in a thermal cycler and run under the
following conditions:
Lid temperature at 105 ˚C.
(a) 95 ˚C for 10 min.
(b) 94 ˚C 30 s. Ramp to annealing temperature at 2.5 ˚C/s.
(c) 1 min at annealing temperature.
l EGFR ex19 ¼ 55 ˚C.
l EGFR L858R ¼ 58 ˚C.
l EGFR T790M ¼ 58 ˚C.
(d) Repeat b and c for 40 cycles.
(e) Hold at 10 ˚C.
13. Set up a plate layout in QuantaSoft.
14. Prime the QX100 reader with droplet reader oil.
15. Transfer the finished PCR plate to the QX100/200 and read
the plate. A general plate layout for 24 samples is shown below:
1 2 3 4 5 6 7 8 9 10 11 12
A Hi Ctrl #1 #1 #1 #9 #9 #9 #17 #17 #17 Hi Ctrl
B Med Ctrl #2 #2 #2 #10 #10 #10 #18 #18 #18 Med ctrl
C Lo Ctrl #3 #3 #3 #11 #11 #11 #19 #19 #19 Lo Ctrl
D NTC #4 #4 #4 #12 #12 #12 #20 #20 #20 NTC
E NTC #5 #5 #5 #13 #13 #13 #21 #21 #21 NTC
F NTC #6 #6 #6 #14 #14 #14 #22 #22 #22 NTC
G NTC #7 #7 #7 #15 #15 #15 #23 #23 #23 NTC
H NTC #8 #8 #8 #16 #16 #16 #24 #24 #24 NTC
3.4 Analysis of EGFR 1. These assays are analyzed using the QuantaSoft 1D droplet
L858R and T790M plots. Set the FAM and VIC thresholds based on the amplitude
Droplet Data of the positive controls (Fig. 3). General observed fluorescence
amplitude ranges for L858R and T790M are listed in Table 2.
2. Sum up the “concentrations/20 μL” of PCR reaction
(provided as results output by QuantaSoft) for each triplicate
belonging to the assay.
3. Results are normalized to copies/mL of plasma by multiplying
by 3.3. This dilution factor accounts for the volume of sample
used for ddPCR, the number of replicates, the volume of plasma
that was extracted, and the volume the extraction was eluted in.
Fig. 3 Analysis of EGFR L858R by 1D gating. The top (WT) and bottom (MUT)
images are VIC and FAM amplitudes of varying EGFR L858R concentrations in a
constant background of 5000 copies of wild type EGFR. Cutoff intervals are
generally established by running positive control samples (e.g., H1975 genomic
DNA for EGFR L858R) along unknowns
Table 2
Typical amplitudes of individual EGFR mutations using ddPCR
Amplitude range
Assay FAM VIC

EGFR del19 >3500 >3000
EGFR L858R >3500 >3000
EGFR T790M >4000 >2000
3.5 Analysis of EGFR See Note 13.

Exon19 Deletion Our EGFR exon19 deletion assay uses a VIC-labeled “refer-
Droplet Data Using ence probe” sequence that is shared by both wild-type and
QuantaSoft 2D Droplet mutant alleles, and thus gives a signal for both, while the
Plots FAM-labeled probe sequence spans the exon19 deletion region
and thus is only detected on wild-type targets [10]. As a conse-
quence, this assay type only gives three discernible clusters.
Figure 4 shows the assay design and 2D droplet plot examples
for the three observed clusters.
1. Select 2D Amplitude in QuantaSoft.
2. Click on lasso function under “Analysis” to the left.
Fig. 4 Analysis of EGFR exon 19 deletion by 2D gating. (a) The assay is designed so that a VIC-labeled
“reference probe” is shared by wild-type and the deletion 19 exon mutants, while a FAM-labeled probe
sequence spans the deletion hotspot. Thus, an EGFR wild-type sample will therefore show double positive
(brown) droplets (b), while an EGFR exon19 mutant will only have VIC-labeled droplets (bottom right) (c).
Negative droplets with no template are shown in d and are in the bottom left corner
3. Lasso the (FAM+VIC+) cluster and the VIC+ only positive

cluster based on the amplitudes of the positive controls. The
former contains the total wild-type EGFR population and the
latter EGFR exon19 deletion population only. An example of a
wild-type only population is shown in Fig. 4b and a sample that
additionally contains EGFR exon19 deletions in Fig. 4c.
4. Sum the copies/20 μL well of PCR of each triplicate for the
FAM+ population, and the copies/20 μL well of PCR of each
triplicate for the VIC+ cluster.
5. Subtract the value of FAM+ from the value of VIC+. This value
estimates the concentration of EGFR exon19 deletion. See
Normal Population
double
F01 occupied F11
double F10
negative F00
3 cluster view of the same

population
double
occupied F11 + F01
double F10
negative F00
Fig. 5 In a nonstandard, three-cluster, drop-off assay, WT targets appear as doubly positive droplets rather than
single positive droplets. As a result, the QuantaSoft algorithms used to compute the channel concentrations
(where normally, one dye channel represents the presence of one species and double positive droplets represent
the presence of both species) must be interpreted differently. The clusters are designated as C1 (F00), C2(F10)
and C3 (F11 + F01) and refer to the number of droplets in that particular cluster. Based on Poisson Statistics the
concentration (in copies/uL) of WT and mutant alleles in the ddPCR reaction are given by: (1) ConcCh1 ¼
ConcWT = {ln [(C1 + C2)/(C1 + C2 + C3)]} [(1000nL/ul)/(0.85nL/droplet partition)] (2) ConcCh2 = ConcWT þ
ConcMUT = {ln [C1/(C1 þ C2 þ C3)]}[(1000nL/ul)/(0.85nL/droplet partition)] (3) ConcCh2 ConcCh1 ¼
ConcMUT ¼ {ln [C1/(C1 þ C2)]} [(1000nL/ul)/(0.85nL/droplet partition)]. These formulas assume that the VIC
(or HEX) only channel is monitoring the mutant allele
Fig. 5 for further explanations how concentrations are calcu-

lated for three-cluster, drop-off assays.
6. Results are normalized to copies/mL of plasma by multiplying
by 3.3.
3.6 Analytical Sensitivity and specificity of ddPCR assays:

Considerations for In our experience the lowest % of mutant allele that can be
ddPCR Assays of reliably detected in a wild type background is 0.05%. We recom-
cfDNA mend that prior to testing plasma cfDNA samples one should test
their technique on low/high copy positive and negative con-
structed samples to help assess the sensitivity and specificity of
EGFR T790M, EGFR exon19 deletion, and L858R in their own
laboratory. Positive control samples can be purchased from Hori-
zon Discovery. Typical acceptance criteria for constructed speci-
mens are:
1. Sensitivity: Seven molecules for all EGFR mutations (T790M,
L858R, exon19 deletion) in a “low” wild-type background
(2000 genome equivalent wt background), or 20 molecules
for all EGFR mutations in a “high” wild-type background
(50,000 genome equivalent wt background)
2. Specificity: 100% for all EGFR mutations in both low and high
wild-type backgrounds
3. No single positive droplets in any no-template controls. Run at
least 5 NTC samples per plate.
3.7 Interpretation of See Note 14.

Plasma Genotyping Interpretation of ddPCR-based plasma genotyping results is
Results highly dependent on clinical context, disease characteristics and
the manner in which the assay is reported. The dynamic range of
plasma ddPCR in advanced NSCLC is wide and may be influenced
by disease burden and tumor biology [11, 13]. As such, reporting
ddPCR assay results as copies per mL of plasma provides informa-
tion on both tumor genotype as well as disease status and prognosis
[14]. The same factors that affect dynamic range are also key
determinants of assay sensitivity with false positive results being
more likely in patients with more limited disease, presumably sec-
ondary to lower rates of cfDNA shed from tumor [13]. The timing
of sample collection with respect to treatment also has a significant
impact on assay sensitivity with maximal assay sensitivity occurring
at initial diagnosis of metastatic disease as well as at subsequent
points of clear disease progression. Changes in quantitative plasma
cfDNA can be expected in response to therapy with decreasing
levels associated with response to effective therapy.
Plasma genotyping using a ddPCR-based assay exhibits exqui-
site specificity for the detection of targetable genomic alterations in
newly diagnosed advanced NSCLC [11, 12]. The detection of
EGFR sensitizing mutations using this method can thus reliably
identify patients that will benefit from targeted therapy. Conversely,
the detection of nontargetable driver mutations such as KRAS
codon 12 mutations can be used as a reliable indicator that target-
able alterations are unlikely to be identified by other methods with
the possible exception of BRAF V600E alterations [15]. Apparent

false positives in newly diagnosed patients should prompt careful
scrutiny of tissue genotyping results and consideration of repeat
biopsy, particularly if high levels of mutant cfDNA are detected.
The specificity of plasma ddPCR for the detection of acquired
resistance mutations such as EGFR T790M is more complex to
interpret. Heterogeneity of acquired resistance mechanisms across
metastatic sites within the same patient is not uncommon. Tissue
genotyping can thus only detect apparent resistance mutations at a
single metastatic site, whereas plasma ddPCR represents an average
of resistance across all metastatic sites. The thresholds of plasma
EGFR T790M copy number, which predict response to EGFR
T790M specific inhibitors represents an active area of investigation.
However, the gold standard comparisons for evaluating plasma
genotyping methods in the acquired resistance setting are challeng-
ing given the limitations of tissue genotyping in this clinical
context.
4 Notes
1. Cheaper pipette tips often contain rough tip edges that can
shed microscopic plastic particulates into sample preps and
ddPCR reactions resulting in shredding of droplets.
2. Time from draw to freezing of plasma should be less than 4 h.
3. During centrifugation, brake switch must be off so the cell–-
plasma interface is not disturbed.
4. When transferring plasma, do not dip the tip of the pipette into
the plasma–cell interface. Leave a thin plasma layer intact over
the interface.
5. Sometimes the color varies from pink to red due to hemolysis,
which will not affect downstream ddPCR genotyping. If
plasma is too viscous for pipetting, it may indicate lysis of
white blood cells, which causes overwhelming copies of wild-
type allele being detected and may affect the ability to detect
rare events.
6. During lysis, make sure that a visible vortex forms in the tube.
To ensure efficient lysis, it is essential that the sample and Buffer
ACL are mixed thoroughly. Do not interrupt the procedure at
this time. Proceed immediately to step 5.
7. When setting up the QIAamp Mini columns over the QIAvac
24 Plus, make sure that the tube extender is firmly inserted into
the column in order to avoid leakage of sample. Keep the
collection tube for the dry spin in step 14.
8. To avoid cross-contamination, be careful not to move the tube
extenders over neighboring QIAamp Mini Columns.
9. Do not use plasma that has been frozen and thawed more than
once to prevent cryoprecipitates from clogging the QIAamp
Mini column. In case cryoprecipitates are clearly visible the
sample may be centrifuged for 5 min at 16,000 g.
10. Set up ddPCR reactions in an amplicon-free PCR work station
that is in an amplicon-free room.
11. The following protocol below is for the manual preparation of
droplet generation. More recently Bio-Rad’s automated drop-
let generator (Bio-Rad #1864101) can also be used.
12. Always carry no-template controls (NTCs) to monitor for
contaminations. If positive droplets are observed in NTC
wells, results from the plate cannot be used.
13. During droplet generation process, both sample wells and oil
wells must contain liquid or the droplet generator will not
generate droplets. Supermix (1) can be used in the sample
wells that are not being used.
14. Using 2 mL of plasma one expects to isolate between 10 and
150 ng DNA (3000–45,000 wild-type events).
References
1. Shaw AT, Kim DW, Nakagawa K et al (2013) 7. Shaw AT, Kim DW, Mehra R et al (2014)
Crizotinib versus chemotherapy in advanced Ceritinib in ALK-rearranged non-small-cell
ALK-positive lung cancer. N Engl J Med lung cancer. N Engl J Med 370:1189–1197
368:2385–2394 8. Diaz LA Jr, Williams RT, Wu J et al (2012) The
2. Mok TS, Wu YL, Thongprasert S et al (2009) molecular evolution of acquired resistance to
Gefitinib or carboplatin-paclitaxel in pulmo- targeted EGFR blockade in colorectal cancers.
nary adenocarcinoma. N Engl J Med Nature 486:537–540
361:947–957 9. Kuang Y, Rogers A, Yeap BY et al (2009) Non-
3. Bergethon K, Shaw AT, Ignatius Ou SH et al invasive detection of EGFR T790M in gefitinib
(2012) ROS1 rearrangements define a unique or erlotinib resistant non-small cell lung cancer.
molecular class of lung cancers. J Clin Oncol Clin Cancer Res 15:2630–2636
30:863–870 10. Yung TKF, Chan KCA, Mok TSK et al (2009)
4. David Planchard JM, Riely GJ, Rudin CM et al Single-molecule detection of epidermal growth
(2013) Interim results of phase II study factor receptor mutations in plasma by micro-
BRF113928 of dabrafenib in BRAF V600E fluidics digital PCR in non–small cell lung can-
mutation–positive non-small cell lung cancer cer patients. Clin Cancer Res 15:2076–2084
(NSCLC) patients. J Clin Oncol 31 11. Oxnard GR, Paweletz CP, Kuang Y et al (2014)
(Supplement):8009 Noninvasive detection of response and resis-
5. Garraway LA, Janne PA (2012) Circumventing tance in EGFR-mutant lung cancer using
cancer drug resistance in the era of persona- quantitative next-generation genotyping of
lized medicine. Cancer Discov 2:214–226 cell-free plasma DNA. Clin Cancer Res
6. Zhou W, Ercan D, Chen L et al (2009) Novel 20:1698–1705
mutant-selective EGFR kinase inhibitors 12. http://www.bio-rad.com/webroot/web/
against EGFR T790M. Nature pdf/lsr/literature/Bulletin_6407.pdf
462:1070–1074
13. Sacher AG, Paweletz CP, Alden R, et al. (2015) survival outcomes in NSCLC patients treated
A prospective study of rapid plasma genotyping with first-line intercalated erlotinib and chemo-
utilizing sequential ddPCR and NGS in newly therapy. Clin Cancer Res 21:3196–3203
diagnosed advanced NSCLC patients. IASLC 15. Sholl LM, Aisner DL, Varella-Garcia M et al
World Conference on Lung Cancer, Denver, (2015) Multi-institutional oncogenic driver
CO mutation analysis in lung adenocarcinoma: the
14. Mok TS, Wu YL, Soo Lee J et al (2015) Detec- lung cancer mutation consortium experience. J
tion and dynamic changes of EGFR mutations Thorac Oncol 10:768–777
from circulating tumor DNA as a predictor of
Chapter 13
Detection of Cancer DNA in Early Stage and Metastatic

Breast Cancer Patients
Arielle J. Medford, Riaz N. Gillani, and Ben Ho Park
Abstract
Breast cancer is the leading cause of cancer in women and the second leading cause of cancer-related death.
There are many subtypes of breast cancer, which can be identified through the process of molecular and
genetic profiling. While the current standard of care utilizes tumor tissue biopsy to subclassify breast cancer,
plasma tumor DNA (ptDNA) can be detected through droplet digital PCR (ddPCR) of plasma obtained
from a simple blood draw. Tissue biopsy is not only more invasive but because tumors exhibit heterogeneity
it can be less accurate. Blood collects DNA shed from normal and cancerous cells alike, thus ddPCR of
plasma offers a broader picture of a cancer’s genetic makeup. This chapter summarizes how patients with
breast cancer can be screened for specific cancerous mutations in both tissue and plasma through the use of
ddPCR.
Key words Breast cancer, Droplet digital, PCR, Liquid biopsy, Biopsy, PIK3CA, Plasma, Blood
1 Introduction
Every year, over 230,000 women are diagnosed with breast cancer,
and nearly 40,000 die from the disease. It is the most commonly
diagnosed cancer in women and the second leading cause of cancer-
related death [1, 2].
Typically, breast cancer is detected via mammography and con-
firmed through tissue sampling. From here, patients undergo spe-
cific therapies targeted toward their class of breast cancer, which is
defined principally by the presence or absence of estrogen receptor
(ER), progesterone receptor (PR), and human epidermal growth
factor receptor 2 (HER-2). Researchers have shown that within
these subgroups, there are specific, idiosyncratic mutations, which
Arielle J. Medford and Riaz N. Gillani contributed equally to this work.
209
210 Arielle J. Medford et al.
can be useful for assessing burden of disease, predicting response

to therapy, and monitoring the latter in a rapid and minimally
invasive approach. These include mutations in PIK3CA, ESR1,
TP53, MAP3K1, GATA3, RPTOR, ERBB2, and ERBB3 [3–6].
Current studies of receptor mutations, however, face a signifi-
cant barrier: they rely on single site biopsy of tumor tissue. This
method is problematic because tumors have been shown to exhibit
heterogeneity, so sampling a single piece of tissue may miss sites of
the specific mutations listed above. This same heterogeneity can
exist within metastatic tumors, and furthermore, there has been
documented heterogeneity between metastases at different sites of
the body, even if the metastases originated from the same primary
tumor [7, 8]. In other words, tissue biopsy limits clinicians from
seeing the whole genetic picture.
In order to circumvent this challenge, researchers can now use
droplet digital PCR (ddPCR) on samples of plasma to screen for
these same mutations, through a procedure colloquially termed
“liquid biopsy.” The body’s blood supply feeds primary tumors as
well as metastases, and preliminary research suggests there is no
selectivity in terms of which tumor cells’ DNA is deposited into the
blood. In fact, normal and cancerous cells deposit into circulation,
which makes the bloodstream an effective repository of cellular
DNA, cancer and otherwise [9–11]. Thus, circulating blood con-
tains traces of all variations of tumor DNA, which makes plasma
tumor DNA (ptDNA) most representative of the varied genetic
makeup of cancer cells. Sampling peripheral blood is also much
easier than tissue biopsy, and the procedure comes with far fewer
risks.
There are many burgeoning applications of ddPCR within
oncology, and breast cancer is poised to greatly benefit from the
approach. The efficacy of ptDNA/ddPCR for detecting PIK3CA
mutations in breast cancer has already been demonstrated with
93.3% sensitivity and 100% specificity in a study of 29 patients
with early stage breast cancer [12]. It is important to clarify that
this approach is specific for plasma, not serum. Research has shown
the latter to not be reliable, whereas plasma has reproducible data
supporting its efficacy [13, 14]. From identifying specific mutations
sensitive to targeted therapy, to monitoring response to treatment,
to detecting residual disease, this highly sensitive, cost effective
approach offers improvement of breast cancer treatment and for-
ward progress in studying the science behind it. This protocol
describes the methodology for using ddPCR to detect tumor
DNA in both plasma and primary tumor tissue.
Detection of Breast Cancer DNA 211
2 Materials
2.1 Design See Subheading 2.5.

and Testing of ddPCR
Primers and Probes
2.2 Blood Handling 1. EDTA or Cell-free DNA BCT tubes (Streck).

and Preparation 2. 23-gauge needle (see Note 1).
2.3 Blood Sample 1. Isolated hood space to prevent contamination.

Processing 2. Bleach wipes (1%).
and Plasma Separation
3. QIAamp Circulating Nucleic Acid Kit (Qiagen cat#: 55114).
4. DNase/RNase-Free distilled water.
5. Centrifuge.
6. DNase/RNase-free filter pipette tips (see Note 2).
2.4 Processing 1. Five to ten-μm thick unstained FFPE histology slides (10–15
and DNA Extraction slides depending on tumor size and the percentage of tumor
from Primary Tumor cells in the slices).
Samples 2. Container for soaking of the histology slides in xylene, ethanol,
and water.
3. Adjacent tissue slide with H&E stain, with tumor tissue demar-
cated from normal by a pathologist.
4. >99% xylene.
5. DNAse/RNAse-free distilled water.
6. 100, 70, and 50% ethanol.
7. Pinpoint Slide DNA Isolation System (Zymo Cat#: D3001).
8. QIAamp DNA FFPE tissue kit (Qiagen Cat#: 56404).
2.5 ddPCR of ptDNA 1. Isolated hood or sterilized space to prevent contamination

or FFPE Samples (separate from blood and plasma processing and preferably in
separate room).
2.5.1 Preamplification
2. Sample to be used as template (prepared from plasma or FFPE
tissue).
3. DNase/RNase-Free distilled water.
4. dNTPs (10 mM concentration).
5. Dimethyl sulfoxide (DMSO).
6. 5 Phusion Buffer: MgCl2, DMSO (NEB Cat#: B0518S).
7. Forward and reverse primers for amplification of locus of inter-
est (50 μM).
8. Phusion High-Fidelity DNA polymerase (NEB Cat#:

M0530S).
9. Thermal cycler with 96-deep well reaction module.
10. 96-well PCR plates.
11. Foil heat seals.
12. DNAse, RNAse-free filter pipette tips.
13. 1.5 mL Eppendorf tubes.
14. QIAquick PCR purification kit (Qiagen Cat#: 28104).
2.5.2 ddPCR 1. ddPCR sterile hood (separate from pre-amp and blood/plasma
prep hoods and preferably in a separate room).
2. Sample prepared in preamplification step.
3. DNase/RNase-free distilled water.
4. Bleach wipes (1%).
5. Sterile troughs.
6. DG8 cartridges for QX200 droplet generator (Bio-Rad Cat#:
1864008).
7. DG8 cartridge holder (Bio-Rad Cat#: 1863051).
8. DG8 gaskets for QX200 droplet generator (Bio-Rad Cat#:
1863009).
9. Pipettes with tips: p10, p20, p100, p50 multichannel, p100
multichannel.
10. 1.5 mL Eppendorf tubes.
11. ddPCR supermix for probes (No dUTP) (Bio-Rad Cat#
186-3023).
12. Droplet generation oil for probes (Bio-Rad Cat# 186-3005).
13. Forward and reverse primers (50 μM)—these may be the same
primers used for locus-specific preamplification (Ordered as
“custom oligos” through IDTDNA.com).
14. Wild-type and mutant probes (100 μM).
15. Promega Female (cat#: Promega G1521) (optional).
16. Four-to-six cutter restriction enzyme (optional).
17. 96-well semiskirted plates.
18. Pierceable foil heat seal (Bio-Rad Cat#: 1814040).
19. QX200 droplet generator (Bio-Rad Cat# 1864002).
20. PX1 PCR plate sealer (Bio-Rad Cat#: 1814000).
21. C1000 Touch Thermal Cycler with 96-Deep Well Reaction
Module (Bio-Rad Cat#: 1851197).
22. QX200 droplet reader (Bio-Rad Cat#: 1864003).
23. QX200 droplet reader oil (Bio-Rad Cat#: 1863004).
24. Laptop computer with Windows operating system.

25. Quantasoft Software.
2.6 Data 1. Laptop computer with windows operating system.

Interpretation 2. Quantasoft Software.
and Statistical
3. Microsoft Excel.
Analysis
3 Methods
The following is the complete protocol for use with blood samples. For
use with FFPE, use the following protocol with the modifications
specified in the addendum at the end of the Methods section.
3.1 Design 1. Primers and probes can be ordered at idtdna.com.

and Testing of ddPCR 2. The ideal amplicon size ranges from 80 to 150 base pairs (see
Primers and Probes Note 3).
3. When designing primers, incorporate the following
parameters:
(a) Maintain a length of 20–28 base pairs.
(b) Melting temperature (Tm) should range from 45 to 55 C.
(c) Do not include more than three repeated base pairs in
a row.
(d) Confirm there is no primer heteroduplexing.
4. Probes are fluorescently labeled oligonucleotides. When
designing these, incorporate the following parameters:
(a) Maintain a length of 20–30 base pairs.
(b) The melting temperature (Tm) should be 5–10 C higher
than that of the primers.
(c) The 50 -end of the probe should NOT include guanine.
(d) The 50 -end should include the fluorophore.
(e) The 30 -end should include the fluorophore quencher.
(f) The probe should start at least 2 base pairs away from
primer binding site.
(g) Wild type probes are typically ordered with the
fluorophore Hex.
(h) Mutant probes are typically ordered with the fluorophore
6-Fam (see Note 4).
(i) Both probes should be quenched with Zen 3’Iowa
Black FQ.
5. When testing probes for sensitivity and specificity, first create a
positive control. This can be from a cell line that carries the
mutation, synthetic double stranded DNA or cloned plasmid

DNA. The positive control will confirm the probe’s capacity to
bind the target mutant allele.
6. Test the sensitivity and specificity of the positive control by first
ordering a “mini-order” of the mutant and wild type probes
and follow the “Subheading 3.5” detailed below, starting with
step 6 of “reaction preparation.”
3.2 Blood Handling 1. Draw the desired quantity of blood from patients (~10 mL per
and Preparation tube) into either an EDTA tube or a Cell-free DNA BCT Tube
(see Notes 5–8).
(a) If using EDTA, extract plasma from the blood within 2 h.
Plasma can be stored at 80 C.
(b) If using Cell-free DNA BCT tubes, blood can be stored
up to 7 days at room temperature. Tubes MUST be stored
at room temperature, and cold and heat should be avoided
(acceptable temperature range: 6–37 C).
2. If transporting blood, assure proper transport temperature (see
step 1a and 1b).
3.3 Blood Sample 1. Put on two layers of gloves and open the biohazard bag con-
Processing taining the patient sample in the PCR hood. Remain in the
and Plasma Separation hood for the remainder of the blood sample processing steps,
except for centrifugation. This precaution minimizes risk of
contamination of this very sensitive assay.
2. Open the EDTA tube, and transfer its contents to a 15 mL
conical tube. After the blood is removed, dispose of the tube
and the outer layer of gloves in the original biohazard bag to
minimize contamination.
3. Centrifuge the 15 mL conical tubes at 1500 150 rcf for
10 min. This step can be done at room temperature or at 4 C.
4. Use a 10 mL serological pipette to transfer the supernatant to a
new 15 mL conical tube. Do not combine plasma from the
same patient at this step. In other words, each tube should
translate to its own conical tube at this step. Do not touch the
buffy coat. It is good practice to leave a bit of plasma volume
behind to ensure minimum contamination of cells (see Note 9).
(a) Optional: Buffy coats can be saved or processed at this
stage for baseline germline DNA (see Note 10).
5. Centrifuge the conical tubes at 3000 150 rcf for 10 min.
6. Use a 5 mL serological pipette to transfer the supernatant to a
new 15 mL conical tube. At this point, plasma from the same
patient can be combined if there is no visible hemolysis. Leave
0.3 mL (~7 mm) on the bottom of the tube to avoid cellular
contamination.
7. Gently mix and record the plasma volume. Expect ~4 mL

plasma per 10 mL blood.
8. Optional: Aliquot 1–1.5 mL plasma to a 1.5 mL vial, and store
at 80 C until further use.
3.4 Plasma DNA 1. Perform all extraction steps in a sterile hood. Prior to begin-
Extraction ning extraction, wipe down all surfaces, equipment, and gloves
with bleach.
2. If plasma was stored at 80 C, thaw it on ice.
(a) Use a pipette to transfer thawed plasma to a 15 mL
conical tube.
(b) Centrifuge at 3600 rpm (3000 rcf) for 15 min at 4 C or
room temperature (see Note 11).
3. Upon returning to the hood, wipe everything again with
bleach. Use a pipette to transfer the supernatant to a new
15 mL conical tube. Leave 0.3 mL (~7 mm), since the pellet
will likely not be visible. The plasma may now be combined in
up to 3 mL per tube (two 1.5 mL vials). This is the maximum
volume specified for the QIAamp® Circulating Nucleic
Acid Kit.
4. Prepare up to 3 mL plasma according to the QIAamp® Circu-
lating Nucleic Acid Kit manufacturer instructions. Elute DNA
in 50 μL of water. They can be used immediately or else stored
at 2–8 C if they are to be used in the next 24 h, or 15 to
30 C until ready for use (see Notes 12–15).
3.5 ddPCR of ptDNA 1. Collect prepared sample to be used as DNA template for
or FFPE Samples ddPCR reaction (from plasma or from FFPE tissue).
3.5.1 Preamplification 2. Genomic DNA (gDNA) from normal cells will serve as an
to Increase the Absolute appropriate negative control in most cases, as gDNA should
Count of Alleles or Genome contain the wild-type locus of interest without mutant alleles.
Equivalents Alternatively, Promega sells a female genomic DNA that should
also have purely wild-type loci.
(a) Prior to use, gDNA or Promega will need to be digested
by a restriction enzyme with ubiquitous restriction sites
throughout the genome, but no restriction site present in
the locus of interest. For most assays, fragments <5 kb are
appropriately sized, though depending on the assay, frag-
ments may need to be as small as <500 bp. For this
protocol, a four- to six-cutter enzyme (e.g., CviQI) gen-
erated appropriately sized fragments (see Note 16).
3. In a bleach-cleaned dry hood, create a 50-μL PCR using with
separate reactions for the sample, positive control, negative
control, and NTC. This is a separate hood from the one to be
used for the droplet generation. Each 50 μL reaction will
comprise the following: 2 μL of template, 10 μL of 5 Phusion

Buffer, 1 μL of dNTPs (10 mM), 1.5 μL of DMSO, 1 μL of
forward primer (50 μM), 1 μL of reverse primer (50 μM),
33 μL of DNAse/RNAse-Free distilled water, and 0.5 μL of
Phusion High-Fidelity DNA polymerase. Phusion polymerase
should be added last to the reaction.
4. Aliquot each 50 μL reaction into a single well of the 96-well
plate. Once all samples have been aliquoted, seal plate with foil
and centrifuge plate for <15 s.
5. Set PCR cycling protocol as follows on thermal cycler: (1) 30 s
at 98 C; (2) 15 s at 98 C; (3) 30 s at an annealing temperature
that is below the melting point of the primers used for amplifi-
cation (see Note 1); (4) 30 s at 72 C. Cycle through step 2–4
for 10 cycles. Then resume with steps: (5) 5 min at 72 C;
(6) infinite hold at 4 C. Ramp setting should be ~2 C/s.
6. After cycling, run each PCR reaction through PCR purification
per “QIAquick PCR purification kit” manufacturer instruc-
tions. Elute end-product into 50 μL of DNAse/RNAse-free
distilled water.
3.5.2 Reaction 1. Prepare for entry into the ddPCR hood by using bleach wipes
Preparation (Perform to clean the exterior, bench top, and any other surfaces that
in Clean Pre-PCR Hood) may be encountered during set up of the assay.
2. Obtain ddPCR Supermix from 20 C freezer and thaw at
room temperature.
3. If product from preamplification has been stored at 20 C,
thaw the product at room temperature.
4. After the product has been thawed, mix well and make serial
tenfold dilutions of the product using distilled water. These
dilutions should span approximately four orders of magnitude
and will be used as the templates for the ddPCR assay to ensure
that at least some dilutions fall within the dynamic range of the
ddPCR system (~5 logs) and give adequate precision. Aliquot
each dilution into a 1.5 mL Eppendorf tube.
5. The same serial tenfold dilutions should be done for positive
and negative DNA controls and NTC. Aliquot distilled water
to be used as a no-template control (NTC) into a 1.5 mL
Eppendorf tube.
6. Make the 20 primer–probe mixture to be used for the ddPCR
reaction according to the proportions below (see Note 18):
Reagent Vol (μL)

Primer F (50 μM) 36
Primer R (50 μM) 36
Probe WT (100 μM) 5
Probe Mut (100 μM) 5
Water 18
Total 100
7. Bring a sterile trough into the ddPCR hood and remove from
the protective plastic cover. Place supermix, primer–probe mix,
and droplet generation oil in the ddPCR hood.
8. Fill the sterile trough with droplet generation oil using a
p1000 pipette with at least 100 μL of oil available for each
reaction.
9. Label an Eppendorf tube for every dilution of the product
being assayed. If the template has been diluted by tenfold
over four orders of magnitude, there should be four Eppendorf
tubes (see Note 19).
10. Within each Eppendorf tube, the reaction contents prior to
droplet generation will contain the supermix, primer–probe
mixture, and appropriate dilution of the template. A single
reaction volume for ddPCR will be 20 μL, divided according
to the table below (Table 1). If each dilution is being run in
triplicate (as is recommended), four volumes of reaction
should be prepared to accommodate for any pipetting error
and to avoid introducing bubbles into the sample well of the
DG8. Therefore, 40 μL of Supermix, 4 μL of primer–probe
mixture, and 36 μL of template should be combined to form an
80 μL reaction mixture in each Eppendorf tube.
3.5.3 Cartridge 1. Place a single DG8 cartridge into the cartridge holder, sliding
Preparation, Droplet the two ends of the cartridge holder inward to lock into place;
Generation, and PCR Plate the cartridge holder can be locked only if the cartridge is
Preparation (Fig. 1) inserted with the proper orientation.
Suggested Plate Layout 2. The cartridge is comprised of 3 rows of 8 wells, with the wells
in each row being of a different diameter than those in the
other rows. If the cartridge holder is oriented so that text may
be read upright by the user, the bottom row of wells will
contain the droplet generation oil (as indicated by the label
on the cartridge holder). Vortex the contents of each PCR
reaction in each Eppendorf tube, and then spin down. In the
middle row of wells, labeled as “sample,” aliquot 20 μL of
reaction into each well. If running each reaction in triplicate,
Table 1
Mastermix/primer probe mix
Reagent 1 (μL)
Template 9
20 primer–probe mix 1
ddPCR Supermix 10
Ch1+Ch2+:0 Ch1+Ch2-:0 Ch1-Ch2+:1096 Ch1-Ch2-:10414

10000
2270
9000
8000
7000
Channel 1 Amplitude
6000
5000 4991
4000
3000
2000
1000
0
0 1000 2000 3000 4000 5000
Channel 2 Amplitude
Template: ESR1 D538G WT plasmid

Probes: ESR1 D538G MUT and WT probe
Fig. 1 Two-dimensional ddPCR amplitude plot showing that the assay detects the D538G gene (X-axis) and
does not erroneously detect the ESR1 wild type (Y-axis) in a D538G mutant plasmid
aliquot 20 μL of each reaction into three adjacent wells. This is

ideally done with an eight-channel pipettor, but if this is not
possible, it is okay to load 1 well at a time.
3. All 8 wells in the middle row of the cartridge should be filled.
Wells remaining in the middle row of the cartridge may be used
for a positive and negative control, or else they may be filled
with water (see Notes 20–23). For the example where each
reaction is run in triplicate, a single cartridge may have the wells
1–3 filled with the reaction contents for one dilution of the
template, wells 4–6 filled with the reaction contents for the
another dilution of the template, well 7 filled with the reaction
contents for the positive control, and well 8 filled with the
reaction contents for the negative control.
4. Use the p100 multichannel pipette to draw up 70 μL of droplet

generation oil per channel; then, aliquot 70 μL of oil into each
well in the bottom row of the cartridge (see Notes 24–26).
5. After the bottom and middle rows of the cartridge are filled,
cover the cartridge with the DG8 gasket, stretching it from one
end of the cartridge over to the other and placing it over the
prongs to secure it in place.
6. Place the cartridge holder containing the filled cartridge with
secured gasket into the QX200 droplet generator. The machine
will detect the cartridge and indicate this by lighting up a green
light. The generator may then be closed with the click of the
top button. Once closed, droplet generation will commence
unless there is some error with cartridge preparation. The
droplet light on the right of the machine will flash during
generation.
7. After several minutes, droplet generation will be complete,
which will be indicated by the droplet light on the far right of
the machine remaining lit. Open the droplet generator using
the button on the sliding green cover, remove the cartridge,
and place it back into the ddPCR hood. Remove the gasket,
and observe that the upper row of the cartridge is now filled
with thousands of droplets individually encapsulating PCR
reactions.
8. Place a 96-well semiskirted plate into the ddPCR hood. Use a p50
multichannel pipette set at 43 μL to slowly withdraw the contents
from the upper row of the cartridge, and deposit these into a
single column of the 96-well semiskirted plate (see Note 27).
9. Repeat steps 1–8 as many times as necessary, depending on the
number of dilutions that are being analyzed.
10. Once all ddPCR reaction contents have been aliquoted into the
96-well semiskirted plate, place a pierceable foil heat seal over
the plate. Turn on the PX1 PCR plate sealer using the power
switch on the back of the machine. The plate sealer will heat up
to 180 C over the course of approximately 2 min.
11. Once heating is complete, press the eject button on the touch
screen to open the plate sealer; insert the plate with foil on top
into the sealer, with the red line on the foil oriented closer to
the machine. Select the “seal” option on the touch screen.
Once plate sealing is complete, the plate may be removed
from the plate sealer, and the plate sealer turned off.
3.5.4 PCR and Plate 1. Set PCR cycling protocol as follows on thermal cycler:
Reading
Step Temp Time Ramp Cycle

1 95 10 min 2 C/s 1
2 94 30 s 2 C/s 40
3 Variable 1 min 2 C/s
4 98 10 min 2 C/s 1
5 4 Hold 1
2. After the PCR cycling protocol is complete, remove plate from

thermal cycler. The QX200 droplet reader should be attached
via a USB cable to a computer with a Windows operating
system and the Quantasoft software installed. Open the droplet
reader using the button on the sliding green cover. Using the
latch on either side, remove the metal cover that will be used to
secure the 96-well plate in place, slot the plate into place, and
then replace and relatch the cover. Close the sliding green
cover.
3. On the attached computer, open an instance of Quantasoft.
Under the “Setup” option on the left side of the screen, open a
prior template or plate, or create a new plate. Label each well
on the plate diagram with the sample name and dilution. Once
the plate has been saved, follow the prompts to assign the
FAM, VIC, and/or HEX probes to the appropriate channels.
4. Select “Run” from the left side of the screen. Indicate the
correct dyes (HEX or VIC) (see Note 28). The droplet reader
will take approximately 2 min to process each well.
5. After the entire plate has been read, the “Analyze” option may
be selected on the left-hand side of the screen to review the data
from the run.
3.6 Data Caption: Example 2D droplet plots.

Interpretation
1. Subsequent data analysis is contingent upon fluorescence
and Statistical
intensity gating for “positive” droplets containing the muta-
Analysis (Figs. 2, 3, tion of interest. The amplitude threshold for a positive droplet
and 4) may be set by selecting the “Analyze” option and viewing the
“2D amplitude” plots for the various samples. This can be done
with the controls first to maximize sensitivity, and then the
experimental samples may be superimposed. Place the thresh-
old as low as possible in order to maximize specificity. The
positive and negative controls should be evaluated with similar
DNA input amounts. An appropriate amplitude cutoff will vary
by experiment and should be set based on the amplitude of
droplets for controls.

10000
2270
9000
8000
7000
Channel 1 Amplitude
6000
5000 4991
4000
3000
2000
1000
0
0 1000 2000 3000 4000 5000
Channel 2 Amplitude
Template: ESR1 D538G MUT plasmid

Fig. 2 Two-dimensional ddPCR amplitude plot showing that the assay detects the D538G gene (Y-axis) and
does not erroneously detect the ESR1 wild type (X-Axis) in a D538G mutant plasmid

10000 2270
9000
8000
7000
Channel 1 Amplitude
6000
5000 4991
4000
3000
2000
1000
0
0 1000 2000 3000 4000 5000
Channel 2 Amplitude
Template: digested MCF7 gDNA

Fig. 3 Two-dimensional ddPCR amplitude plot showing that the assay detects the wild type ESR1 gene (X-axis)
and does not erroneously detect the D538G mutation (Y-axis) in digested MCF7 gDNA

9000
2455
8000
7000
Channel 1 Amplitude
6000
5000
4000 3967
3000
2000
1000
0
0 1000 2000 3000 4000 5000
Channel 2 Amplitude
Template: patient cfDNA with known ESR1 D538G mutation

Fig. 4 Two-dimensional ddPCR amplitude plot showing that the assay detects the wild type ESR1 gene (X-axis)
and the D538G mutation (Y-axis) in patient ctDNA
2. Once this threshold has been set, subsequent analyses may be

done. One such analysis is the determination of concentration,
which takes a Poisson distribution of occupied and unoccupied
droplets into account. Under “Analyze”, the “Concentration”
option allows for the visualization of concentration of wild-
type and mutant amplicons as copies/μL. Under this same
option, “Fractional Abundance” may be selected, allowing for
the visualization of concentration of mutant as percentage of
total.
3. Another analysis under the “Events” option allows for a visual-
ization of total number of droplets, as well as droplets that are
occupied by mutant and/or wild-type amplicons (see Notes
29–30).
*For analysis of FFPE, perform the below DNA extraction and
then begin above protocol at Subheading 3.5.
3.7 Processing 1. Heat a histology slide on a heating plate having a 5 to 10-μm

and DNA Extraction thick, unstained FFPE section at ~60 C for ~5–10 min until
from Primary Tumor paraffin wax melts (see Note 30).
Samples (FFPE) 2. Submerge slide into >99% xylene at room temperature for
~30 min.
3. After 30 min, discard this xylene and resubmerge in new
xylene; incubate at room temperature for another 30 min.
4. Discard this second batch of xylene, and then submerge the

slide in 100% ethanol for ~2 min. The slide should then be
submerged in 70% ethanol, 50% ethanol, and distilled water for
~2 min each, with the goal of gradual hydration.
5. Next, air-dry the slide for ~30–60 min; the slide will be dry
when it becomes visibly white.
6. Apply the pinpoint solution from the “Pinpoint Slide DNA
Isolation System” to the area of tumor tissue per the manufac-
turer’s protocol, using another prepared H&E slide of an
adjacent cross section of tissue as a guide (see Note 31).
7. Remove embedded tissue from the slide with a clean and sharp
scalpel, and transfer tissue to a clean Eppendorf tube; centri-
fuge briefly to bring the blue tissue pieces to the bottom.
8. Follow step 10 onward from the QIAamp DNA FFPE tissue
kit protocol to isolate gDNA from this tissue sample, ultimately
eluting into 20–100 μL of distilled water.
4 Notes
1. When drawing blood for use with ptDNA, it is best to use a

23 gauge needle. Smaller sizes risk shearing cells, and larger
sizes increases patient discomfort.
2. Bio-Rad protocol specifies Rainin® tips.
3. In general, the smaller the amplicon, the more sensitive the
ddPCR detection. Larger amplicons may be used to enable
multiple mutation sites to be detected by a single amplicon,
but large sizes (>500 bp) begin to lose fluorescence with
increasing length.
4. This is the convention. Most importantly, mutant and wild type
should have different fluorophores.
5. For this original protocol, researchers drew three tubes of
blood (~30 mL).
6. For every 10 mL of blood, expect 4 mL plasma to be isolated.
7. Store EDTA or Cell-free DNA BCT tubes horizontalFly to
minimize lysis. When lysed samples are processed, they will
have a pink-to-red tinge to the supernatant. Nonlysed samples
should have clear-to-yellow supernatants.
8. Tumor DNA is detected in plasma.
9. When transferring the supernatant during blood sample prepa-
ration, avoid using bulbs for suction, which may inadvertently
lead to suctioning too quickly and disruption of layers.
10. The buffy coat consists of white blood cells, which can be used
to determine the underlying germline DNA.
11. Centrifugation is necessary for thawed plasma due to cryopre-

cipitate, which could clog the DNA extraction columns.
12. During the cell lysis reaction of the QIAamp® Circulating
Nucleic Acid Kit protocol, make sure the contents are vortexed
to become homogenous
13. Do not pause the experiment at the lysis step of the nucleic acid
isolation.
14. At the last elution step of the QIAamp® Circulating Nucleic
Acid Kit protocol, centrifuge for 2 min (the protocol calls for
1 min).
15. During centrifugation, if the cap in the Eppendorf tube is left
open, it risks picking up contaminants from other materials
spun in the same centrifuge. Thus, remove the cap before
spinning, and place a new cap on the tubes after the spin.
16. Sanger sequencing may be completed on FFPE and plasma
samples prior to ddPCR to confirm that the amplicon of inter-
est is present. This is a test to confirm that the amplicon is there
but will likely detect only wild type. ddPCR will detect specific
mutants, which will be present in much lower abundance.
17. Exact annealing temperature varies with primers. First, test the
primers with a full-cycle temperature gradient PCR on geno-
mic DNA to determine optimal annealing temperature.
18. Primer probe mix can be stored at 20 C for later use.
19. Serial tenfold dilutions of the template allow for the creation of
at least one concentration that appropriately saturates droplets
to a concentration in the “digital” range, which may vary
greatly depending on the target, but in this assay ranges from
~1000 to 5000 template-filled droplets out of ~15,000 total
droplets. In prior experiences of working with patient samples,
a dilution of approximately 103 has led to concentrations in
the digital range. The ideal dilution will vary by sample as well
as each instance of the protocol, and is contingent on final
elution volumes.
20. When aliquoting samples for ddPCR, start with water, then
wild type, then mutant. This will minimize contamination.
21. Thoroughly vortex the ddPCR supermix before pipetting. Do
the same before pipetting the mix into the ddPCR cartridge.
The contents of the mix tend to settle out of solution, so this
step is important for adequate distribution of contents.
22. All wells in the middle and bottom row must be filled in order
for droplets to be generated.
23. Avoid air bubbles when aliquoting into ddPCR cartridge. This
will disrupt adequate droplet formation.
24. Oil should be added only after sample has been added to all
8 wells—adding oil first will compromise droplet quality.
25. All oil must be at the same level. If air gets into the system, the
droplet generation terminates.
26. After adding oil, proceed to droplet generation within 2 min.
27. When suctioning droplets from the uppermost row of the
cartridge into the 96-well semiskirted plate, take the utmost
care to suction slowly and evenly. This will help to prevent the
shearing of droplets, maximizing the number of droplets per
reaction well, and therefore improving the utility of the ddPCR
assay. Once thermocycled, droplets are no longer fragile.
28. If the incorrect dyes are selected before the run, it is possible to
correct the information after the run.
29. Poisson statistics are used to calculate the copies/μL concentra-
tion metric based on the formula:
#of copies ¼ ln 1 Total Positives
counted Total counted. This for-
mula yields the expected value of copies in the portion of a
sample that was counted based on the number of droplets filled
with the amplicon of interest (positives) and the number of total
droplets counted (total counted). In the QuantaSoft analysis
software, this result is expressed as copies/μL which can be
related back to the amount of sample placed into the 20 μL
ddPCR reaction to determine the number of copies of mutant
or wildtype targets quantified in the reaction.
30. The figures below demonstrate the effect of varying levels of
sample abundance on the calculated concentration of the
amplicon of interest. The top graph shows total formed dro-
plets for ddPCR in green, with droplets filled with the ampli-
con of interest in blue, at three concentrations of template
DNA (serial twofold dilutions) and with water control. The
bottom graph illustrates the corresponding calculated concen-
trations (copies/μL) of the amplicon of interest, in this case the
ligand-binding domain of ESR1 ex10. As can be seen, there is
concordance between an approximate 50% reduction in the
concentration of template DNA at dilutions of 1.25 106
to 6.25 107, and the resulting approximate 50% reduction
in calculated concentration of the amplicon of interest. How-
ever, at a sample abundance where there is near full saturation
of generated droplets, such as a dilution of 2.5 106 in this
experiment, the calculated concentration of the amplicon of
interest is less reliable. This may be explained by the fact that
the former dilutions are closer to the digital range, defined as
saturation of a few thousand droplets of the approximately
15,000–20,000 that are generated in total, where the concen-
tration of the amplicon of interest may be more reliably
measured (Fig. 5).
20000
17406 17478 17315 17830
16309 16462
15650
15000
14721
# of Events
10150
10000 9197 9109
8540
5661
4895
5000
4 0
0
-6 ·· ·· ···
2.5 x 10 1.25 x 10-6 6.25 x 10-7 Water control
Droplets filled with amplicon of interest All formed droplets
3500
amplicon of interest (copies / uL)
Calculated concentration of
3100
3000 2900
2500
2000
1500
1000 899 962
500 430 388

0
0 0.254
· ··
2.5 x 10-6 -6
1.25 x 10 6.25 x 10-7 Water control
Genomic DNA template sample (concentration)
Fig. 5 Top: Bar graph of ddPCR analysis of varying concentrations of samples. The number of droplets positive
is indicated at the top of each bar (blue–droplets containing mutated DNA, green–total number of droplets
analyzed). Water is used as a negative control. Bottom: Increasingly diluted concentrations of genomic DNA
entered into assay
31. Tissues are prone to DNA contamination due to the normal

processing of FFPE blocks. Fresh microtome blades should be
used for each specimen, and if possible, it is preferable to
discard the first 2–3 slices.
32. When using the “Pinpoint Slide DNA Isolation System,” use a
pipette tip to swab and spread the pinpoint solution over the
area of interest.
References
1. American Cancer Society (2014) Breast cancer 4. Cancer Genome Atlas Network (2012) Com-
facts & figures 2013-2014. American Cancer prehensive molecular portraits of human breast
Society Inc., Atlanta tumours. Nature 490:61–70
2. Eckhardt BL, Francis PA, Parker BS, Anderson 5. Robinson DR, Wu YM, Vats P et al (2013)
RL (2012) Strategies for the discovery and Activating ESR1 mutations in hormone-
development of therapies for metastatic breast resistant metastatic breast cancer. Nat Genet
cancer. Nat Rev Drug Discov 11:479–497 45:1446–1451
3. Toy W, Shen Y, Won H et al (2013) ESR1 6. Bose R, Kavuri SM, Searleman AC et al (2013)
ligand-binding domain mutations in Activating HER2 mutations in HER2 gene
hormone-resistant breast cancer. Nat Genet amplification negative breast cancer. Cancer
45:1439–1445 Discov 3:224–237
7. Gerlinger M, Rowan AJ, Horswell S et al 11. Casciano I, Vinci AD, Banelli B et al (2010)
(2012) Intratumor heterogeneity and Circulating tumor nucleic acids: perspective in
branched evolution revealed by multiregion breast cancer. Breast Care (Basel, Switzerland)
sequencing. N Engl J Med 366:883–892 5:75–80
8. Higgins MJ, Jelovac D, Barnathan E et al 12. Beaver JA, Jelovac D, Balukrishna S et al
(2012) Detection of tumor PIK3CA status in (2014) Detection of cancer DNA in plasma of
metastatic breast cancer using peripheral patients with early-stage breast cancer. Clin
blood. Clin Cancer Res 18:3462–3469 Cancer Res 20:2643–2650
9. Swarup V, Rajeswari MR (2007) Circulating 13. El Messaoudi S, Rolet F, Mouliere F, Thierry
(cell-free) nucleic acids – a promising, AR (2013) Circulating cell free DNA: preana-
non-invasive tool for early detection of several lytical considerations. Clin Chim Acta
human diseases. FEBS Lett 581:795–799 424:222–230
10. Hodgson DR, Wellings R, Orr MC et al (2010) 14. Oshiro C, Kagara N, Naoi Y et al (2015)
Circulating tumour-derived predictive biomar- PIK3CA mutations in serum DNA are predic-
kers in oncology. Drug Discov Today tive of recurrence in primary breast cancer
15:98–101 patients. Breast Cancer Res Treat 150:299–307
Chapter 14
Droplet Digital PCR for Minimal Residual Disease Detection

in Mature Lymphoproliferative Disorders
Daniela Drandi, Simone Ferrero, and Marco Ladetto
Abstract
Minimal residual disease (MRD) detection has a powerful prognostic relevance for response evaluation and
prediction of relapse in hematological malignancies. Real-time quantitative PCR (qPCR) has become the
settled and standardized method for MRD assessment in lymphoid disorders. However, qPCR is a relative
quantification approach, since it requires a reference standard curve. Droplet digitalTM PCR (ddPCRTM)
allows a reliable absolute tumor burden quantification withdrawing the need for preparing, for each
experiment, a tumor-specific standard curve. We have recently shown that ddPCR has a good concordance
with qPCR and could be a feasible and reliable tool for MRD monitoring in mature lymphoproliferative
disorders. In this chapter we describe the experimental workflow, from the detection of the clonal molecular
marker to the MRD monitoring by ddPCR, in patients affected by multiple myeloma, mantle cell
lymphoma and follicular lymphoma. However, standardization programs among different laboratories
are needed in order to ensure the reliability and reproducibility of ddPCR-based MRD results.
Key words Minimal residual disease (MRD), Follicular lymphoma (FL), Mantle cell lymphoma
(MCL), Multiple myeloma (MM), Immunoglobulin heavy-chain gene (IGH), t(14;18) translocation,
t(11;14) translocation, BCL2/IGH, BCL1/IGH, Droplet digital PCR
1 Introduction
Minimal residual disease (MRD) monitoring in hematology is

defined as any approach aimed at detecting and possibly quantifying
residual tumor cells beyond the sensitivity level of routine imaging
and laboratory techniques. Whenever a patient achieves complete
clinical remission a number of different scenarios might actually
take place including full eradication of the neoplastic clone, long-
term persistence of quiescent or nonclonogenic or immunologi-
cally regulated tumor cells or persistence of clonogenic cells capable
of giving rise to a full clinical relapse within months or years. MRD
analysis is currently a valuable early predictor of outcome in several
lymphoid malignancies, having made a considerable impact on
clinical research. Actually, the persistence of detectable MRD after
229
230 Daniela Drandi et al.
a successful treatment heralds the disease recurrence and thus MRD

monitoring might drive preemptive treatments, aimed at avoiding
or delaying the onset of clinical relapse [1–8].
Different technologies have been used for MRD study. Molec-
ular methods—including polymerase chain reaction (PCR), real-
time quantitative PCR (qPCR), and the most recently introduced
next-generation sequencing (NGS)—are widely employed for
MRD monitoring in acute lymphoblastic leukemia (ALL) and lym-
phomas [9–19], while multicolor flow cytometry (MFC) has a
major role for MRD detection in chronic lymphocitic leukemia
(CLL) and multiple myeloma (MM) [20–22]. Although nowadays
qPCR is a well-established tool for MRD monitoring in lympho-
proliferative disorders, it has the main disadvantage of relying upon
a serial dilution standard curve for target quantification. One of the
latest evolutions of PCR, droplet digital PCR (ddPCR), has intro-
duced several practical advantages, compared to qPCR, most nota-
bly its absolute quantification nature does not require a reference
standard curve [23–33]. Moreover, we have recently shown that
ddPCR has a good concordance with qPCR with respect to sensi-
tivity, accuracy, and reproducibility, in the context of MRD evalua-
tion in mature lymphoproliferative disorders [34]. The greater
applicability and reduced labor intensiveness, compared to qPCR,
recommend ddPCR as an attractive alternative method for MRD
assessment.
In this chapter, we firstly provide a technical roadmap for a
suitable clonal-specific tumor marker identification from circulating
tumor cells. Then we describe our ddPCR-based MRD detection
approach based on different target types: (1) the immunoglobulin
heavy chain gene rearrangement (IGH) for MM and mantle cell
lymphoma (MCL), (2) the BCL1/IGH fusion gene, derived from
the chromosomal translocation t(11;14), for MCL, and (3) the
BCL2-MBR/IGH gene, arising from the chromosomal transloca-
tion t(14;18), for follicular lymphoma (FL).
Finally, we point out that ddPCR prognostic value still needs to
be validated in the context of prospective clinical trials. Most
importantly, before entering into the clinical practice, the promise
of ddPCR, compared to qPCR and MFC, needs to be proved in the
context of standardization programs. It is part of the current activ-
ities of the Euro-MRD group, a division of the European Scientific
foundation for Laboratory Hemato Oncology (ESLHO), to test
and validate these ddPCR assay at the inter-laboratory level, as has
already been done several years ago for qPCR. Since this is currently
a work in progress, the ddPCR procedures described in this manu-
script are likely to be updated in the not too distant future.
ddPCR for MRD Monitoring in Mature Lymphoid Malignancies 231
2 Materials
2.1 Lab Workflow PCR protocols need appropriate caution to safeguard from con-
Requirements taminations. This aspect is particularly important in ddPCR, in
which even a single contaminant molecule could be detected. To
avoid contamination it is crucial to use clean supplies (i.e., gloves,
racks, tubes, plates, etc.) and dedicated pipettes (see Note 1).
Therefore, before PCR experiment setup, verify the following
workspace availability to prevent the risk of DNA contamination:
l Pre-PCR zone: dedicated to mix preparation. Keep any DNA or
PCR products away from this area.
l DNA zone: dedicated to adding template and droplet genera-
tion. Placed in a room separate from the pre-PCR zone.
l Post-PCR zone: dedicated to PCR amplification, PCR product
analysis and droplet reading.
2.2 Instruments 1. Fume hood for agarose gel setup.

and Software 2. Vortexer and minicentrifuge.
3. Nanodrop UV-Vis spectrophotometer (Fisher Thermo Scien-
tific, Waltham, MA, USA).
4. Benchtop centrifuge with microplate carriers and
microcentrifuge.
5. Thermal Cycler with heated lid, i.e., T100 (Bio-Rad, Labora-
tories, Hercules, CA, USA) or thermal cycler with similar
characteristics able to reach a ramp rate of 2.5 C/s.
6. Horizontal electrophoresis apparatus and power supply
(2–300 V/4–500 mA output voltage/current range).
7. Digital camera or any gel documentation and imaging systems.
8. DNA Sequencing facility for Sanger sequencing analysis.
9. Thermomixer (i.e., Eppendorf ThermoMixer C).
10. QX100TM or QX200TM ddPCRTM System (Bio-Rad, Labora-
tories, Hercules, CA, USA).
11. PX1TM PCR plate sealer (Bio-Rad, Laboratories, Hercules,
CA, USA).
12. Multichannel pipettes and filtered pipette tips.
13. Chromas (free licenced software) for electropherogram
visualization.
14. IMGT/V-QUEST tool (www.imgt.org) for IGH genes
sequence analysis.
15. Blastn (www.blast.ncbi.nlm.nih.gov) for BCL2-MBR/IGH
and BCL1/IGH genes sequence analysis.
16. Primerquest and Oligo Analyzer 3.1 tools (www.eu.idtdna.

com) or Primer3Plus (www.primer3plus.com) for forward
and reverse allele-specific oligonucleotide (ASO) primers and
probes design.
17. QuantaSoftTM ddPCR analysis software (Bio-Rad, Labora-
tories, Hercules, CA, USA).
2.3 Reagents 1. Erythrocyte lysis buffer (0.155 M NH4Cl, 10 mM KHCO3,

and Kits 0.1 mM ethylenediamine tetraacetic acid (EDTA),
pH 7.2–7.4).
2.3.1 Bone Marrow
(BM) and Peripheral Blood 2. Genomic DNA (gDNA) extraction from circulating tumor
(PB) Sample Processing cells can be performed by different methods such as DNAzol
(Thermo Fisher Scientific, Waltham, MA, USA), NucleoSpin
Tissue (Macherey-Nagel, Bethlehem, PA, USA), or automated
systems, such as the Maxwell RSC (Promega, Madison, WI,
USA).
2.3.2 Tumor-Specific 1. Forward and reverse consensus primers for IGH [16, 35, 36],
Molecular Marker BCL1/IGH [11] and BCL2-MBR/IGH [9] PCR screening
Assessment and sequencing (Table 1).
2. PCR reagents: GoTaq Flexi (5 U/μL), 5 Green GoTaq Flexi
buffer, 25 mM MgCl2 (Promega, Madison, WI, USA), dNTPs
mix (2 mM) (nonspecific company), and nuclease-free water.
3. 1 TAE buffer (40 mM Tris–acetate, 1 mM EDTA, in distilled
water, pH 8.0).
4. Agarose gel (2%).
5. Ethidium bromide 1% or RealSafe staining solution (Durviz,
Spain).
6. 100 bp DNA mass ladder (i.e., φX174-HaeIII digest).
7. QIAquick gel extraction kit (Qiagen GmbH, Germany).
8. QIAquick PCR purification kit (Qiagen GmbH, Germany).
9. TOPO-TA cloning Kit (Thermo Fisher Scientific, Waltham,
MA, USA).
10. ImMedia AmpBlue and ImMedia AMPLiquid agar for IPTG/
X-gal plates and low-salt LB liquid medium for bacteria grow-
ing, respectively (Thermo Fisher Scientific, Waltham, MA,
USA).
11. Wizard Plus SV minipreps DNA purification System (Promega,
Madison, WI, USA).
12. gDNA (100 ng/μL) from a BCL2-MBR/IGH-positive cell
line (such as DOHH-2 or RL).
13. gDNA (100 ng/μL) from a BCL1/IGH-positive cell line
(such as JVM2).
Table 1
Primer sequences for IGH, BCL1/IGH and BCL2-MBR/IGH screening by PCR or nested-PCR. Reverse
primers are italicized
Target Primer Sequence (50 -30 )

IGH VH1Fs CAGGTGCAGCTGGTGCARYCTG
VH2Fs CAGRTCACCTTGAAGGAGTCTG
VH3Fs GAGGTGCAGCTGGTGSAGTCYG
VH4aFs CAGSTGCAGCTGCAGGAGTCSG
VH4bFs CAGGTGCAGCTACARCAGTGGG
VH5Fs GAGGTGCAGCTGKTGCAGTCTG
VH6Fs CAGGTACAGCTGCAGCAGTCAG
VH1D CCTCAGTGAAGGTCTCCTGCAAGG
VH2D TCCTGCGCTGGTGAAAGCCACACA
VH3D GGTCCCTGAGACTCTCCTGTGCA
VH4aD TCGGAGACCCTGTCCCTCACCTGCA
VH4bD CGCTGTCTCTGGTTACTCCATCAG
VH5D GAAAAAGCCCGGGGAGTCTCTGAA
VH6D CCTGTGCCATCTCCGGGGACAGTG
JHD ACCTGAGGAGACGGTGACCAGGGT
VH1-FR2 CTGGGTGCGACAGGCCCCTGGACAA
VH2-FR2 TGGATCCGTCAGCCCCCAGGGAAGG
VH3-FR2 GGTCCGCCAGGCTCCAGGGAA
VH4-FR2 TGGATCCGCCAGCCCCCAGGGAAGG
VH5-FR2 GGGTGCGCCAGATGCCCGGGAAAGG
VH6-FR2 TGGATCAGGCAGTCCCCATCGAGAG
VH7-FR2 TTGGGTGCGACAGGCCCCTGGACAA
JHC CTTACCTGAGGAGACGGTGACC
BCL1/IGH BCL1-P2 GAAGGACTTGTGGGTTGC
BCL1-P4 GCTGCTGTACACATCGGT
JH3 ACCTGAGGAGACGGTGACC
BCL2-MBR/IGH MBR2 CAGCCTTGAAACATTGATGG
JH3 ACCTGAGGAGACGGTGACC
MBR3 TATGGTGGTTTGACCTTTAG
JH4 ACCAGGGTCCCTTGGCCCCA
2.3.3 MRD Detection by 1. Forward and reverse allele-specific oligonucleotide (ASO) pri-
ddPCR mers and probes for individual IGH regions and for BCL1/
IGH translocations. Consensus forward and reverse primers
and probes for BCL2-MBR/IGH. See Table 2 and Fig. 6.
2. 2 ddPCR Supermix for Probes (No dUTP) (cat. 186-3024),
DG8 cartridges (cat. 186-4008), DG8 gaskets (cat.
186-3009), ddPCR droplet generation oil for probes (cat.
186-3005), pierceable foil heat seals (cat. 181-4040), ddPCR
droplet reader oil (cat. 186-3004) (Bio-Rad Laboratories, Her-
cules, CA, USA).
Table 2
Primers and probes sequences used in MRD based ddPCR
Target Oligos type NAME 50 -SEQUENCE-30

IGH PRIMER forward Pt code-F Pt specific sequence derived from CDR2 region
PRIMER reverse Pt code-R Pt specific sequence derived from CDR3 region
LVH1 GCACAGCCTACATGGAGCTGAGCAG
MVH2 ACCACCTGGTTTTTGGAGGTGTCCTT
LVH3 TCCTCGGCTCTCAGGC
MVH3TER CTCTGGAGATGGTGAATCGGC
PROBES MVH4 TGTCTGCAGCGGTCACAGA
MVH4TER TGTCCGCGGCGGTCACAGA
LVH5 CTTCAGGCTGCTCCACTGCAGGTAG
LVH1bis TACATGGAGCTGAGCAGCCTGA
LVH4 CAACCCCTCCCTCAAAGAGTC
VH3DD3 GATTCACCATCTCCAGAGACA
DVH4 TGTTTGCGGCGGTCACAGA
VH3MG CGGCCCCTCACGGAGTCT
PRIMER forward Pt code-F Pt specific sequence derived from “N” region
BCL1/IGH PRIMER consensus JH3 ACCTGAGGAGACGGTGACC
PROBES JH1-4-5 ACCCTGGTCACCGTCTCCTCAGGTG
JHDD1 ACGTCTGGGGCAAAGGGACCACGG
JHG2 CGATCTCTGGGGCCGTGGCAC
BCL2-MBR/ PRIMER forward Mbr2/Q CTATGGTGGTTTGACCTTTAGAG
IGH PRIMER reverse JH32-short CCTGAGGAGACGGTGACC
PROBE BCL2 CTGTTTCAACACAGACCCACCCAGAG
3. 96-well PCR plates and optical adhesive films or 0.2 mL strip

tubes with cups (used for mix preparation and collection before
droplets generation).
4. Scotch tape.
5. Hard-shell high-profile 96-well semiskirted PCR plates (cat.
0030 128.575) (Eppendorf, Hamburg, Germany).
6. gDNA (100 ng/μL) from a BCL2-MBR/IGH-positive cell
line (such as DOHH-2 or RL).
7. gDNA (100 ng/μL) from buffy coat (BC), pooled from 5 to
10 healthy donors, to use as negative control in IGH and
BCL1/IGH ddPCR reactions.
8. gDNA (100 ng/μL) from a BCL2-MBR/IGH-negative sam-
ple as MCF-7 human breast cancer cell line, a chemotreated
patients or any subject previously assessed by PCR to be BCL2-
MBR/IGH negative, since this translocation can be also pres-
ent in healthy donors without lymphoma [37].
3 Methods
Methods described below provide a technical support for clonal-

specific tumor marker identification and ddPCR-based MRD mon-
itoring in MM, MCL and FL. Figure 1 shows the schematic experi-
mental workflow, for each disease, summarizing the steps from
sample acquisition through patient’s unique tumor marker deter-
mination to MRD quantification by ddPCR.
3.1 Bone Marrow 1. For sample collection, the use of sodium citrate or EDTA is
and Peripheral Blood preferred (see Note 2) over Heparin.
Samples Processing 2. Resuspend blood samples in erythrocyte lysis buffer (NH4Cl)
(dilute BM 1:4 and PB 1:2). Leave blood tubes for 15 min at
room temperature (lying flat in the dark) then centrifuge for
10 min at 450 g at room temperature. Discard the superna-
tant wash in NH4Cl and centrifuge for 10 min at 450 g at
room temperature. Remove supernatant, resuspend cell pellets
in PBS or 0.9% NaCl count and dispense 5–10 106 cells in
1.5 mL tubes; centrifuge for 1 min at 13000 g and discard
supernatant. Cells can now be stored indefinitely, as dried
pellets, at 80 C for further gDNA extraction.
3. Perform gDNA extraction with a commonly used method
or kit.
4. Estimate the gDNA quality and concentration before experi-
mental use (see Note 3).
5. Prepare 200–300 μL of 100 ng/μL stock solution and store at
20 C (see Note 4).
3.2 Tumor-Specific Be aware that samples containing <5% of tumor cells present a
Molecular Marker lower rate of success in marker identification (especially for IGH
Identification sequencing) by Sanger sequencing. Therefore, diagnostic samples
containing a high percentage of tumor cells (>5%) are required.
Multiple strategies to obtain the tumor-specific marker are
currently available [15–17].
Here we present a PCR based approach (Fig. 2a–c), established
as previously published [9, 11, 36], to identify and to direct
sequencing:
1. IGH rearrangements in MM and MCL.
2. BCL1/IGH gene translocation, in MCL (if IGH rearrange-
ment is not successfully identified).
3. BCL2-MBR/IGH gene translocation, in FL.
3.2.1 IGH Clonal 1. The strategy used for IGH screening adopts three sets of
Rearrangement Screening consensus forward primers (VH-FS, VH-D and VH-FR2)
at Baseline derived from framework regions (FR1, FR2), each set is
236
Daniela Drandi et al.
Fig. 1 Experimental workflow for MRD monitoring by ddPCR. For each type of disease, steps from sample acquisition, tumor marker screening and MRD
quantification by ddPCR are described. Albeit MCL could harbor both IGH and BCL1/IGH tumor markers, the acceptance of only one target for MRD monitoring
depends on the protocol and clinical trial purpose (ref. 11, 16). Here we describe the approach for MRD monitoring on patient’s unique tumor marker where BCL1/
IGH is performed only if IGH rearrangement is not successfully identified. Samples in which all steps fail to identify a clonal target sequence (“no marker” patients),
can not be monitored for MRD studies. gDNA (genomic DNA), BCL1 (BCL1/IGH), BCL2 (BCL2/IGH), ASO P/P (Allele Specific Oligonucleotide Primers/ Probe),
multiple myeloma (MM), mantle cell lymphoma (MCL), and follicular lymphoma (FL)
A V D J
Rearranged IgH (VDJ)

VH-FR2 JHC
VH-FS VH-D JHD
L FR1 CDR1 FR2 CDR2 FR3 CDR3 FR4
B #11q13 BCL1 GENE (UPSTREAM MTC SITE) IGH GENE # 14q32 (JH SEGMENT)
BCL1-P4 JH3
BCL1-P2 JH3
#18q21 BCL2 GENE (EXON3) IGH GENE # 14q32 (JH SEGMENT)
C
MBR3 JH4
MBR2 JH3
Fig. 2 Schematic representation of PCR strategies used, for marker identification, for: (a) IGH rearrangements
screening in MM and MCL, (b) BCL1/IGH gene translocation, in MCL and (c) BCL2-MBR/IGH gene translocation
in FL, performed on diagnostic sample. IGH rearrangement derives from the combination of V-(D)-J segments
that code for the variable domains of the IGH molecules. In both BCL1/IGH and BCL2-MBR/IGH translocations
the IGH locus at 14q32.3 is involved and the BCL1 gene from 11q13 and BCL2 region from 18q21 are
translocated, respectively, to the IGH locus. IGH gene regions: L (leader), F (framework), CD (complementary
determining)
combined separately with the appropriate JH antisense consen-

sus primer (JHD or JHC; Fig. 2a). We suggest performing a
first PCR screening by individually combining each of the
VH-FS family primers with the reverse JHD primer. In case
of no amplification, switch to the VH-D and then to VH-FR2
set of primers combined one-at-a-time with the JHD or JHC
primers, respectively (Table 1) [10, 16, 35, 36].
2. In the Pre-PCR zone, prepare a 45 μL PCR mix solution in
0.2 mL tubes. Figure 3 shows an example of how to prepare the
PCR reaction mix for the IGH screening with VH-FS primers.
3. In the DNA zone, add 5 μL of gDNA (100 ng/μL) in each of
the patient and control tubes. Before adding the gDNA to the
mix, ensure that the DNA is thoroughly mixed. Vortex, spin
down, and then pipet the gDNA sample up and down a few
times.
IGH REARRANGEMENT
Reagents 1 Reaction (µl)
Sample gDNA code VH-F primer
dNTPs mix (2mM) 5 1 CTR+ gDNA or PL VH- specific
5x GoTaq Flexi Green Buffer 10 2 pt1 VH-FS1
MgCl2 (20mM) 5 3 pt1 VH-FS2
GoTaq Flexi 5U/µl (Promega) 0,25 4 pt1 VH-FS3
VH-FS (10pmol/µl) 1 5 pt1 VH-FS4a
JHD (10pmol/µl) 1 6 pt1 VH-FS4b
H2O 22,75 7 pt1 VH-F5
TOT mix 45 8 pt1 VH-F6
9 CTR- H2O all VH-F primers
gDNA 5
TOT 50
Initial denaturation: 94°C for 1 minute

Denaturation: 94°C for 30 seconds
Annealing: 62°C for 30 seconds repeated for 33 cycles
Extension: 72°C for 30 seconds
Final extension: 72°C for 10 minutes
Fig. 3 PCR reaction mix and setup for the IGH screening with VH-FS primers. The IGH strategy adopts three
sets of consensus forward primers (VH-FS, VH-D, and VH-FR2) each set combined separately with the
appropriate JH antisense consensus primer (JHD or JHC). The PCR reaction mix here described combines
each of the VH-FS family primers with the reverse JHD primer. In case of no amplification or no monoclonal
IGH gene identification, the VH-D and then the VH-FR2 set of primers are combined one-at-a-time with the
JHD or JHC primers, respectively, see Table 1 for primers sequence. (VH-FS (VH-FS forward primer), gDNA
(genomic DNA), CTR+ (positive control), PL (plasmid), CTR- (negative control), pt (patient)
4. Perform PCR with thermal cycling conditions and control

DNA’s as shown in Fig. 3.
5. Analyze the amplification product by running 10 μL of PCR
product on a 2% agarose gel, together with a DNA molecular
weight marker (see Note 5).
6. Visualize with UV light and capture the image by a digital
photo camera, in order to have documentation of the experi-
ment result (see Note 6). The IGH product size must be
between 300 and 500 bp.
7. In case of lack of amplification, repeat the PCR using a new set
of primers (i.e., VH-D and VH-FR2; see Note 7). If no ampli-
fication is observed with all three sets of primers, no MRD
analysis based on this molecular marker can be performed. In
case of MCL sample, move to Subheading 3.2.3 for BCL1/
IGH screening. In case of MM, the patient is marked as “IGH
negative: no marker available for MRD analysis”.
8. If the PCR product appears on the gel as a unique band, the
whole volume of amplified gDNA can be directly purified by
“QIAquick PCR purification kit,” following the manufac-
turer’s instructions. Otherwise, use the “QIAquick gel extrac-
tion kit.” In this case, DNA containing the proper amplification
band must by sliced from the gel, weighed on a scale, put in a
1.5 mL microcentrifuge tube and stored at 4 C until the

extraction. (It is recommended to store no longer than for
1 day).
9. Direct Sanger sequencing from both strands, with the same
primers used in the successful PCR reaction described above
can now be performed. To check the validity of the sequence,
visualize the electropherogram by “Chromas” or other known
software (see Note 8).
10. In case of identification of a monoclonal sequence, proceed
with sequence analysis using the IMGT/V-QUEST database
(www.imgt.org). IMGT allows you to verify that your sequence
is an IGH and to display the gene regions where primers and
probe must be depicted (Subheading 3.3.1).
11. If the sequence is polyclonal, proceed with IGH cloning (Sub-
heading 3.2.2; see Fig. 1). This circumstance is more frequent
in MM samples, mostly due to their high rate of somatic
hypermutation, which may significantly reduce primer
efficiency.
3.2.2 Cloning of Patient- To be used when direct Sanger sequencing does not allow identifi-
Specific Rearrangement cation of a clonal sequence. For MCL patients, cloning should be
performed only in cases where both IGH and BCL1/IGH have
been screened and a polyclonal sequence has been identified.
1. Set up the ligation reaction on ice using the TA-TOPO cloning
kit: Combine 1 μL of TA-TOPO Cloning vector, 1 μL of Salt
solution, 4 μL of fresh patient-specific PCR product. Mix
together well. Incubate at room temperature for 30 min and,
if not immediately used for bacterial transformation, store at
20 C.
2. Thaw competent E. coli (stored 80 C) on ice, add 3 μL of
ligation product to competent cells, shake gently and incubate
for 30 min, on ice.
3. Heat-shock at 42 C for 30 s on a Thermomixer (without
shaking) and immediately put on ice for 2 min.
4. Add 250 μL of S.O.C Medium (supplied with the TA-TOPO
cloning kit), incubate for 60 min at 37 C in a thermomixer
under vigorous shaking (800–1000 rpm).
5. Spread 50–70 μL of transformed cells on agar petri dish, previ-
ously prepared with “ImMedia Amp blue” (prepared following
the vendor’s instructions). Air-dry for few minutes. Incubate at
37 C overnight (see Note 9).
6. Grow up the transformed cells (10–15 white colonies) into
1.5 mL of “ImMedia Amp Liquid” incubating overnight at
37 C under vigorous shaking (800–1000 rpm) on a thermo-
mixer. The next day, perform a PCR amplification using 2 μL of
the overnight grown transformed cells, adjusting the volume of

PCR mix with 25.75 μL of water, instead of 22.75 μL. In order
to verify the presence of the cloned rearrangement the same
primers, and the same conditions previously used at the screen-
ing, must be used. Run on a 2% gel and verify which plasmids
contain product with the desired molecular weight band.
7. Perform plasmid DNA isolation and purification (Wizard Plus
SV Miniprep DNA purification system, Promega, Madison,
WI) from those plasmids showing similar amplification size.
8. Check the sequence of at least ten different plasmids. Sequenc-
ing only one strand from the forward side is usually enough for
good sequence comparison. If an identical sequence shows up
in at least three different plasmids, this sequence can be consid-
ered to represent the tumor clone [12]. Then, proceed with
sequence analyses by IMGT (for IGH) or Blastn (for BCL1/
IGH) and the design of clone-specific primers and probe (see
Subheading 3.3.1).
3.2.3 BCL1/IGH MCL patients not successfully screened for an IGH rearrangement
Translocation Screening at are then screened for the BCL1/IGH translocation, by a semi-
Diagnosis nested PCR approach (see Fig. 2b). This consists of a first PCR
amplification using a sense 50 primer (BCL1-P2) located in the
major translocation cluster (MTC) upstream of the BCL1 gene,
followed by a second PCR reaction performed with a sense 50
primer (BCL1-P4) located 43 bp downstream from the BCL1-P2
primer. Both amplifications employ an antisense 30 joining region
(JH3) consensus primer [11].
1. In the Pre-PCR zone, prepare a 45 μL mix solution in 0.2 mL
tubes, as shown in Fig. 4, using the 1st amplification primer
pair specified (BCL1-P2/JH3).
2. In the DNA zone, add 5 μL of gDNA (100 ng/μL). Before
adding the gDNA to the mix, ensure that the DNA is thor-
oughly mixed. Vortex, spin down, and then pipet the gDNA
sample a few times.
3. Run 1st round PCR under the conditions shown in Fig. 4.
4. Set up the 2nd round PCR: in the Pre-PCR zone, prepare a
48 μL mix solution as shown in Fig. 4, (adjusting the mix with
25.75 μL of water, instead of 22.75 μL), using the 2nd ampli-
fication primer pair (BCL1-P4/JH3).
5. In the post PCR-zone, add 2 μL of 1st amplification PCR
product.
6. Run the 2nd round PCR at the thermal cycling conditions
shown in Fig. 4.
BCL1/IGH
Reagents µ l)
1st round PCR (µ 2nd round PCR (µl)
Sample gDNA code
dNTPs mix (2mM) 5 5 1 CTR+ JVM-2 (10-2)
5x GoTaq Flexi Green Buffer 10 10 2 Pt1
MgCl2 (15mM) 5 5 … …
GoTaq Flexi 5U/ µl (Promega) 0.25 0.25 … Ptn
BCL1-MTC (10pmol/µl) 1 1 n CTR- H2O
JH (10pmol/µl) 1 1
H2O 22.75 25.75
TOT mix 45 48
gDNA 5 2
TOT 50 50
Primers BCL1-P2/JH3 BCL1-P4 /JH3

Annealing: 58°C for 30 seconds repeated for 33 cycles (1st round PCR)
Extension: 72°C for 30 seconds repeated for 30 cycles (2nd round PCR)
Fig. 4 BCL1/IGH rearrengement reaction and setup. BCL1-MTC (forward primer, BCL1-P2 or BCL1-P4), JH:
reverse primer (JH3 or JH4), gDNA (genomic DNA), CTR+ (positive control), CTR- (negative control),
pt. (patient), JVM-2 (BCL1/IGH positive cell line)
7. Follow the steps 5, 6, 8, and 9 described in Subheading 3.2.1.

BCL1/IGH PCR product size should be in the size range
between 100 and 300 bp (see Note 7).
8. If the sequence electropherogram is monoclonal, proceed with
sequence analysis using Blastn (www.blast.ncbi.nlm.nih.gov)
to verify the BCL1/IGH translocation sequence and proceed
with primers and probe design as described in Subheading
3.3.1, otherwise proceed with cloning as above (see Subheading
3.2.2).
3.2.4 BCL2-MBR/IGH A number of variant translocations involving the BCL2 gene have
Translocation Screening at been described; however only MBR translocations (Major Break-
Diagnosis point Region), and to a lesser extent mcr (minor cluster region)
translocations, have been extensively exploited for MRD analysis
[6, 9, 38]. Thus, only a subset of FL patients, ranging from 50% to
65%, can currently be assessed using the BCL2/IGH nested PCR
approach [6, 39, 40]. The procedure here described is focused only
on the most common BCL2-MBR/IGH translocation (Fig. 2c).
However, if no clonal signal is observed for the BCL2-MBR/IGH
translocation, we suggest trying with other primers sets for minor
breakpoint region detection [38].
1. In the Pre-PCR zone, prepare a 45 μL mix solution, in 0.2 mL

tubes, as shown in Fig. 5, using the BCL2-MBR2/JH3
primer pair.
2. In the DNA zone, add 5 μL of gDNA (100 ng/μL). Before
adding the gDNA to the mix, ensure that the DNA is thor-
oughly mixed. Vortex, spin down, and then pipet the gDNA
sample few times.
3. Follow 1st round PCR thermal cycler conditions as shown in
Fig. 5 (27 cycles).
4. In the Pre-PCR zone, set up the 2nd round PCR: prepare a
45 μL mix solution, as shown in Fig. 5, using the BCL2-
MBR3/JH4 primer pair.
5. In the post PCR-zone, add 5 μL of 1st amplification product
and run the 2nd round PCR (Fig. 5).
6. Follow the steps 5, 6, 8, and 9 as described in Subheading
3.2.1. BCL2-MBR/IGH PCR product size should range
between 150 and 400 bp (see Note7).
7. Proceed with Blastn (www.blast.ncbi.nlm.nih.gov) to verify the
BCL2-MBR/IGH translocation sequence.
3.3 MRD Monitoring ddPCR monitoring is performed based on the nucleotide sequence
by ddPCR of the clonal tumor-specific marker. Here we propose one strategy
for marker identification albeit other approaches, as gene scanning
3.3.1 Clone-Specific
or next generation sequencing, are now available [16, 17].
Primers and Probe Design
The junctional “N” regions, located in IGH rearrangements
(Fig. 6a), as well as in the BCL1/IGH (Fig. 6b) and BCL2-MBR/
IGH rearrangements (Fig. 6c), represent a “fingerprint” of the
lymphoma cell, and therefore their correct characterization is cru-
cial for tumor-specific primers and probe design and for patient-
specific MRD monitoring by ddPCR.
Particularly for the IGH and BCL1/IGH gene rearrenge-
ments, based on the sequences of interest, primers and probe
should be designed following the widely known common recom-
mendations [12, 15] (i.e., optimal melting temperature
(Tm) between 57 and 62 C, primers length between 13 and
26 mer, % CG content between 30 and 70%, probes with a Tm
6–8 C higher than the primers etc.). Moreover, we suggest Pri-
merQuest and Oligo Analyzer 3.1 (www.eu.idtdna.com) or Pri-
mer3Plus (www.primer3plus.com) as helpful tools for assay
design and optimization.
Different primer and probe design strategies need to be per-
formed for each target:
1. The strategy for the IGH rearrangement-based MRD monitor-
ing, here described, is named “ASO primer” and consists in as
follows: based on the IGH sequence, tumor-specific primers
BCL2-MBR/IGH
Reagents 1 Reaction (ml)
dNTPs mix (2mM) 5 Sample gDNA code

5x GoTaq Flexi Green Buffer 10 1 CTR+ DOHH2 (10-2)
MgCl2 (20mM) 5 2 Pt1
GoTaq Flexi 5U/ ml (Promega) 0.25 … …
BCL2-MBR (20pmol/ml) 1 … Ptn
JH (20pmol/ml) 1 n CTR- H2O
H2O 22.75
TOT mix 45
1st amplification BCL2-MBR2/JH3
gDNA (1st PCR) or 1st PCR product (2ndPCR) 5 2nd amplification BCL2-MBR3/JH4
TOT 50
1st PCR reaction

2nd PCR reaction

Fig. 5 BCL2-MBR/IGH rearrengement reaction and setup. BCL2-MBR (forward primer, BCL2-MBR3 or BCL2-
MBR4), JH (reverse primer JH3 or JH4), gDNA (genomic DNA), CTR+ (positive control), CTR- (negative control),
pt (patient), DOHH2 (BCL2-MBR/IGH positive cell line)
are positioned on the most hypervariable areas located in the

IGH-V second and third complementary-determining region
(CDR2-CDR3), while the consensus probe is designed on the
most constant FR3 region (Fig. 6a) and can be employed for a
large proportion of patients [2, 7, 8, 12, 41] (see Note 10). In
our ongoing MRD studies, we routinely and successfully use a
panel of consensus probes listed in Table 2 that are suitable for
a high proportion of patients. However, in cases with frequent
somatic mutations in the IGH gene rearrangements, a
completely sequence-based design of primers and probe is
recommended.
2. For BCL1/IGH primer and probe design (Fig. 6b): a forward
patient-specific ASO primer, depicted within the junctional
“N” insertion or overlapping the breakpoint in the BCL1
sequence and should be used in combination with the same
JH primer used in the seminested PCR for IGH screening
A V N D N J
Rearranged IgH (VDJ)
ASO 5’ probe ASO 3’
L FR1 CDR1 FR2 CDR2 FR3 CDR3 FR4
ASO 5’ probe JH3
B N
#11q13 BCL1 GENE (UPSTREAM MTC SITE) IGH GENE # 14q32 (JH SEGMENT)
CONSENSUS 5’ MBR2/Q probe JH32-SHORT CONSENSUS 3’
C N
#18q21 BCL2 GENE (EXON3) IGH GENE # 14q32 (JH SEGMENT)
FAM
MGB OR BHQ1
Fig. 6 Schematic representation of strategies used for designing: (a) IGH, (b) BCL1/IGH, and (c) BCL2-MBR/IGH
primers and probes for MRD analysis by ddPCR. N (N region with target specific nucleotide insertion), IGH gene
regions: L (leader), F (framework), CD (complementary determining), all strategies use FAM or BHQ1 probes
(JH3). In order to reach high specificity, the probe is posi-

tioned in the JH region adjacent to the forward primer. Since
the translocation involves different JH genes (JH1 to JH6),
probes are not consensus for all patients but are represented by
a panel of probes (Table 2), depending on the JH sequence
rearranged. In this manner, the same probe can be suitable for
different patients sharing the same JH gene (see Note 11).
3. For BCL2-MBR/IGH translocation (Fig. 6c) the strategy uses
consensus primer-probe, previously described [6, 40], univer-
sally suitable for different FL patients, making MRD analysis in
this type of lymphoma much cheaper and faster compared to
ASO primers approach (see Note 12).
3.3.2 ddPCR Reaction Major concepts to be aware of, before planning a ddPCR
Setup experiment:
1. No standard curve generation is needed for ddPCR experiment
setup, and therefore no determination of percentage of tumor
cell, by flow cytometry, is required. Tumor burden quantifica-

tion is calculated as number of copies of target in 1 105 cells
(corresponding approximately to 500 ng of gDNA).
2. As for any kind of PCR experiments, a positive control is
mandatory (see Note 13). The majority of BM or PB diagnostic
samples for MCL contain less than 50% of tumor cells infiltra-
tion. However, in order not to waste precious diagnostic
gDNA and to avoid system overloading, in case of highly
infiltrated samples we suggest loading a 101 or even a 102
diagnostic gDNA dilution point as positive control.
3. Since nonspecific amplification can be observed, negative con-
trols are mandatory. Nonspecific amplification (buffy coat)
should be run in six replicates and no template control
(NTC) in triplicates, for each specific marker quantification.
4. As the load of less template will result in loss of sensitivity,
particular attention must be used in gDNA quantification of
follow-up samples. So far, no consensus has been reached on
reference gene usage. Thus, in the context of the Italian MRD
lymphoma network, supported by the Italian Lymphoma
Foundation (FIL), an ongoing study is currently evaluating
several reference genes to establish which is the more stable
and reliable for gDNA quality/quantity assay. In the meantime,
we suggest to test, especially for follow-up samples, a reference
gene (i.e., albumin or RNAseP) on 100 ng of gDNA, in only
one replicate, preferably in the same ddPCR reaction as the
target gene.
(a) In the Pre-PCR zone, prepare the 20 primer and probe
mix and the ddPCR reaction mix, as shown in Fig. 7 (see
Note 14).
(b) Dispense the reaction mix in one well, based on the num-
ber of technical replicates needed, as described in Fig. 7
(i.e., for 1 replicate 16.5 μL and for 3 replicates 49.5 μL of
mix).
(c) In the DNA zone, load the amount of gDNA (100 ng/μ
L) depending on the number of planned replicates per
well, as indicated in Fig. 7 (i.e., for 1 replicate ¼ 16.5 μL
of mix, add 5.5 μL of gDNA; for 3 replicates ¼ 49.5 μL
(16.5 3) of mix, add 16.5 μL (5.5 3) of gDNA). From
this well, 20 μL of mix will be taken for each replicate
droplet generation. To ensure that the gDNA is thor-
oughly mixed: vortex, spin down, and then pipet the
sample a few times before adding the gDNA to the mix
(see Note 15).
(d) Seal carefully the plate or the strips with optical adhesive
film or caps, mix and spin down briefly.
Fig. 7 ddPCR reaction and setup
(e) Proceed with droplet generation, loading 20 μL of reac-

tion mix and 70 μL of droplet generation oil into the
proper DG8 cartridge wells, following manufacturer’s
instruction (see Note 16).
(f) Transfer 40 μL of generated droplets to the hard-shell
high-profile, 96-well semiskirted PCR plates and seal
immediately with a strip of scotch tape.
(g) After all samples are loaded, remove the tape and seal with
a pierceable foil on PX1 PCR plate sealer (Bio-Rad,
Laboratories, Hercules, CA, USA). Run on thermocycler
using the default Bio-Rad thermal cycling protocol
(95 C, 10 min; 94 C, x 30 s, and (optimized or gradient)
Tm C x 1 min, for 40 cycles; 98 C for 10 min) adjusting
the Tm according to the ASO primers for MM and MCL
and to 59 C for BCL2-MBR/IGH (Fig. 7).
(h) Load the post-PCR 96-wells plate in the QX100-QX200
droplet reader.Before run on the droplet reader, PCR
plate can also be stored at 4 C, no longer than 24 h.
For data interpretation, MRD tumor burden quantification is
expressed as total copies of target gene generated by performing
the merge of replicates by QuantaSoft (Bio-Rad Inc.) (see Notes 17

and 18). Examples of data quality and analysis, for each target, are
shown in Figs. 8, 9, and 10.
3.4 Final Remarks Recent advances in gene quantification strategies have led to
increased sensitivity and reduced variability and risk of contamina-
tion. ddPCR appears to be a feasible and attractive alternative
method for MRD assessment in lymphoproliferative disorders and
might complement or even substitute qPCR in routine clinical
laboratories. One milestone of MRD monitoring by qPCR has
been the multilaboratory standardization established by large coop-
erative efforts such as the Euro-MRD group (see Note 19). Cur-
rently ddPCR still has to undergo such a critical validation step.
However, considering the reduced labor intensiveness of the exper-
imental setup and the current robust structure of the interlabora-
tory quality control groups, we believe that ddPCR standardization
would be definitely easier and faster than qPCR. However, the real
advantages and predictive value of ddPCR still need to be further
investigated in the context of prospective clinical trials.
4 Notes
1. Keep PCR reagents in a room where template is not isolated or

stored. Use sterile filter tips for pipetting to minimize contam-
ination from aerosols. We suggest to verify the quality of tips
used for droplet loading or to use what recommended by
Bio-Rad in the instruction manual. We recommend having
separate pipettes and pipette tips for PCR setup, DNA loading,
droplet generation, and post-PCR. Moreover, we suggest to
regularly decontaminate PCR work areas (with 10% bleach) as
well as pipettes (under UV).
2. Based on personal experimental observations (unpublished
data), we suggest collecting samples in EDTA or sodium citrate
instead of Heparin.
3. For assessment of gDNA concentration, the NanoDrop
method is quite accurate. However, always check the absor-
bance spectrum and not only the concentration value.
260/230 and 260/280 ratios and the shape of the curve can
give you important information about your sample quality.
When measuring your gDNA on the NanoDrop, firstly ensure
that the DNA is thoroughly mixed, preferably using vortex and
pipetting several times. If amount of gDNA is below 100 ng/μ
L or 260/280 ratio is not 1.8–1.9, sample must be repurified.
4. We suggest to estimate gDNA concentration after 100 ng/μL
stock preparation and before any PCR reaction, especially if
sample is stored at 4 C.
Fig. 8 MRD monitoring of IGH rearrengement in an MCL patient: ddPCR was performed with patient-specific
ASO primers and LVH5 probe. In each pannel samples are reported from diagnostic to last follow-up. (a) 1D
5. Always include a DNA molecular weight marker such as ϕX174

DNA, cleaved with Hae III, ranging from 72 to 1353 bps, to
check the correct PCR product size. We suggest to use the 5
GoTaqFlexi Green Buffer which allows loading the amplifica-
tion reaction directly onto the gel, without need of loading
buffer and which is compatible with “QIAquick PCR purifica-
tion kit.” Run the gel at 120 V for 15–20 min, or until the two
color dye markers are well separated on the gel. One percent
ethidium bromide can be substituted by RealSafe Nucleic acid
staining solution at the concentration suggested by the manu-
facturer (Durviz, S.L., Spain).
6. Each qualitative PCR reaction must include a positive control
(to verify the quality of amplification) and a “no template
control” (NTC) (to check for contamination). If the positive
control fails or in case of PCR contamination, if the NTC is
positive, review all the procedures and change all reagents.
7. In case of lack of amplification, but good performance of PCR
confirmed by positive control sample, always verify that you
have good sample gDNA quantity/quality by performing a
PCR reaction for a housekeeping gene (e.g., Albumin,
RNase P, and p53).
8. For electropherogram analysis “Chromas” is ideal for basic
analysis of sequences and has no multiple sequences alignment
capability. To create a sequence database for routinely checking
for sequence contaminations, we suggest to gather all the
sequences by a multiple sequence alignment software such as
free CLC sequences viewer software (http://www.clcbio.com).
9. The efficiency of the ligation is determined by the comparison
of the growth of the experimental versus negative control
plates. If problems with cloning are encountered, verify that:
(a) you used fresh PCR product for ligation, since TOPO-TA
cloning requires fresh PCR; (b) competent cells are stored
properly at 80 C, since they are very sensitive to
Fig. 8 (continued) plot for patient coded “166” shows: bone marrow (BM), peripheral blood (PB) at diagnosis
(DIA), and post therapy follow-up (FU) at three time points (FLK1, FU2, and FU3). A single threshold has been
set in between the highest amplitude value of BC and the positive control (DIA). (b) Plot for positive events
(shorter bars) and total droplets (taller bars) for each sample analysed. Of note one event occurs in one of the
BC replicates and it is omitted from the analysis since it shows to be a nonspecific signal (overly high
amplitude value 11.000, data not shown in plot A since outside the amplitude range for graphical). (c)
Representation of concentration in terms of copies/μL
Fig. 9 MRD monitoring of BCL1/IGH translocation in an MCL patient: ddPCR performed with a patient-specific
ASO forward primer, JH1-4-5 probe and JH3 reverse primer. (a) 1D plot for patient samples at diagnosis
temperature; (c) thermomixer temperature for heat-shock pro-

cedure is correct; (d) SOC medium has not been stored for
too long.
10. Different groups use an “ASO probe” approach, derived from
the Euro-MRD guidelines for ALL, a neoplasia that is not
affected by somatic hypermutation of IGH sequences. The
“ASO primer” approach has shown to be a more powerful
strategy in the context of B-cell lymphoproliferative disorders,
characterized by a variable mutational load of IGH rearrange-
ment [40]. Moreover, “ASO primer” strategy has the advan-
tage to be much cheaper since no patient-specific probes are
needed.
11. Within the Euro-MRD network, the Lymphoma MRD group
recently planned to compare the specificity of different strategies
for BCL1/IGH primers and probe design, for MRD monitor-
ing (24th Euro-MRD meeting, Zurich 6–7 November 2015).
12. Clonal BCL2-MBR/IGH translocation is quantified by
ddPCR with a consensus primer-probe system (Table 2) that
does not require information about the sequence. However,
we recommend to always check the BCL2-MBR/IGH
sequence at diagnosis, in order to confirm the patient’s specific
clone and for its further monitoring in follow-up samples, as
well as to monitor for possible contaminations.
13. With respect to qPCR, ddPCR has the advantage of being an
absolute quantification method. However, although a standard
curve is not needed, in each experiment a positive control is
required. For IGH rearrangements and BCL1/IGH based
ddPCR, each patient requires his own baseline (or positive)
sample as positive control, while for BCL2/IGH based ddPCR
reactions gDNA from DOHH2 cell line is used as positive
control. The annealing temperature for each ASO primer set
can be different. For ASO primers test, use a thermal gradient
(2 C from the calculated primers Tm) to optimize ddPCR
results. If the positive control shows positive droplets at greater
than one amplitude level (graphically represented by two
clouds in 2D droplet plots or 2 bands in 1D droplet plots),
the annealing temperature can be increased in order to improve
the specificity; alternatively new ASO primers must be designed
until a single positive cluster is achieved (and the concentration
Fig. 9 (continued) (BASELINE) and at three follow-up time points. A single threshold has been set in between
the highest amplitude value of BC and the positive control (b) Positive events (shorter bars) and total droplets
(taller bars) for each sample analysed (c) Representation of concentration in terms of copies/μL
Fig. 10 MRD monitoring of BCL2-MBR/IGH translocation in follow-up samples from six FL patients: ddPCR
performed with consensus primers and probe (see Table 2). (a) 1D plot for diluted DOHH2 positive control cell
is consistent over a range of temperatures) even while the

separation between negative and positive clusters may vary.
14. We observed that for all assays, final primer and probe concen-
trations of 500 nM and 200 nM, respectively, work properly in
the ddPCR reaction (for 25 μL of 20 probe primer mix: 1 μL
probe (100 μM), 2.5 μL each primers (100 μM), 19 μL H2O).
For ddPCR analysis, MGB or BHQ1 probes are required.
TAMRA probes must be avoided since they lead to high back-
ground and noisy signals. As shown in Fig. 7, we normally
prepare the ddPCR mix for a volume increased by 10% to be
sure to have enough reaction mix volume for each replicate,
and to not risk introducing air bubbles into the DG8 cartridges
when loading the samples.
15. A total of 500 ng of gDNA is loaded in each replicate. Be aware
that, depending on the DNA extraction protocol, gDNA can
be more viscous and this characteristic can affect droplet for-
mation. For this reason, in case of sticky DNA, it is essential to
use restriction enzymes to decrease the size of the input DNA.
We observed that 2 U/μL of restriction enzyme (HINFI)
directly added into the reaction mix are sufficient for a good
ddPCR performance and proper droplet generation. However,
before using the enzyme, verify that primers and probe
sequences will not be damaged. To prepare a ddPCR reaction
mix containing the enzyme, add 1.1 μL of 2 U/μL of enzyme
to the 22 μL of ddPCR reaction mix, adjusting properly with
water.
16. We suggest: (a) to load the 20 μL of ddPCR mix in to the DG8
cartridge using a multichannel pipet with filtered tips; (b) to
paying attention when removing the DG8 gasket and always
remove it from the NTC, or BC well position, to the positive
control sample; (c) to load the ddPCR reaction mix into the
cartridge always before the oil; (d) to transfer the 40 μL of
droplets/well to the hard-shell, high-profile, 96-well semi-
skirted PCR plates using a multichannel pipet with no filter
tips and immediately seal the wells with a scotch tape; (e) to
Fig. 10 (continued) line (first three replicates), six follow-up samples from FL patients (PT) (three replicates
each) and negative controls (MCF7 and NTC). Of note, three different gDNA samples form MCF7 cell line, (two
replicates each) had been used as negative control. Variabilty in negative droplets amplitude depends on
sample type and purity and only in one patient (PT13, wells E04-E06) it is between technical replicates.
Moreover, this assay uses consensus primers and probe for different patients and different amplitude could
represents differencies in assay efficiency. A single threshold has been set in between the highest value of
amplitude of the background and the positive control. (b) Positive events (shorter bars) and total droplets (taller
bars) for each sample analysed (c) Representation of concentration in terms of copies/μL
load the technical replicates in different DG8 cartridges to

avoid losing a full sample reaction, due to technical error that
can happen during pipetting or the droplet generation proce-
dure; (f) to remove carefully, any bubble created into the DG8
cartridge “sample” well during sample loading.
17. If you find a contamination (positive events in the range of the
positive control amplitude) in your NTC sample: (a) replace all
reagents and thoroughly clean PCR preparation areas;
(b) check whether the probe is degraded, since free dye can
give rise to signal and/or high background.
18. Only replicates with more than 9000 droplets must be consid-
ered for the analysis. We define as MRD positive by ddPCR
those samples that have all positive replicates and negative
those that have all negative replicates. Based on the higher
variability observed at low target concentrations and applying
the Euro-MRD guidelines, cases showing alternatively positive
or negative replicates must be judged with caution, especially if
less than totally three positive events are detected in the three
technical replicates. In such a case, ddPCR of the sample
should be performed again on more replicates (at least six) to
see if the result appears consistent. One critical limitation of
qPCR is its drawback to providing a reliable target quantifica-
tion for a proportion of samples with tumor burden between
the sensitivity and the quantitative range of the method. Sam-
ples falling in this window of unreliable quantification, by
qPCR, are currently defined as “positive nonquantifiable”
(qPNQ). ddPCR showed to be able to partially recover these
qPNQ samples [34]. However, specific guidelines for data
interpretation still need to be established and the potential
advantages and predictive value of ddPCR results, expecially
for these cases, need to be further validated.
19. The Euro-MRD group, (http://www.euromrd.org) is com-
posed by international MRD-PCR laboratories, under the
umbrella of ESLHO consortium (European Scientific founda-
tion for Laboratory Hemato Oncology). Euro-MRD group is
responsible for developing of new strategies and techniques, as
well as for developing guidelines for the interpretation of
qPCR based MRD data and in the near future also for devel-
oping guidelines for ddPCR based MRD.
Acknowledgments
We gratefully acknowledge Elisa Genuardi, Barbara Mantoan, Mar-

tina Ferrante, Luigia Monitillo, Manuela Gambella, Daniela Bar-
bero, Irene Della Starza, Elena Ciabatti, Nadia Dani, and Marta
Varotto for their excellent technical support. Moreover, we are
grateful to the Italian Lymphoma Foundation (FIL) that is sup-
porting our ongoing research projects on MCL and FL.
References
1. Andersen NS, Pedersen LB, Laurell A et al action BMH4-CT98-3936. Leukemia
(2009) Pre-emptive treatment with rituximab 17:2257–2317
of molecular relapse after autologous stem cell 11. Rimokh R, Berger F, Delsol G et al (1994)
transplantation in mantle cell lymphoma. J Clin Detection of the chromosomal translocation t
Oncol 27:4365–4370 (11;14) by polymerase chain reaction in mantle
2. Ladetto M, Pagliano G, Ferrero S et al (2010) cell lymphomas. Blood 83(7):1871–1875
Major tumor shrinking and persistent molecu- 12. Ladetto M, Donovan JW, Harig S et al (2000)
lar remissions after consolidation with bortezo- Real-time polymerase chain reaction of immu-
mib, thalidomide, and dexamethasone in noglobulin rearrangements for quantitative
patients with autografted myeloma. J Clin evaluation of minimal residual disease in multi-
Oncol 28:2077–2084 ple myeloma. Biol Blood Marrow Transplant
3. Pott C, Hoster E, Delfau-Larue MH et al 6:241–253
(2010) Molecular remission is an independent 13. Brüggemann M, Droese J, Bolz I et al (2000)
predictor of clinical outcome in patients with Improved assessment of minimal residual dis-
mantle cell lymphoma after combined immu- ease in B cell malignancies using fluorogenic
nochemotherapy: a European MCL intergroup consensus probes for real-time quantitative
study. Blood 115:3215–3223 PCR. Leukemia 14:1419–1425
4. Pott C (2011) Minimal residual disease detec- 14. van der Velden VH, Cazzaniga G, Schrauder A
tion in mantle cell lymphoma: technical aspects et al (2007) Analysis of minimal residual disease
and clinical relevance. Semin Hematol by Ig/TCR gene rearrangements: guidelines
48:172–184 for interpretation of real-time quantitative
5. Korthals M, Sehnke N, Kronenwett R et al PCR data. Leukemia 21:604–611
(2012) The level of minimal residual disease 15. van der Velden VH, Hochhaus A, Cazzaniga G
in the bone marrow of patients with multiple et al (2003) Detection of minimal residual disease
myeloma before high-dose therapy and autolo- in hematologic malignancies by real-time quanti-
gous blood stem cell transplantation is an inde- tative PCR: principles, approaches, and labora-
pendent predictive parameter. Biol Blood tory aspects. Leukemia 17(6):1013–1034
Marrow Transplant 18:423–431 16. Pott C, Brüggemann M, Ritgen M et al (2013)
6. Ladetto M, Lobetti-Bodoni C, Mantoan B et al MRD detection in B-cell non-Hodgkin lym-
(2013) Persistence of minimal residual disease phomas using Ig gene rearrangements and
in bone marrow predicts outcome in follicular chromosomal translocations as targets for
lymphomas treated with a rituximab-intensive real-time quantitative PCR. Methods Mol
program. Blood 122:3759–3766 Biol 971:175–200
7. Ferrero S, Ladetto M, Drandi D et al (2015) 17. Ladetto M, Brüggemann M, Monitillo L et al
Long-term results of the GIMEMA VEL-03- (2014) Next-generation sequencing and real-
096 trial in MM patients receiving VTD con- time quantitative PCR for minimal residual dis-
solidation after ASCT: MRD kinetics’ impact ease detection in B-cell disorders. Leukemia
on survival. Leukemia 29(3):689–695 28:1299–1307
8. Ladetto M, Ferrero S, Drandi D et al (2016) 18. Martinez-Lopez J, Lahuerta JJ, Pepin F et al
Prospective molecular monitoring of minimal (2014) Prognostic value of deep sequencing
residual disease after non-myeloablative allo- method for minimal residual disease detection
grafting in newly diagnosed multiple myeloma. in multiple myeloma. Blood 123:3073–3079
Leukemia 30(5):1211–1214 19. van Dongen JJ, van der Velden VH, Brügge-
9. Gribben JG, Freedman A, Woo SD et al (1991) mann M et al (2015) Minimal residual disease
All advanced stage non-Hodgkin’s lymphomas diagnostics in acute lymphoblastic leukemia:
with a polymerase chain reaction amplifiable need for sensitive, fast, and standardized tech-
breakpoint of bcl-2 have residual cells contain- nologies. Blood 125(26):3996–4009
ing the bcl-2 rearrangement at evaluation and 20. Böttcher S, Ritgen M, Fischer K et al (2012)
after treatment. Blood 78(12):3275–3280 Minimal residual disease quantification is an
10. van Dongen JJ, Langerak AW, Bruggemann M independent predictor of progression-free and
et al (2003) Design and standardization of overall survival in chronic lymphocytic leuke-
PCR primers and protocols for detection of mia: a multivariate analysis from the rando-
clonal immunoglobulin and T-cell receptor mized GCLLSG CLL8 trial. J Clin Oncol
gene recombinations in suspect lymphoproli- 30:980–988
ferations: report of the BIOMED-2 concerted
21. Rawstron AC, Child JA, de Tute RM et al 33. Huggett JF, Cowen S, Foy CA (2015) Con-
(2013) Minimal residual disease assessed by siderations for digital PCR as an accurate
multiparameter flow cytometry in multiple molecular diagnostic tool. Clin Chem 61
myeloma: impact on outcome in the Medical (1):79–88
Research Council Myeloma IX Study. J Clin 34. Drandi D, Kubiczkova-Besse L, Ferrero S et al
Oncol 31:2540–2547 (2015) Minimal residual disease detection by
22. Paiva B, Gutiérrez NC, Rosiñol L et al (2012) droplet digital PCR in multiple myeloma, man-
High-risk cytogenetics and persistent minimal tle cell lymphoma, and follicular lymphoma. A
residual disease by multiparameter flow cyto- comparison with real-time PCR. J Mol Diagn
metry predict unsustained complete response 17(6):652–660
after autologous stem cell transplantation in 35. Deane M, McCarthy KP, Wiedemann LM et al
multiple myeloma. Blood 119:687–691 (1991) An improved method for detection of
23. Markey AL, Mohr S, Day PJ (2010) High- B-lymphoid clonality by polimerase chain reac-
throughput droplet PCR. Methods tion. Leukemia 5:726
50:277–281 36. Voena C, Ladetto M, Astolfi M et al (1997) A
24. Hindson BJ, Ness KD, Masquelier DA et al novel nested-PCR strategy for the detection of
(2011) High-throughput droplet digital PCR rearranged immunoglobulin heavy-chain genes
system for absolute quantitation of DNA copy in B cell tumors. Leukemia 11:1793
number. Anal Chem 83:8604–8610 37. Summers KE, Goff LK, Wilson AG et al (2001)
25. Pinheiro LB, Coleman VA, Hindson CM et al Frequency of the Bcl-2/IgH rearrangement in
(2012) Evaluation of a droplet digital polymer- normal individuals: implications for the moni-
ase chain reaction format for DNA copy num- toring of disease in patients with follicular lym-
ber quantification. Anal Chem 84:1003–1011 phoma. J Clin Oncol 19:420–424
26. Day E, Dear PH, McCaughan F (2013) Digital 38. Albinger-Hegyi A, Hochreutener B, Abdou
PCR strategies in the development and analysis MT et al (2002) High frequency of t(14;18)
of molecular biomarkers for personalized med- translocation breakpoints outside of major
icine. Methods 59(1):101–107 breakpoint and minor cluster regions in follic-
27. Didelot A, Kotsopoulos SK, Lupo A et al ular lymphomas improved polymerase chain
(2013) Multiplex picoliter-droplet digital reaction protocols for their detection. Am J
PCR for quantitative assessment of DNA integ- Pathol 160:823–832
rity in clinical samples. Clin Chem 59:815–823 39. Ladetto M, De Marco F, Benedetti F et al
28. Dingle TC, Sedlak RH, Cook L et al (2013) (2008) Prospective, multicenter randomized
Tolerance of droplet-digital PCR vs real-time GITMO/IIL trial comparing intensive
quantitative PCR to inhibitory substances. Clin (R-HDS) versus conventional (CHOP-R) che-
Chem 59:1670–1672 moimmunotherapy in high-risk follicular lym-
29. Hayden RT, Gu Z, Ingersoll J, Abdul-Ali D phoma at diagnosis: the superior disease
(2013) Comparison of droplet digital PCR to control of R-HDS does not translate into an
real-time PCR for quantitative detection of overall survival advantage. Blood 111
cytomegalovirus. J Clin Microbiol 51:540–546 (8):4004–4013
30. Hindson CM, Chevillet JR, Briggs HA et al 40. Ladetto M, Sametti S, Donovan JW et al
(2013) Absolute quantification by droplet dig- (2001) A validated real-time quantitative PCR
ital PCR versus analog real-time PCR. Nat approach shows a correlation between tumor
Methods 10:1003–1005 burden and successful ex vivo purging in follic-
ular lymphoma patients. Exp Hematol 29
31. Huggett JF, Whale A (2013) Digital PCR as a (2):183–193
novel technology and its potential implications
for molecular diagnostics. Clin Chem 41. Della Starza I, Cavalli M, Del Giudice I et al
59:1691–1693 (2014) Comparison of two real-time quantita-
tive polymerase chain reaction strategies for
32. Nixon G, Garson JA, Grant P et al (2014) minimal residual disease evaluation in lympho-
Comparative study of sensitivity, linearity, and proliferative disorders: correlation between
resistance to inhibition of digital and nondigital immunoglobulin gene mutation load and real-
polymerase chain reaction and loop mediated time quantitative polymerase chain reaction
isothermal amplification assays for quantifica- performance. Hematol Oncol 32(3):133–138
tion of human cytomegalovirus. Anal Chem
86:4387–4394
Chapter 15
Quantitation of JAK2 V617F Allele Burden by Using

the QuantStudio™ 3D Digital PCR System
Elena Kinz and Axel Muendlein
Abstract
The JAK2 V617F mutation is highly prevalent in patients with myeloproliferative neoplasms (MPN).
Furthermore, it has been shown that its allelic burden correlates with hematologic characteristics, drug
response, and clinical endpoints in MPN patients. Digital PCR is an emerging technology for sensitive
mutation detection and quantitation, based on dilution and high-grade partitioning of a sample. Here, we
describe the use of the nanofluidic chip-based QuantStudio™ 3D Digital PCR System for quantitation of
the JAK2 V617F mutation.
Key words Digital PCR, Nanofluidic chip, JAK2 V617F, Allele burden, Myeloproliferative neoplasms
1 Introduction
The detection of the acquired mutation JAK2 V617F in patients

with Philadelphia-negative myeloproliferative neoplasms (MPNs),
including polycythemia vera (PV), essential thrombocythemia
(ET), and primary myelofibrosis (PMF) in 2005 [1–4], has revolu-
tionized MPN diagnostics. The mutation occurs in more than 95%
of PV cases and about 50–60% of those with ET or PMF. Conse-
quently, presence of the JAK2 V617F mutation has been included
as a major diagnostic criterion for these diseases in the World
Health Organization (WHO) classification [5].
Subsequent studies indicated that the different MPN pheno-
types had different mutational loads [6, 7]: Patients with ET have
the lowest allele burden (<50%), those with PV and PMF interme-
diate levels, and those with post-PV myelofibrosis the highest bur-
den [8]. Therefore, at MPN diagnosis, measurement of the allele
burden may help to discriminate between different MPN pheno-
types: when an allele burden is greater than 50%, the likelihood of a
masked PV or myelofibrotic evolution exists [9]. JAK2 allele bur-
den may have prognostic significance as well, since it correlates with
clinical endpoints in MPN patients [10]: A high allele burden in
257
258 Elena Kinz and Axel Muendlein
patients with ET or PV has been associated with increased risk of

thrombotic events [11] and transformation to myelofibrosis
[12, 13]. On the other hand, in PMF patients, a low allele burden
(25%) has been correlated with poorer survival, probably due to
increased leukemic transformation [14] or systemic infections
[15]. Furthermore, JAK2 V617F quantitation has been used to
assess efficacy of treatment with significant reductions in allele
burden reported in PV patients following pegylated interferon
α-2a treatment [16, 17]. Therefore, quantitation of JAK2 V617F
allele burden may be used in stratification of patients with MPN
regarding prognosis as well as drug treatment.
Several techniques have been described to quantify JAK2
V617F allele burden; out of those, until recently, allele specific
real-time quantitative polymerase chain reaction (AS-qPCR) has
been proposed to be the most reliable and sensitive one [18, 19].
Recently, we showed a high correlation between a AS-qPCR kit
for quantitation of JAK2 allele burden, registered for in vitro diag-
nostic use, and chip-based digital PCR (dPCR) using the Quant-
Studio™ 3D Digital PCR System (Thermo Fisher Scientific,
Waltham, Massachusetts, USA) providing sensitivity to detect less
than 0.1% JAK2 V617F allelic burden [20]. Accuracy and repro-
ducibility of results obtained by AS-qPCR strongly depend on used
standards needed for quantitation of target molecules. In contrast,
chip-based dPCR allows assessment of the total copy number of
target molecules without the need for standard curves facilitating
standardization of quantitation. Particularly, chip-based digital
PCR demonstrates an appropriate technique for the detection of
mutations present at low allele frequencies because partitioning of
sample reduces wild-type background signals and therefore
enhances accuracy of detection. Therefore, chip-based dPCR
appears very suitable for measurement of JAK2 V617F allelic
burden.
The QuantStudio™ 3D Digital PCR System platform is com-
posed of a GeneAmp 9700 thermal cycler (including a chip adapter
kit), an automatic chip loader, and the QuantStudio™ 3D Instru-
ment. The QuantStudio™ 3D Digital PCR 20K Chip consists of
individual 20,000 partitions and each partition is a chamber for an
individual PCR reaction. Genomic DNA (gDNA) samples are
diluted down to a limiting quantity, such that most individual
PCR reactions contain either zero or one DNA molecule. Digital
PCR is an endpoint analysis based on a 50 -exonuclease assay using
fluorogenic TaqMan® probes targeting the mutation. The absolute
quantification of target is calculated by Poisson statistics, based on
the number of negative partitions, which contain no DNA
[21]. After mixing diluted gDNA sample with the premixed primer
and probe assay and a PCR master mix, the PCR reaction mix is
loaded onto the chip, which in turn is placed in the PCR machine to
run the reaction. Next, the chip is placed in the QuantStudio™ 3D
Quantitation of JAK2 V617F Allele Burden by Using the QuantStudioTM 3D Digital. . . 259
Instrument to read out fluorogenic signals. Data are analyzed using

QuantStudio™ 3D AnalysisSuite™ Software.
According to the specifications of the manufacturer as well as to
our observations, QuantStudio™ 3D Digital PCR System platform
is able to quantitate JAK2 V617F mutants at a prevalence as low as
0.1%, reaching sensitivity recommended for residual disease
monitoring [9].
Importantly, sensitivity can be further increased using addi-
tional chips, which data are aggregated by the software and ana-
lyzed as one chip. Therefore, chip-based digital PCR represents an
appropriate technique for detection and quantitation of the JAK2
V617F allelic burden.
2 Materials
2.1 Sample 1. Peripheral whole blood, purified blood granulocytes and bone
Collection and DNA marrow aspirates, respectively, is suitable as a source for geno-
Preparation mic DNA used in JAK2 V617F quantitation (see Note 1).
2. Kit for isolation of genomic DNA providing high DNA quality
e.g.: QIAamp DNA Blood Mini Kit (Qiagen, Hilden, Ger-
many), peqGOLD Blood DNA Mini kit (VWR, Vienna, Aus-
tria), or others (see Note 2).
3. Fluorometer (e.g. Qubit™, Thermo Fisher) or spectropho-
tometer (e.g. Nano-Drop™ Lite, VWR) to determine nucleic
acid concentration (see Note 3).
2.2 PCR Mix 1. Genomic DNA (see above). Store at 20 C.

2. QuantStudio™ 3D Digital PCR Mastermix (Thermo Fisher)
Store at 20 C
3. Twenty-fold premixed Primers and TaqMan®-probe mix
obtained by the TaqMan® Assays-by-Design service from
Thermo Fisher. Primer and probe sequences are as follows:
JAK2 V617F forward primer: 50 -AAGCTTTCTCACAAG-
CATTTGGTTT; JAK2 V617F reverse primer: 50 --
AGAAAGGCATTAGAAAGCCTGTAGTT; JAK2 V617-
probe (wild-type allele): VIC®-TCTCCACAGACACATAC-
MGB; and JAK2 F617-probe (mutant allele) 6-FAM™-T
CCACAGAAACATAC-MGB. Store at 20 C (see Note 4).
4. Nuclease-free water.
2.3 Chip-Loading, 1. QuantStudio™ 3D Digital PCR 20K Chip (v2) (Thermo

Automated Fisher) (see Note 5).
2. QuantStudio™ 3D Digital PCR Chip Case Lid.
3. QuantStudio™ 3D Digital PCR Chip Loader (Thermo
Fisher).
4. QuantStudio™ 3D Digital Sample Loading Blades (Thermo

Fisher).
5. Immersion Fluid (Thermo Fisher).
2.4 dPCR Run 1. QuantStudio™ 3D Digital PCR Chip Carriers (Thermo

Fisher).
2. QuantStudio™ 3D Digital PCR Thermal Pads (Thermo
Fisher).
3. Dual Flat Block GeneAmp® PCR System 9700 (Thermo
Fisher).
4. QuantStudio™ 3D Tilt Base.
2.5 dPCR Signal 1. QuantStudio™ 3D Instrument (Thermo Fisher).

Read Out and Data 2. QuantStudio™ 3D AnalysisSuite™ Software (Thermo Fisher).
Analysis
2.6 General 1. Isopropanol.

Consumables 2. Pipette and pipette tips.
and Other Equipment
3. Additional syringe tips (10 and 20 μL pipette tips without
filter).
4. Powder free nitrile gloves.
5. Low-lint wipes.
6. Microcentrifuge.
7. Vortexer.
3 Methods
3.1 gDNA Prepare gDNA from peripheral blood, purified granulocytes or

Preparation bone marrow aspirates (see Note 1) using a commercial available
DNA preparation kit (see Note 2) following instructions of the
manufacturer. Accurately assess DNA concentration using a fluo-
rometer or a spectrophotometer (see Note 3).
3.2 Copy Number The procedure of determining the number of target copies per
Calculation genome (genomic equivalents) is important for appropriate chip
load and quality. The DNA molecule of interest must be present in
a defined quantity, so that individual PCR reactions contain zero or
one target molecule. To calculate DNA concentration, it necessary
to estimate target copy number of genome:
1. Convert DNA concentration from ng/μL into copies/μL,
assuming 3.5 pg/copy for the JAK2 gene, or other gene,
present at two copies per diploid genome (see Note 6).
2. According to the recommendations of the manufacturer, the

concentration of the target sequence in the dPCR reaction may
range between 200 and 2000 copies/μL (corresponding to
3000–30,000 haploid genomic equivalents in the final reaction
on the 20K v2 chip; see Note 7). In case of quantitation of
JAK2 V617F allelic burden, a DNA concentration of 1000–-
2000 copies/μL in the PCR reaction shows good results (see
Note 8).
3. Positive and negative template as well as wild type controls
(if available) should be integrated into dPCR experiments.
One may use plasmid constructs, which have to been linearized
before use, to estimate sensitivity of dPCR application. Calcu-
lation of copy number must be adjusted to respective
plasmid size.
3.3 dPCR Setup Bring QuantStudio™ 3D Digital PCR Mastermix and 20-fold
premixed JAK2 TaqMan® assay to room temperature and prepare
nuclease free water and DNA samples. Set up dPCR reaction mix
for two chips per sample as shown in Table 1.
Depending on your estimated sample concentration, add
gDNA ( y) and water (x) to obtain a final concentration between
1000 and 2000 copies/μL. Place on ice until use and warm it to
room temperature
3.4 Chip Loading Proper loading of QuantStudio™ 3D Digital PCR 20K Chip
(v2) with PCR reaction mix is essential for an accurate quantitation
of target gene copy number. We strongly recommend using the
QuantStudio™ 3D Digital PCR Chip Loader for loading proce-
dure since we experienced that differences in handling can influence
the outcome of analysis, especially the precision of target gene
estimation in technical duplicates. The procedure of automated
Table 1
dPCR reaction mix for two chips per sample
Volume [μL]
Material Two chips
Master Mix (v2) 2 17.4
®
TaqMan Assay 20 1.74
Nuclease-free water a
x
gDNAa y
Total volume 34.8b
a
For a desired template concentration of 1000 copies/μL, use 12.2 μL gDNA with a concentration of 10 ng/μL and
4.3 μL water
b
Total volume is calculated upon an application volume of 14.5 μL per chip. The application volume might be increased if
the chip quality reviewed by software is not sufficient
chip loading is given in detail in the manual of the QuantStudio™

3D Digital PCR system (see Note 9). In brief, automated chip
loading procedure is as follows:
1. Plug in the chip loader. Prepare immersion fluid syringe by
gently pulling back the plunger to release air and then unscrew
the cap and replace it by a tip by pushing it into place carefully
(see Note 10).
2. Load the chip in the chip nest face according to the picture next
to the nest. Apply the loading blade by pressing the loading
blade lever with the one hand and pressing the loading blade
into the loader head with the other (see Note 11).
3. Peel away the red protective film from the back of the chip lid
(see Note 12) and place it with the sticky side up in the lid nest
while pressing the lid nest button (see Note 13).
4. Transfer 14.5 μL of the dPCR reaction mixture into the sample
loading port of the sample blade (see Note 14).
5. Press the black loading button on the top of the chip loader to
start chip loading.
6. After loading, slowly apply immersion fluid on the chip surface
without contact. Fifteen drops should be sufficient to cover the
whole surface (see Note 15).
7. To apply the chip lid on the chip, rotate the arm loader and
press it down for 15 s to confirm tight sealing. Each flash of the
status light illustrates 1 s intervals (see Note 16).
8. Press the lid nest button to release the lid and put back the arm
in the original position. Take care that the chip stays in the chip
nest and does not remain in the lid nest.
9. To fill the chip with immersion fluid through the chip port,
hold the chip in a 45 angle with the port in the top right
corner. A small air bubble (<2–3 mm in diameter), slightly
larger than the fill port, should remain. Rotate the chip to see
whether hidden bubbles have remained (see Note 17).
10. For sealing, pull back the top half of the label, remove the label
backing and press the label firmly over the fill port for 5 s to
confirm tight sealing (see Note 18). Place chips in a dark and
clean location until cycling. Cycling should be started within
2 h after preparation.
3.5 dPCR Thermal 1. Prepare chip adapter and thermal pads and make sure that they
Cycling are clean. Otherwise clean them with isopropanol sprayed on a
low-lint wipe.
2. Confirm that the tilt base is installed beneath thermal cycler.
3. Install chip adapters on both samples blocks, no matter how

many samples you are going to cycle, it is necessary to have
installed both adapters.
4. Check that you use the correct firmware version for the appro-
priate thermal cycler. (see user guide)
5. Program PCR method before starting. Following cycling con-
dition are used for JAK2 V617F dPCR thermocycling on the
Gene Amp 9700 PCR System: 96 C for 10 min, 39 cycles at
56 C for 2 min and at 98 C for 30 s, followed by a final
extension step at 60 C for 2 min (see Note 19).
6. Open the heated cover of the thermal cycler and place the chips
onto the sample blocks so that the fill ports are positioned
toward the front of the cycler. Confirm that chips are clean
and free from contaminants (see Note 20).
7. Lay clean QuantStudio™ 3D Digital PCR thermal pads on the
chip adapter accurately. Do not use other thermal pads.
8. Close and engage heated cover of the thermal cycler.
9. Select preprogrammed run and start cycler.
10. Unload cycler after cycling programme is finished (see
Note 21).
11. Inspect it for leaks, let it adapt to room temperature and clean it
with isopropanol before imaging. If you do not image it imme-
diately, place it in a clean and dark location (see Note 22).
3.6 Imaging The QuantStudio™ 3D Instrument processes imaging of chips

and Primary Analysis individually and creates a single experiment file (.eds) for every
on QuantStudio™ 3D chip analyzed, which is named after the ID number on the chip,
Instrument detected by the system.
1. Switch on the instrument and wait until the instrument
requests chip insertion (“Insert chip to continue”)
2. You may need to specify data destination (QuantStudio™ 3D
software cloud, personal network or USB stick) and well vol-
ume prior to analysis according to user guide.
3. Open the chip tray and insert the chip face up into the tray. The
Chip Barcode need to be orientated toward the front of the
instrument.
4. Close the tray, the detection will start automatically. You can
monitor the progress on the screen. If an error occurs, inspect
the chip for problems and refer to the troubleshooting guide if
no obvious problems are visible.
5. You may insert another chip after imaging is done and the
instrument displays “Analyzing data” (see Note 23).
6. If you want to reread chips afterward, make sure that you store
them in a dark place until you have analyzed the data with the
software (see Note 24).
7. The results are assessed in copies/μL of nucleic acid sequences
targeted by VIC® or FAM™ dye. Flags of different colors
illustrate a first quality assessment in primary analysis. These
flags allow direct estimation of chip quality after primary analy-
sis as green colors display that data meets quality thresholds,
yellow that analysis was not ideal and should be reviewed, and
red that analysis failed.
8. Further information about quality and results of detection can
be estimated in secondary analysis with cloud based QuantStu-
dio™ 3D analysis software.
4 Analysis
4.1 Chip Analysis QuantStudio™ 3D Digital PCR AnalysisSuite™ Software is a

Using QuantStudio™ cloud based software suite for analysis and administration of
3D Software dPCR experiment data. It calculates absolute and relative number
of target nucleic acid of sample and performs statistics.
1. Make sure that your computer has network connection.
2. Open a browser window and go to https://apps.thermofisher.
com/quantstudio3d/ (see Note 25). Sign into the software or
create a new account (see Note 26).
3. You can create and edit projects in the Projects Home screen.
Create a project and fill in a distinct project name. Import chip
data (.eds files) from your computer or directly from the instru-
ment (see user guide). Experiment data from up to 100 chips
can be included in one project.
4. Define chips by entering sample name, assays used and fill in
dilution factor of chip (see Note 27).
Quality of the chip can be reviewed in “Review Data” tab
where all chips are displayed in the left corner of the screen.
Chip of interest can be selected by double-click. Choose “col-
our by quality” (at Action Box) and threshold—a measure of
individual well quality will be displayed from low (red) to high
(green) quality on a continuous scale from 0 to 1. The default
quality threshold is 0.5. The quality flag allows you to directly
determine which chip to review. If a user has modified the
analysis results originally generated by the instrument software,
the flag appears faded (see Note 28). If you observe a yellow
flag although the chip loading quality appears proper, separa-
tion of digital dye signals might be insufficient. In that case
manual editing of dye signals appears necessary (see step 6).
5. To verify the uniformity of calls across the chip, review color

data by call by choosing “colour by call” (at Action Box) after
the quality threshold has been set. In this option each data
point is displayed based on FAM™ and VIC® or ROX™ target
dye signal detected.
6. In the scatter plot next to the chip view window, reporter dye
signals are shown: FAM™ (blue) signals are plotted on Y-axis
against VIC® (red) signals at the X-axis. Green signals illustrate
both FAM™ and VIC® reporter dye signals while yellow
ROX™ dye signal, used as internal standard, illustrates “no
amplification” signals where the fluorescence intensity did not
reach threshold. In the scatter plot you can now edit single data
points and exclude those which are not reliable using the lasso
tool (see Note 29).
7. Using multiple chips for one sample is required to improve
precision of rare allele detection (as shown in Fig. 1). It is
possible to omit chips, which did not reach standards with the
“Omit Chip” option. Chosen chips are then excluded from
result calculation.
8. Final results can be reviewed in the “See results” tab. Aggregate
results of multiple chips and/or different dilutions of each
sample (in case of identical name) are displayed in a bar
graph. Ratio of mutant to wild-type target [%] or absolute
quantification [copies/μL] is displayed at the Y-axis and data
group (sample assay) on the X-axis.
9. The final results will give you an overview of relative and
absolute results for total target and for each dye as well as
confidence levels. Precision will give you information about
Fig. 1 Impact of employed sample concentration and number of chips per sample, respectively, on precision.
Blue line shows precision of FAM (mutant allele) and red line shows precision of VIC (wild type allele). (a)
Impact of employed sample concentration on precision: Precision improves using higher template concentra-
tion, displayed as copies per reaction. (b) Impact of number of chips on precision: Precision improves with
increasing number of chips per sample
the confidence of the analysis defined as size of confidence

interval for distinguishing between two sample concentrations
at the same confidence level. The default confidence level is
95% and the default precision is 10%. Both can be edited by the
user (see Note 30).
10. It is recommended to review single chip results in “replicates”
tab to identify outliers and “digital calls” to identify insufficient
chip load (see Note 31).
4.2 Examples 1. Figure 2 shows typical results for quantitation of JAK2 allele
for JAK2 V617F Allele burden using QuantStudio™ 3D Digital PCR 20K Chip and
Quantitation Using QuantStudio™ 3D Digital PCR Analysis Suite™ Software.
QuantStudio™ 3D JAK2 V617F allele burden is automatically determined by the
Digital PCR software given as target/total (%).
2. Figure 3 shows results for quantitation of JAK2 allele burden
obtained from chips with different template concentrations.
Both chips are of adequate quality and show comparable quan-
titation results for target DNA, despite differences in initial
template concentration. Of note, precision is better in the
chip using higher template concentrations compared to the
chip using lower template concentrations in PCR.
3. Figure 4 shows chip images and results obtained from chips of
different quality. Although chips exhibit different quality and a
reduction of precision from high quality to low quality, estima-
tion of target DNA delivers comparable results.
5 Notes
1. Detection of the JAK2 V617F mutation and quantitation of its

allele burden can be performed using peripheral whole blood,
purified granulocytes, or bone marrow samples. It should be
noted, however, that JAK2 V617F allele burden measured in
peripheral whole blood and bone marrow samples is approxi-
mately equivalent because granulocytes constitute the predom-
inant population in both specimens [22, 23]. Furthermore, it
has been shown that granulocytic isolation achieves, on aver-
age, only 15% higher JAK2 allele burden compared with that of
whole white blood cells [24]. Therefore, total white blood cells
from the peripheral blood are the preferred type of cell popula-
tion for these analyses [25]
2. Selected DNA preparation kit should provide genomic DNA
free from PCR inhibitors and with a A260/280 ratio between
1.7 and 1.9. Be aware that gDNA concentration is sufficient
(see Notes 7 and 8) for subsequent dPCR analysis, especially in
case of low allele burden. Reduce elution volume if necessary.
Fig. 2 Typical results for quantitation of JAK2 allele burden using QuantStudio™ 3D Digital PCR 20K Chip and
QuantStudio™ 3D Digital PCR AnalysisSuite™ Software. (a) Chip image. Green flag in upper right corner
denotes adequate separation of digital dye signals. (b) Scatter plot: In the scatter plot, fluorescence signals are
shown for each well of the chip. Wells with JAK2 V671F mutant alleles are represented by FAM™-signals
(blue); wells with wild type alleles are represented by VIC®-signals (red); wells with both mutant and wild-type
alleles are represented by FAM™ and VIC®-signals (green); and wells without any alleles (passive reference)
are represented by ROX®-signals (yellow). FAM™, 6-carboxyfluorescein; VIC®, 4,7,20 -trichloro-70 -phenyl-6-
carboxyfluorescein; ROX®, 6-carboxy-X-rhodamine. (c) Results obtained by QuantStudio™ 3D Digital PCR
AnalysisSuite™ Software, given for displayed chip-image. Target/total % gives JAK2 V617F allele burden
(23.4%). Copies per reaction (rxn) gives mean number of copies per chip partition and copies per microliter
gives quantity of sample on chip for mutant and wild type allele, FAM™ and VIC®-signals, respectively.
Additionally, a dilution factor can be entered into the software for assessing quantity of sample stock. Validity
of results is given by predefined 95% confidence interval (CI) and precision, respectively, where the latter is
defined as spread of CI around two sample concentrations at a given CI. Software shows no data for VIC
concerning total percentage together with its CI
3. Accurate gDNA quantification is recommended in assessing

optimal starting copy number in dPCR. A fluorescence based
DNA quantification method is more accurate compared to
spectrophotometry, and, therefore, should be preferred, at
least in samples with a low DNA concentration.
4. Thermo Fisher offers a wet lab-validated TaqMan® SNP Gen-
otyping Assay for JAK2 V617F quantitation, also.
5. Thermo Fisher introduced in 2015 a new generation of chips,
named v2, which, in contrast to the previous v1 chip genera-
tion, enables improved UV adhesive-free sealing. Digital PCR
Fig. 3 Results for quantitation of JAK2 allele burden obtained from chips with different sample dilutions.
Figure shows chip images, scatter plots, and major results from two chips of the same sample, but from
different DNA dilutions (a: 300 copies/μL; b: 1000 copies/μL). Both chips are of adequate quality and show
comparable quantitation results for target DNA, despite differences in initial template concentration. Of note,
precision is better in chip B, using higher template concentrations in PCR compared to chip A
employing v2 chips requires QuantStudio 3D™ Digital PCR

Master Mix v2. However, v1 chips together with Master Mix
v1 are no more available.
6. The mass of one haploid human genome is given between 3.30
and 3.51 pg in the literature, depending on assumed human
genome length (3.0 billion base pairs versus 3.2 billion base
pairs), the assumed average molecular weight of a nucleotide
pair (650 g/mol versus 660 g/mol), and rounding differences
of the avogadro constant, respectively. Differences in the con-
version factor, however, will not significantly affect validity of
dPCR results. Alternatively, use a copy number calculator
(available for example on http://www.thermofisher.com/at/
en/home/brands/thermo-scientific/molecular-biology/
molecular-biology-learning-center/molecular-biology-
resource-library/thermo-scientific-web-tools/dna-copy-num
ber-calculator.html).
7. Be aware that the specification of the gDNA concentration
refers to the gDNA concentration in the dPCR reaction and
not to in the gDNA stock solution. For instance, assuming a
PCR-mix volume of 34.8 μL (in case of using two chips for one
sample to perform duplicates) and a desired template
Fig. 4 (a–c) Results for quantitation of JAK2 allele burden obtained from chips with different quality.
Figure shows chip images and major results from three chips of the same sample using same template
concentration in PCR, but differ in quality. Concentration was about 300 copies/μL and therefore at the lower
end of optimal concentration range (200–2000 copies/μL). Although chips exhibit different quality and a
reduction of precision from high quality (a) to low quality (c), estimation of target DNA delivers comparable
results
concentration of 1000 copies/μL, PCR mix contains 34,800

copies or 122 ng gDNA in total.
8. In case of low allele burden (<1%) higher DNA concentrations
(up to 6000 copies/μL) may be used, which results in an
improved precision of calculated copy number of the rare allele.
If you have limited sample concentrations, precision can be

optimized using multiple chips. Impact of employed sample
concentration and number of chips, respectively, on precision is
shown in Fig. 1.
9. A survey of the workflow including chip loading is demon-
strated in the YouTube video “QuantStudio 3D Digital PCR
Workflow Video” (https://www.youtube.com/watch?
v¼PjgwDhN63Zc).
10. To avoid spreading to much immersion fluid on the chip later
on, depress the plunger and release some immersion fluid to
get a feeling for pressure ratio. Note that the fluid should be
used within 1 h after uncapping the syringe.
11. Confirm that the blade is properly seated by pressing firmly
against the loader head and checking its position on the back-
side of the loader head as well. A correct position of the loader
head and blade is essential since it is responsible for the appli-
cation of PCR reaction mixture on the chip. However, it is
possible to install the blade after filling in the sample.
12. To avoid contact with the sticky film beneath, gently grasp only
the outside margin of the lid. Remove the protective film
quickly. Otherwise, it is likely that you will remove the sticker
as well. V2 Chips may not fit properly in first generation chip
loaders. In that case, Life Technologies provides a “chip grip-
per upgrade kit.”
13. Another possibility to properly apply the lid in the lid nest is to
remove the sticky film after insertion of the lid into the nest.
Therefore, you will need to tightly press the lid nest button
while removing the film so that the lid stays fixed.
14. Filling the loading port appropriately is utterly important for
chip quality. Therefore it will need precise handling and
requires some practice. First of all, it is important to distribute
the solution evenly in the port so that each side of the chip will
contain the same amount of solution later on. Further, one
should avoid discarding some solution by deflecting the blades
through applying too much pressure on the tip. To avoid air
bubbles in the port, do not depress the pipette to the second
stop. However, if a bubble appears, try to remove it by carefully
taping the port from both sides. Make sure that the port is in
the right position afterward and do not tap too hard to avoid
shifting the loading head or discarding solution. Note that the
exact adjustment of the loading head is essential for proper chip
load. If you recognize that the loading head does not apply
solution properly, contact service support. Another option is to
fill the loading blade prior to installation into the loading head,
which might ease filling procedure, since one can choose opti-
mal filling angle. In this particular case, take care to not discard
any dPCR solution during installation of the blade.
15. It is important to apply the immersion fluid about 20 s after

loading to avoid evaporation of reaction mixture. Make sure
that the immersion fluid covers the surface of the chip but does
not flood the whole chip case. When applying the sticky film of
the lid later on, make sure that there is no immersion fluid on
the edges of the chip case. In case there is immersion fluid on
the edges, you can remove it with a wipe that has been sprayed
with some isopropanol.
16. A pressure of >20 pounds (lb) which approximates about
9 kilograms (kg) is recommended for tight sealing. Though it
is difficult to estimate the correct pressure, it can be recom-
mended to press 20 s with moderate pressure and in a 90 angle
with the chip. Make sure that the chip loader is in an appropri-
ate position close to the operator’s body and cannot tilt
forward.
17. Choosing the correct angle is important to release air bubbles
between the chip and the chip lid. Remaining bubbles can
interfere with PCR and impair analysis of the chip later
on. The 45 angle provides a good starting position although
the operator will need to adjust the chip position in order to
position the bubble beneath the chip port.
It is important to pour the immersion fluid in slowly, so that
the filling can be stopped easily when the chip is nearly filled
(remaining a small air bubbles <2–3 mm at the port).
18. Make sure not to touch chip window too firmly after the chip is
sealed. Pressure on the chip window might expel reaction
mixture out of the wells.
19. For quantitation of JAK2 V617F allele burden PCR conditions
are modified to 56 C annealing temperature instead of 60 C
annealing temperature recommended by the manufacturer as
default annealing temperature. The lower annealing tempera-
ture results in an improved cluster separation.
20. The elevated fill pot makes sure that small bubbles will float to
the top of the case, not impairing the amplification.
21. You can remove chips after final extension step, if the cycler has
cooled down to <25 C. Avoid fast changes in temperature to
prevent condensation of chips. According to specifications of
the manufacturer it is possible to leave chips on 10 C for up to
24 h (see Note 22, also). Chips must be brought to room
temperature prior to imaging.
22. To our experience, chips can be stored several days at room
temperature in a clean and dark place with consistent quality.
23. There is no need to wait until the analysis is completed. To start
with the next run, just remove the prior chip and load the
next one.
24. It is recommended in the manufacturer’s manual, that a chip

should be reread within 1 h after the initial read. In fact, the
chip can be reread much longer (e.g., on the next day) with
consistent quality if it is properly protected from light. Chip
storage at room temperature is sufficient.
25. The software can be run using any compatible operating system
and web browser but it is optimized for Microsoft® Windows®
operating system and use on Google® Chrome™. Please refer
to user guide to check requirement since performance may vary
depending on operation system and configuration.
26. We recommend using the help-tool to get started with analysis.
It contains demonstration tools, i.e., video demonstrations,
helping you to make a start.
27. It is important to define rare allele dye, displaying the fluores-
cence dye, which detects the rare allele. Note that the identical
sample names must be used in case of duplicates.
28. It is possible to include or exclude wells from analysis by
adjusting the quality threshold. This assessment can be applied
to all chips of the assay. The different colors of the quality flags
are based on well quality threshold and other data characteris-
tic. The chip quality assessment can help to detect problems
with application of dPCR solution by visualizing the spatial
distribution of data across the chip.
29. There are several tools, which help you to adjust data points.
You can adjust the scatter plot axis and size of data points as
well as the zoom. After you defined optimal conditions, you
may use the lasso tool to choose data points to include or
exclude (“Undetermined”) from your data analysis. Separate
data clouds (FAM™; VIC®, ROX™) accurately to improve
precision of results. Especially VIC® and ROX™ clouds often
overlap, concerning the huge amount of data points. Loss of
some data points through exclusion is readily tolerated.
30. One can lower and therefore improve the calculated precision
by measuring multiple chips for the same sample and combin-
ing them in one chip (see Fig. 1). The software automatically
combines chips with the same name in one virtual chip. The
number of chips is listed in results (“Chips”). Furthermore, it is
recommended to fill in dilution factor as the calculated sample
quantity (copies/μL) will not display your stock concentration,
but the gDNA concentration in dPCR reaction.
31. Outliers can be excluded from calculation with the “omit chip”
option in the “Review Data” tab. Digital calls contain data of
negative, positive or undetermined calls and may help you
identify problems with chip load or detection quality. For
optimal output, >17,000 calls should have been qualified by
quality threshold.
References
1. Levine RL et al (2005) Activating mutation in status, is associated with inferior overall and
the tyrosine kinase JAK2 in polycythemia vera, leukemia-free survival. Leukemia 22:756–761
essential thrombocythemia, and myeloid meta- 15. Guglielmelli P et al (2009) JAK2V617F muta-
plasia with myelofibrosis. Cancer Cell tional status and allele burden have little influ-
7:387–397 ence on clinical phenotype and prognosis in
2. Baxter EJ et al (2005) Acquired mutation of patients with post-polycythemia vera and
the tyrosine kinase JAK2 in human myelopro- post-essential thrombocythemia myelofibrosis.
liferative disorders. Lancet 365:1054–1061 Haematologica 94:144–146
3. Kralovics R et al (2005) A gain-of-function 16. Quintás-Cardama A et al (2009) Pegylated
mutation of JAK2 in myeloproliferative disor- interferon alfa-2a yields high rates of hemato-
ders. N Engl J Med 352:1779–1790 logic and molecular response in patients with
4. James C et al (2005) A unique clonal JAK2 advanced essential thrombocythemia and poly-
mutation leading to constitutive signalling cythemia vera. J Clin Oncol 27:5418–5424
causes polycythaemia vera. Nature 17. Kiladjian J-J et al (2008) Pegylated interferon-
434:1144–1148 alfa-2a induces complete hematologic and
5. Vardiman JW et al (2009) The 2008 revision of molecular responses with low toxicity in poly-
the World Health Organization (WHO) classi- cythemia vera. Blood 112:3065–3072
fication of myeloid neoplasms and acute leuke- 18. Jovanovic J (2013) V et al. Establishing optimal
mia: rationale and important changes. Blood quantitative-polymerase chain reaction assays
114:937–951 for routine diagnosis and tracking of minimal
6. Passamonti F et al (2006) Relation between residual disease in JAK2-V617F-associated
JAK2 (V617F) mutation status, granulocyte myeloproliferative neoplasms: a joint
activation, and constitutive mobilization of European LeukemiaNet/MPN&MPNr-Euro-
CD34þ cells into peripheral blood in myelo- Net (COST action BM0902) stu. Leukemia
proliferative disorders. Blood 107:3676–3682 27:2032–2039
7. Lippert E et al (2006) The JAK2-V617F muta- 19. Lippert E et al (2009) Concordance of assays
tion is frequently present at diagnosis in designed for the quantification of JAK2V617F:
patients with essential thrombocythemia and a multicenter study. Haematologica 94:38–45
polycythemia vera. Blood 108:1865–1867 20. Kinz E, Leiherer A, Lang AH, Drexel H,
8. Passamonti F, Rumi E (2009) Clinical rele- Muendlein A (2015) Accurate quantitation of
vance of JAK2 (V617F) mutant allele burden. JAK2 V617F allele burden by array-based digi-
Haematologica 94:7–10 tal PCR. Int J Lab Hematol 37:217–224
9. Langabeer SE et al (2015) Molecular diagnos- 21. Majumdar N, Wessel T, Marks J (2015) Digital
tics of myeloproliferative neoplasms. Eur J PCR modeling for maximal sensitivity, dynamic
Haematol 95:270–279 range and measurement precision. PLoS One
10. Vannucchi AM, Pieri L, Guglielmelli P (2010) 10:e0118833
JAK2 allele burden in the myeloproliferative 22. Larsen TS, Pallisgaard N, Møller MB, Hassel-
neoplasms: effects on phenotype, prognosis balch HC (2008) Quantitative assessment of
and change with treatment. Ther Adv Hematol the JAK2 V617F allele burden: equivalent
2:21–32 levels in peripheral blood and bone marrow.
11. Vannucchi AM, Guglielmelli P (2013) JAK2 Leukemia 22(1):194–195. http://www.ncbi.
mutation-related disease and thrombosis. nlm.nih.gov/pubmed/?term¼17625603
Semin Thromb Hemost 39:496–506 23. Takahashi K et al (2013) JAK2 p.V617F detec-
12. Koren-Michowitz M et al (2012) JAK2V617F tion and allele burden measurement in periph-
allele burden is associated with transformation eral blood and bone marrow aspirates in
to myelofibrosis. Leuk Lymphoma patients with myeloproliferative neoplasms.
53:2210–2213 Blood 122:3784–3786
13. Alvarez-Larrán A et al (2014) JAK2 V617F 24. Hermouet S et al (2007) Comparison of whole
monitoring in polycythemia vera and essential blood vs purified blood granulocytes for the
thrombocythemia: clinical usefulness for pre- detection and quantitation of JAK2(V617F).
dicting myelofibrotic transformation and Leukemia 21:1128–1130
thrombotic events. Am J Hematol 89:517–523 25. Gong JZ et al (2013) Laboratory Practice
14. Tefferi A et al (2008) Low JAK2V617F allele Guidelines for Detecting and Reporting JAK2
burden in primary myelofibrosis, compared to and MPL Mutations in Myeloproliferative
either a higher allele burden or unmutated Neoplasms. J Mol Diagn 15:733–744
Chapter 16
Novel Multiplexing Strategies for Quantification of Rare

Alleles Using ddPCR
Miguel Alcaide and Ryan D. Morin
Abstract
Droplet digital PCR (ddPCR) has come to be regarded as the gold standard for the ultrasensitive detection
and absolute quantification of closely related DNA sequences within complex mixtures. Most ddPCR assays
to date, however, rely on sets of hydrolysis probes conjugated with dyes having different emission spectra to
allow independent counting of rare mutant and wild-type alleles. Here, we describe a set of novel strategies
that leverage the simultaneous detection and quantification of both mutant and wild-type alleles with a
single hydrolysis probe. Variants of these strategies empower multiplexing and a more cost-effective
approach for concurrent screening of multiple genetic variants.
Key words Digital PCR, Genetic testing, Noninvasive genetic profiling, Rare allele detection, Cell-
free DNA, Clonal evolution, Pathogen detection
1 Introduction
The power to detect and quantify rare genetic variants within

complex DNA mixtures with both high sensitivity and specificity
has profound implications both in healthcare and environmental
analysis. Noninvasive genetic profiling of cell-free fetal DNA in the
maternal bloodstream for prenatal testing [1, 2] and the investiga-
tion of circulating tumor DNA (ctDNA) in cancer diagnosis and
monitoring by means of liquid biopsies [3–5] are among the most
cited applications. Low-level pathogen detection is also critical to
prevent waterborne or foodborne illnesses and has also recently
shown great utility to monitor infections in anatomic fluids, the
emergence of antimicrobial resistance, or the proliferation of
genetically modified organisms [6–10]. Most of the standard
molecular methods such as quantitative PCR, traditional Sanger
sequencing, microarrays, fluorescence in-situ hybridization
(FISH), immunology-based methods or even generic next genera-
tion sequencing (NGS) of PCR products cannot reliably discern
between real and putatively false variants when rare allele
275
276 Miguel Alcaide and Ryan D. Morin
frequencies are around or below 1% [5]. The advent of digital PCR

and novel NGS methods involving molecule barcoding strategies
have pushed the limits of detection for rare alleles to one in several
thousand or even hundreds of thousands [11–14].
Digital PCR involves the partition of a single PCR reaction into
thousands or even millions of parallel microreactions for the con-
finement of single (or low numbers of) copies of the locus of
interest within prefabricated chambers or nanoliter/picoliter dro-
plets generated by water-in-oil emulsions [15, 16]. When com-
pared to other PCR-based methods, digital PCR is less sensitive
than conventional qPCR to the inhibitors that can be present, for
example in crude samples such as water, soil, or anatomical fluids,
thus offering a more robust testing platform [17, 18]. Digital PCR
enables the straightforward and absolute quantification of rare
alleles without the need of reference samples to construct standard
curves and can discriminate among closely related DNA sequences
with exquisite specificity. Single nucleotide polymorphisms and
small indels underlie numerous inherited genetic disorders and
somatic alterations of this variety have been recurrently reported
as mutation “hotspots” in many human cancers. Such simple
genetic alterations can similarly drive the adaptive evolution of
disease-associated microorganisms, offering a variety of avenues
by which sensitive genotyping can provide potential clinical utility
[19–21].
Hydrolysis probes conjugated with dyes of different emission
spectra allow the simultaneous detection and quantification of
highly similar DNA sequences at single base pair resolution. Duplex
digital PCR is an appealing technology because it reduces technical
errors associated with parallel reactions and offers significant sav-
ings in time, overall costs and the amount of clinically or environ-
mentally precious samples needed for a given assay. Even though
currently available digital PCR systems are often restricted to two
optical detection channels (with exceptions, for instance [22]),
higher order multiplexing is nonetheless enabled in these systems
by the use of different concentrations of probes labeled with the
same dye but against different targets, or by the use of variable
ratios of two probes labeled with different dyes targeting the same
sequence, thus giving them a unique location in 2D fluorescence
space [23]. The accumulation of unquenched fluorophores asso-
ciated with probe hydrolysis after end-point PCR permits measur-
ing the amount and type of fluorescence emitted by each individual
partition and distinguishing whose which are positive and those
which are negative for a target of interest. Established statistical
models relying on the Poisson distribution are then applied to
undertake absolute quantification of different DNA templates in
the initial sample [23]. Until recently, nondiscriminating assays to
detect multiple mutations such as “wild-type negative” or “drop-
off” assays required a minimum of two hydrolysis probes labeled
Novel Multiplexing Strategies for Quantification of Rare Alleles Using ddPCR 277
with different fluorophores [23–25]. Similarly, discriminating

amplitude-based or ratio-based multiplexed assays specifically tar-
geted rare and wild-type alleles with probes conjugated with differ-
ent dyes [23, 26].
In this chapter, we describe assay design strategies which reduce
the number of probes necessary for successful multiplex detection
and quantification of both rare and wild-type alleles. These employ
either discriminating approaches that exclusively use one or more
mutant-specific hydrolysis probes, which also quantify the wild-type
allele(s), or nondiscriminating approaches exclusively using one or
more probes against wild-type DNA sequences (“inverted” ddPCR
assays), yet which also quantify mutant alleles. In addition, these
can be designed to detect and quantify one or several loci in a single
reaction by the presence of one or two primer pairs targeting them.
These cost-effective and straightforward assays are possible owing
to the less efficient hydrolysis of imperfectly annealed mutant or
wild-type probes—where single base mismatches between the
probe and the DNA template occur—and where the consequences
on each target molecule can be captured in digitized reactions as
droplets with lowered fluorescence amplitude(s) [27]. The funda-
mentals of other multiplex digital PCR strategies for rare allele
detection have been recently reviewed in detail [23] and therefore
will not be addressed in the present chapter.
2 Materials
1. ddPCR™ Supermix for Probes (No dUTP) (Bio-Rad, Hercu-

les, CA, USA; Cat. No. # 186-3023, 186-3024 or 186-3025)
(see Note 1).
2. Automated Droplet Generation Oil for Probes (Bio-Rad, Her-
cules, CA, USA; Cat. No. #1864110) (see Note 2).
3. ddPCR™ Droplet Reader Oil (Bio-Rad, Hercules, CA, USA;
Cat. No. #1863004) (see Note 3).
4. DG32™ Automated Droplet Generator Cartridges (Bio-Rad,
Hercules, CA, USA; Cat. No. #1864108 and #1864108) (see
Note 4).
5. Pipet Tips for the AutoDG™ System (Bio-Rad, Hercules, CA,
USA; Cat. No. #1864108 and #1864120).
6. Pipet Tip Waste Bins for the AutoDG™ System (Bio-Rad,
Hercules, CA, USA; Cat. No. #1864108 and #1864125) (see
Note 5).
7. ddPCR™ 96-Well PCR Plates (Bio-Rad, Hercules, CA, USA;
Cat. No. #1864108 and #12001925) or Eppendorf twin. tec®
PCR Plate 96, semiskirted (Eppendorf, Hamburg, Germany;
Cat. No: #951020303).
8. PCR Plate Heat Seal, foil, pierceable (Bio-Rad, Hercules, CA,

USA; Cat. No. #1864108 and #1814040) (see Note 6).
9. PCR-Cooler (Eppendorf, Hamburg, Germany; Cat. No:
#022510541) (see Note 7).
10. Automated Droplet Generator (AutoDG™ System; Bio-Rad,
Hercules, CA, USA; Cat. No. #1864101) (see Note 8).
11. QX200™ Droplet Reader (Bio-Rad, Hercules, CA, USA; Cat.
No. #1864003) (see Note 9).
12. PX1™ PCR Plate Sealer (Bio-Rad, Hercules, CA, USA; Cat.
No. #1814000) (see Note 10).
13. C1000 Touch™ Thermal Cycler with 96-Deep Well Reaction
Module (Bio-Rad, Hercules, CA, USA; Cat. No. #1814000)
(see Note 11).
14. Assay-specific hydrolysis probes: ZEN™ Double-Quenched or
LNA PrimeTime® probes (Integrated DNA technologies, Cor-
alville, IA, USA) conjugated with 6-FAM™ or HEX™ dyes
and 30 Iowa Black® FQ dark quenchers (see Note 12).
15. QuantaSoft™ software, Regulatory edition (Bio-Rad, Hercu-
les, CA, USA; Cat. No. #1864011).
16. Rainin (Mettler-Toledo, Giessen, Germany) or Eppendorf fil-
ter tips (Eppendorf, Hamburg, Germany; Cat. No:
#951020401) (see Note 13).
17. ddPCR™ Buffer control for probes (Bio-Rad, Hercules, CA,
USA; Cat. No. #1863052).
18. Ultra-pure water.
19. 10 mM Tris–HCl (pH ¼ 8.0) or 10 mM Tris–HCl, 0.1 mM
EDTA (pH ¼ 8.0).
20. Buffer AVE (Qiagen, Venlo, Netherlands).
21. Buffer ATE (Qiagen, Venlo, Netherlands).
22. Restriction endonucleases (optional, see Note 14).
23. Uracil DNA Glycosylase (optional, see Note 15).
24. Wild-type DNA to be used as negative control
25. Covaris M220 ultrafocused sonicator or similar (optional; Cov-
aris, Woburn, MA, USA).
3 Methods
3.1 Sample This protocol has been validated and optimized using genomic
Preparation DNA extracts from cells or fresh tissues, prepared with various
Qiagen purification kits according to the manufacturer’s instruc-
tions, and eluted either in 10 mM Tris–HCl (pH ¼ 8.0) or 10 mM
Tris–HCl, 0.1 mM EDTA; Cell-free DNA (cfDNA) extracts eluted
in buffer AVE (Qiagen, Venlo, Netherlands); and FFPE-derived

DNA extracts (see Note 15) eluted in buffer ATE (Qiagen, Venlo,
Netherlands) (see examples, including FFPE samples, in [27]).
Cell-free DNA does not need any treatment prior to digital PCR
and can be extracted from plasma using any number of commercial
kits. Preanalytical handling of blood for cfDNA processing is
reviewed elsewhere [5]. Restriction endonuclease digestion or
mechanical shearing is particularly recommended when the con-
centration of intact genomic DNA exceeds 72 ng per 22 μl ddPCR
reaction in order to facilitate proper droplet formation. Enzymatic
digestion can be accomplished with a restriction endonuclease that
does not cut target or reference amplicons (see Note 14). Mechani-
cal fragmentation can be performed using a Covaris ultrafocused
sonicator (Covaris, Woburn, MA, USA). Fragmentation of sample
DNA not only reduces sample viscosity but may also improve
accessibility of assays to certain target regions even at lower DNA
concentrations.
3.2 Assay Design In designing primers, the PCR amplicon size should be kept to a
Considerations minimum, ideally between 60 and 80 bp, with shorter amplicons
preferred. This is particularly relevant when using fragmented
3.2.1 Amplicon Size
DNA, for example cell-free DNA, as template. Long PCR ampli-
cons are typically observed to have lower fluorescence amplitudes
and higher background noise. In addition, as reviewed elsewhere
[5], longer PCR amplicons can compromise assay sensitivity in
cfDNA as a substantial fraction of the typically short DNA tem-
plates could go undetected because they do not contain binding
sites for both assay primers [28, 29]. Below is a list of considera-
tions for the design of assays with good performance. However, it is
noted that complying with the full list of recommendations listed
can be challenging due to the sequence characteristics of the target
and its genomic context. For useful assay design tools that generate
candidate primer and probe sequences for the intended assays, see
below (and see Note 16).
3.2.2 PCR Primers 1. Run BLAST (https://blast.ncbi.nlm.nih.gov/Blast.cgi) or

BLAT (https://genome.ucsc.edu/cgi-bin/hgBlat?com
mand¼start) on potential primer sequences to avoid cross
reactivity with other regions of the genome
2. The two primers must have a similar melting temperature
(Tm 2 C) and GC contents generally ranging between
35% and 65%
3. Primer length must not exceed 30 nucleotides
4. Avoid sequences folding into strong secondary structures, self-
dimers or heterodimers as well as long homopolymer runs,
particularly polyG >4 bases long
5. Ensure that there are no common SNPs overlapping the 30

ends of the primers, as these could decrease amplification
efficiency.
3.2.3 Hydrolysis Probes 1. Use BLAST or BLAT to screen potential probe sequences to
avoid cross reactivity with other regions of the genome.
2. Avoid G bases at the 50 end as they have a well-demonstrated
quenching effect over some fluorophores.
3. Probes are commonly intended to have a Tm 6–10 C warmer
than that of primers during traditional qPCR experiments.
Although hydrolysis probe-based assays designed following
these criteria can work well in ddPCR reactions, Tm values
just 3 C higher have been observed to perform better. Special
modifications such as LNA bases (Integrated DNA Technolo-
gies, Coralville, IA, USA) or MGB modifications (minor
groove binder, Applied Biosystems, Foster City, CA, USA)
can be used to increase the Tm of probes without extending
their length.
4. Avoid long homopolymer G runs (>4 bases), secondary struc-
tures and the formation of homodimers or heterodimers with
the PCR primers
5. Place the genomic position to be interrogated as centered as
possible within the sequence of the probe
6. Select the sense or antisense strand to leave at least a five
nucleotide distance with respect to the 30 end of the primer
that anneals to the same strand. Probes designed to one strand
may be less prone than the other to form secondary structures
or dimers.
7. Conjugate the probes with 6-FAM™ (Channel 1) or HEX™
(Channel 2) fluorescent dyes and add ZEN double-quenchers
or Iowa Black® Dark quenchers (Integrated DNA technolo-
gies, Coralville, IA, USA). Although not evaluated for this
particular application, Black Hole Quenchers® (LGC Biosearch
Technologies, Novato, CA, USA) and TaqMan® minor groove
binder (MGB) probes (Life Technologies, Carlsbad, CA,
USA), to name a few alternative probe designs, are also
expected to perform well for this particular application (see
also Note 12).
3.3 Setting Up In this section, general instructions applicable to each of the uniplex
and Running ddPCR or multiplexing strategies described in this chapter are given for
Experiments reaction setup, droplet generation, thermocycling, and droplet
reading/data acquisition. Refer to the sections following this one
(Subheadings 3.5 through 3.9) for detailed guidance on the rele-
vant uniplex or multiplexing approach(es) being followed.
Table 1
Suggested volumes and final concentrations during the preparation of the different discriminating
and nondiscriminating hydrolysis probe-based ddPCR experiments described in this chapter
Uniplex
(Single- Uniplex Multiplex (Single
locus) (Dual-locus) locus)
2 SuperMix ddPCR for Probes 11 μL (1) 11 μL (1) 11 μL (1)
(no dUTP)
Primers Premix—10 μM (each) stock 2.2 μL 2.2 μL (1 μM) 2.2 μL (1 μM)
(1 μM) (2 Pairs) (1 Pair)
(1 Pair)
6-FAM™ Hydrolysis Probe (Mutant- or 1.5 μL 1.5 μL (0.34 μM) 2.2 μL (0.5 μM)
Wild type Specific)—5 μM stock (0.34 μM) Mutant-specific
only
6-FAM™ Hydrolysis Probe (Mutant- 0.8 μL (0.18 μM)
Specific)—5 μM stock
6-FAM™ Hydrolysis Probe (Mutant- 0.33 μL (0.075 μM)
Specific)—5 μM stock
HEX Hydrolysis Probe (Mutant- or Wild 1.5 μL (0.34 μM) 1.4 μL (0.3 μM)
type-Specific)—5 μM stock Mutant-specific
only
Dark probes—5 μM (each) stock 0–1.5 μL 0–1.5 μL (variable)
(optional) (variable) (0, 1, or 2 locus)
DNA + Ultra Pure Water 5.8–7.3 μL 2.8–5.8 μL 4.07 μL
Total 22 μL 22 μL 22 μL
All reactions are made to a final volume of 22 μL/sample
1. Droplet digital PCR reactions must be set to a minimum final

volume of 22 μL to avoid mechanical errors during automated
droplet generation. This volume should contain 11 μL of
2 ddPCR™ Supermix for Probes (no dUTP). The remaining
11 μL encompass PCR primers (1 μM final concentrations),
hydrolysis probes (0.075–0.5 μM final concentrations, see
Note 17), and a variable amount of input DNA (see Note 18
and Table 1). All reagents should be thawed at room tempera-
ture, vortexed and briefly spun to eliminate concentration
gradients formed during storage. After preparing, the desired
reactions according to Table 1, individual reactions are dis-
pensed at room temperature into Bio-Rad ddPCR™ or Eppen-
dorf twin. tec® 96-well semiskirted PCR plates.
2. For droplet generation, follow the QX100 or QX200 Droplet
Generator Instruction Manual (#10026322 or 10031907,
respectively) for manual droplet generation if an automated
droplet generator is not available in your laboratory. If you
Table 2
Recommended cycling protocol for ddPCR reactions relying on hydrolysis probes
Cycling step Temperature ( C) Time Number of cycles

Enzyme activation 95 C 10 min 1
Denaturation 94 C 30 s 40
Annealing/Extension Optimized for individual assays (52–62 C) 1 min

Enzyme deactivation 98 C 10 min 1

Hold 4 C Infinite 1
choose to assemble the reactions on ice, it is important to

equilibrate reactions at room temperature for at least 3 min
before droplet generation.
3. A minimum of 8 samples dispensed in one single column is
required for automated droplet generation (see Note 19). Place
your sample plate in the designated area within the automated
droplet generator (left plate compartment) and a clean plate
onto a plate cooler in its corresponding spot (right plate com-
partment) (see also Notes 2, 4 and 5).
4. Click on Configure Sample Plate button and select the specific
columns of the plate containing samples. Confirm and initiate
automated droplet generation (<5 min per 8 samples or col-
umn of samples).
5. Check for error messages in the Automated Droplet Generator
screen (see Note 20) and proceed if the operation completed
successfully. Refer to Chapter 4 of the Automated Droplet
Generator Instruction Manual (http://www.bio-rad.com/
webroot/web/pdf/lsr/literature/10043138.pdf) and/or
contact Tech support for troubleshooting.
6. After all reactions have been converted to droplets, cover the
collection plate with pierceable foil and seal the foil with heat
(180 C) in the PX1™ PCR Plate Sealer (see Note 6).
7. Introduce the plate inside the C1000 Touch™ Thermal Cycler
and run the preloaded program described in Table 2, with the
lid preheated at 105 C, setting the volume of the reaction at
40 μL and the ramp rate for each step of the cycle at 2 C/s (see
Note 21).
3.4 Data Acquisition 1. Once the PCR amplification is complete, place the sealed plate
and Analysis containing samples in the QX200™ Droplet Digital PCR
reader (see Note 3).
2. Open the QuantaSoft™ software and set up a new plate layout.
Double click on a particular well to open the well-editor
dialog box.
3. For each well, introduce sample name, experimental type

(“Absolute quantification” for target counts >100 copies per
well, or “Rare Event Detection” for <100 counts per well), the
type of supermix should be entered as ddPCR™ Supermix for
probes (no dUTP), name of targets, optical channel associated
with each target (6-FAM™ in channel 1; HEX™ or VIC™
in channel 2) and type of sample analyzed (e.g., unknown,
reference, positive control, negative control, no-template
control). Click “Apply” to label the wells and click “OK”
when finished.
4. Click on the Run button to start the droplet reading process.
5. After the data collection completes, you should inspect your
results by clicking on the Analysis tab. To perform quantifica-
tion, you must first set the thresholds and clusters manually
across the 2-D amplitude data plots, as guided by the cluster
boundaries observed in reference samples, positive controls,
negative controls, and no-template DNA controls. Specific
instructions for target quantification are detailed below for
each of the ddPCR assays described in this chapter.
6. Use copies/μl and allele ratios to estimate the abundance of
targeted alleles in your initial samples.
3.5 Uniplex “Single- We recently demonstrated that ddPCR enables the simultaneous
Locus” Assay discrimination of mutant and wild-type alleles at single base pair
Validation resolution with the use of a unique hydrolysis probe [27]. We
and Optimization Prior observed that imperfectly annealed probes can be degraded, if a
to Use in Multiplexing single mismatch between the probe and the DNA template occurs,
(One Pair of Primers when mutant and wild-type probes are not competing for hybridi-
and One Mutant Allele- zation in the same microreaction. This phenomenon generates
Specific Probe) droplets with at least three clearly differentiated fluorescence ampli-
tudes. The lowest fluorescence amplitude relates to the imperfect
quenching of the fluorophore attached to the unhydrolyzed probe
(“double-negative” droplets, i.e., droplets that do not contain
either wild-type or mutant alleles for a given locus) and the highest
fluorescence amplitude (“positive” droplets) can be explained by
the specific and highly efficient hydrolysis of probes after a perfectly
complementary match with the DNA template. Notably, we
observed an intermediate fluorescence amplitude band (“low posi-
tive” droplets), clearly distinguishable from “double-negative”
droplets and “positive” droplets, which we attribute to the less
efficient hydrolysis of imperfectly annealed probes (see Fig. 1).
Absolute quantification of both mutant and wild-type alleles for a
given locus using assays that forego, for example wild-type specific
probes, is made possible by this feature. Lower cost per assay and
enhanced possibilities for multiplexing are the most direct benefits
of using such single hydrolysis probes (see Note 22). An important
preliminary step to leverage the potential of this novel approach
Fig. 1 Effect of annealing temperature—tested in a single well—for two uniplex ddPCR assays, targeting two
mutations at two different loci. The first assay (top panel) exhibits optimal separation between droplets
carrying the allele specifically targeted by the probe (“positive” droplets), droplets containing alternative
alleles differing in one single base pair (“low positive” droplets), and “double-negative” droplets at 57.5 C
(top, middle, and bottom boxed regions). For the second assay (bottom panel), the optimal annealing
temperature is 60 C, with lower annealing temperatures showing poor differentiation between “positive”
and “low positive” droplets. Thus, if these assays were to be run together in a single well, the higher annealing
temperature of 60 C would need to be used. These assays were performed using genomic DNA extracted
from fresh tumor biopsies, previously known to carry mutations at the two loci investigated, as DNA template.
The first assay relies on a single hydrolysis probe labeled with 6-FAM™ (Channel 1) and the second assay
relies on a single hydrolysis probe labeled with HEX™ (Channel 2)
(and the additional ones below) is to determine the best set of

conditions and parameters that generate optimal fluorescence
amplitude differentiation between different target alleles.
3.5.1 Positive Samples carrying specific mutations of interest (positive controls)

and Negative Controls and wild-type DNA (negative controls) are critical to validate uni-
plex ddPCR assays prior to use in multiplexing. In the absence of
relevant biological samples containing the alleles of interest, syn-
thetic double-stranded and sequence-verified DNA fragments such
as gBlocks® Gene Fragments (Integrated DNA technologies, Cor-
alville, IA, USA) can be used for control templates (e.g., spiked into
wild-type background genomic DNA).
3.5.2 Annealing Annealing temperature during emulsion PCR is one of the factors
Temperature with the greatest impact on assay resolution. Certain annealing
temperatures provide the best separation between “double-nega-
tive,” “positive,” and “low positive” droplets for a given locus (see
Fig. 1). As an aside, where multiple probes will be present in a single
assay reaction (e.g., see Subheading 3.7 on uniplex dual-locus reac-
tions), if the tested probes show very different optimum annealing
temperatures, it is recommended to redesign one of the them prior
to multiplexing in order to achieve a more similar temperature
optimum for both that gives good discrimination of clusters for
each of the independent loci (see Note 23).
3.5.3 Inclusion of “Dark The inclusion of an oligonucleotide that matches the sequence of
Probes” abundant wild-type allele but lacking a conjugated fluorescenct dye
(i.e., “dark probes,” see for example [30]) can enhance the separa-
tion between the fluorescence amplitudes associated with rare and
wild-type alleles (Fig. 2) (see Note 24).
3.5.4 Running Standard setup and preparation of discriminating uniplex ddPCR

and Analyzing the Assay assays is detailed in Table 1. After thermocycling the reactions and
then reading them in the droplet reader, quantification of mutant
and wild-type alleles is accomplished following manual identifica-
tion of clusters in the QuantaSoft™ software (see for example
Fig. 3, Panel a).
Under the Setup mode, select wells for unknown samples,
positive controls, negative controls, and no-template DNA controls
(if available), all run with the same assay. For the determination of
allele counts and mutant fractional abundance, ensure that the two
optical channels have been associated to the proper target (i.e.,
allele) in the Setup menu and display the data on the 2D Amplitude
plot after clicking on Analyze. You will need to associate “positive
droplets” carrying mutant alleles with the FAM channel and “low
positive droplets” carrying wild-type alleles with the HEX channel.
To achieve this, use the rectangular (or circular or freehand) Multi-
Fig. 2 (a) The inclusion of a wild-type oligonucleotide not conjugated with a fluorescent dye (i.e., a “dark
probe” as illustrated in panel b) enhances the separation of clusters of droplets carrying the rare or wild-type
alleles when running a uniplex ddPCR assays. Suggested manual thresholds for the independent quantification
of rare (magenta line, blue droplets) and wild-type alleles (black line, intermediate band of black droplets) are
also indicated. We used a DNA extract derived from a fresh tumor biopsy sample carrying the mutation
specifically targeted by the probe and a commercially available wild-type DNA sample as positive and negative
controls, respectively
well thresholding selection tool, click on the / black button and
enclose all droplets within the shape chosen to initially designate all
droplets as “double-negative” droplets. Then, click on the /+
green button (HEX/VIC channel) and using the selection tool,
enclose and define the cluster of “low positive” droplets carrying
wild-type alleles. Next, click on the +/ blue button (FAM chan-
nel) to similarly define the cluster of “positive” droplets carrying
mutant alleles. Hovering the mouse over the positive and negative
controls in the well-editor (which highlights their clusters in the 2D
plot) may assist you in defining the boundaries between the differ-
ent fluorescence clusters when multiple samples are concurrently
displayed in the 2D amplitude plots (see Fig. 3, Panel a). Absolute
counts of mutant alleles per μL of sample are displayed in the top
table of the QuantaSoft™ software and can also be visualized as
exportable charts in the Concentration and Ratio tabs or exported,
if desired, as a “.csv” file for additional data analysis. After cluster
identification in 2D plots, data may also be displayed in 1D Ampli-
tude plots which is particularly useful for a side-by-side comparison
showing the quality of cluster separation (appearing as “bands” of
differing amplitudes in 1D plots) when the same assay is performed
at different annealing temperatures (Fig. 1) or supplemented with
“dark probes” (Fig. 2).
Fig. 3 (a) Steps for the calculation of mutant allele abundance and allele ratios during ddPCR experiments
using uniplex single-locus assays. Under “Setup,” select all unknown samples, positive controls, negative
controls, and no-template controls (if available) analyzed with the same assay. Then, in the 2D Amplitude view
under Analyze, use the rectangular Multi-well thresholding selection tool to initially enclose and thereby
designate all droplets as “double-negative” in the 2D plot (panel 1, where all clusters will appear as grey). To
properly identify and define each of the positive and negative clusters, first hover your mouse over the wild-
type negative control(s) in the well-editor which will highlight their clusters in the 2D plot. Then, after clicking
on the green “/+” button, use any of the Multi-well thresholding selection tools to enclose droplets
harboring wild-type alleles (WT) within the selected shape (rectangle, circle, or freehand shape), which will
cause them to be associated with HEX/VIC fluorescence (panel 2). Similarly, after clicking on the blue “+/”
button, hover your mouse over the mutant positive control(s) in the well-editor to help define the cluster of
droplets harboring the mutant allele (MUT), which after they are enclosed within the chosen shape will be
associated with FAM fluorescence (panel 3). (b) Noninvasive monitoring of somatic mutations in the plasma of
cancer patients using uniplex single-locus ddPCR assays. These assays have the potential to simultaneously
quantify, with high precision and sensitivity, somatic mutations (MUT) and wild-type (WT) alleles from liquid
biopsies. One patient showed a significant decrease in circulating tumour DNA levels during the course of
therapeutic treatment (left panel). A different patient (right panel) was refractory to the treatment and ctDNA
levels did not decrease over time. VAF stands for the “variant allele frequency” of rare alleles. For the
calculation of allele fractions, “positive droplets” carrying mutant alleles are assigned to the FAM channel and
“low positive droplets” carrying wild-type alleles are assigned to the HEX channel in 2D Ampitude plots, as
depicted in Panel (a). 1D Amplitude plots, as shown here, are nonetheless useful to compare, side-by-side, the
absolute number of mutant molecules reported across different samples. Please note that the fractional
abundance of each allele can be more informative in cases where different volumes of sample have been used
to perform assay (see the example of the nonresponder patient)
Fig. 4 “Inverted” ddPCR assays enable the uncovering of multiple genetic aberrations affecting a mutation
hotspot by using one single hydrolysis probe matching the sequence of the wild-type allele. WT stands for
droplets containing wild-type alleles of the target of interest. In the example shown below, as many as four
different single nucleotide polymorphisms (MUT1 to MUT4) can be detected and quantified with one single
hydrolysis probe. For this experiment, we used four different DNA samples extracted from fresh tumor
biopsies. Each sample was previously known to carry a different rare allele within the same mutation hotspot.
Fluorescence amplitude signals associated with each particular mutant allele may slightly vary (see for
instance MUT1 versus MUT2). This finding implies that optimum annealing temperatures during thermocycling
may slightly differ depending on the specific mutant allele carried by the sample. For the calculation of allele
fractions “positive droplets” carrying wild-type alleles are assigned to the HEX channel and “low positive
droplets” carrying mutant alleles are assigned to the FAM channel; 2D plots not shown in this figure)
3.6 Nondiscri- Hydrolysis probes with the wild-type sequence of a given locus can
minating, Uniplex be used to detect multiple small genetic variations such as single
“Single-Locus” nucleotide polymorphisms at a hotspot or even single or dual base
Screening Reactions indels. The hydrolysis of imperfectly annealed wild-type
(One Pair of Primers probes generates a secondary fluorescence amplitude band that is
and One Wild-Type- distinguishable from the lower fluorescence amplitude band
Specific Hydrolysis emitted by “double-negative” droplets (i.e., droplets lacking copies
Probe) of the locus of interest). This approach is well suited to screening
for hotspot mutations characterized by multiple genetic abnorm-
alities when the only information required from the assay is
whether a certain hotspot is mutated or not (Fig. 4). These
“inverted” ddPCR assays [27] are ideal to screen for germline
mutations associated with known genetic disorders as well as for
recurrent somatic mutations with allele frequencies above 10%,
though potentially as low as 1%, from fresh tumor biopsies (see
Note 25).
3.6.1 Running Volumes and suggested final concentrations of reagents for the
and Analyzing the Assay preparation of nondiscriminating uniplex “single-locus” ddPCR
assays are displayed in Table 1. Analogous to the procedure used
for Uniplex assays with mutant-specific probe(s) in Subheading 3.5,
quantification of mutant and wild-type alleles here is accomplished
after manually setting thresholds in the QuantaSoft™ software.
Select samples, positive controls, negative controls, and
no-template DNA controls (if available) in the Setup mode. Ensure
that the FAM optical channel has been associated to mutant alleles
and the HEX/VIC optical channel has been associated to the wild-
type allele in the Setup menu and then display the data on the 2D
Amplitude plot after clicking on Analysis. In essence, this approach
assigns “low positive droplets” carrying mutant alleles to the FAM
channel and “positive droplets” carrying wild-type alleles to the
HEX/VIC channel. After clicking on the “/” black button, use
the rectangular Multi-well thresholding selection tool to initially des-
ignate the entire set of droplets as “double-negative” droplets.
Click on the green “/+” button (HEX/VIC channel) to define
the clusters of “positive” droplets for wild-type alleles (i.e., droplets
emitting the most intense fluorescence signal). Click on the blue
“+/” button (FAM channel) to define the clusters of “low posi-
tive” droplets (i.e., droplets containing rare alleles). The process for
allele quantification relies in the same series of steps depicted in
Fig. 3 (Panel a), with the exception that wild-type alleles linked to
the HEX channel will be now represented in the cluster of highest
fluorescence signal and mutant alleles linked to the FAM channel
will be now represented in the cluster of intermediate fluorescence
signal. Allele counts per well and estimated fractional abundances
for each mutant allele will be given on the top table and can be
visualized in the Concentration and Ratio charts and exported as “.
csv” file.
3.7 Noncompeting The possibility to simultaneously quantify mutant and wild-type

Discriminatory Uniplex alleles with one single hydrolysis probe opens the possibility to use
“Dual-Locus” two probes conjugated with dyes of different emission spectra to
Reactions (Two Primer interrogate two different loci in one single assay well (Fig. 5). The
Pairs and Two Rare most direct benefit of such assay formats, beyond savings in overall
Allele-Specific costs, is that splitting precious and scarce samples to run indepen-
Hydrolysis Probes) dent assays for each individual locus can be avoided. Tracking more
than one mutation has also important implications, for example,
when monitoring tumor evolution during the course of therapeutic
interventions in cancer patients [31, 32].
and Analyzing the Assay preparation of discriminating “dual-locus” reactions are displayed
in Table 1. The assay designs, optimization, and ddPCR workflow
are carried out as described in Subheadings 3.2–3.4 above. Quan-
tification of mutant and wild-type alleles is again accomplished after
Fig. 5 Uniplex “Dual-locus” ddPCR reactions using two mutant only hydrolysis probes conjugated with dyes of
different emission spectra to report the presence of rare alleles at two independent genomic locations. The
efficient hydrolysis of mutant-specific probes generates high fluorescence amplitude either in the optical
Channel 1 or Channel 2 for those droplets containing rare alleles. The hydrolysis of imperfectly annealed
probes within droplets harboring wild-type alleles only (green circle) or together with mutant alleles (orange
circles) allows the independent quantification of wild-type alleles. This assay was conducted on 5 ng (roughly)
of a DNA sample extracted from fresh tumour biopsies previously known to harbor rare alleles for each one of
the two interrogated loci. Calculation of allele copies per well and allele fractions at each independent locus, is
done sequentially for the two loci where in each iteration, “positive droplets” carrying mutant alleles are
assigned to the FAM channel (blue droplets) and “low positive droplets” carrying wild-type alleles (green
droplets) are assigned to the HEX channel. This operation is carried out independently for each locus
(illustrated in panels a and b, respectively)
manually setting thresholds in the QuantaSoft™ software. Select

samples, positive controls, negative controls, and no-template
DNA controls in the Setup or Analyze mode and display data on
the 2D Amplitude plot. However, in contrast to uniplex ddPCR
assays, each locus in a dual-locus ddPCR assay interrogating two
independent genomic positions needs to be analyzed
independently and sequentially, if using QuantaSoft™ v1.7.4.0917

or earlier versions, with the data being exported as graphs and/or
as a “.csv” file after analyzing each genomic locus separately. (See
Note 26 for an alternative analysis using newer software.)
Ensure that the two optical channels have been associated to
their corresponding allele type (e.g., wild type or mutant) in the
Setup menu and display the data on the 2D Amplitude plot after
clicking on Analysis. As with uniplex assays above, you will need to
associate “positive droplets” carrying mutant alleles with the FAM
channel and “low positive droplets” carrying wild-type alleles with
the HEX/VIC channel, doing so iteratively for each individual
locus. For the locus 1 target analysis, first set the contents of the
entire plot as “/” double negative (i.e., grey) droplets. Use the
Multi-well thresholding selection tool and click on the green “/+”
button (HEX/VIC channel) to define the clusters of “low positive”
droplets carrying wild-type alleles of locus 1. Include here droplets
with mixed content (i.e., locus 1 “low positive”/locus 2 “low
positive” or locus 1 “low positive”/locus 2 “positive,” see Fig. 5).
Click on the blue “+/” button (FAM channel) to define the
clusters of “positive” droplets for locus 1, also including the dro-
plets of mixed content (i.e., locus 1 “positive”/locus 2 “low posi-
tive” or locus 1 “positive”/locus 2 “positive”). Allele counts and
estimated allele frequencies for locus 1 mutations will be given on
the top table and in the Concentration and Ratio charts.
After exporting locus 1 data, repeat the same set of operations
to estimate the copies/well and fractional abundance of mutant and
wild-type alleles of locus 2. You must redefine the new target ID
associated with locus 2 in the Setup Menu unless generic “Mutant”
or “Wild type” names are always associated with the FAM or
HEX/VIC channel, respectively. Locus ID should be included in
the name of the file exported from QuantaSoft™ in order not to
confuse from which locus the exported “.csv” file data was derived.
3.8 Inverted, Two different mutation hotspots can be screened for single nucle-
Nondiscriminating otide genetic variants by using two wild-type-specific probes conju-
Uniplex “Dual-Locus” gated with dyes of different emission spectra. Hence, a substantial
Reactions (Two Primer number of mutations can be investigated in one single assay
Pairs and Two Wild- (Fig. 6). Nondiscriminating “dual-locus” “inverted” ddPCR assays
Type-Specific have the same limitations in terms of sensitivity and specificity as
Hydrolysis Probes) uniplex “inverted” ddPCR assays (see Note 25).
and Analyzing the Assay preparation of nondiscriminating “dual-locus” reactions are dis-
played in Table 1. Assay designs, optimization, and workflow are
as described in Subheadings 3.2–3.4.
Quantification of mutant and wild-type alleles is accomplished
after manually setting thresholds in the QuantaSoft™ software.
Fig. 6 Uniplex “Dual-locus” “inverted” ddPCR reactions with two hydrolysis probes matching only the wild-
type alleles of two independent loci. These assays allow the simultaneous quantification of multiple mutant
and the wild-type alleles using hydrolysis probes conjugated with two dyes (one per locus) detected in either
the optical Channel 1 or Channel 2 in a single well. As many as nine different fluorescent clusters can be
observed in the example shown in this figure. This assay was conducted on 10 ng (roughly) of a DNA sample
extracted from fresh tumour biopsies with a priori known mutations for each one of the two mutation hotpots
interrogated. For the calculation of allele copies and fractions at each independent locus, “positive droplets”
carrying wild-type alleles (green droplets) are assigned to the HEX/VIC channel and “low positive droplets”
carrying mutant alleles are assigned to the FAM channel (blue droplets). This operation is carried out
independently for each locus (illustrated in panels a and b, for locus 1 and 2 respectively). Droplets containing
copies of multiple alleles (for the same or different loci) are indicated by orange circles
Select samples, positive controls, negative controls, and

no-template DNA controls in the Setup or Analyze mode and
display data on the 2D Amplitude plot. Nondiscriminating “dual-
locus” ddPCR assays need to be analyzed for each locus indepen-
dently and sequentially, when using QuantaSoft™ v1.7.4.0917 or
earlier versions. For each sequentially analyzed locus, ensure that
the two optical channels have been have been associated to its
corresponding allele type (e.g., wild type or mutant) in the Setup
menu and display the data on the 2D Amplitude plot after clicking
on Analysis. For this approach, assign “positive droplets” carrying
wild-type alleles to the HEX channel and “low positive droplets”
carrying mutant alleles to the FAM channel. First, however, set the
contents of the entire plot as “/” double negative (i.e., grey)
droplets. Use the Multi-well thresholding selection tool and click on
the green “/+” button (HEX channel) to define the clusters of
“positive droplets” carrying wild-type alleles of locus 1. Include
droplets with mixed content containing either wild-type or mutant
alleles of locus 1 (see Fig. 6). Click on the blue “+/” button (FAM
channel) to define the clusters of “low positive” droplets for locus
1 (i.e., droplets carrying rare alleles), also including the droplets of
mixed content carrying mutant alleles for locus 1. Allele counts and
estimated fractional abundances for locus 1 will be provided on the
top table and in the Concentration and Ratio charts, as well as in the
exportable “csv” file. Repeat the same set of operations to estimate
the copies/well and fractional abundance of mutant and wild-type
alleles pertaining to locus 2. You must redefine the new target ID
associated with locus 2 in the Setup Menu unless generic “Mutant”
or “Wild type” names are always associated with the FAM or
HEX/VIC channel, respectively. Locus ID should be included in
the name of the file exported from QuantaSoft™ in order not to
confuse from which locus the exported “.csv” file data was derived.
See Note 26 for an alternative analysis using newer software.
3.9 Higher Order, Mutations within a given hotspot can be individually distinguished
Single-Locus by using exclusively mutant-specific hydrolysis probes conjugated
“Amplitude” with one or the other of two dyes (here, FAM and HEX) applied at
Multiplexing (One Pair different final concentrations in the end-point PCR. Allele multi-
of Primers and Several plexing in this way has advantages over conventional duplex rare
Mutant Allele-Specific, mutation detection approaches due to the fact that wild-type-spe-
But No Wild Type, cific hydrolysis probes are not required. Yet as previously explained,
Hydrolysis Probe) these wild-type alleles can still be detected and quantified due to
their imperfect hybridization to mutant probe(s) yielding droplets
with an intermediate dual fluorescence signal (green clusters in
Fig. 7).
3.9.1 Running Table 1 shows recommended starting concentration ratios for up to

and Analyzing the Assay three mutant-specific probes conjugated with the same fluorescent
dye (typically 6-FAM™). These recommendations are only sugges-
tive and we recommend slightly altering the final concentration of
each probe and conducting a temperature gradient until a satisfac-
tory separation of each individual cluster is achieved. Quantification
of mutant and wild-type alleles is accomplished after manually
setting thresholds in the QuantaSoft™ software. First, select sam-
ples, positive controls, negative controls, and no-template DNA
controls in the Setup view. Ensure that the two optical channels
have been associated to their corresponding target types (i.e.,
mutant or wild type) in the Setup menu and then after clicking on
Analysis, display the data on the 2D Amplitude plot. All mutant
alleles, regardless of the final probe concentration or dye used to
Fig. 7 Multiplex detection of rare alleles using a single pair of primers and several mutant hydrolysis probes
labeled with two different fluorophores, used at customizable final concentrations. In the upper panel, four
mutant-specific probes were used together to interrogate a mixture of genomic DNA extracted from four
independent tumor biopsies, each of them harboring a different mutation. Three hydrolysis probes were
labeled with 6-FAM™ and a fourth probe was labeled with HEX™. The three probes conjugated with
6-FAM™ can be associated with three well-defined fluorescent clusters (blue), according to the final
concentration of each individual probe in the ddPCR reaction. The hydrolysis of the fourth probe labeled
with HEX generates high fluorescence in Channel 2, but it is also assigned to FAM fluorescence for proper
quantification purposes. The wild-type allele (green cluster) can be detected and independently quantified
owing to the nonspecific hydrolysis of both 6-FAM™ and HEX™ probes within droplets exclusively containing
wild-type alleles. When applied to a sample carrying one single mutation (bottom panels), these assays
generate three different fluorescence clusters (one for “positive” droplets (blue clusters), one for “low
positive,” intermediate droplets (green clusters), and one for “double-negative” droplets (grey clusters);
samples negative for mutations (not shown) only exhibit two fluorescence clusters: “low positive,” intermedi-
ate droplets plus “double-negative” droplets)
target each particular allele, will be associated in the Setup menu

with the FAM channel. Only wild-type alleles (i.e., those contained
within droplets in the green cluster emitting a mixture of FAM and
HEX fluorescence signals as seen in Fig. 7) will be associated with
the HEX channel (in contrast to conventional QuantaSoft use
where a dual-dye cluster would be labeled in orange droplets).
To define clusters, first, set the contents of the entire plot as
“/” double negative (i.e., grey) droplets. Use the freehand lasso
Multi-well thresholding selection tool and click on the blue “+/”
button (FAM channel) to define the cluster(s) of “positive” dro-
plets carrying mutant alleles. (Typically, there will only be one such
mutant cluster in a given sample; here, multiple samples each with a
different mutant allele were combined). Click on the green “/+”

button (HEX channel) to define the clusters of “low positive,
intermediate” droplets carrying just wild-type alleles. Mutant and
wild-type allele counts per well and estimated fractional abundance
are given on the top table and in the Concentration and Ratio
charts. Note that this assay strategy does not require iterative analy-
sis as with the dual locus assay approaches above (Subheadings 3.7
and 3.8).
4 Notes
1. It is critical to not confuse this with the ddPCR™ Supermix for

Probes (Bio-Rad, Hercules, CA, USA; Cat. No. # 186-3010,
186-3026, 186-3027, 186-3028), which contains dUTP. We
have observed that the latter does not provide proper cluster
resolution for the discriminating and nondiscriminating assays
here described. Consistently using fresh ddPCR™ Supermix
for Probes (no dUTP) is also critical for optimal assay resolu-
tion. The ddPCR™ Supermix for Probes (no dUTP) is stable,
according to the manufacturer, for at least 6 months from the
time of receipt when stored at 20 C. Store at 4 C for a
maximum of 2 weeks after thawing. The manufacturer does not
recommend multiple freeze/thaw cycles as this may affect assay
resolution, or using reagents beyond their expiration date.
2. As with ddPCR™ Supermix for Probes (no dUTP), old or
expired digital PCR droplet generation oil for probes can result
in nonoptimal assay resolution. Always check the oil bottle for
any signs of microbial proliferation or turbidity.
3. Spoiled or expired droplet reader oil can also generate subopti-
mal results.
4. Place the cartridges with the green sections facing to the right
of the automated droplet generator. Ensure not to lift the
DG32™ cartridges from their compartment while still in use.
If lifted and placed back again, the automated droplet genera-
tor will consider it as a new cartridge and will reuse some of the
sections. This will cause droplet generation errors and/or sam-
ple cross-contamination.
5. Empty waste bin when necessary.
6. Place the foil seal on top of the 96-well plate to be covered with
the red line facing upward.
7. Ensure that the color of the precooled 20 C plate cooler
indicates the right temperature. These plates can maintain the
temperature of samples at 0 C for at least 1 h and then change
their color when their temperature exceeds 7 C.
8. Ensure that the levels and quality of the droplet generation oil
for probes, pipette tips, and DG32™ cartridges are ready and
well positioned to start the run. Be sure to use Droplet Gener-
ation Oil for Probes and, if necessary, replace any Automated
Droplet Generation Oil for EvaGreen (suited for a different
application) that might have been left over in the Auto DG
from previous uses. It is critical to set up droplet digital PCR
reactions in a final volume of 22 μL (as opposed to the 20 μL
recommended in some of the manufacturer’s protocols) to
minimize the occurrence of errors during automated droplet
generation. The later can result in low droplet counts and/or
sample loss.
9. Inspect the levels of the droplet reader oil and waste bins before
starting a run and replace if necessary. For example, a blinking
green light indicates that the waste bin is about to be
completely filled. A steady orange light indicates that the
waste bin needs to be immediately replaced. If the instrument
has not been used for more than 2 weeks it is recommended to
run a prime-flush-prime cycle to remove old oil from the
internal fluidic channels of the instrument.
10. Preheat the instrument during droplet generation and do not
use it until the temperature has reached 180 C. Remove the
block from inside the instrument before the prewarming step
starts.
11. Thermal cycling programs should preferably be preloaded in
the instrument before starting a run. Other thermocyclers
fitting the 96-well plates and specifications described above
may also be compatible with this protocol, but should be tested
for temperature uniformity and accuracy.
12. The methods described in this chapter have been validated with
hydrolysis probes that were manufactured as described above.
We recently evaluated the performance of custom ddPCR
probe assays manufactured by Bio-Rad (Bio-Rad, Hercules,
CA, USA; Cat. No. #1864011) and results were also satisfac-
tory. The employ of other types of fluorescence dyes or dark
quenchers has not been evaluated, nor the suitability of other
fluorophore-labelled oligonucleotides such as Scorpion or
Amplifluor™ primer-probes or hybridization probes such as
molecular beacons.
13. When pipetting DNA extracts, it is important to use high-
quality tips that do not shed microparticles. Microparticles
are known to clog the microfluidic chambers of the DG32™
droplet generation cartridge and hamper droplet generation.
14. Four-base cutters and high fidelity restriction endonucleases
insensitive to DNA methylation are preferred. Approximately
2–5 units of enzyme previously diluted in an appropriate
enzyme dilution buffer can often be directly added to the

ddPCR reaction. Enzymatic fragmentation can be also carried
out prior to ddPCR according to the manufacturer’s protocol.
Heat inactivation or postdigestion cleanup is not required in
this case, but it is important to not heat up the sample above
65 C and to ensure at least a tenfold dilution of the restriction
enzyme buffer in the final ddPCR reaction since high salt
content can inhibit the ddPCR reaction and should be avoided.
15. Uracil glycosylase treatment is recommended to reduce false
positives if working with FFPE samples.
16. Some manufacturers provide useful tools for the design of
hydrolysis probe-based assays, such as PrimerQuest®
(Integrated DNA technologies, Coralville, IA, USA; http://
www.idtdna.com/Primerquest/Home/Index), and Oli-
goArchitect™ (Sigma-Aldrich, St. Louis, MO, USA; https://
www.sigmaaldrich.com/technical-documents/articles/biology/
probe-design-services.html) or Multiple Primer Analyzer
(Thermo Fisher Scientific, Waltham, MA, USA). Primer3Plus
(http://www.bioinformatics.nl/cgi-bin/primer3plus/primer
$ 3plus.cgi/), Oligo Calc (http://biotools.nubic.northwest
ern.edu/OligoCalc.html) and e-PCR (https://www.ncbi.
nlm.nih.gov/tools/epcr/) are also popular tools to evaluate
oligonucleotide properties and assay specificity. It must be
noted that some of these tools might be too stringent to
nominate a candidate assay. A design that takes into account
the vast majority (but perhaps not all) of the abovementioned
recommendations could therefore arise as the only alternative
in such situations.
17. Variable final concentrations of individual probes conjugated
with the same fluorophore in the ddPCR reaction enable dis-
crimination of several alleles, thus facilitating high order
multiplexing.
18. There are no lower limits for the amount of input DNA used
during a single ddPCR assay but this amount will determine
both the precision and sensitivity of the assay. One may antici-
pate the expected number of copies for a given locus by con-
sidering the mass of the haploid human genome (~3.3 pg). Up
to 150 ng of human genomic DNA can be added in a single
well for a final concentration of 2000 target copies/μL but
particular care must be paid to the accurate classification of a
substantial number of droplets with mixed content. Assays
more affected by droplet misclassification and background
noise, such as inverted ddPCR assays or assays conducted
over damaged FFPE samples, may benefit from lower amounts
of input DNA per well and the use of multiple replicate wells to
increase the total number of genome equivalents analyzed per

sample.
19. If using less than eight samples for a given run, do not leave any
column wells empty or with a volume lower than 22 μL. Fill
such wells with 22 μL of the 1 ddPCR™ Supermix for probes
(no dUTP) or 1 ddPCR™ Buffer Control for Probes. The
same applies for any multiple of 8 samples, i.e., do not leave any
well empty for a given column selected for droplet generation
as this will introduce air into the instrument’s fluidics and
immediately abort the process. Similarly, any air bubbles
observed in DG8 sample wells should be dislodged and
removed with a pipette tip to avoid their interfering with
proper droplet formation.
20. Successful droplet generation can be visually inspected by the
observation of a cloudy layer (droplets) on top of a transparent
layer (oil). Failures during droplet generation translate into low
numbers of droplets, small volumes in the collection plates
(mostly oil) and usually result in sample loss.
21. Thermocycled emulsions of droplets can be kept at 4 C for at
least 24 h.
22. Our previous study targeting recurrent point mutations in
B-cell lymphomas [27] showed a high concordance between
uniplex and standard duplex ddPCR assays when applied, for
example, to the same serial dilution of cell line genomic DNA.
We also observed a high concordance between the allele fre-
quencies inferred via uniplex ddPCR assays and those inferred
by NGS methods like deep amplicon sequencing or targeted
hybridization capture in fresh tumor biopsies. Uniplex ddPCR
assays demonstrate exquisite sensitivity as well, being capable of
detecting one single mutant DNA copy diluted in a back-
ground of 10,000 wild-type DNA molecules in one single
well. Such extreme situations must nonetheless be confirmed
by conducting assay replicates in both the interrogated sample
and negative controls (see examples in ref. 27 and Chapter 3 by
Tzonev, this volume). Uniplex discriminating ddPCR assays
can therefore be used, for example, to screen for rare genetic
variants associated with disease and hold promise for the non-
invasive detection and quantification of rare genetic variants in
cell-free DNA samples (Fig. 3). One of the downsides of these
assays, however, relates to confidently identifying droplets con-
taining both mutant and wild-type alleles. This phenomenon
may lead to the underestimation of wild-type alleles at higher
DNA input amounts due to the increased frequency of double-
positive droplets which may not be distinguishable from
mutant-only positive droplets. However, this should not affect
the total count of mutant alleles per volume of sample. Serial
dilutions of cell line genomic DNA have also demonstrated a

strong concordance between observed and expected allele fre-
quencies during uniplex ddPCR assays, suggesting that the
possible underestimation of wild-type alleles because of dro-
plets of mixed content should not have a dramatic effect on the
calculation of allele ratios when DNA input amounts are, for
example, below 10 ng (equivalent to around 3333 human
haploid genome equivalents per well). Another important lim-
itation is that other rare but closely related alleles present in the
sample might be confounded with wild-type alleles if there is a
mismatch(es) between the mutant-specific probe and the DNA
sequence of the second mutant allele.
23. It is strongly recommended to test optimum annealing tem-
peratures for cluster differentiation on positive controls with a
priori known allele ratios to determine if the allele counts
associated with a certain fluorescence cluster are excessive
(e.g., at lower temperatures) or deficient (e.g., at higher tem-
peratures), possibly suggesting underly or overly stringent
annealing conditions, respectively.
24. In some instances, it can be helpful to add a 30 spacer modifica-
tion to the “dark probe” to avoid the emergence of additional
fluoresce amplitude bands nearby the cluster of empty droplets.
30 C3 spacers (Integrated DNA technologies, Coralville, IA,
USA) would represent an example of this type of chemical
modifications and their main role is to avoid the extension of
the probe by DNA polymerases.
25. The sensitivity and specificity of “inverted” ddPCR assays on
samples with very low frequencies of mutant DNA, such as
those expected in liquid biopsies, is compromised by the
ddPCR “rain” effect (i.e., droplets emitting intermediate levels
of fluorescence that are actually wild-type positive) and is
therefore not recommended in situations where high sensitivity
is needed. The quality of DNA across FFPE samples is also
expected to be highly variable and associated with different
levels of rain, thus precluding the use of “inverted” assays, for
instance, in samples where DNA is too damaged or contains a
significant amount of PCR inhibitors. Assay redesign, extended
cycling regimes and DNA shearing have nonetheless been
shown to lessen “rain” during ddPCR analyses [33] and the
effect of these parameters on “inverted” ddPCR assays should
be further explored. The analysis of positive, negative (wild
type), and no-template DNA controls are also crucial here, as
in Subheading 3.5.4, for setting up appropriate thresholds and
manual clusters and for adequately identifying and classifying
“rain” droplets. The concurrent presence of wild-type and
mutant alleles within the same droplet may obscure the fluo-
rescent amplitude signal associated with mutant alleles in these
assays, a feature that also limits the sensitivity and maximum

useful DNA input per well of “inverted” ddPCR when com-
pared to assays relying on mutant-specific probes.
26. Simultaneous quantification of more than one genomic locus
(and >4 clusters at a time) can now be performed using the
“Advanced User Options” of the new QuantaSoft™ Analysis
Pro software, available from http://www.bio-rad.com/en-us/
sku/quantasoft-analysis-pro-v1-quantasoft-analysis-pro-soft
ware. See Bio-Rad Bulletin #6827 for more details.
Acknowledgments
This study was supported by the Canadian Institute for Health

Research (CIHR) (New Investigator award and operating grant
300738), the Terry Fox Research Institute (projects #1043 and
#1021), the Natural Sciences and Engineering Research Council of
Canada (Research Tools and Instruments program EQPEQ 1501),
and the BC Cancer Foundation.
References
1. Everett TR, Chitty LS (2015) Cell-free fetal molecular diagnosis directly from blood. Clin
DNA: the new tool in fetal medicine. Ultra- Microbiol Infect 21:323–331
sound Obstet Gynecol 45:499–507 9. Prachayangprecha S et al (2014) Exploring the
2. Liao GJW, Gronowski AM, Zhao Z (2014) potential of next-generation sequencing in
Non-invasive prenatal testing using cell-free detection of respiratory viruses. J Clin Micro-
fetal DNA in maternal circulation. Clin Chim biol 52:3722–3730
Acta 428:44–50 10. Fu W et al (2015) A highly sensitive and specific
3. Bettegowda C et al (2014) Detection of circu- method for the screening detection of geneti-
lating tumor DNA in early- and late-stage cally modified organisms based on digital PCR
human malignancies. Sci Transl Med without pretreatment. Sci Rep 5:12715
6:224ra24 11. Reid AL, Freeman JB, Millward M, Ziman M,
4. Heitzer E, Ulz P, Geigl JB (2015) Circulating Gray ES (2015) Detection of BRAF-V600E
tumor DNA as a liquid biopsy for cancer. Clin and V600K in melanoma circulating tumour
Chem 61:112–123 cells by droplet digital PCR. Clin Biochem
5. Volik S, Alcaide M, Morin RD, Collins CC 48:999–1002
(2016) Cell-free DNA (cfDNA): clinical signif- 12. Mukaide M et al (2014) High-throughput and
icance and utility in cancer shaped by emerging sensitive next-generation droplet digital PCR
technologies. Mol Cancer Res 14(10):898. assay for the quantitation of the hepatitis C
https://doi.org/10.1158/1541-7786.MCR- virus mutation at core amino acid 70. J Virol
16-0044 Methods 207:169–177
6. Ramı́rez-Castillo FY et al (2015) Waterborne 13. Kennedy SR et al (2014) Detecting ultralow-
pathogens: detection methods and challenges. frequency mutations by Duplex Sequencing.
Pathogens 4:307–334 Nat Protoc 9:2586–2606
7. Whale AS et al (2016) Detection of rare drug 14. Newman AM et al (2016) Integrated digital
resistance mutations by digital PCR in a human error suppression for improved detection of
influenza A virus model system and clinical circulating tumor DNA. Nat Biotechnol
samples. J Clin Microbiol 54:392–400 34:547–555
8. Opota O, Jaton K, Greub G (2015) Microbial 15. Vogelstein B, Kinzler KW (1999) Digital PCR.
diagnosis of bloodstream infection: towards Proc Natl Acad Sci U S A 96:9236–9241
16. Dong L et al (2015) Comparison of four digi- 25. Findlay SD, Vincent KM, Berman JR, Postovit
tal PCR platforms for accurate quantification of L-M (2016) A digital PCR-based method for
DNA copy number of a certified plasmid DNA efficient and highly specific screening of
reference material. Sci Rep 5:13174 genome edited cells. PLoS One 11:e0153901
17. Dingle TC, Sedlak RH, Cook L, Jerome KR 26. Taly V et al (2013) Multiplex picodroplet digi-
(2013) Tolerance of droplet-digital PCR versus tal PCR to detect KRAS mutations in circulat-
real-time quantitative PCR to inhibitory sub- ing DNA from the plasma of colorectal cancer
stances. Clin Chem 59:1670–1672 patients. Clin Chem 59:1722–1731
18. Rački N, Dreo T, Gutierrez-Aguirre I, Blejec A, 27. Alcaide M et al (2016) Multiplex droplet digital
Ravnikar M (2014) Reverse transcriptase drop- PCR quantification of recurrent somatic muta-
let digital PCR shows high resilience to PCR tions in diffuse large B-cell and follicular lym-
inhibitors from plant, soil and water samples. phoma. Clin Chem 62(9):1238–1247. https://
Plant Methods 10:42 doi.org/10.1373/clinchem.2016.255315
19. Hirschhorn JN, Lohmueller K, Byrne E, 28. Sikora A et al (2010) Detection of increased
Hirschhorn K (2002) A comprehensive review amounts of cell-free fetal DNA with short PCR
of genetic association studies. Genet Med amplicons. Clin Chem 56:136–138
4:45–61 29. Andersen RF, Spindler K-LG, Brandslund I,
20. Chang MT et al (2016) Identifying recurrent Jakobsen A, Pallisgaard N (2015) Improved
mutations in cancer reveals widespread lineage sensitivity of circulating tumor DNA measure-
diversity and mutational specificity. Nat Bio- ment using short PCR amplicons. Clin Chim
technol 34:155–163 Acta 439:97–101
21. Hawkey PM (2008) The growing burden of 30. Miyaoka Y et al (2016) Systematic quantifica-
antimicrobial resistance. J Antimicrob Che- tion of HDR and NHEJ reveals effects of locus,
mother 62(Suppl 1):i1–i9 nuclease, and cell type on genome-editing. Sci
22. Madic J et al (2016) Three-color crystal digital Rep 6:23549
PCR. Biomol Detect Quantif 10:34. https:// 31. Pantel K, Diaz LAJ, Polyak K (2013) Tracking
doi.org/10.1016/j.bdq.2016.10.002 tumor resistance using ‘liquid biopsies. Nat
23. Whale AS, Huggett JF, Tzonev S (2016) Fun- Med 19:676–677
damentals of multiplexing with digital PCR. 32. Barber LJ, Davies MN, Gerlinger M (2015)
Biomol Detect Quantif 10:15. https://doi. Dissecting cancer evolution at the macro-
org/10.1016/j.bdq.2016.05.002 heterogeneity and micro-heterogeneity scale.
24. Bidshahri R et al (2016) Quantitative detection Curr Opin Genet Dev 30:1–6
and resolution of BRAF V600 status in colo- 33. Lievens A, Jacchia S, Kagkli D, Savini C, Querci
rectal cancer using droplet digital PCR and a M (2016) Measuring digital PCR quality: per-
novel wildtype negative assay. J Mol Diagn formance parameters and their optimization.
18:190–204 PLoS One 11:e0153317
Chapter 17
Identification and Use of Personalized Genomic Markers

for Monitoring Circulating Tumor DNA
Yilun Chen, Anthony M. George, Eleonor Olsson, and Lao H. Saal
Abstract
Digital PCR techniques are ideally suited for accurately quantifying trace amounts of target DNA
sequences, such as tumor-derived mutant DNA that is present in the blood circulation of patients with
cancer. Here, we describe an approach marrying low-coverage whole-genome sequencing of tumor tissues,
to enumerate chromosomal rearrangement breakpoints, together with droplet digital PCR (ddPCR)-based
personalized rearrangement assays to cost-effectively monitor circulating tumor DNA levels at multiple
time-points during the clinical course. The method is generally applicable to essentially any cancer patient,
as all cancers harbor unstable genomes, and may have uses for measuring minimal residual disease, response
to therapy, and early detection of metastasis.
Key words Cell-free circulating tumor DNA, Personalized medicine, Liquid biopsy, Noninvasive
diagnosis, Whole-genome sequencing, Droplet digital PCR
1 Introduction
Nuclear and mitochondrial DNA originating from normal and

diseased cells can enter the blood stream by several processes and
be found in plasma or serum as short (<200 bp) fragments, termed
cell-free circulating DNA (cfDNA), at concentrations ranging from
5 to 20 ng/ml plasma in healthy persons to >1000 ng/ml plasma
in advanced cancer patients. Although tumor-derived DNA (circu-
lating tumor DNA; ctDNA) was first described decades ago, due to
technological limitations, ctDNA has only recently emerged as a
powerful biomarker in cancer research and clinical oncology
[1–3]. The half-life of ctDNA can be measured in minutes to
hours; therefore the quantity of ctDNA measured at any given
time can be used as a snapshot of the cancer burden. In several
cancer types, the levels of ctDNA has been shown to be associated
with tumor progression [4–7] and risk of metastasis and death
[8, 9]. Furthermore, quantification of ctDNA may be used to
303
304 Yilun Chen et al.
monitor response to therapy, identify resistance mutations, and as a

noninvasive companion diagnostic.
A number of features can be used to distinguish ctDNA from
normal wild-type DNA in cfDNA, the most specific being somatic
mutations and genome rearrangements [5, 10]. The fraction of
cfDNA that originates from cancer cells can vary from less than
1% in early stage cancer to 50% in advanced metastatic disease.
Irrespective of the ctDNA fraction, any given mutated or rear-
ranged fragment may be present at exceedingly low abundancy
due to tumor heterogeneity/subclonality (e.g., mutant allele frac-
tions <0.01%) [4, 11]. Moreover, the volume of blood plasma
available for analysis may be limiting factor. For these reasons,
highly sensitive and specific quantitative methods are required for
accurate ctDNA analysis. Because most detection approaches use
polymerase enzymes with inherent base misincorporation error
rates that can introduce a single nucleotide sequence variant of
interest by chance, genome rearrangements such as chromosomal
rearrangements are currently the most exquisitely specific choice of
ctDNA biomarker. Additionally, chromosomal rearrangements are
excellent tumor-tracking biomarkers for breast cancer because
many are formed early in tumorigenesis during a period of telomere
crisis and breakage-fusion-bridge cycling in the originating neo-
plastic clone, and therefore these rearrangements are “trunk” mar-
kers that remain present in the majority of subsequent subclones
[12–14].
Here, we describe an approach marrying low-coverage whole-
genome sequencing of tumor tissues, to enumerate chromosomal
rearrangement breakpoints, together with droplet digital PCR
(ddPCR)-based personalized rearrangement assays to cost-
effectively monitor circulating tumor DNA levels at multiple
time-points [8].
2 Materials
2.1 Tissue 1. Tumor tissue (fresh-frozen or frozen after RNAlater

Procurement preservation).
and Preparation 2. RNAlater (Ambion).
Components
3. Disposable sterile scalpels.
4. Plastic petri dishes.
5. Normal tissue sample, e.g., peripheral blood lymphocytes.
2.2 Tumor 1. AllPrep DNA/RNA Mini kit (Qiagen).

and Normal DNA 2. QIAshredder columns (Qiagen).
Isolation Components
3. 2-mercaptoethanol.
4. DX-Antifoaming reagent (Qiagen).
Monitoring ctDNA using Personalized ddPCR Biomarkers 305
5. QIAcube instrument (Qiagen).

6. QIAcube Rotor adapters (Qiagen).
7. QIAcube Reagent bottles (Qiagen).
8. Wizard Genomic DNA Purification kit (Promega) or QIAamp
DNA Blood Mini Kit (Qiagen).
2.3 Sequencing 1. Covaris tubes (Covaris).

Library Preparation 2. Covaris S220 Focused ultrasonicator (Covaris).
Components
3. 2100 Bioanalyzer instrument (Agilent).
4. High Sensitivity DNA Analysis Kit (Agilent).
5. Illumina TruSeq DNA Sample Preparation Kit (Illumina).
6. Agencourt AMPure XP Beads (Beckman Coulter).
7. Freshly Prepared 80% ethanol in water.
8. PCR grade water.
9. Thermal cycler.
10. Magnetic stand (Life Technologies).
11. 50 Tris–acetate–EDTA Buffer (Prepare 1 TAE buffer
[40 mM Tris, 20 mM acetic acid, 1 mM EDTA, pH 8.0]
with distilled water before use).
12. 100 bp DNA Ladder (Life Technologies).
13. 50 bp DNA Ladder (Life Technologies).
14. Disposable sterile scalpels.
15. Ultrapure Agarose 1000 (Life Technologies).
16. 6 DNA Gel Loading Dye (Thermo Fisher Scientific).
17. MinElute Gel Extraction Kit (Qiagen).
18. Benchtop vortex mixer.
19. SyBr Gold Nucleic Acid Gel Stain (Life Technologies).
20. Qubit Fluorometer (Thermo Fisher Scientific).
21. Qubit dsDNA HS Assay Kit (Thermo Fisher Scientific).
22. HiSeq 2000 or 2500 sequencing instrument (Illumina).
23. cBot instrument (Illumina).
24. TruSeq DNA Sample Preparation Kit (Illumina).
25. TruSeq Cluster Kit cBot HS (Illumina).
26. PE Flow Cells (Illumina).
27. TruSeq SBS Kit HS (Illumina).
2.4 Cell-Free DNA 1. Purple K2EDTA Vacutainer (Becton Dickinson), Cell-Free

Collection DNA BCT (Streck), or other suitable blood collection tube.
and Extraction 2. Disposable Pasteur pipettes.
Components
3. DPBS without Mg2+/Ca2+.
4. QIAamp UltraSens Virus Kit (Qiagen).

5. Benchtop vortex mixer.
6. One variable-speed thermomixer at 60 C, 1.5 ml.
7. One variable-speed thermomixer at 40 C, 2.0 ml.
8. Refrigerated centrifuge at 4 C.
9. Centrifuge at room temperature.
10. DNA LoBind tubes, 1.5 and 2.0 ml (Eppendorf).
2.5 Circulating 1. Phusion Master Mix (Thermo Fisher Scientific).

Tumor DNA Detection 2. Caliper LabChip XT System (PerkinElmer).
Components
3. QX100/200 Droplet Digital PCR system (Bio-Rad).
3 Methods
3.1 Tissue The methods for tissue handling and preparation can vary. Primary
Procurement tumor tissue may be obtained from a biopsy or surgical specimen.
and Preparation For biopsies, fine needle aspirates may not always yield sufficient
tissue and therefore core biopsy or larger is recommended. Highest
quality nucleic acids are obtained from well-handled tissues with
minimal ischemic time prior to fresh-freezing to at least 80 C or
preserved in RNAlater prior to freezing. We recommend obtaining
paired normal tissue (see Note 1), such as peripheral blood lym-
phocytes from the buffy coat fraction of a blood collection tube.
Tissue preserved by formalin fixation and paraffin embedding
(FFPE) can be used, but will result in lower-yield and quality of
DNA, as well as increased potential for sequencing artifacts (typi-
cally C > T base substitutions).
1. Complete details of our routine tumor tissue preparation is
described elsewhere [15]. In brief, we typically use fresh-frozen
(to at least 80 C) tumor tissue or tumor specimens preserved
in RNAlater prior to freezing. Using a disposable scalpel and
petri dish, the tumor sample is partitioned into three parts, one
10–30 mg piece for isolation of nucleic acids, one ~10 mg piece
for construction of tissue macroarray, and if remaining, the
remainder piece is saved for future research. Samples are frozen
at 80 C until use.
2. For normal tissue sample, we typically use peripheral blood
lymphocytes. The buffy coat fraction can be obtained following
centrifugation of a standard EDTA blood collection tube at
2000 g for 10 min at 4 C, transfer of the top plasma fraction
to separate 1.5 ml aliquots using a sterile Pasteur pipette,
followed by transfer of the buffy coat fraction to a separate
tube (see Note 2). The remaining red blood cell fraction may be
discarded. Store fractions at 80 C.
3.2 Tumor Any standard method for purification of genomic DNA may be
and Normal DNA used. As described elsewhere [8, 15], we prefer the AllPrep
Isolation DNA/RNA Kit (Qiagen) for simultaneous purification of genomic
DNA and total RNA from tumor tissues, and the Wizard Genomic
DNA Purification Kit (Promega) or QIAamp DNA Blood Mini Kit
(Qiagen) for isolation of normal blood DNA from peripheral blood
lymphocytes. Standard manufacturer protocols should be followed,
and the isolated nucleic acids stored frozen at 80 C until use.
3.3 Sample The sample library preparation for sequencing is performed using
Preparation the TruSeq DNA Sample Preparation Kit (Illumina) according to
for Sequencing the TruSeq Sample preparation guide (Part # 15005180 Rev. A)
using the low-throughput (LT) protocol (see Note 3). We follow
the standard protocol, with the following changes as detailed in
Subheadings 3.3.1–3.3.5 below.
3.3.1 Fragmentation 1. Add 2.4 μg per sample of genomic DNA to individual Covaris
of DNA tubes in a total volume of 120 μl.
2. Fragment the genomic DNA using the S220 Focused Ultra-
sonicator instrument (Covaris). For improved physical
sequence coverage, we shear the genomic DNA to ~700 bp
average length using the following settings: duty cycle 5%,
intensity 3, cycles per burst 200, time 30 s at 5 C. Samples
may be stored at 20 C.
3. Analyze the distribution of fragment lengths after shearing on a
2100 Bioanalyzer instrument (Agilent) using the High Sensi-
tivity DNA Analysis Kit (Agilent) before continuing with
library preparation. Successfully sheared genomic DNA will
have a size distribution from 300 to 2000 bp with peak at
900 to 1000 bp and a concentration >500 pg/μl.
4. End repair, purification, adenylation, ligation of adaptors, and
purification after ligation are performed following standard
Illumina TruSeq protocol (Part # 15005180 Rev. A). Libraries
may be stored at 20 C after either purification step.
3.3.2 Size Separation 1. Ligated sequencing libraries are size separated by electropho-
After Ligation resis using a 2% agarose gel made with Ultrapure Agarose 1000
(Life Technologies) and 1 Tris–acetate–EDTA (TAE) buffer.
Heat gel mixture in microwave, let cool for 5 min, and add 1 μl
Sybr Gold Nucleic Acid Gel Stain (Life Technologies) per
10 ml gel and swirl to mix before pouring; let gel set.
2. Prepare samples, 20 μl after purification, by adding 4 μl of 6
DNA Gel Loading Dye (Thermo Fisher Scientific).
3. Prepare ladder: for each desired lane (typically two or three
lanes, the first and last and one in the center), mix 3 μl of 1:10
diluted 50 or 100 bp DNA Ladder (Life Technologies), 17 μl
resuspension buffer RSB (TruSeq kit; Illumina), and 4 μl of 6

DNA Gel Loading Dye.
4. Load 24 μl of each sample and ladder mix with at least one
empty lane between each sample/ladder lanes.
5. Run gel at 120 V constant voltage for 120–145 min.
6. View the gel on a UV transilluminator. Using a fresh scalpel
blade for each sample, excise a band from the gel spanning the
width of the lane and ranging in size from 550 to 950 bp and
place in a tube (see Note 4). Use the DNA ladders as a guide.
Minimize time under transillumination to avoid DNA damage.
Determine weight of each gel piece by measuring tube without
and with gel piece. Gel pieces may be stored at 20 C.
3.3.3 Purification After 1. Follow the protocol for the MinElute Gel Extraction Kit (Qia-
Size Separation gen) to purify each sample. Incubate the gel slices in the QG
solution at room temperature (a change from the Qiagen pro-
tocol, as recommended by Illumina) until the gel slices have
completely dissolved (10–20 min), while vortexing every
2 min.
2. Elute the samples in 25 μl Buffer EB (if more than one column
is needed per sample, the eluted DNA should be pooled after
purification).
3. Store the samples at 20 C or continue with PCR enrichment.
3.3.4 PCR Enrichment 1. Thaw the PCR Master Mix and PCR Primer Cocktail at room
of DNA Fragments temperature. Once thawed, keep the tubes on ice.
2. Prepare PCR reactions using 12 μl of DNA (~1 μg) per sample,
8 μl of H2O (PCR grade), 5 μl of PCR Primer Cocktail, and
25 μl of PCR Master Mix (total volume is 50 μl per reaction).
3. Use the following program for PCR enrichment (with heated
lid):
98 C for 30 s
12 cycles of: (modification of Illumina protocol)
98 C for 60 s (modification of Illumina protocol)
60 C for 30 s
72 C for 30 s
72 C for 5 min
Hold at 4 C
4. Purification after PCR is performed using AMPure XP
Beads (Beckman Coulter) following the Illumina protocol (see
Note 5).
5. Analyze the distribution of post-PCR fragment lengths by
diluting 1 μl library in 49 μl water and running on a 2100
Bioanalyzer instrument (Agilent) using the High Sensitivity

DNA Analysis Kit (Agilent) and the concentration should be
measured using the Qubit Fluorometer (Thermo Fisher Scien-
tific). Successful libraries will have a fragment peak at approxi-
mately 800–900 bp and an undiluted concentration >10 ng/μl
(typically 30–50 ng/μl).
3.3.5 Cluster Generation Sequencing clusters can be generated on a cBot instrument (Illu-
and Sequencing mina) using TruSeq Cluster Kit cBot HS (Illumina) and PE Flow
Cells (Illumina).
Paired-end sequencing of 2 50 bp, 2 100 bp, or
2 150 bp plus index read can be performed on a HiSeq 2000
or HiSeq 2500 sequencer using TruSeq SBS Kit HS. For enumera-
tion of chromosomal rearrangements using our SplitSeq bioinfor-
matics pipeline, >60 million read-pairs (2 50 bp or 2 100 bp)
to >9 physical coverage (>2 sequence coverage) is generally
sufficient.
3.4 Sequencing Bioinformatics steps are described in detail elsewhere [8]. In short:
Bioinformatics
1. The paired-end reads are aligned to a reference human
genome, e.g., GRCh37, using Novoalign (Novocraft Technol-
ogies) with soft-clipped read alignments (option –o Softclip)
(see Note 6).
2. Potential chromosomal aberrations are first identified with
high sensitivity using BreakDancer [16] with default options
for discordant read-pairs.
3. The rearrangement-supporting discordant read pairs are rea-
ligned to the reference genome using Novoalign with the 1000
top-scoring alignments above a moderate alignment score
being reported (options –r Exhaustive 1000 –t 250). Initially
discordant read pairs that become concordant after this step
should be excluded. The purpose of his step is to reduce the
false-positive rate due to misalignment of paralogous
sequences.
4. DNA copy number is determined across the genome in win-
dows of 50 kb using FREEC version 5.6 with default para-
meters [17]. DNA copy number gains at putative
rearrangements may be utilized to prioritize candidates.
5. Identified rearrangement breakpoint ends are annotated with
RefSeq genes, sequence gaps (gaps track), and repetitive ele-
ments (RepeatMasker track) for the human reference genome
(hg19), all obtained from the UCSC Table Browser (http://
genome.ucsc.edu/cgi-bin/hgTables), and with entries from
the Database of Genomic Variants (http://dgv.tcag.ca/).
6. To deplete the list of predicted chromosomal rearrangements
for potential nonspecific rearrangements and false-positives
before experimental validation, the following filtering criteria

are applied:
(a) At least two discordant read-pairs supporting the
rearrangement.
(b) No satellite DNA (RepeatMasker class “Satellite”) present
within 1 kb of any of the two breakpoint ends of the
rearrangement.
(c) No sequence gap (UCSC track “gaps”) present within
1 kb of any of the two breakpoint ends.
(d) No matching rearrangement in other tumor samples
within 1 kb (both breakpoint ends matched).
(e) No matching rearrangement in other normal samples
within 1 kb (both breakpoint ends matched).
(f) For intrachromosomal rearrangements, size of the rear-
rangement (distance between the two breakpoint ends)
greater than 1 kb.
(g) Both breakpoint ends on chromosomes 1–22 or X (i.e.,
no involvement of nonstandard sequence contigs present
in the human genome reference).
7. SplitSeq software was developed to reconstruct the exact
sequence of breakpoints from low-coverage whole-genome
sequencing data. The software and code will be described in a
forthcoming manuscript. Briefly, SplitSeq uses discordant read-
pairs as anchors and searches the regions around the break-
points for split-reads with soft-clipped bases that do not match
the reference (soft-clipping in Novoalign can occur either due
to mismatches or due to low base qualities at any end of a read).
The same regions are also searched for reads with unmapped
mates (i.e., read-pairs where only one read of the pair is
aligned), and alignment of the unmapped mates to the break-
point regions is attempted. Alignment is done with a reduced
gap extension penalty (parameter -x 1) starting with the full
read length, followed by two rounds of trimming the read to
2/3 of its previous length on either end. Any mapped split-read
identified with this procedure is added to the list of split reads.
Next, all soft-clipped reads from this list are sorted by their
clipping position in the reference genome in order to identify
putative exact breakpoint positions. Reference genome posi-
tions with support of less than a total of X clipped bases
(typically 6) may be discarded. For all other positions, starting
with those with the highest number of clipped-base support, all
pairs connecting the two breakpoints (defined by the discor-
dant read-pair anchors) are searched for perfect matches of all
clipped bases when aligned to a fusion sequence reconstructed
from the two breakpoint positions.
3.5 Cell-Free Plasma 1. Fill blood collection tube.

DNA Extraction 2. Centrifuge the blood samples at 2000 g for 10 min at 4 C
3.5.1 Fractionation (see Notes 2 and 7).
of the Whole Blood 3. Transfer the upper plasma fraction with a Pasteur pipette into
1.5 ml aliquots, and process immediately or store at 80 C.
3.6 Preparation 1. Before starting the cell-free DNA extraction protocol, prepare
for Isolation of cfDNA the following reagents and devices. The reagents are all from
the Qiagen UltraSens Virus Kit unless otherwise specified.
(a) Add 310 μl Buffer AVE to one tube of lyophilized carrier
RNA to obtain a 1 μg/μl solution. Use 5.6 μl of this
solution per extraction. Store the carrier RNA aliquots at
20 C.
(b) Add 96–100% ethanol to the Buffer AB, AW1, and AW2
according to the manufacturer’s instructions.
(c) In heating block, bring Buffer AR to 60 C (330 μl is
required per sample).
(d) If working with frozen stored plasma, thaw on ice.
3.7 Isolation Cell-free DNA extraction is performed using the UltraSens Virus
of cfDNA Kit protocol with modifications:
1. Centrifuge the plasma samples at 10,000 g for 10 min at
4 C.
2. Transfer the supernatant to 2 ml DNA LoBind tubes, and store
any excess aliquots at 80 C (see Note 8).
3. If isolating cfDNA from less than 1 ml plasma, use DPBS
(without Mg2+/Ca2+) to bring the volume of each plasma
sample to 1 ml (see Note 9).
4. Pipet 5.6 μl carrier RNA (prepared in Subheading 3.6 above)
into the dry lid of each sample tube (see Note 10).
5. Add 800 μl Buffer AC directly to each sample.
6. Close the lid, invert tube five times and mix thoroughly by
vortexing for 10 s.
7. Incubate the tubes at room temperature for 15 min to allow
protein-nucleic acid complexes to form. Regular mixing by
inverting the tube several times every 5 min or less during the
period of incubation is suggested.
8. Centrifuge the samples at 1200 g for 3 min at room temper-
ature (see Note 11).
9. Remove and discard all but ~50 μl of the supernatant from the
samples (see Note 12).
10. Flick the tubes forcefully to loosen the pellets.
11. Add 20 μl Proteinase K to the lid of each tube (see Note 13).
12. Add 300 μl of preheated Buffer AR to each tube and invert and
flick the tube directly after adding the buffer. Do not use the
vortex mixer here.
13. Incubate the tubes for 10 min at 40 C in a thermomixer at
400 revolutions per minute or until all pellets are completely
dissolved. A brief centrifuge is helpful to remove drops from
the lid.
14. Add 400 μl Buffer AB, mix thoroughly with vortex mixer, and
centrifuge briefly.
15. Carefully apply 700 μl lysate to a QIAamp spin column, centri-
fuge at 2000 g for 3 min, and discard the flow-through (see
Note 14).
16. Place the QIAamp spin column in a new 2 ml collection tube,
add 500 μl Buffer AW1, and centrifuge at 6000 g at room
temperature for 1 min.
add 500 μl Buffer AW2, and centrifuge at the maximum speed
(13,000–16,000 g) at room temperature for 3 min.
and centrifuge at the maximum speed for 1 min to completely
dry the membrane (see Note 15).
19. Elute cfDNA using two rounds of Buffer EB. Place the
QIAamp spin column in a clean 1.5 ml DNA LoBind tube,
apply 50 μl elution buffer directly to the membrane, incubate
1 min at room temperature, and centrifuge at 6000 g at
room temperature for 1 min. Repeat this step and combine the
eluate to obtain ~100 μl cfDNA solution (see Note 16).
3.8 Detection 1. Select a subset of the predicted fusion sequences for assay
of Tumor-Specific design. It is advisable to inspect the breakpoint positions with
Genomic a genome browser such as IGV. For example, inspect individual
Rearrangements reads that lay across the breakpoint (split-reads). We typically
in cfDNA select 10 rearrangements per tumor, from which ~5 will be
analyzed in cfDNA samples per patient.
3.8.1 Rearrangement
Selection
3.9 ddPCR Assay 1. Design primers and probes. The forward and reverse primers
Design should be on opposite sides of the breakpoint location with a
hydrolysis probe positioned between them. The primers should
be positioned as close to the probe as possible, but the primers
and probe should not overlap each other or the breakpoint
position (see Note 17). The amplicons should be designed to
be as short as possible due to the highly fragmented nature of
cfDNA.
2. The Tm and sequence parameters for the primers and probes

should adhere to the Applied Biosystems Primer Express Soft-
ware guidelines on “Quantification TaqMan MGB Probe
Design Guidelines and Considerations” and “Primer Design
Guidelines and Considerations” as well as the guidelines from
the Bio-Rad “Droplet Digital PCR Applications Guide”.
(a) Primer Guidelines:
l Tm between 50 and 65 C, preferably between 58 and
60 C and should have similar Tm (e.g., within 1–2 C
of each other).
l GC content between 30 and 80%, preferably between
50 and 60%.
l Avoid polynucleotide repeats, in particular avoid
repeats of four or more consecutive Gs or Cs.
l The terminal 30 base should be a G or C if possible, the
final three bases should not all be Gs or Cs, and the final
five nucleotides should include no more than three G
and/or C bases.
l The primers should be positioned as close to the probe
as possible without overlapping.
lAvoid hairpin loops, self-dimerization, and cross-
dimerization.
(b) Probe Guidelines:
l Tm of the probe should 68–70 C, and greater than
3 C higher than the Tm of the primers, preferably
8–12 C higher.
l Probe length should be 13–30 bases.
l GC content between 30 and 80%.
l The terminal 50 end should not be a G, and when using
a FAM dye-labeled probe, the second base from the 50
should also not be a G.
l Avoid polynucleotide repeats, in particular avoid
repeats of four or more consecutive Gs or Cs. Also
avoid repeats of six or more consecutive As.
l Nonfluorescent quenchers, such as Black Hole
Quencher, are recommended.
l Consider selecting probes with more Cs than Gs to
minimize fluorescence quenching.
l Avoid hairpin loops, self-dimerization, and cross-
dimerization.
3. Label the hydrolysis probe with fluorescence group of FAM,
HEX, or VIC and we recommended quenching using an inter-
nal ZEN and 30 -IBFQ molecule (Integrated DNA
Technologies).
3.10 Rearrangement 1. Order the primer pairs (but not the hydrolysis probes) and
and Primer Validation perform standard PCR using tumor DNA as positive control
template to validate the structural variants of interest, and using
matched normal DNA to distinguish tumor-specific somatic
rearrangements from germline variants. Any standard PCR
set-up may be employed here; typically we perform:
(a) PCR Reactions (each 10 μl total volume):
Phusion Master Mix 1 (Thermo Scientific)
250 nM of each primer
2% DMSO
10 ng of template DNA
(b) PCR Cycling Conditions:
Initial denaturation: 98 C for 2 min
11 cycles of:
98 C for 10 s
70 C (1 C/cycle) for 30 s (70–60 C in 1 C
intervals)
72 C for 15 s
29 cycles of:
98 C for 10 s
60 C for 30 s
72 C for 15 s
Final elongation at 72 C for 5 min
2. Analyze the PCR products to validate the fragment size and the
germline or somatic status of the alterations, e.g., using the
Caliper LabChip XT System (PerkinElmer) (see Note 18).
3. For the structural variants that were confirmed to be somatic by
touchdown PCR, order the TaqMan probesv for the
corresponding assays.
3.11 ddPCR Assay 1. Validate the complete assay (probe and primer pair) in droplet
Validation digital PCR using the QX100/200 Droplet Digital PCR Sys-
tem (Bio-Rad) with tumor DNA as the template for positive
control and matched normal DNA as the template for negative
control. In addition to negative control DNA template, a
no-template control (NTC) consisting of PCR-grade water
can be run. The following ddPCR setup was optimized for
this protocol:
ddPCR Reactions (each 20 μl total volume) (see Notes 19–21):
1 digital PCR supermix for probes (Bio-Rad)
Fig. 1 Example personalized genomic marker analysis of circulating DNA at multiple time points during clinical
follow-up. Panels a–c illustrate ddPCR results for the 2p14 wild-type assay for breast cancer patient EM9 and
panels d–f show results for the inv(12)(q13.13q21.31) assay for the same patient. Wells are ordered, from left
to right: tumor positive control (þ Ctrl), blood plasma prior to surgery (PreOp), plasma at 5-months (5m),
1 year (1y), 2 years (2y), and 3 years (3y) after surgery, and negative control ( Ctrl). QuantaSoft 2D droplet
plots are shown in a and d with wells separated by vertical yellow lines. Panels b and c, and e and f illustrate
the transformation steps for assay-specific data normalization and thresholding on the exported intensity
measurements (see Subheading 3.12). In b and e, the blue line “2*negMax” is drawn for a given ddPCR assay
at two times the maximum droplet intensity of the negative control (no-template control, NTC; or matched
normal DNA, Normal) reaction well, defining the data from the given assay into an upper and lower portion.
Yellow horizontal lines are drawn at the median value for the droplets in the lower portion within each well
(“lowerMed”), and the green line is drawn across at the median value for the upper portion of the positive
control primary tumor DNA well (“posUpperMed”). Data are then scaled to 0 and 1 (c, f) corresponding to these
lowerMed and posUpperMed values, respectively. Red dashed lines are drawn for nine example threshold
values, from 0.1 to 0.9, and numbers given within the lines indicate the count of droplets above the threshold
for the respective well. Summary statistics for each well is provided below each plot. Figure adapted from
Olsson et al. [8]
250 nM of each primer

250 nM of probe
10 ng of template DNA
2. Mix the ddPCR reactions well before transferring 20 μl to the

wells of the sample row (middle row) of a DG8 cartridge (the
DG8 cartridge should already be loaded into the DG8 car-
tridge holder before transfer) (see Note 22). Empty sample
wells should be filled with 1 ddPCR buffer control. Do not
mix the reactions in the DG8 cartridge.
3. Add 70 μl of droplet generation oil into the wells of the oil row
(bottom row). Always add the sample before the oil, as the
opposite can result in fewer generated droplets.
4. Put the rubber gasket into place on the DG8 cartridge holder
and then carefully load it into and start the QX100/200 Drop-
let Generator. The two liquids combined in the microfluidic
channels and an emulsion is produced in the droplet row (top
row) of the cartridge.
5. Slowly and carefully (to avoid shearing of the droplets) pipette
40 μl from the droplets row to a 96-well PCR plate (Eppendorf
twin.tec PCR plates 96, semiskirted). Consider using an elec-
tronic multichannel pipette adjusted to the slowest aspirate and
dispense settings.
6. Place the 96-well PCR plate in a thermal cycler (e.g., Bio-Rad
C1000 Touch Thermal Cycler) and run the ddPCR cycling
protocol as specified here:
ddPCR Cycling Conditions (see Note 23):
Initial denaturation/enzyme activation: 95 C for 10 min
10 touchdown cycles of:
94 C for 30 s
65 C for 60 s (adjusted 0.7 C/cycle)
35 cycles of:
94 C for 30 s
58 C for 60 s
Enzyme deactivation: 98 C for 10 min.
Hold at 10 C.
7. Transfer the 96-well PCR plate to the Bio-Rad (QX100/200)
Droplet Reader and set the QuantaSoft (Bio-Rad) software to
ABS (absolute quantification) mode and to interrogate the
appropriate fluorescent signals (FAM, HEX, or VIC) (see
Note 24).
8. For ddPCR assays that validate successfully, ~5 assays can be
used to analyze circulating DNA isolated from patient blood
plasma (Fig. 1). Follow the same ddPCR setup and instructions
as above (see Note 25). For the template input volume and
concentration, consider the number of assays you wish to run,
the number of replicates, and, without exceeding 100,000
haploid genome copies/well (<330 ng/well), we recommend

using the maximum amount of template per ddPCR reaction
that can be afforded per reaction well (typically, 8 μl of reaction
volume is available for the template).
3.12 Droplet Digital 1. Determine a threshold for each assay based on the intensity
PCR Intensity values from positive and negative control samples (Fig. 1). The
Thresholding threshold is a fixed amplitude value that reliably separates true-
positive droplets from true-negative droplets. Even with a good
quality assay, however, some droplet intensities will fall some-
where between the positive and negative clouds. Because of
this, careful consideration should be taken when defining
exactly where to place the threshold. If the threshold is set
too low, the frequency of potential false-positives is higher
whereas alternatively if the threshold is set too high, the fre-
quency of potential false-negatives increases. When detecting a
target molecule that occurs as an extremely rare event, as can be
the case with circulating tumor DNA, it is critical that both
false-positives and false-negatives are avoided.
2. Various methods can be used to define where to draw the
threshold. One method is to export the droplet intensity values
from QuantaSoft in order to be able to apply automatic, unbi-
ased, reproducible, and operator independent thresholding
that is based on normalized intensity signals from positive and
negative controls (Fig. 1). Normalizing the data facilitates
employing one intensity threshold for all samples in order to
define negative droplets (below threshold) and positive dro-
plets (above threshold). Using centering and scaling, this pro-
cedure normalized each individual assay’s droplet fluorescent
intensity data to a new scale where 0 corresponds to no inten-
sity (at the median intensity of negative control droplets) and
1 corresponds to full intensity (at the median intensity of
positive control droplets) and was carried out as described
below. For each rearrangement assay:
(a) Determine negMax ¼ maximum intensity value in the
negative control well.
(b) For each ddPCR reaction well, define a lower group of
droplets with intensity <¼2*negMax and an upper group
of droplets with intensity >2*negMax.
(c) For each well, determine lowerMedian ¼ median intensity
value of the lower group of droplets.
(d) For each well, subtract lowerMedian from each intensity
value (bringing all medians to zero).
(e) Determine posUpperMedian ¼ median intensity value in
the upper group of droplets in the positive control well.
(f) For each well, divide each intensity value by posUpperMe-

dian (setting the median of the upper droplets in the
positive control to 1).
(g) Set threshold t to 0.5 to define negative droplets (below
threshold) and positive droplets (greater than or equal to
threshold).
3. We typically measure the concentration of a normal nonrear-
ranged 132 bp region of chromosome 2p14 as a surrogate for
total cfDNA This 2p14 region very infrequently undergoes
copy number change in breast cancer. The assay sequences
are, 2p14-Fwd GCTGAATTGCTTTGAGCTTGCAG; 2p14-
Rev CATGTCACCTCCAGTTACAGGGAA; 2p14-Probe
TGTCACCAGTTGTCTGCTGTTCTTTCGGG.
4. The number of fragments per μl input purified circulating
DNA (CVi) is calculated from the number of positive droplets
P, total number of droplets analyzed T, droplet volume Vd
(0.91 103 μl or 0.85 103 μl depending on hardware/
software versions; we recommend checking with the manufac-
turer, Bio-Rad, if unsure), ddPCR reaction volume Vr (20 μl;
includes PCR mix, primers, probe, input DNA), and volume of
purified circulating DNA input into the reaction Vi, using the

ln ð1TP Þ V r
formula: C V i ¼ Vd Vi .
5. Quantification may be expressed as copies/ml plasma by multi-

plying by a conversion factor accounting for starting plasma
volume and the cfDNA elution volume. Tumor-specific rear-
rangements may also be expressed as percent ctDNA by divid-
ing by the counts of the normal control fragment of 2p14.
4 Notes
1. In order to filter germline events, both tumor and normal

DNA from the same patient should be analyzed. In the event
that matched normal DNA sample is unavailable, unmatched
normal DNAs may be used to filter the most common germline
events in the population.
2. Fractionate blood as soon as possible after collection. For
standard EDTA tubes, within 2 h is ideal. Streck Cell-Free
BCT tubes may be kept at room temperature or 4 C for up
to 14 days (refer to manufacturer instructions).
3. If more than 48 samples will be prepared for sequencing, the
high-throughput (HT) protocol should be used.
4. Cutting a band between 550 and 950 bp will result in an insert
size of approximately 450–850 bp, accounting for the size of
the adapters. Adapters add approximately 120 bp to each

fragment.
5. The AMPure XP beads must be acclimated to room tempera-
ture before use (at least 30 min).
6. Data from samples run on multiple lanes of a flow cell should
be merged into one BAM file using Novosort (Novocraft
Technologies). Duplicate read-pairs in the BAM files can be
flagged with Picard MarkDuplicates and ignored in subsequent
analyses.
7. Whole blood contains approximately 50% plasma. Plasma is the
preferred source of cfDNA since, during the formation of
serum, normal cells lyse and release their genomic DNA into
solution, thereby increasing background and obscuring true
cfDNA quantities.
8. Free circulating DNA in blood plasma is sensitive to multiple
freeze and thaw cycles; therefore aliquot your plasma samples.
9. Optionally, a volume of exogenous DNA reference fragments,
with the size between 100 and 200 nucleotides and at known
concentration, can be added to the sample prior to compensa-
tion with DPBS. Reference exogenous DNA fragments may be
used to calculate the extraction efficiency.
10. It is important that the carrier RNA does not mix with Buffer
AC or with the sample until the two have been mixed together
as this can affect the yield.
11. A pellet should be at the bottom of each tube containing the
cfDNA and the supernatant should be mostly clear. Centrifuge
force and time may be adjusted, but prolonged and/or a high
centrifuge force can compact the pellet and make resuspension
difficult.
12. Leaving a small volume of supernatant greatly aids
resuspension.
13. We recommend keeping Proteinase K solution at 4 C.
14. Excessive centrifuge force will reduce binding of nucleic acids
to the membrane. If the total volume of the lysate is larger than
700 μl, repeat this step with the same column until all lysate has
been loaded.
15. Residual Buffer AW2 often remains on the plastic “lip” above
the membrane. This may be avoided by removing excess solu-
tion around the lip using a pipette prior to this second
centrifugation.
16. The elution buffer and volumes may be changed depending on
the downstream application, however the total yield may
decrease if the elution volumes are small.
17. There are two generations of Droplet Digital PCR systems

from Bio-Rad. The QX100 Droplet Digital PCR system sup-
ports hydrolysis probe assays (TaqMan) while the QX200
Droplet Digital PCR system supports both hydrolysis probe
and DNA binding dye (EvaGreen) assays. This protocol is
optimized to use hydrolysis probe assay method.
18. Analysis of the PCR products can be done with a number of
instruments including the 2100 Bioanalyzer (Agilent) or by
standard gel electrophoresis. The Caliper system is ideal for
high throughput.
19. Prepare reactions at a slightly higher volume than 20 μl, i.e.,
22 μl. This is to ensure that a full 20 μl is available for droplet
generation.
20. For the positive control reactions, the amount of template
tumor DNA used can vary but should be less than 66 ng. For
reference, 10 ng of human genomic DNA (which is roughly
3000 haploid genome copies) will result in thousands of posi-
tive reaction droplets if the assay is working properly, so much
less tumor template can be used if desired.
21. For the negative control reactions, it is advisable to run a large
amount of nonspecific DNA template (e.g., >20 ng matched
normal DNA). Since circulating tumor DNA can exist at
extremely low concentrations in extracted plasma, it is impor-
tant to have very specific assays since the number of true-
positive droplets can be low. Ideally no false-positive droplets
should occur even after analyzing very many thousands of
normal genomes as template. An advantage of structural var-
iants as tumor biomarkers, as compared to other types of
variations such as point mutations, is that they result in very
unique sequences, and thus very specific assays should be
attainable.
22. Plan ahead the ddPCR cartridge and plate layout. For a given
assay, it is recommended to place test samples first, followed by
the positive and then negative controls (with regards to the
order in which the droplets are generated and the plate is read).
23. If possible, the ramp rate between any two consecutive steps
should be set to ~2.5 C/s to ensure reliable thermal control.
24. The droplet reader measures the fluorescence intensity, size,
and shape of each droplet, excludes droplets that do not pass
quality metrics, and then outputs the intensities of the passing
droplets for each of the measured dyes. For an assay to be
considered validated, the following criteria should be met:
(a) In the positive control wells, there should be two clusters
(appear as clouds or bands in 1D plot) of droplets that are
clearly separate from one another. The cluster with the
higher amplitude represents droplets that contained the

target template and the cluster with the lower amplitude
represents droplets that contained no target template.
There two bands should be relatively compact (tight
around their average intensity).
(b) In the negative control wells, there should be a single
cluster that has the same average intensity as the lower
amplitude cluster of the positive control reactions for the
same assay. The negative control cluster should be a tight
band with no stray droplets in the positive amplitude
direction. If this is observed, the assay should be excluded
as it could result in false-positives.
25. In breast cancer, metastases typically retain 89% (range
61–100%) of the chromosomal rearrangements present in
their matched primary tumor [14]; using the lower 61% figure,
the odds are 99/100 that one or more rearrangements out of
five will be retained in a metastatic clone and thus be informa-
tive for minimal residual disease, tracking response to therapy,
or for early detection of occult metastatic disease.
Acknowledgments
We thank members of the Translational Oncogenomics Unit, Divi-

sion of Oncology and Pathology for assistance, and in particular to
Christof Winter and Robert Rigo for bioinformatics work. This
work was supported by the Swedish Cancer Society, Swedish
Research Council, Swedish Foundation for Strategic Research,
Knut and Alice Wallenberg Foundation, VINNOVA, and Govern-
mental Funding of Clinical Research within National Health Ser-
vice, Swedish Breast Cancer Group, Crafoord Foundation, Lund
University Medical Faculty, Gunnar Nilsson Cancer Foundation,
Skåne University Hospital Foundation, BioCARE Research Pro-
gram, King Gustav Vth Jubilee Foundation, Krapperup Founda-
tion, and the Mrs. Berta Kamprad Foundation.
References
1. Haber DA, Velculescu VE (2014) Blood-based 4. Bettegowda C, Sausen M, Leary RJ, Kinde I,
analyses of cancer: circulating tumor cells and Wang Y, Agrawal N, Bartlett BR, Wang H,
circulating tumor DNA. Cancer Discov Luber B, Alani RM, Antonarakis ES, Azad
4:650–661 NS, Bardelli A, Brem H, Cameron JL, Lee
2. Ignatiadis M, Dawson SJ (2014) Circulating CC, Fecher LA, Gallia GL, Gibbs P, Le D,
tumor cells and circulating tumor DNA for Giuntoli RL, Goggins M, Hogarty MD,
precision medicine: dream or reality? Ann Holdhoff M, Hong SM, Jiao Y, Juhl HH,
Oncol 25:2304–2313 Kim JJ, Siravegna G, Laheru DA,
3. Schwarzenbach H, Hoon DS, Pantel K (2011) Lauricella C, Lim M, Lipson EJ, Marie SK,
Cell-free nucleic acids as biomarkers in cancer Netto GJ, Oliner KS, Olivi A, Olsson L, Rig-
patients. Nat Rev Cancer 11:426–437 gins GJ, Sartore-Bianchi A, Schmidt K, Shih
l M, Oba-Shinjo SM, Siena S, Theodorescu D, Detection and quantification of mutations in

Tie J, Harkins TT, Veronese S, Wang TL, the plasma of patients with colorectal tumors.
Weingart JD, Wolfgang CL, Wood LD, Proc Natl Acad Sci U S A 102:16368–16373
Xing D, Hruban RH, Wu J, Allen PJ, Schmidt 12. Chin K, de Solorzano CO, Knowles D,
CM, Choti MA, Velculescu VE, Kinzler KW, Jones A, Chou W, Rodriguez EG, Kuo WL,
Vogelstein B, Papadopoulos N, Diaz LA Jr Ljung BM, Chew K, Myambo K, Miranda M,
(2014) Detection of circulating tumor DNA Krig S, Garbe J, Stampfer M, Yaswen P, Gray
in early- and late-stage human malignancies. JW, Lockett SJ (2004) In situ analyses of
Sci Transl Med 6:224ra224 genome instability in breast cancer. Nat Genet
5. Dawson SJ, Tsui DW, Murtaza M, Biggs H, 36:984–988
Rueda OM, Chin SF, Dunning MJ, Gale D, 13. Alkner S, Tang MH, Brueffer C, Dahlgren M,
Forshew T, Mahler-Araujo B, Rajan S, Chen Y, Olsson E, Winter C, Baker S,
Humphray S, Becq J, Halsall D, Wallis M, Ehinger A, Ryden L, Saal LH, Ferno M,
Bentley D, Caldas C, Rosenfeld N (2013) Gruvberger-Saal SK (2015) Contralateral
Analysis of circulating tumor DNA to monitor breast cancer can represent a metastatic spread
metastatic breast cancer. N Engl J Med of the first primary tumor: determination of
368:1199–1209 clonal relationship between contralateral breast
6. Jung K, Fleischhacker M, Rabien A (2010) cancers using next-generation whole genome
Cell-free DNA in the blood as a solid tumor sequencing. Breast Cancer Res 17:102
biomarker—a critical appraisal of the literature. 14. Tang MH, Dahlgren M, Brueffer C,
Clin Chim Acta 411:1611–1624 Tjitrowirjo T, Winter C, Chen Y, Olsson E,
7. Diehl F, Schmidt K, Choti MA, Romans K, Wang K, Torngren T, Sjostrom M, Grabau D,
Goodman S, Li M, Thornton K, Agrawal N, Bendahl PO, Ryden L, Nimeus E, Saal LH,
Sokoll L, Szabo SA, Kinzler KW, Vogelstein B, Borg A, Gruvberger-Saal SK (2015) Remark-
Diaz LA Jr (2008) Circulating mutant DNA to able similarities of chromosomal rearrange-
assess tumor dynamics. Nat Med 14:985–990 ments between primary human breast cancers
8. Olsson E, Winter C, George A, Chen Y, and matched distant metastases as revealed by
Howlin J, Tang M-HE, Dahlgren M, whole-genome sequencing. Oncotarget
Schulz R, Grabau D, van Westen D, Ferno M, 6:37169–37184
Ingvar C, Rose C, Bendahl PO, Ryden L, 15. Saal LH, Vallon-Christersson J, Hakkinen J,
Borg A, Gruvberger-Saal SK, Jernstrom H, Hegardt C, Grabau D, Winter C, Brueffer C,
Saal LH (2015) Serial monitoring of circulat- Tang MH, Reutersward C, Schulz R,
ing tumor DNA in patients with primary breast Karlsson A, Ehinger A, Malina J, Manjer J,
cancer for detection of occult metastatic dis- Malmberg M, Larsson C, Ryden L, Loman N,
ease. EMBO Mol Med 7:1034–1047 Borg A (2015) The Sweden Cancerome Analy-
9. Garcia-Murillas I, Schiavon G, Weigelt B, sis Network - Breast (SCAN-B) Initiative: a
Ng C, Hrebien S, Cutts RJ, Cheang M, large-scale multicenter infrastructure towards
Osin P, Nerurkar A, Kozarewa I, Garrido JA, implementation of breast cancer genomic ana-
Dowsett M, Reis-Filho JS, Smith IE, Turner lyses in the clinical routine. Genome Med 7:20
NC (2015) Mutation tracking in circulating 16. Chen K, Wallis JW, McLellan MD, Larson DE,
tumor DNA predicts relapse in early breast Kalicki JM, Pohl CS, McGrath SD, Wendl MC,
cancer. Sci Transl Med 7:302ra133 Zhang Q, Locke DP, Shi X, Fulton RS, Ley TJ,
10. Leary RJ, Kinde I, Diehl F, Schmidt K, Wilson RK, Ding L, Mardis ER (2009) Break-
Clouser C, Duncan C, Antipova A, Lee C, Dancer: an algorithm for high-resolution
McKernan K, De La Vega FM, Kinzler KW, mapping of genomic structural variation. Nat
Vogelstein B, Diaz LA Jr, Velculescu VE Methods 6:677–681
(2010) Development of personalized tumor 17. Boeva V, Zinovyev A, Bleakley K, Vert JP,
biomarkers using massively parallel sequencing. Janoueix-Lerosey I, Delattre O, Barillot E
Sci Transl Med 2:20ra14 (2011) Control-free calling of copy number
11. Diehl F, Li M, Dressman D, He Y, Shen D, alterations in deep-sequencing data using
Szabo S, Diaz LA Jr, Goodman SN, David KA, GC-content normalization. Bioinformatics
Juhl H, Kinzler KW, Vogelstein B (2005) 27:268–269
Chapter 18
Single Color Multiplexed ddPCR Copy Number

Measurements and Single Nucleotide Variant Genotyping
Christina M. Wood-Bouwens and Hanlee P. Ji
Abstract
Droplet digital PCR (ddPCR) allows for accurate quantification of genetic events such as copy number
variation and single nucleotide variants. Probe-based assays represent the current “gold-standard” for
detection and quantification of these genetic events. Here, we introduce a cost-effective single color
ddPCR assay that allows for single genome resolution quantification of copy number and single nucleotide
variation.
Key words Digital PCR, Single color, Mutation detection, Single nucleotide variant (SNV), Copy
number variation (CNV), Circulating cell free DNA (ccfDNA)
1 Introduction
This novel molecular assay utilizes BioRad’s QX200 Droplet Digi-

tal PCR (ddPCR) system to detect and quantify clinically actionable
mutations within poor quality and limited quantity samples such as
FFPE DNA and circulating tumor DNA (ctDNA). This new
molecular assay technology has been optimized for quantitatively
measuring clinically actionable mutations, is highly sensitive, and is
capable of detecting the targeted mutant fraction at an individual
DNA molecule resolution. DNA shed from tumors can be
extracted from the plasma fraction of routine blood draws; the
resulting ctDNA is extremely variable and low in concentration,
typically yielding well below 5 ng DNA per μL of plasma [1], thus
making the ctDNA a prime candidate for this assay. The ddPCR
single-color system allows for accurate and absolute quantification
of the mutant and wild-type allelic fractions within a given patient
sample using easy to design standard PCR primers; this means that
any cancer mutation can have a customized assay generated and
tested efficiently.
Allele-specific mutation quantitation relies on two DNA primer
sets that are identical with the exception of the mutant or wild type
323
324 Christina M. Wood-Bouwens and Hanlee P. Ji
specific base at the 30 end of the “detecting” primer sets, to amplify

the genomic region of interest. Through the addition of artificial 50
noncomplementary tails to our mutant and wild type-specific
“detection” primers, we are able to consistently differentiate
between droplets that contain the mutant or wild type alleles
based upon their differential amplicon lengths. The artificial tails
minimize bias within the PCR reaction by allowing our primers to
target an identical region of genomic DNA, with the exception of
the SNP specific base at the 30 end of the detecting primer. The
ddPCR technology allows the standard PCR reaction to be parti-
tioned into approximately 20,000 individual emulsions. These can
be effectively treated as 20,000 unique reactions that are cycled in
parallel within a single well of a PCR plate. We can then assay each
individual droplet to assess whether the individual DNA molecule
partitioned into the droplet is “wild-type” or the target
“mutant”—giving an absolute count of mutant and wild-type tem-
plates in a given patient sample. This assay is capable of detecting
seven mutant genome equivalents among a total of 7000 genome
equivalents using standard nonmodified DNA primers. The sensi-
tivity of this assay also allows it to handle the low concentration
samples typically gathered from the extraction of circulating cell
free DNA from the plasma fraction of a standard blood draw.
2 Materials
Pre- and post-PCR workspaces should be in physically separate

locations with designated equipment and consumables (*Denotes
an item that should be kept exclusively in post-PCR areas). All
reagents/primers should be divided into smaller one-time-use
volumes to reduce the risk of contamination.
2.1 Instruments/ 1. QX200™ Droplet Generator (Bio-Rad, 1864002).

Equipment 2. QX200™ Droplet Reader* (Bio-Rad, 1864003).
3. QuantaSoft™ Software (Bio-Rad).
4. Veriti® 96-well Thermal Cycler* (Applied Biosystems,
4375786).
5. DG8™ Cartridge Holder (Bio-Rad, 1863051).
6. PX1™ Plate Sealer (Bio-Rad, 1814000).
7. AirClean 600 Combination PCR Workstation.
8. Centrifuge.
9. Tabletop centrifuge.
10. Promega Maxwell® RSC.
Single Color Multiplexed ddPCR Copy Number Measurements and Single. . . 325
2.2 Consumables 1. QX200™ Droplet Generation Oil for EvaGreen (Bio-Rad,

1864005).
tor (Bio-Rad, 1863008).
(Bio-Rad, 1863009).
4. ddPCR™ Droplet Reader Oil (Bio-Rad, 1863004).
5. Eppendorf 96-Well twin tec PCR Plates; 250 μL semiskirted
(Eppendorf, 951020389).
6. Pierceable Foil Heat Seal (Bio-Rad, 1814040).
7. Eppendorf DNA LoBind Microcentrifuge Tubes; Safe-Lock
Tube 1.5 mL; DNA; PCR Clean; (Eppendorf, 22431021).
8. Rainin P20 Low-Retention Filter Tips (Rainin, RT-L10FLR).
9. Rainin P200 Low-Retention Filter Tips (Rainin,
RT-L200FLR).
10. Rainin P1000 Low-Retention Filter Tips (Rainin,
RT-L1000FLR).
2.3 ddPCR Reagents 1. QX200™ ddPCR™ EvaGreen Supermix (Bio-Rad, 1864034).

2. Qx200™ 2 Buffer Control for EvaGreen (Bio-Rad,
1864052).
3. UltraPure™ DNase/RNase-Free Distilled Water (Invitrogen,
10977-023).
4. Elution Buffer, 10 mM Tris–HCl pH 8.0: Prepare elution
buffer solution by diluting UltraPure™ 1 M Tris–HCI,
pH 8.0 (Invitrogen, 15568-025) to a final concentration of
10 mM with UltraPure™ Water (Invitrogen, 10977-023).
5. Restriction endonuclease (TBD by user, see Subheading 3.1).
6. Primers: Resuspend all primers in elution buffer to a final
concentration of 100 μM; vortex primers thoroughly and
briefly spin down. Dilute primer to a final working concentra-
tion of 10 μM in UltraPure™ Water and store resuspended
primers at 20 C).
2.4 Sample 1. Maxwell® RSC DNA FFPE Kit (Promega, AS1450).

Extraction Reagents 2. EDTA Blood Collection Vials.
3. Maxwell® RSC ccfDNA Plasma Kit (Promega, AS1480).
3 Methods
All reagent preparation and droplet generation should be carried

out in a PCR workstation to reduce the risk of contamination (see
Note 1). Use Rainin Low Retention Filter-Tips when preparing all
reagents and master mixes.
3.1 Sample DNA 1. Obtain 5–10 μm sections of FFPE block containing tissue of
Extraction interest.
3.1.1 Extraction of DNA 2. Proceed with the standard protocol included in the Maxwell®
from FFPET Tissue RSC DNA FFPE Kit.
3.1.2 Extraction 1. Collect two EDTA vials of blood, keep at 4 C until ready to
of ccfDNA from Blood process. Note: Processing must take place within 2 h of
collection.
2. Spin blood containing vials for 10 min at 2000 g in a
centrifuge.
3. Transfer plasma fraction to a fresh 1.5 or 2.0 mL LoBind
Eppendorf tube.
4. Spin the plasma containing tube for 10 min at 2000 g.
5. Transfer supernatant a fresh 1.5 or 2.0 mL LoBind Eppendorf
tube using care not to disturb the pellet.
Proceed to step 6 for long-term storage or skip step 6 and proceed to
step 7 to proceed and extract ccfDNA.
6. Flash freeze the tube containing plasma in liquid nitrogen until
frozen, store at 80 C until ready to extract ccfDNA.
7. Proceed with the standard protocol included in the Maxwell®
RSC ccfDNA Plasma Kit.
8. Store extracted ccfDNA at 20 C for short term (1–2 days) or
at 80 C for long-term storage. Avoid excessive freeze–thaw
cycles.
3.2 Designing SNV CNV assay primers should target approximately 20 bp of genomic
Primers sequence and may be designed using Primer3. The primers’ speci-
ficity can be confirmed using the UCSC Genome Brower Blat tool
3.2.1 Copy Number
[2]. CNV assay primers targeting the genomic region of interest
Variation (CNV) Assay
should be multiplexed with an off target reference gene as a copy
Primer Design
number control (multiplexing suggestions Subheading 3.2.3). To
determine the copy number of the gene of interest (GoI), you will
count the number of double positive droplets, or droplets that
contain both the reference and GoI, plus the droplets that contain
the GoI and dividing by the double positive droplets plus the
droplets containing the reference. Then multiply the resulting
value by 2 (Fig. 1).
3.2.2 Single Nucleotide Three primers are designed for each SNV genotyping assay: wild
Variation (SNV) Assay type and mutant genotyping primers and one common primer
Primer Design upstream or downstream of the genotyping primer. The genotyp-
ing primers should be approximately 20 bp and identical to one
another except for the one with the SNV specific position at the last
30 base position (Fig. 2). IDT’s primer quest is a useful tool for
designing these genotyping assays. The specificity of any primer
Fig. 1 Example CNV Assay: Quanta Soft generated 2D plot of a copy number variation assay targeting one gene
of interest (smaller amplicon) and one reference gene that has been artificially tailed in order to generate a
larger amplicon. Gene of Interest copy number (GoI CN) is determined by counting the number double positive
droplets plus the droplets that contain the GoI and dividing by the double positive droplets plus the droplets
containing the Ref then multiplying by 2
Fig. 2 SNV Genotyping primer design and assay structure. Black arrow represents the shared genomic primer,
the blue arrow represents a wild type allele specific nontailed primer, and the green arrow represents an
artificially tailed mutant allele specific primer
designed without PrimerQuest should be verified using the Blat

tool [2]. Artificial 50 noncomplementary tails to one genotyping
primer allows for differentiation between droplets that contain the
mutant or wild type alleles based upon their differential amplicon
lengths (and consequent difference in fluorescence amplitude due

to the difference in amount of dye binding per target molecule).
The artificial tails minimize bias within the PCR reaction by allow-
ing our primers to target an identical region of genomic DNA (see
Subheading 3.2.3 for multiplexing suggestions).
3.2.3 Modifying Single- In contrast to typical single-color multiplexing that involves two
Color SNV Assays unique amplicons [3], we utilize the addition of noncomplemen-
for Multiplexing tary 50 tails artificially increases amplicon length. This allows us to
target a single genomic region that only differs if the targeted
mutation is present; the presence of the mutation will result in
2 PCR products with unique lengths. Long tail sequences should
be “AT” rich while short tails should be “GC” rich in order to
minimize melting temperature variation. To multiplex the SNV
assays, artificial tails should be added to either the wild type or
mutant genotyping primer (see table below for examples). For
example, to optimize a multiplexed SNV genotyping assay the
user should create multiple iterations of tailed primers by adding
repeats of “AAAT” to the 50 end of the mutant or reference specific
primer in order to achieve 12, 24, 36, 48, 60, 80, and 100 non-
complementary bases. All seven possible tailed primers sets (com-
mon forward + tailed reverse) should be individually multiplexed
with the nontailed reference allele-specific primer set in a ddPCR
assay. When designing multiplexed assays testing a variety of
increasing tail lengths helps to ensure that the resulting reference
and mutant amplicons provide a distinct ddPCR signals. Figure 3
shows an improvement on distinct signal between the mutant and
reference populations that results from a 12 bp and a 100 bp tail
added to the mutant specific primer. Adding long tails can signifi-
cantly increase the melting temperature of primers; the user may
add short “G” or “C” rich tails to the common genomic primer and
the non-long tailed detecting primer to bring the melting tempera-
tures of each individual primer within 3 C of one another while
maintaining amplicon length uniqueness.
3.3 Preparing Forward and reverse primers may be combined when diluting to
Primers and Template 10 μM and then added to the PCR master-mix. Ensure that both
DNA for ddPCR Assays primers exist at 10 μM each in the final primer mix.
3.3.1 Preparation
of Primers
3.3.2 Restriction Digest Template DNA should be digested using a restriction endonuclease
Prior to ddPCR according to the manufacturer’s protocol prior to use in ddPCR.
Ensure that the cut site of the enzyme is outside of any targeted
amplicons. High quality DNA and DNA extracted from FFPE
tissues should be digested with a restriction digest; however
ccfDNA extracted from plasma does not require restriction
Fig. 3 Separation of SNV populations for multiplexed mutation detection. KRAS G12V multiplexed genotyping
assay of a 50% KRAS G12V mutant DNA template. Repeats of “AAAT” are used to artificially “tail” the 50 end of
the mutant detecting primer (reverse) with either 12 or 100 bp. The overall amplitude of the mutant population
increases in channel 1 (y-axis) as the length of the 50 artificial tail increases (left versus right) resulting in
better mutant versus reference population separation. Representative green and blue ovals placed over
mutant and wild type populations respectively
digestion prior to use. The digest should be performed prior to

ddPCR so that the digested template may be quantified before
running SNV or CNV assays. This quantification allows the user
to create a standard curve at a concentration equivalent to that of
the patient sample before each run.
3.4 ddPCR Assay (Adapted from Bio-Rad Product Insert # 10028376 Rev C.)
Preparation
1. All High Quality DNA and DNA extracted from formalin-
fixed, paraffin-embedded tissue (FFPET) should be digested
with a restriction endonuclease that does not cut within the
targeted amplicon.
2. Digested template DNA may be added to the ddPCR master-
mix directly, but the final amount should not exceed 20 ng per
22 μL reaction. See Note 2 regarding optimum template con-
centrations for SNV and CNV reactions.
3. Thaw all reagents at room temperature. Vortex all primers and
spin down in a table-top microcentrifuge for 5–10 s. Mix by
pipetting the QX200™ ddPCR™ EvaGreen Supermix
15 times with a P1000; spin down briefly.
4. Prepare the ddPCR PCR master-mix by combinin QX200™
Evagreen Supermix, shared (common) primer, reference allele
Table 1
Example SNV assay master-mix preparation
Reagent 1 reaction Final concentration

QX200 ddPCR™ EvaGreen Supermix 11.0 μL 1
10 μM Common Primer Variable 100–400 nM
10 μM Reference Allele Primer 50–100 nM
10 μM Mutant Allele Primer 50–100 nM
Template DNA Variable Up to 20 ng
UltraPure™ Water Remaining volume to 22 μL –
primer, mutant allele primer, and template DNA and water to a

final volume of 22 μL per reaction as suggested in Table 1 (see
Note 3 about multiple samples and buffer controls).
(a) The final concentration of the common primer in the
reaction should be equivalent to the concentration of the
reference and wild type allele primers. (i.e., 100 nM com-
mon primer, 100 nM reference allele primer, 100 nM wild
type allele primer).
3.5 Generating (Adapted from Bio-Rad Manual #10031907)

Droplets
1. Add 20 μL of the thoroughly mixed master-mix to the middle
row of wells on the DG8™ cartridge that is snapped into the
cartridge holder. Pipette only to the first stop to ensure that no
bubbles form in the well (see Note 4).
2. Pipette 70 μL of QX200™ Droplet Generation Oil for Eva-
Green into the wells of row 1 into the wells labeled “oil.”
3. Affix gasket over the cartridge holder and place into Droplet
Generator, press the button to close door and begin droplet
generation.
4. Using a P50 multichannel pipette aspirate 41 μL of droplets
from the outlet well row using the slowest aspiration rate
possible. This should take between 5–10 s.
5. Dispense 41 μL of the droplets into a eppendorf 96-well twin-
tech semiskirted plate using the slowest dispense rate possible.
This should take between 5 and 10 s.
6. Place one pierceable foil heat seal on top of the plate with the
red strip facing up.
7. Place plate with heat seal in the plate sealer set to 180 C for 5 s.
8. Carefully transfer the sealed plate to a thermal cycler.
Table 2
Thermal cycling conditions for ddPCR reactions
Cycling step Temperature ( C) Time (mm:ss) Ramp rate Number of cycles

Set lid temperature to 105 C
Enzyme activation 95 5:00 2 /s 1
Denaturation 95 0:30 40
Annealing/extension Variablea 1:00–2:00
Signal stabilization 4 5:00 1
90 5:00 1
Hold 4 1 1
a
Each primer set’s annealing/extension temperature should be individually optimized using a thermal gradient within the
primer’s expected melting temperature. The annealing/extension time may also be adjusted to improve ddPCR signal.
Additional information may be found in Bio-Rad’s Droplet Digital PCR Applications Guide
3.6 Thermal Cycling 1. The thermal cycling conditions for both the SNV and CNV
Conditions assays retain a majority of the same steps, however there is
variation from primer set to primer set based on primer anneal-
ing temperatures. Table 2 outlines a general ddPCR protocol.
3.7 Reading Droplets (Adapted from Bio-Rad Manual #10031906)

and Data Analysis
1. Create a new template using QuantaSoft. For both CNV and
3.7.1 Preparing Template SNV assays be sure to select SuperMix: QX200 Evagreen
File and Reading Droplets ddPCR Supermix then select Assay Type: CNV1, and ensure
that data will be collected for both channel 1 and channel 2 by
selecting their individual boxes. Apply these settings for each
well that contains an assay. Add experiment specific informa-
tion into the provided boxes before saving the template.
2. After opening the Droplet Reader door, unsnap the top secur-
ing plate and insert the cycled-plate onto the QX200™ Drop-
let Reader in the correct orientation. Return the top securing
plate and snap in place before closing the droplet reader door.
3. Open the appropriate template and click “Run” on the left
hand side of the QuantaSoft program. Specify the preferred
data collection direction (across rows or down columns) for the
range of wells specified in step 1 to be read.
3.7.2 QuantaSoft (Bio-Rad Manual #10031906)

Analysis
1. After the run has completed select “OK” to return to the
and Exporting Data
template main page.
2. Select wells to be analyzed and click “Analyze” on the left hand
side of the screen.
3. The automated clustering provided by the QuantaSoft software
will not differentiate between multiple positive populations for
the single-color assays. All Clustering on the program should

be done manually, and you may use the provided selection tools
to cluster populations.
4. To use any other clustering algorithms, you may export raw
amplitude data select all wells with data and return to the
template main screen. Select options and then export cluster
and amplitude data. The resulting file will be a “.csv” that
contains the channel 1 and channel 2 amplitudes as well as the
cluster that was either automatically or manually assigned in the
QuantaSoft software to each droplet.
3.7.3 Analyzing CNV Data Copy number is determined by counting the total number of
positive droplets amplified by the region of interest primer set and
comparing this to the total number of positive droplets amplified by
the reference primer set. The clustering algorithm used to deter-
mine gene copy number analyzes each data point individually in
order to locate the center of a cluster [4]. The user may then define
a threshold that is used by the algorithm to assign a cluster number
to each data point.
3.7.4 Analyzing SNV Data The multiplexed single-color ddPCR genotyping assay can be used
to determine the absolute quantity of reference and variant alleles
present within a given sample. Similar to the CNV assay, the num-
ber of reference or variant alleles can be determined by feeding the
raw amplitude data into a standard clustering algorithm that
accounts for Poisson distribution. The user may then count the
number of positive events in either the mutant or wild-type positive
clusters. In order to generate confidence intervals for calling
mutant and wild type events, we run a series of standard curves
parallel to each patient sample containing assay (Fig. 4).
Fig. 4 Example In-Line Standard Curve. 2-Dimensional plot of raw channel 1 and channel 2 amplitude data for
the KRAS G12V standard curve assay at mutant fractional amounts of 50, 10, 1, and 0%
4 Notes
1. In addition to the use of a PCR workstation and to reduce the

risk of contamination pre and post-PCR. Pre- and post-PCR
workspaces should be in physically separate locations with
designated equipment and consumables. All workspaces and
instruments should be thoroughly decontaminated once a
week with a 10% bleach solution for 10 min and a subsequent
wipe down with 70% ethanol solution.
2. We have found that the positive populations have the best
resolution when using approximately 10 ng (~3000 genome
equivalents) of template DNA per reaction in CNV assays, and
1–20 ng (~300–6000 genome equivalents) of template DNA
for SNV assays. For SNV assays multiple replicates should be
run in order to increase the total number of templates being
assayed. This will help boost signal and reduce noise.
3. Reaction mixes may be made in batches depending on the
variable being tested. Pool as many common reagents as possi-
ble to reduce variability of results. If <8 samples are to be run,
fill all remaining sample wells with 20 μL of 1
QX200™ ddPCR™ Buffer control for EvaGreen.
4. Gently dislodge any bubble with a clean P20 pipette tip prior to
starting droplet generation.
Acknowledgments
This work was supported by the National Institutes of Health

(NIH) grant NHGR1 P01 HG00020526 (H.P.J., C.M.W.B).
References
1. Gray ES, Rizos H, Reid AL, Boyd SC, Pereira 3. McDermott GP, Do D, Litterst CM, Maar D,
MR, Lo J et al (2015) Circulating tumor DNA Hindson CM, Steenblock ER et al (2013) Mul-
to monitor treatment response and detect tiplexed target detection using DNA-binding
acquired resistance in patients with metastatic dye chemistry in droplet digital PCR. Anal
melanoma. Oncotarget 6(39):42008–42018. Chem 85(23):11619–11627. https://doi.org/
https://doi.org/10.18632/oncotarget.5788 10.1021/ac403061n
2. Miotke L, Lau BT, Rumma RT, Ji HP (2014) 4. Rodriguez A, Laio A (2014) Machine learning.
High sensitivity detection and quantitation of Clustering by fast search and find of density
DNA copy number and single nucleotide var- peaks. Science 344(6191):1492–1496. https://
iants with single color droplet digital PCR. doi.org/10.1126/science.1242072
Anal Chem 86(5):2618–2624. https://doi.
org/10.1021/ac403843j
Chapter 19
A Universal Droplet Digital PCR Approach for Monitoring

of Graft Health After Transplantation Using a Preselected
SNP Set
Julia Beck, Michael Oellerich, and Ekkehard Schütz
Abstract
Transplanted organs release cell-free DNA into the bloodstream of the recipient. This graft-derived cell-free
DNA (GcfDNA) is a sensitive biomarker for organ health, since higher GcfDNA levels are indicative of
increased cell-death in the graft. This protocol describes a method to measure relative GcfDNA concentra-
tions by ddPCR assays. The method uses a set of preselected SNP assays from which the informative SNPs
for each recipient–donor combination are selected in a straightforward two-step procedure that requires
only one blood draw. Sampling of donor tissue and separate genotyping is not required, rendering the
technique applicable also to patients, whose transplantation was not recent. In these patients there will be
mostly no access to donor DNA anymore.
Key words Transplantation, Rejection monitoring
1 Introduction
Graft-derived cell-free DNA is present in the circulation of patients,

who have received organ transplants and it has gained high interest
as specific biomarker for graft health and injury [1]. Besides the low
amount of the total cell-free DNA that is extracted from the plasma,
the GcfDNA is only a minor fraction with as little as ten genomic
copies per mL of plasma. Because of the said minute amounts of the
analyte, the accurate quantification of GcfDNA by any method
relies on the specific detection of graft-specific genetic markers
and was first restricted to the detection of Y-chromosome
sequences in gender-mismatched donor–recipient pairs [2]. The
use of HLA-mismatches enables GcfDNA testing [3], but requires
upfront knowledge of the HLA genotypes and laborious assay
optimization for every donor–recipient pair [4] for several hundred
known HLA genotypes. With the advance and wide availability of
single molecule counting techniques, such as high-throughput
335
336 Julia Beck et al.
Fig. 1 The first screening step is performed in a real-time PCR using the recipient’s genomic DNA extracted
from buffy coat. In this step all SNP assays for which the recipient has a heterozygous genotype are
eliminated, because they cannot be used in the quantification ddPCR. The next assay selection step uses
preamplified cfDNA as template in ddPCR defining the final informative assay set for the individual patient. An
informative assay detects an SNP that is homozygous in the recipient and for which the graft carries the
heterologous allele, either in heterozygous or in (preferred) homozygous state. The percentages and numbers
of assays given for each selection step are calculated for a minor allele frequency of 0.5 and vary between
individual patients
sequencing (HTS) and droplet digital PCR (ddPCR), GcfDNA

quantification became feasible as biomarker for graft health [5–7]
under routine conditions. Both aforementioned techniques target
informative single nucleotide polymorphisms (SNPs) (i.e., SNPs
with different alleles in donor and recipient; Fig. 1), for which the
two distinct alleles are directly counted, resulting in a percentage
measure of GcfDNA. Higher levels indicate increased cell death in
the donor organ. The thresholds below which transplanted organs
are considered stable are about 10% for livers [7], 0.25–0.5% for
hearts [5, 8], and about 1% for kidneys [8]. Compared to HTS, the
quantification by ddPCR is less expensive (less than $200 per test),
enables shorter turnaround times (about 24 h) and can be per-
formed with even one single sample at the same costs. The ddPCR
method described in Beck et al. [7], which is explained in more
detail herein, is based on a set of predefined and optimized assays,
which are targeting SNPs that were preselected for having high
A Universal Droplet Digital PCR Approach for Monitoring of Graft Health. . . 337
minor allele frequencies (MAF). Based on Hardy–Weinberg equi-

librium calculations a SNP with a MAF between 0.4 and 0.5 has a
nearly equal distribution of both alleles in the given population and
therefore has a chance of 11.5–12.5% of having a different (homo-
zygous) genotype in two individuals of this population (e.g., donor
and recipient) (Fig. 1). To identify at least three such informative
SNPs for each donor–recipient combination a set of only 30–35
need to be tested once upfront. By using such a preselected set of
SNPs in conjunction with the described screening approach for the
individual patient, it is possible to interrogate the informative SNP
set using just one blood-draw from the recipient. Sampling of
donor tissue and separate genotyping is not required, rendering
the technique applicable also to patients, whose transplantation was
not recent. In these patients there will be mostly no access to donor
DNA anymore. The method has already been proven useful to
confirm the lower therapeutic tacrolimus ranges in a set of
10 liver transplanted patients [9] and to closely monitor a marginal
donor liver [10] and has been used in more than 500 samples from
107 liver transplant recipients as part of a prospective multicenter
trial [11]. The described method includes a preamplification of the
total extracted cfDNA has been shown to result in CVs < 15% for
GcfDNA concentrations of 2% [7]. Also, the workflow for infor-
mative assays selection is described.
2 Materials
2.1 Sample 1. Either: EDTA-anticoagulated blood in collection tubes (any

Collection vendor) Or: Cell-Free DNA BCT Tubes (Streck, Omaha,
USA) (see Note 1).
2.2 Extraction 1. 15 mL Falcon® tubes (any vendor).

of cfDNA/Cell 2. High Pure Viral Nucleic Acid Large Volume Kit (Roche
Associated DNA Applied Science, Mannheim, Germany).
3. DNeasy Blood and Tissue Kit (QIAGEN, Hilden, Germany).
4. DNA LoBind Tubes (Eppendorf, Hamburg, Germany).
5. 0.1 Tris–EDTA, (Low TE) pH 8.0.
2.3 Real-Time PCR 1. FastStart Taq DNA Polymerase, dNTP Pack (includes: 10
on Buffy Coat DNA PCR reaction Buffer with 20 mM MgCl2, MgCl2 stock solu-
(Recipient Germline) tion (25 mM) and PCR grade dNTP mix; Roche Applied
Science, Mannheim, Germany).
2. Primers and probes acc. to Table 1 (any vendor).
3. Real-time PCR instrument (any vendor).
Table 1
338
Assay
name Primer.F Primer.R Probe.A (50 -FAM/30 -BHQ1) Probe.B (50 -HEX/30 -BHQ1) MgCl2
S38 TCAATCCTCACAACTTCCCTAAGGG AGTGGGAGGGAGGTACAGTGA aaaagggggtggtgtCaatgtc agggactgacattCacaccacc Yes
S43 GTCTCTGGGGGTCTGTTGGCC AGAGGAAGGACTCCCAGGGGG tggagacgggtccgCagag tggcacaggtgctctCcgg No
S46 TCCAGCAGAGGAAATAGTACTTGC AGCCACCTGGTCTCCTTTCA ctgggagagaaagaacaaaCagcat catttccccaaatgctCtttgttct Yes

Julia Beck et al.
S48 GATCAACTCCTGAAGAGACTCCGT AGGGAGGGATGGAGAGGGAC cgggagccctgcgCtttg tttccatgacaaaCcgcaggg No
S50 TCTTGTCGAGGCTGCCCTGAAAGG ACAGAGCCGGCCGGTCGC cggttttcgctcCcgtgaa agtccatttcacgCgagcg Yes
S53 CAAAGAGCTCRAACCCCAAG TGTGGGCAAGGCAGGACT aggCctgggtggagaagt ccagCccttgtctcaaaagcc Yes
S54 AAGTAGGTCTAATTTCAAGAATATCTAA TTTATATACTTTAAACCTGTATTTCTAATTAC atgaaaccaagcagtaCtgtggaat accaacaaattccacaCtactgct Yes
S55 TGGTTAAACTGTAGTACATCCATGGA ACCTTTTGGGACTGGCTTTCT acttctcagcaacagCctgga ctctggaaattcatccagCctgt Yes
S57 CAGCCTCTGGTTCCAGGCCT GGAGAATCCCAGAAGCAGGCTGA cactcacgtttgggatacttCgtttc cccagtaaggaatggagaaacCaagta Yes
S58 GCCACCTTAGCCTCCCAAAG AGGGTGACTGTATTAATTATTGTTCAAACT attacaggcatgagCcaccg caaggcacggtgCctcat Yes
S59 AGAAAGAAAGAAGCAGGGAAGGGAC TGGAGCTAAAATGAGCCTGCGT attacatagcttatcaCttgcagagcc actcctggctctgcaaCtgat Yes
S63 GCTGTTGCTGCCTCACAGGT AGGGCAAAGGCAAATGCACCA aactggaagtaacacCtgcacca cttgactcttggtgcaCgtgt No
S66 ACCCTGACCCTCAGTTCCTT AAGAGCCCTTATAAGGTGTGAGAAA aggatattgctagagtggagtCagaac accactgttatttgttctCactccact No
S67 ATGAAGAGTAAGCGGGGCCG CGGACCCATTTCACCCACCA cccgacccttaacCtcccc tggagagggttggggaCgtta No
S68 GCCTCTGCCATATCCTCAC AGGTCGGATGTTGGAAAGG agacaCttgtgggactcagaagg acaaCtgtctcctgctgtcct Yes
S70 TGGCCCAGTTAGAAGGTGTGGA CGGCCACCCATCCTGGAGAT accctcctgtactgCgcac acagtgaaggtgtgcCcagt No
S77 GGGCCTCAGTTCTAGACGAGT GTTTCCGTGAAGTAGGCGCT atgctcagcacacAgggga cactgcttccccCgtgtg Yes
S78 AGGCAGAACTAAACGTTGGCTT TGCGGAACAGTGACAATTTGTTC atgcAgctttggcatgaggt atgccaaagcCgcatattttctct No
S79 CAGGGAGTGCTTTACTGAGGC ACTCAAACACGGAGCTGGGC ggcagcaggtgccAagca aggcattactgctCggcacc Yes
S82 TTTGCACTTGACGCACCAGC CCGAGGCAGAGGAAGGAAGTG tgcAatgagagcagaggcct catCgcagccctcctgca No
S83 GGTTTTGCTTCTGATGATCCCTCT AGCATTGTGTAGGGACTGGTAAATT atacActctgttgttgagtgccac cagagCgtatgtatgaagtccagagt No

S84 CCTAATCACTCGTGAGGAGTG ATCCACCATGATGCTCACAA cccaCgggaggaatgtctttg cccAtgggacttctggcc Yes
S85 GGAAACATCTAGACGCGCA CGGAATCGGGAGGCCAC acacaAagtggcctcccg acaGagtggcctcccgat Yes
S86 GTCTCCCTCCCCAAAGGTGC GCCAACCTCAAGGGGCAGTT aggaaagaaacctttcAgatgtcagt tgaggattaactgacatcCgaaaggt Yes
S87 GGCATCTGAATTCAAGCTTTGGTC TTCTTCTAGTTGGTCTGGTAGGCT aggcttgtacactCtccccc acactgggatgggggaAagt Yes
S88 TGGTTATTGTTACTAGGTCCCCACC AGAATAAGCAAGATGTTGGCAGTGAG aggactttattggggaggCtgac ctggaagccaaagtcaCcctc Yes
S90 AAGGAAAAGGAATTCCTGGC CCTGCCTATGCTCAGGCA cagTgccctctgccaggaa gggcCctgcctgagcatag Yes
S92 TTTATTTAAATGACTGTCCAGGTC TTTCACAGACCTTCAAACCAC ccCgcagttgcacagcttg actgcAggccacaaggtg Yes
S94 CTGGGGCAGAGTGGAGAGTC ATCCACCTCTGAACCCAGCC aggacActgcagctgtgg cagCgtcctctgtgctacct No
S96 TCCCAGGCTCCAGGTCAGAT GGATCAATGTGGCTGCTCCCT tctcCgcccttctgagatgc agggcAgagactctggaact No
S97 AGCCCTGCACACTCACTTACC TGGCATTCAGATCATCAGGCTTCT ccatcaggtgctggcActc tgcagggaagagCgccag Yes
S99 GGCAAAGTGGGCAAGGGTCT GCCTCCTAAAGCTTGAGCCACA ttggggccaGgtacctgg tggggccaAgtacctggt No
S102 AACAGTGGCAGCCCTCTTGT ACACTTGGTTCATGGGGTTGTG tggccttatctttggccctaaCatg aggcacatcctacatCttagggc Yes
S103 TTCTATATGAATTTCTCTTCTATTTCTAAA AAGCAGTCAGGGAAGTATCC ccctggggccatcaGgtt ccctggggccatcaAgttt Yes
S105 ACCCCAAGAGGCTTTATAGGGG CCTTCCCAACGGGTTTGACC ccactgggctggCccctc agtggaggagggAccagc Yes
S108 ACACTCCTGCTGCGTGTCTG TTCCTCCCCACCACTCCCAT ggtcccagctggtCgtgg atgctccccacAaccagct Yes
S110 GGTCCTACCGAGGTGGGTGA CATTGCCAAGGACAGAGGGAGA tttggtagggaaggaactcCcaat atcagtggccattgTgagttcc No

Capital letters in probe sequences indicate the position of the inferred SNP
A Universal Droplet Digital PCR Approach for Monitoring of Graft Health. . .
339
2.4 Preamplification 1. NEBNext Ultra II DNA Library Prep Kit (New England Bio-
of cfDNA labs, Ipswich, USA).
2. NEBNext Singleplex Oligos for Illumina (New England Bio-
labs, Ipswich, USA).
3. 80% ethanol (freshly prepared).
4. Nuclease-free water.
5. 0.1 Tris–EDTA, (Low TE) pH 8.0.
6. EvaGreen® dye 20 (any vendor).
7. P5-Primer-Transplant (50 -CCTACACTCTTTCCCTACACG
ACGCTCTTCCGATCT-30 ).
8. P7-Primer-Transplant (50 -GTGACTGGAGTTCAGACGTG
TGCTCTTCCGATCT-30 ).
9. AMPure XP Beads (Beckman Coulter, Brea, USA).
10. DNA LoBind Tubes (Eppendorf, Hamburg, Germany).
11. Magnetic rack/stand (any vendor).
12. Real-time cycler (any vendor).
13. PCR cycler (any vendor).
14. PCR plates according to recommendations of instrument
vendor.
2.5 Droplet 1. QX200 ddPCR System (Bio-Rad, Hercules, USA).

Digital PCR 2. PCR Plate Sealer (any vendor).
3. PCR Machine (any vendor).
4. ddPCR Supermix for Probes (Bio-Rad, Hercules, USA).
5. Primer and probes acc. to Table 1 (any vendor).
6. Droplet Generation Oil for Probes (Bio-Rad, Hercules, USA).
7. DG8 Cartridges for QX200 Droplet Generator (Bio-Rad, Her-
cules, USA).
8. DG8 Gaskets for QX200 Droplet Generator (Bio-Rad, Hercu-
les, USA).
9. Piercable Foil Heat Seal (Bio-Rad, Hercules, USA).
10. 96-well PCR plates (twin.tec semiskirted, Eppendorf, Ham-
burg, Germany).
2.6 SNP Probe 1. A set of assays targeting 38 different SNPs is used.

Hydrolysis Assays 2. All SNPs comply with the following criteria:
(a) Known and validated MAF of 43% in Caucasian whites
and over all reported ethnicities (public databases: Hap-
map or 1000Genomes).
(b) Not located within or directly adjacent to an annotated

repetitive element.
3. SNP-specific probes were designed using thermodynamic
nearest-neighbor model calculations with a desired Tm of
65 C at standard PCR buffer conditions (e.g., 0.18 mol/L
monovalent cation equivalents [12] and 500 nmol/L DNA/-
Primer) and were selected for maximum Gibbs free energy
difference between allele binding (match/mismatch) at the
given conditions. For each SNP one allele-specific probe carries
a FAM fluorophore while the other SNP carries a HEX fluor-
ophore in conjunction with BHQ1 as quencher. Respective
PCR primers were designed to exhibit a melting temperature
of 68 C and a binding efficacy of 95% at 60 C, thermody-
namically assessed using published formulas [13]. All probes
and primers are given in Table 1.
2.7 Analyses 1. Quantasoft software (Bio-Rad, Hercules, USA).

2. Excel (Microsoft, Seattle, USA).
3 Methods
3.1 Informative Figure 1 depicts an overview of the procedure employed to deter-

Assay Selection mine the informative assays, personalized for each recipient–donor
combination. To select those SNPs with a homozygous genotype in
the recipient, the patient’s germline DNA—extracted from white
blood cells—is tested for all SNPs in real-time PCR format. The
following step uses preamplified cfDNA as input in a ddPCR for
testing only those assays, in which a homozygous recipient geno-
type was detected in the first step. Every assay showing an allele that
is heterologous to the recipient is an informative assay. SNPs with
heterozygous alleles in the graft show half of the percentages as
compared to homozygous alleles. The assay selection procedure is
applied only once for each patient, informative assays are stored in a
database (Excel, Filemaker, SQL, etc.), and all subsequent samples
are tested only with the patient’s personal informative assay set.
3.2 Sample 1. Collect 10–20 mL blood per patient in either EDTA-

Collection and Storage anticoagulated blood tubes or Cell-Free DNA BCT tubes
referring to respective protocols for proper handling. For
EDTA blood, the plasma should be separated from blood
cells within 1 h after collection. Blood samples collected in
Cell-Free DNA BCT for cfDNA analysis are stable for 14 days
(according to manufacturer), when stored between 6 and
37 C.
2. Separate plasma from blood cells by centrifugation at 2500 g

for 10 min at 4 C. Remove the plasma carefully without
disturbing buffy coat layer.
3. Plasma should be aliquoted in 1–2 mL portions.
4. Collect buffy coat for extraction of recipients’ germline DNA.
Collection of buffy coat is needed for first time contact
patients, only.
5. Collected plasma and buffy coat should be stored frozen at
20 C if not used immediately.
3.3 Extraction 1. Subject 1–2 mL plasma aliquots to a second centrifugation at

of cfDNA/Cell 4000 g for 20 min at 4 C to remove any remaining cells/cell
Associated DNA debris.
2. Transfer supernatant to a fresh 15 mL tube. Extract cfDNA
using the High Pure Viral Nucleic Acid Large Volume kit
according to the manufacturer’s instructions, but without the
use of poly-A carrier RNA.
3. Make sure to collect final eluate (48–50 μL) in DNA Lobind
tubes with low TE buffer. Store at 20 C until further proces-
sing (see Note 2).
4. For first-time contact patients extract the buffy coat DNA
using the DNeasy Blood and Tissue Kit (QIAGEN, Hilden)
according to the manufacturer’s protocol.
3.4 Real-Time PCR 1. Prepare real-time PCR in a total volume of 20 μL per reaction.
on Buffy Coat DNA 2. Each reaction should contain 2 μL FastStart 10 buffer with
MgCl2, 200 μmol/L each dNTP, 2 units FastStart Taq,
900 nmol/L each primer, 250 nmol/L each probe (Table 1;
if the use of MgCl2 is indicated add 1.8 μL of the 25 mM PCR
grade MgCl2 stock solution to a final total concentration of
4.3 mmol/L).
3. Cycling conditions using a Roche LightCycler 480 are: 95 C
for 10 min, 50 (95 C for 30 s, 65 C for 1 min) with
fluorescence detection (FAM/HEX) in each cycle at 80 C
and ramp rates of 4.4 C/s for heating and 2.2 C for cooling
(see Note 3).
3.5 Preamplification Use the NEBNext II Ultra DNA Library Prep Kit for Illumina
of cfDNA according to the manual with adherences to the following
instructions:
1. Since cfDNA is already fragmented, no initial fragmentation of
the extracted DNA is required. Use the 50 μL cfDNA elution
directly for library preparation.
2. Follow the protocol for 5 ng–1 μg fragmented input DNA.
3. For ligation DO NOT dilute the adaptor tenfold, although this

is suggested for input amounts of <100 ng. Use the NEBNext
Singleplex adaptor (NEB, #E7350).
4. Perform the option 1.3B “Cleanup of Adapter-ligated DNA
without Size Selection” for purification after adapter
ligation step.
5. To the PCR amplification reaction add 1.5 μL EvaGreen dye
20 and 500 nmol/L of each primer P5-Primer-Transplant
and P7-Primer-Transplant instead of the Universal PCR
Primer/i05 and Index Primer/i07 (see Note 4). Perform
PCR amplification in a real-time PCR instrument while closely
monitoring the amplification using the FAM (green) channel
of a real-time PCR instrument. Stop PCR just after the ampli-
fication curves leave the linear phase (see Note 5) and proceed
to “Cleanup of PCR” according to NEBNext protocol.
6. Determine concentration by any appropriate method. Concen-
trations should be between 20 and 100 ng/μL (see Note 6).
Preamplification of library can be completed within <3.5 h.
3.6 Droplet 1. Prepare ddPCR reactions using 11 μL of the 2 ddPCR Super-

Digital PCR mix for Probes (Bio-Rad).
2. Each reaction should contain 30–100 ng preamplified cfDNA
as template, 900 nmol/L of each primer, and 250 nmol/L of
each probe in a total volume of 22 μL.
3. Generate droplets using the QX200 droplet generator or auto-
mated droplet generator.
4. Cycling conditions using a Biometra 3000 Gradient Cycler are:
95 C for 10 min, 50 (94 C for 30 s, 61 C for 1 min), then
98 C for 10 min, ramp rate 2.5 C/s (see Note 3).
5. Some assays are supplemented with additional 2.3 mmol/L
MgCl2 (see Table 1).
6. Assays can be grouped according to their respective annealing
temperature, so that several assays can be run on the same
96-well plate.
7. Positive controls with known heterozygote genotypes as well as
nontemplate controls for each assay should be included in every
run (see Note 6). After cycling droplets are read in a QX200
droplet reader instrument (Bio-Rad) selecting “rare event
detection” and “FAM/HEX assay format.”
3.7 qPCR Analysis 1. Genotypes are called using the “Endpoint Genotyping” mod-
of Candidate Assays ule of the LightCycler 480 software.
2. Manual inspection and recalling might be necessary for some
reactions.
3. Refer to Fig. 2 for example.
Fig. 2 Examplary qPCR assay run on a LightCycler480. A scatterplot of the “Endpoint Genotyping” analysis
module (LightCycler480 Software, version: 1.5.0 SP3) is shown on the left (x-axis: FAM signal intensity, y-axis:
HEX signal intensity). Genotypes called as AA are shown in blue, genotypes called as BB in green and AB
genotypes in red. One of the samples (grey circle) shows lower FAM signal intensity, but is still unequivocally
called as AA genotype
3.8 Droplet Digital 1. Primary analysis is conducted using the Quantsoft instrument
PCR Analysis software.
2. Set the proper thresholds for each assay ensuring the best
separation of the four droplet fractions FAMþ/HEX,
FAM/HEXþ, FAMþ/HEXþ, and FAM/HEX (see
Note 7, Fig. 3).
3. Check nontemplate controls for negativity (see Note 6). Select
“Events” tab and check the box for display of total droplets.
4. Reactions should achieve a minimum of 10,000 total droplets.
5. Check the box for display of positive droplets.
6. Reactions should contain at least 100 positive droplets for the
minor fraction in order to achieve a maximum CV of ~10% (see
Note 6).
7. Select the “Ratio” tab in Quantasoft and check the box for
display of “Fractional Abundance.”
8. Check the heterozygous positive controls for achieving the
expected ratio of 50/50 for the two alleles (see Note 6).
9. Export results to comma separated values (.csv) file.
16000 4513 14000
14000 12000
12000
Channel 1 Amplitude
10000
Channel 1 Amplitude
10000
8000
8000
6093
6000 5486
6000
4000
4000
2000 2000
0 0
0 1000 2000 3000 4000 5000 6000 7000 8000 1 1000 2000 3000 4000 5000 6000 7000 8000 9000
Channel 2 Amplitude Channel 2 Amplitude
10000 2139 9000 1905
9000 8000
8000
7000
Channel 1 Amplitude
7000
Channel 1 Amplitude 6000
6000
5000
5000
4000
4000 3413
3287
3000 3000
2000 2000
1000 1000
0 0
0 1000 2000 3000 4000 5000 0 500 1000 1500 2000 2500 3000 3500 4000
Channel 2 Amplitude Channel 2 Amplitude
Fig. 3 Four examples for ddPCR assays as 2D-plots (Quantasoft, version: 1.7.4) from upper left in clockwise
order: S59 with informative B-allele (FAM channel) (17%); S63 with informative A-allele (FAM channel)
(1.43%); S43 with informative B-allele (HEX channel) (8%); S38 with informative B-allele (HEX channel) (19%)
10. Open .csv file in Microsoft Excel and group all assays per
sample.
11. All fractional abundances (FA) are displayed as A/
(A + B) 100.
12. Correct the assays for which the informative allele is B by
calculating 100-FA.
13. Correct the heterozygote informative alleles by calculating
FA 2.
14. Refer to Table 2 for example.
15. Calculate the mean and standard deviation over all corrected
FA for each sample.
4 Notes
1. Cell-free DNA BCT Tubes contain, in addition to EDTA, a

preservative reagent that prevents DNA release from blood
cells during storage for up to 14 days, when stored at
Table 2
Example calculation for initial ddPCR screen for informative assays
FA as exported 100-FA for FA*2 for Tested Inferred

from informative heterozygous recipient graft
Quantasoft B allele informative alleles Final FA genotype genotype
0 Not BB BB
informative
0 Not BB BB
informative
5.5 11.0 11.0 BB AB
12.5 12.5 BB AA
86.3 13.7 13.7 AA BB
86.9 13.1 13.1 AA BB
88.5 11.5 11.5 AA BB
92.4 7.6 15.2 15.2 AA AB
93.1 6.9 13.8 13.8 AA AB
94.5 5.5 11.0 11.0 AA AB
94.7 5.3 10.6 10.6 AA AB
100 Not AA AA
informative
Mean 12.5
STDEV 1.6
Every row depicts the values for one assay. Final FA at different calculation steps are shown in bold italics. Recipient
genotype results are obtained from real-time PCR using DNA extracted from buffy coat. Note that informative SNPs
have a higher chance of being in heterozygous rather than being in homozygous state in the graft
FA fractional abundance
6–37 C (according to manufacturer). These tubes should be

used if samples need to be shipped or cannot be centrifuged
within 1 h after blood draw.
2. cfDNA concentrations in the eluate show wide individual varia-
bility with high concentrations seen for example soon after
surgery. The median yield from 1 mL plasma was 30 ng (min
0.8 ng/mL, max: 280 ng/mL) in 270 samples obtained from
liver transplanted patients. Due to the low cfDNA concentra-
tions adequate techniques; for example, droplet PCR or at least
fluorometric quantification should be used for accurate mea-
surements. The protocol described here, however, does not
require routine quantification of the cfDNA elutions, instead
the samples can be subjected to preamplification directly. After
a maximum amplification for 15 cycles a sample should then
yield >20 ng library material (also see Note 6).
3. Annealing/extension temperature and/or ramp rates may

need to be adjusted for different thermocyclers. 2 C/s is
recommended by Bio-Rad.
4. Shorter primers than in the original sequencing library proto-
col are used, since the libraries are not intended for sequencing.
5. Amplification reaction should be stopped at the end of the
linear PCR phase in order to avoid over-amplification. Library
amplifications with different amounts of input DNA will have
different stopping points. Discontinue the real-time PCR when
the first libraries reach their inflection point and then estimate
the number of additional cycles for the lower concentration
libraries. Remove the first libraries from the PCR plate and add
more cycles to the others (see Note 6).
6. Quality Controls throughout the whole procedure are:
(a) Amplification curves and final concentration of preampli-
fied libraries: less than 15 cycles needed; yield >20 ng/μL.
(b) Number of total droplets and positive droplets in ddPCR:
>10,000 total; >100 positive for donor allele.
(c) Fractional abundance of heterozygote control DNA in
ddPCR: target: 50%, accepted range: 47–53%.
(d) Positive droplets in ddPCR NTC control: <2 events.
(e) The analysis of at least four informative assays per patient
sample is recommended.
If sample fails quality control checkpoint #1 it is recom-
mended to reextract the plasma cfDNA and repeat library
preparation. The initial extracted plasma volume might be
raised. In case of sample failure for checkpoints #2–#4, repeat
ddPCR.
7. The positions of droplet clusters in the 2D plot of the hetero-
zygous positive control can help to set the thresholds in the
patient samples. Heterozygous genotypes can be identified
from the real-time PCR on patients’ buffy coat DNA and
then used as controls in the ddPCR.
Acknowledgments
The authors thank Sarah Bierau and Stefan Balzer for their excellent
technical assistance.
References
1. Gielis EM, Ledeganck KJ, De Winter BY, Del DNA: an upcoming biomarker in transplanta-
Favero J, Bosmans JL, Claas FH, tion. Am J Transplant. https://doi.org/10.
Abramowicz D, Eikmans M (2015) Cell-Free 1111/ajt.13387
2. Zhang J, Tong KL, Li PK, Chan AY, Yeung E (2015) Donor-derived cell-free DNA is a
CK, Pang CC, Wong TY, Lee KC, Lo YM novel universal biomarker for allograft rejec-
(1999) Presence of donor- and recipient- tion in solid organ transplantation. Transplant
derived DNA in cell-free urine samples of Proc 47(8):2400–2403. https://doi.org/10.
renal transplantation recipients: urinary DNA 1016/j.transproceed.2015.08.035
chimerism. Clin Chem 45(10):1741–1746 9. Oellerich M, Schutz E, Kanzow P, Schmitz J,
3. Gadi VK, Nelson JL, Boespflug ND, Guthrie Beck J, Kollmar O, Streit F, Walson PD (2014)
KA, Kuhr CS (2006) Soluble donor DNA con- Use of graft-derived cell-free DNA as an organ
centrations in recipient serum correlate with integrity biomarker to reexamine effective
pancreas-kidney rejection. Clin Chem 52 tacrolimus trough concentrations after liver
(3):379–382. https://doi.org/10.1373/ transplantation. Ther Drug Monit 36
clinchem.2005.058974 (2):136–140. https://doi.org/10.1097/
4. Lo YM (2011) Transplantation monitoring by FTD.0000000000000044
plasma DNA sequencing. Clin Chem 57 10. Kanzow P, Kollmar O, Schutz E, Oellerich M,
(7):941–942. https://doi.org/10.1373/ Schmitz J, Beck J, Walson PD, Slotta JE (2014)
clinchem.2011.166686 Graft-derived cell-free DNA as an early organ
5. De Vlaminck I, Valantine HA, Snyder TM, integrity biomarker after transplantation of a
Strehl C, Cohen G, Luikart H, Neff NF, marginal HELLP syndrome donor liver. Trans-
Okamoto J, Bernstein D, Weisshaar D, Quake plantation 98(5):e43–e45. https://doi.org/
SR, Khush KK (2014) Circulating cell-free 10.1097/TP.0000000000000303
DNA enables noninvasive diagnosis of heart 11. Schütz E, Fischer A, Beck J, Harden M, Koch
transplant rejection. Sci Transl Med 6 M, Wuensch T, Stockmann M, Nashan B, Koll-
(241):241ra277. https://doi.org/10.1126/ mar O, Matthaei J, Kanzow P, Walson PD,
scitranslmed.3007803 Brockmöller J, Oellerich M, Singal A (2017)
6. Snyder TM, Khush KK, Valantine HA, Quake Graft-derived cell-free DNA, a noninvasive
SR (2011) Universal noninvasive detection of early rejection and graft damage marker in
solid organ transplant rejection. Proc Natl liver transplantation: a prospective, observa-
Acad Sci U S A 108(15):6229–6234. https:// tional, multicenter cohort study. PLoS Med
doi.org/10.1073/pnas.1013924108 14(4):e1002286
7. Beck J, Bierau S, Balzer S, Andag R, Kanzow P, 12. von Ahsen N, Wittwer CT, Schutz E (2001)
Schmitz J, Gaedcke J, Moerer O, Slotta JE, Oligonucleotide melting temperatures under
Walson P, Kollmar O, Oellerich M, Schutz E PCR conditions: nearest-neighbor corrections
(2013) Digital droplet PCR for rapid quantifi- for Mg2+, deoxynucleotide triphosphate, and
cation of donor DNA in the circulation of dimethyl sulfoxide concentrations with com-
transplant recipients as a potential universal parison to alternative empirical formulas. Clin
biomarker of graft injury. Clin Chem 59 Chem 47(11):1956–1961
(12):1732–1741. https://doi.org/10.1373/ 13. Schütz E, von Ahsen N (2009) Influencing
clinchem.2013.210328 factors of dsDNA dye (high-resolution) melt-
8. Beck J, Oellerich M, Schulz U, Schauerte V, ing curves and improved genotype call based
Reinhard L, Fuchs U, Knabbe C, on thermodynamic considerations. Anal Bio-
Zittermann A, Olbricht C, Gummert JF, chem 385(1):143–152
Shipkova M, Birschmann I, Wieland E, Schutz
Chapter 20
Detection and Quantification of HDR and NHEJ Induced

by Genome Editing at Endogenous Gene Loci Using
Droplet Digital PCR
Yuichiro Miyaoka, Steven J. Mayerl, Amanda H. Chan,
and Bruce R. Conklin
Abstract
Genome editing holds great promise for experimental biology and potential clinical use. To successfully
utilize genome editing, it is critical to sensitively detect and quantify its outcomes: homology-directed
repair (HDR) and nonhomologous end joining (NHEJ). This has been difficult at endogenous gene loci
and instead is frequently done using artificial reporter systems. Here, we describe a droplet digital PCR
(ddPCR)-based method to simultaneously measure HDR and NHEJ at endogenous gene loci. This highly
sensitive and quantitative method may significantly contribute to a better understanding of DNA repair
mechanisms underlying genome editing and to the improvement of genome editing technology by allow-
ing for efficient and systematic testing of many genome editing conditions in parallel.
Key words Genome editing, TALEN, CRISPR/Cas9, HDR, NHEJ, ddPCR
1 Introduction
Genome editing allows for manipulation of genomes at will in any

cell type. This can be used to precisely alter a single base, to insert a
sequence of interest at a precise location or to introduce random
insertions and deletions which can disrupt gene function. The tools
for genome editing are sequence-specific nucleases that induce a
double strand break (DSB) or nicks at targeted genomic regions.
The induced DNA damage activates two major DNA repair path-
ways to repair the damage. One is nonhomologous end-joining
(NHEJ), in which the broken DNA ends are joined together with-
out any templates. NHEJ often generates random insertions and
deletions at the site of repair. The other pathway is homology-
directed repair (HDR), in which cells use homologous DNA as a
repair template to precisely repair DNA. By delivering homologous
donor DNA with a new DNA sequence along with editing
349
350 Yuichiro Miyaoka et al.
reagents, the genome can be precisely modified via donor-

dependent repair [1].
When precise donor-dependent HDR is required, it is impor-
tant to measure genome editing outcomes—HDR and NHEJ—for
successful application of genome editing, since NHEJ often pre-
dominates. This task has typically been done by using artificial
reporter systems, deep sequencing, or gel-based assays, or by
detecting only NHEJ as a surrogate for HDR because of technical
limitations [2–5]. None of these methods are suitable for high-
throughput screening. To overcome this hurdle, the method we
describe here utilizes a combination of allele-specific hydrolysis
probes and a new digital PCR technology called Droplet Digital™
PCR (ddPCR™) [6]. Our original ddPCR-based assay with probes
for the wild-type (WT) and HDR alleles was designed to detect
only HDR-mediated point mutagenesis without measuring NHEJ
[7]. (In this protocol, we describe the original allele before genome
editing as “WT,” but the same strategy can be applied to correct
mutations.) The method we describe here [8] contains additional
NHEJ probes to allow simultaneous quantification of both
HDR-mediated point mutagenesis and NHEJ-mediated introduc-
tion of insertions and deletions (Figs. 1 and 2). ddPCR partitions a
reaction into 20,000 nanoliter-scale water-in-oil droplets, so that
each droplet contains from zero to ~10 copies of the genome
targets which are analyzed by HDR and NHEJ allele-specific
hydrolysis probes in individual droplet reactions. The allele-specific
combination of fluorescent signals results in droplets that contain
either HDR, NHEJ, or WT alleles, or combinations of these; these
droplets in turn occupy distinct locations on the two-dimensional
plot, allowing absolute quantification of discrete alleles (e.g.,
Fig. 1c). Therefore, HDR and NHEJ events can be detected in a
highly sensitive and quantitative manner.
Measuring genome editing outcomes is critical in evaluating
genome editing tools (ZFN, TALEN, CRISPR/Cas9, etc.) and
conditions (concentration of nucleases, target sequences, donor
types, etc.). This protocol provides researchers with a rapid way to
measure HDR- and NHEJ-inducing activities of genome editing
conditions.
2 Materials
Prepare all solutions using ultrapure water (prepared by purifying

deionized water to attain a sensitivity of 18 MΩ cm at 25 C) and
analytical grade reagents. Use filtered tips and compatible pipetters
(such as RANIN) to avoid contamination of reagents in sample
preparation and droplet generation. Prepare and store all reagents
at 20 C (unless indicated otherwise). Diligently follow all waste
disposal regulations when disposing of waste materials.
Detection and Quantification of HDR and NHEJ Induced by Genome. . . 351
Fig. 1 Separated cut and mutation sites: Assay design, example target sequence, and 2D droplet plot of edited
genomic DNA. Depending on the editing strategy and the relative positions of cut site and edit site, the
positions of assay probes and the need for a competitive blocking (dark) probe vary. The reference and HDR
probes are constant regardless of the location or number of cut sites. (a) Assay design when cut site and
mutation site are separated. Outcomes for three different alleles are shown (retention of WT sequence,
alteration of a single base via HDR, or indel via NHEJ). The HDR and NHEJ probes are located on nonoverlap-
ping mutation and cut sites, respectively. To prevent nonspecific binding of the HDR probe to the original “WT”
sequence, a nonfluorescent (dark) “WT” probe is included that competes with the HDR probe for WT allele
binding (see Figs. 2, 4, and 5 for other cases). (b) Assay design when cut site and mutation site overlap.
See Fig. 4 legend for details. (c) Assay design when two cut sites are introduced. See Fig. 5 legend for details
2.1 Reagents 1. 20 FAM and HEX hydrolysis probe and primer assay mix-
tures in water. Components are shown in Table 1.
2. WT genomic DNA in water adjusted to 100–150 ng/μl.
3. Synthetic double-stranded DNA (gBlocks, Integrated DNA
Technologies) that contains the point mutation at the desired
edit site (HDR control) or a 2-bp deletion at the predicted
nuclease cut site (NHEJ control). Lyophilized gBlocks are
Fig. 2 Comparison of three ddPCR assay designs for three different editing designs. (a) Assay design when cut
site and mutation site are separated. See Fig. 1a legend for details. (b) Assay design when two cut sites are
introduced. See Fig. 4 legend for details. (c) Assay design when cut site and mutation site overlap. See Fig. 5
legend for details
Table 1
Components of 20 assay mixtures
Component Concentration (μM)

Forward primer 18
Reverse primer 18
Reference probe (FAM) 5
HDR probe (FAM) 5
NHEJ probe(s) (HEX/VIC) 5 (each, if multiple probes are used)
0
Dark probe (3 phosphate is added) 10 (depending on the assay)
resuspended in 250 μl TE + 100 ng/μl polyA carrier. Two

additional 200-fold dilutions in TE + polyA result in a master
stock of around 40,000 copies/μl that is maintained in LoBind
tubes (Eppendorf) (see Note 1). Dilute the master stock by
20-fold to make a solution of around 2000 copies/μl for use.
Note that these are concentrations estimates and can be off by

tenfold in either direction.
4. Restriction enzyme that does not cut within the amplicon.
HindIII-HF, CviQI, MseI, AluI, HaeIII restriction enzymes
(NEB) have all been validated to work directly in the ddPCR
Supermix. 2–4 U restriction enzyme can be directly added to a
20 μl ddPCR reaction without additional incubation.
5. ddPCR™ Supermix for Probes (No dUTP) (Bio-Rad
186-3024).
6. ddPCR™ Buffer Control Kit (Bio-Rad 186-3052).
7. Droplet Generation Oil for Probes (stored at room tempera-
ture) (Bio-Rad 186-3005).
8. DG8™ Gaskets for QX200™ Droplet Generator (Bio-Rad
186-3009).
9. DG8™ Cartridges for Droplet Generator (Bio-Rad
186-4008).
10. Pierceable Foil Heat Seal (Bio-Rad 181-4040).
11. Eppendorf twin.tec 96-well plates, semiskirted (Fisher
951020346).
12. Rainin filter pipette tips (see Note 2).
13. Genomic DNA in distilled water or TE isolated from cells
treated with genome editing tools to induce HDR
and/or NHEJ.
2.2 Equipment 1. QX100™ or QX200™ Droplet Digital™ PCR system

(Bio-Rad 186-4001).
2. PX1™ PCR Plate Sealer (Bio-Rad 181-4000).
3. DG8™ Cartridge Holder (Bio-Rad 186-3051).
4. 96-well thermocycler (such as Bio-Rad C1000 that has a
uniform block with deep wells, 2 C/s ramp rate, and temper-
ature gradient capability).
5. Rainin 20 μl eight-channel pipette.
6. Rainin 50 μl eight-channel pipette (see Note 3).
3 Methods
To detect HDR and NHEJ at the same time, four different kinds of
probes are designed within a single amplicon [8]. The first is a FAM
(reference) probe that is nonoverlapping with the cut site and
which always binds to the genomic DNA. It provides a positive
FAM amplitude for counting total genome copies in the sample
(contributing to a FAM+ or FAM++ signal, Fig. 1). The second is a
HEX (NHEJ) probe that binds at the cut or nick site of the
genomic locus. It has a wild type sequence, so in the case where
NHEJ occurs and results in either insertion or deletion of
sequences at the cut site, the probe will no longer be able to bind
resulting in a loss of HEX signal. Loss of HEX signal with retention
of the reference probe FAM signal (FAM+, HEX) identifies
molecules that have undergone NHEJ (Fig. 1b, c). The third
probe is another FAM probe (HDR probe), which will bind to
the DNA when precise edits have occurred indicating an HDR
event. In the case of HDR, the FAM amplitude will further increase
separating the HDR population from the wild type (FAM++, HEX
+). The fourth probe is a nonfluorescent (“dark”) probe having
“WT” sequence which prevents non-specific-binding of the HDR
probe to the unaltered “WT” allele (see Fig. 2, assay designs a
and c).
3.1 Design 1. Modify the settings of Primer3Plus (http://primer3plus.com)

of Hydrolysis Probes compatible with the master mix: 50 mM monovalent cations,
and Primers 3.0 mM divalent cations, 0 mM dNTPs, SantaLucia 1998
thermodynamic and salt correction parameters (see ref. 9 and
further guidance below).
2. For the target region of interest, position the predicted nucle-
ase cut sites mid-amplicon, with 75–125 bp flanking either side
up to and including the primer binding sites (see Note 4).
3. Set melting temperatures for primers, reference probes, NHEJ
probes, HDR probes, and Dark probes as 55 1 C, 60 C,
57 1 C, 55 C, and 57 C, respectively.
4. At least one primer must be positioned outside the donor
molecule sequence to detect integrated edits (Fig. 1b).
5. Reference probe and primers are designed distant from the cut
site to avoid loss of binding sites by NHEJ.
6. In some cases, a dark, nonextendible oligonucleotide (30 phos-
phorylation) is designed to block cross-reactivity of the HDR
probe and the WT sequence. This is generally required if the
HDR edit site and nuclease cut site do not closely overlap
(Figs. 1a and 2a, c). If they do directly overlap, the NHEJ
probe can serve as the competitive blocker (Fig. 2b). When
measuring NHEJ induced by dual nuclease systems, two NHEJ
probes should be designed for the two cut/nick sites (Fig. 2c).
7. Probe position and number vary depending on the relative
positions of the cut site(s) and edit site (Figs. 1 and 2). Our
assay for RBM20 is shown as an example (Fig. 1b).
3.2 Assay Validation 1. Mix the reagents shown in Table 2 in a well of a 96-well plate to
make a 25-μl reaction (see Note 5).
Table 2
Reagents for assay validation
Reagent Amount per well

ddPCR Supermix for Probes (no dUTP) 12.5 μl
20 assay mixture 1.25 μl
Restriction enzyme of choice 2–4 U
WT genomic DNA in water Up to 150 ng
HDR and NHEJ gBlock control solutions with 2000 copies/μl 1 μl each
Water To give 25 μl final reaction volume
2. Carefully apply 20 μl of the mixture into each of the 8 “sample”

wells of a DG8 Droplet Generator Cartridge (see Note 6).
3. Apply 70 μl of Droplet Generation Oil for Probes into each of
the 8 “oil” wells of the DG8 Droplet Generator Cartridge (see
Note 7).
4. Hook a DG8 Gasket for QX200 Droplet Generator onto the
DG8™ Cartridge Holder.
5. Put the holder in the droplet generator to generate droplets in
8 wells (in ~2 min).
6. Transfer droplets into a semiskirted Eppendorf twin.tec
96-well plate by using an eight-channel pipette set to 45 μl
(see Note 8).
7. Seal the plate with a Pierceable Foil Heat Seal using the PX1™
PCR Plate Sealer set to 180 C, 5 s.
8. For assay validation, a two-step thermal cycling with a 50–60 C
gradient is recommended (Table 3). If assay amplicon exceeds
150 bp in length, a three step thermal cycling protocol with a
discrete 72 C extension step can be helpful (Table 4).
9. Analyze the droplets by the Droplet Reader.
10. To determine the limit of detection of an assay, analyze
100–150 ng of WT genomic DNA samples spiked with differ-
ent amounts of HDR or NHEJ gBlock. The limit of detection
is the lowest amount of gBlock that gives positive signal signif-
icantly higher than the background noise by comparison of
results with and without gBlock.
3.3 Digital PCR 1. Dilute genomic DNA samples to 100–150 ng/μl in distilled
Detection of HDR water or TE (see Note 9).
and NHEJ Events 2. Assemble the master mix below on ice by mixing the reagents
in Genomic DNA shown in Table 5 (volumes shown are for one reaction to which
Samples DNA must still be added). Multiply these by the number of
Table 3
Two-step thermal cycling for ddPCR (all the steps ramped by 2 ˚C/s)
Step Temperature ( C) Duration and repeat

1 95 10 min
2 94 30 s
3 59 1 min, (repeat steps 2 and 3, 39 more times)
4 98 10 min
5 12 Hold
Table 4
Three-step thermal cycling for ddPCR (all the steps ramped by 2 ˚C/s)
Step Temperature ( C) Duration and repeat

1 95 10 min
2 94 30 s
3 58 1 min
4 72 2 min, (repeat steps 2, 3, and 4, 39 more times)
5 98 10 min
6 12 Hold
Table 5
Reagents for quantification of HDR and NHEJ in genomic DNA samples (X μl (100–150 ng) of a
genomic DNA sample later)
Reagent Volume per well

ddPCR Supermix for Probes (no dUTP) 12.5 μl
20 assay mixture 1.25 μl
Restriction enzyme of choice 2–4 U
Water Up to 25—X μl
samples, a distilled water only negative control, and the positive

control mixture (WT genomic DNA with synthetic alleles as
made in Subheading 3.2, above).
3. Aliquot 25—X μl of master mix into eight-tube PCR strips or a
96-well plate (any 96-well plate).
4. Add X μl (for a total of 100–150 ng) of genomic DNA per
sample. Also, add controls to tubes or wells designated for the
controls.
5. If the total number of samples to be analyzed is not a multiple

of 8, the remaining empty tubes of the PCR strip must be filled
with 25 μl of ddPCR™ Buffer Control Kit diluted to 1 by
distilled water.
6. Briefly spin the PCR tubes or plates down.
7. Gently pipette up and down 7–8 times to mix reactions by a
20 μl eight-channel pipette and carefully apply the mixtures
into “sample” wells.
8. Generate droplets as in Subheading 3.2, steps 2–6.
9. Repeat steps 4–8 until droplets are generated for all the
samples.
10. Seal the plate with foil using the PX1™ PCR Plate Sealer set to
180 C, 5 s.
11. Perform a thermal cycling with the best temperature found in
Subheading 3.2.
12. Analyze the droplets by the ddPCR analyzer in the same way as
Subheading 3.2.
3.4 Analysis 1. Go to “Analyze” and then “2-D Amplitude” in the QuantaSoft

of Droplet Digital software.
PCR Data 2. In a successful assay, distinct negative, WT, NHEJ, and HDR
populations should be seen in the 2D-plot (Figs. 3, 4, and 5).
3. If four distinct populations cannot be detected, the probes
and/or primers must be redesigned.
4. By using the lariat tool, gate the droplets without any template
(NEmpty), droplets only with NHEJ allele (NNHEJ), droplets
Fig. 3 Separated cut and mutation sites, with one NHEJ probe: Assay design and 2D droplet plot with droplet
group definitions for analysis. Details as in Fig. 1a, c
Fig. 4 Cut site and mutation site overlap. (a) Assay design when cut site and mutation site overlap. In these
cases, the NHEJ probe overlaps with and is on the same strand as the HDR probe. The NHEJ probe thus
competes with the HDR probe for WT binding and a dark probe is not necessary. Also, the HDR allele becomes
FAM++ and HEX. (b) Other droplet populations defined as in Fig. 1c
with WT allele and both WT and NHEJ alleles (NWT+), and

other droplets with HDR allele (NHDR+) as black, blue, green,
and orange populations, respectively (Figs. 3, 4, and 5).
5. To obtain the numbers of droplets in each population, export
the data as a CSV file by clicking “Export CSV” icon in the
QuantaSoft software. In the CSV file, the Ch1+Ch2+, Ch1
+Ch2, Ch1Ch2+, and Ch1Ch2 columns are NHDR+,
NNHEJ, NWT+, and NEmpty, respectively. Use formulas below
to calculate the frequencies of WT, HDR, and NHEJ alleles
from these numbers.
6. The standard formula for ddPCR quantification is:

c ¼ ln N neg =N total =V droplet
where
c ¼ copies per microliter of initial reaction Nneg ¼ the number
of droplets that do not contain the species of interest.
Ntotal ¼ the total number of droplets.
Vdroplet ¼ the volume of an individual droplet (0.85 nl with
current reaction conditions).
7. In this ddPCR-based assay, some of the droplet populations
cannot easily be separated, such as the droplet group contain-
ing WT and NHEJ + WT droplets (NWT+) or the droplet group
containing HDR and HDR + WT droplets (NHDR+). To
Fig. 5 Two cut sites are introduced: ddPCR assay design and 2D droplet plot. (a) Assay design when two cut
sites are introduced. Because dual Cas9 systems introduce two cuts, two NHEJ probes are included in the
assay to detect possible indels on either side of the mutation site. When neither NHEJ probe competes with the
HDR probe—as shown here—a dark probe is also designed to avoid binding of the HDR probe to the WT
allele. Because two NHEJ probes are included in this assay, the WT allele is detected as FAM+ and HEX++,
and the NHEJ alleles are detected as FAM+ and HEX+, or FAM+ and HEX. (b) Two-dimensional plot of an
assay with two NHEJ probes. Definitions of NEmpty and NHDR+ populations are the same as in Fig. 1c. However,
because two NHEJ probes are included in this assay, there are two droplet populations containing only NHEJ
alleles—one that lost one of the two NHEJ probe binding sites of NHEJ probes (FAM+ HEX+) and one that lost
both (FAM+ HEX). All other droplets were gated as NWT+ population (FAM+ HEX++). These definitions are
used to calculate the HDR and NHEJ allelic frequencies (see Subheading 3.4)
quantify in this case, an appropriate subset of droplets is used to

calculate Nneg and Ntotal (see Note 10).
Definitions
l Nempty ¼ number of droplets in the double-negative cluster
labeled “empty.”
l NNHEJ ¼ number of droplets in the cluster labeled
“NHEJ.”
l NWT+ ¼ number of droplets in the clusters labeled “WT+”.
l NHDR+ ¼ number of droplets in the clusters labeled
“HDR+”.
8. For NHEJ quantification

l Nneg ¼ Nempty
l Ntotal ¼ Nempty + NNHEJ
9. For HDR quantification
l Nneg ¼ Nempty + NNHEJ + NWT+
l Ntotal ¼ Nempty + NNHEJ + NWT+ + NHDR+
10. For WT quantification
l Nneg ¼ Nempty + NNHEJ + NHDR+
l Ntotal ¼ Nempty + NNHEJ + NWT+ + NHDR+
11. Frequencies of WT, HDR, and NHEJ are calculated as:
F WT ¼ c WT 100=ðc WT þ c NHEJ þ c HDR Þ
F HDR ¼ c HDR 100=ðc WT þ c NHEJ þ c HDR Þ
F NHEJ ¼ c NEHJ 100=ðc WT þ c NHEJ þ c HDR Þ
where
F ¼ allelic frequency (%).
cWT ¼ the WT allelic concentration.
cHDR ¼ the HDR allelic concentration.
cNHEJ ¼ the NEHJ allelic concentration.
An example of analysis is shown in Fig. 6.
Fig. 6 Different HDR and NHEJ frequencies induced by TALEN in RBM20 in HEK293 cells and human induced
pluripotent stem cells. (a and b) HDR and NHEJ allelic frequencies induced by the same TALENs targeting
RBM20 in HEK293 cells (a) and human induced pluripotent stem cells (b). The NHEJ and HDR probes directly
compete to each other, so NHDR+ population was FAM++ HEX as in Fig. 4. The frequencies are shown in the
two-dimensional plots. The two different cell types showed different HDR and NHEJ frequencies
4 Notes
1. Master high-copy gBlock stocks should be kept in a post-PCR

environment to avoid contamination. The control gBlock
stocks must be handled with utmost care to avoid contamina-
tion, only bringing highly dilute solutions into the assay
setup area.
2. Filtered tips must be used to avoid contamination.
3. In order to slowly pipette droplets to avoid shearing of them, a
50 μl pipette is recommended rather than a 200 μl pipette.
4. Predicted cut sites are 3 bp upstream of PAM for CRISPR and
equidistant between DNA binding domains for TALEN or
FokI-dCas9.
5. Handle positive control mixtures carefully (e.g., careful open-
ing of the control stock tubes and changing gloves after
handling the control stock). If assay contamination is sus-
pected, remake the 20 assay.
6. Avoid creating bubbles in the DG8 sample wells. Bubbles
floating on the surface of the sample may not affect droplet
generation, but bubbles in the bottom of the well will disrupt
droplet generation. Spinning prepared samples before droplet
generation in tubes or plates removes bubbles. Take note of
orientation of the cartridge with respect to sample order to
ensure correct loading of the final Eppendorf twin.tec 96-well
plate.
7. Samples should be loaded onto the chip prior to droplet
generation oil.
8. Do not press the pipette tightly to the bottom of the cartridge
or pipette too vigorously as this will shear the droplets. Cover
the PCR plate with the foil sheet immediately after transfer to
reduce the risk of contamination.
9. With too much input DNA, each droplet would have too many
genomic DNA copies, which interferes with proper separation
of the four distinct populations and thus accurate estimation of
allelic concentrations. With too little input DNA, the number
of genomic copies analyzed by the assay would not be enough
to have high sensitivity. Therefore, the amount of genomic
DNA input should not exceed 150 ng per ddPCR reaction.
10. For NHEJ quantification, only NHEJ single positives and the
Empty (double-negative) droplets were used. For WT and for
HDR quantification, all droplets are used.
Acknowledgment
We thank Jennifer R. Berman, Samantha B. Cooper, Bin Zhang,

and George A. Karlin-Neumann (Bio-Rad) for technical help and
helpful discussions. This work was supported by the National Insti-
tutes of Health (U01-HL100406, U01-GM09614,
R01-HL108677, U01-HL098179, U01-HL099997,
P01-HL089707, and R01-HL060664 to B.R.C.); the UCSF
Liver Center to B.R.C., the Bluefield Project to Cure Frontotem-
poral Dementia to B.R.C., the Uehara Memorial Foundation
Research Fellowship to Y.M., and Gladstone-CIRM Fellowship to
Y.M.
References
1. Gaj T, Gersbach CA, Barbas CF III (2013) 6. Hindson BJ, Ness KD, Masquelier DA,
ZFN, TALEN, and CRISPR/Cas-based meth- Belgrader P, Heredia NJ, Makarewicz AJ, Bright
ods for genome engineering. Trends Biotechnol IJ, Lucero MY, Hiddessen AL, Legler TC,
31(7):397–405. https://doi.org/10.1016/j. Kitano TK, Hodel MR, Petersen JF, Wyatt PW,
tibtech.2013.04.004 Steenblock ER, Shah PH, Bousse LJ, Troup CB,
2. Certo MT, Ryu BY, Annis JE, Garibov M, Mellen JC, Wittmann DK, Erndt NG, Cauley
Jarjour J, Rawlings DJ, Scharenberg AM TH, Koehler RT, So AP, Dube S, Rose KA,
(2011) Tracking genome engineering outcome Montesclaros L, Wang S, Stumbo DP, Hodges
at individual DNA breakpoints. Nat Methods 8 SP, Romine S, Milanovich FP, White HE, Regan
(8):671–676. https://doi.org/10.1038/ JF, Karlin-Neumann GA, Hindson CM,
nmeth.1648 Saxonov S, Colston BW (2011) High-
3. Ran FA, Hsu PD, Lin CY, Gootenberg JS, throughput droplet digital PCR system for abso-
Konermann S, Trevino AE, Scott DA, Inoue A, lute quantitation of DNA copy number. Anal
Matoba S, Zhang Y, Zhang F (2013) Double Chem 83(22):8604–8610. https://doi.org/10.
nicking by RNA-guided CRISPR Cas9 for 1021/ac202028g
enhanced genome editing specificity. Cell 154 7. Miyaoka Y, Chan AH, Judge LM, Yoo J,
(6):1380–1389. https://doi.org/10.1016/j. Huang M, Nguyen TD, Lizarraga PP, So PL,
cell.2013.08.021 Conklin BR (2014) Isolation of single-base
4. Hendel A, Kildebeck EJ, Fine EJ, Clark JT, genome-edited human iPS cells without antibi-
Punjya N, Sebastiano V, Bao G, Porteus MH otic selection. Nat Methods 11(3):291–293.
(2014) Quantifying genome-editing outcomes https://doi.org/10.1038/nmeth.2840
at endogenous loci with SMRT sequencing. 8. Miyaoka Y, Berman JR, Cooper SB, Mayerl SJ,
Cell Rep 7(1):293–305. https://doi.org/10. Chan AH, Zhang B, Karlin-Neumann GA, Con-
1016/j.celrep.2014.02.040 klin BR (2016) Systematic quantification of
5. Lin S, Staahl BT, Alla RK, Doudna JA (2014) HDR and NHEJ reveals effects of locus, nucle-
Enhanced homology-directed human genome ase, and cell type on genome-editing. Sci Rep
engineering by controlled timing of CRISPR/ Mar 31(6):23549. PMID: 27030102,
Cas9 delivery. elife 3:e04766. https://doi.org/ PMC4814844
10.7554/eLife.04766 9. Bio-Rad Bulletin 6407, Droplet Digital Applica-
tions guide
Chapter 21
DNA Methylation Analysis Using Droplet Digital PCR

Ming Yu, Tai J. Heinzerling, and William M. Grady
Abstract
Droplet digital (ddPCR) is a recent advance in PCR technology that enables the precise detection and
absolute quantification of nucleic acid target sequences and that has a range of applications for both research
and clinical diagnostic studies. Here, we discuss the parameters important in the design and performance of
ddPCR for the detection and quantification of methylated DNA. We provide explicit instructions for
conducting methylation specific ddPCR (aka MethyLight ddPCR). We also present an example that
demonstrates the sensitivity and precision of the method for detecting methylated DNA in the promoter
region of mir342/EVL, a potential DNA methylation biomarker for colon cancer risk. Common technical
problems and troubleshooting for conducting successful MethyLight ddPCR assays are also discussed.
Key words Droplet digital PCR, MethyLight, DNA methylation, Colorectal cancer, Risk biomarkers,
MethyLight ddPCR
1 Introduction
DNA methylation is a commonly found epigenetic modification of

mammalian DNA that results from the addition of a methyl group
to the 50 carbon of a cytosine in a CpG dinucleotide sequence. The
majority of the mammalian genome is heavily methylated, and
DNA methylation has been shown to play an important role in
development and disease by regulating gene expression in cells.
Aberrant DNA methylation, which entails global hypomethylation
throughout the genome and focal hypermethylation of CpG islands
located in gene promoter regions and intergenic DNA, is a com-
mon epigenetic alteration in many human malignancies, including
colorectal cancer (CRC). The study of DNA methylation altera-
tions in CRC has provided the basis for many principles that have
guided our understanding of epigenetic alterations in cancer in
general. Intriguingly, DNA methylation alterations have been
observed in normal colon tissue that is predisposed to develop
polyps or cancer, a phenomenon called “field cancerization”
[1–3], albeit at a much lower frequency than in diseased tissue
363
364 Ming Yu et al.
itself. Thus, these infrequently methylated DNA genes, for example

EVL [4], have the potential to serve as biomarkers to identify
individuals at high risk for developing CRC. It is likely that the
field cancerization process is relevant for other cancers. MethyLight
ddPCR is an ideal method for the study of field cancerization given
its precision for detecting low levels of infrequently methylated
alleles.
PCR-based methods have been applied extensively for the
assessment of DNA methylation. The first generation of PCR for
detecting methylated alleles was based on PCR amplification of
DNA after digestion with methylation-sensitive restriction enzymes
[5]. This was followed by a more flexible method called
methylation-specific PCR (MSP), for which primers are designed
specifically against DNA that has been treated with sodium bisul-
fite, which converts unmethylated cytosine to uracil through deam-
ination [6]. In its initial iteration, the bisulfite-converted sequences
of DNA in MSP are subjected to end-point analysis by UV transil-
lumination after horizontal gel electrophoresis generating qualita-
tive results. However, due to its qualitative nature, this original
version of the MSP method proved suboptimal for applications
that require quantitative assessment of the amount of methylated
DNA present, which is often needed in epigenetic studies as well as
in biomarker development. The development of MethyLight, a
fluorescence-based quantitative PCR (TaqMan®) assay to measure
DNA methylation, has considerably enhanced our ability to detect
and quantify methylated alleles [7]. In MethyLight, the primers
and/or fluorescent probes are designed against bisulfite-converted
DNA sequence and quantitative information is obtained in real-
time. MethyLight PCR typically is designed to detect methylated
alleles and can quantitate the amount of the methylated allele in a
relative fashion following correction for DNA loading and normal-
ization to a methylated DNA standard. Despite its advantages over
endpoint MSP, MethyLight is still susceptible to the presence of
PCR inhibitors, has limited sensitivity for detecting rare methylated
alleles in a background of unmethylated alleles, and can be skewed
through the normalization process, which can lead to inconsistent
results, especially in the detection of low levels of DNA methyla-
tion. Thus, conventional quantitative methylation specific PCR is
often unable to reach the degree of precision and sensitivity needed
to accurately detect low levels of methylated DNA and thus fully
utilize DNA methylation as a field cancerization effect marker or
cancer risk biomarker.
Here, we describe the development of a novel and highly
sensitive assay for detecting methylated DNA based on droplet
digital PCR, called the MethyLight ddPCR assay, We put emphasis
on how to detect and perform absolute quantification of rare
methylation events in low-concentration DNA samples with
MethyLight ddPCR. An overview of the steps involved in a typical
MethyLight ddPCR 365
Fig. 1 Workflow diagram of the procedure for the MethyLight ddPCR, which is
used for DNA methylation analysis
MethyLight ddPCR assay is shown in Fig. 1 and described in more

detail in the text in this chapter. Our group and others have shown
that this method allows the precise detection of infrequently
methylated alleles that cannot be detected by conventional Methy-
Light, [8, 9].
2 Materials
The quality of DNA preparation from the biological samples of

interest can impact DNA methylation analysis using ddPCR
MethyLight. Acceptable sample sources include those that are
acceptable for conventional PCR applications and include, but are
not limited to, fresh frozen tissue, formalin-fixed, paraffin-embed-
ded (FFPE) tissue, cell lines, and cell-free DNA in the plasma/
serum. It is also important to be aware that genomic DNA will be
subjected to bisulfite conversion under harsh chemical conditions,
leading to fragmented DNA. Therefore, there is no need to do
restriction digestion of DNA samples, as recommended for intact
DNA. Although ddPCR is less sensitive to PCR inhibitors than
366 Ming Yu et al.
other technologies, we recommend thoroughly removing PCR

inhibitors during DNA purification. Consider reducing the impact
of PCR inhibitors on reactions by diluting samples 1:10. The
recommended range of input DNA of the QX100/200 system is
from 3 to 106060.6 copies/20 μL reaction. Assuming 3.3 pg/
human haploid genome based on the Celera Genomics estimate),
the amount of input DNA is estimated to range from 10 pg to
350 ng of human DNA/20 μL reaction. To estimate the number of
copies/ng of DNA of a different organism, one needs to know the
mass and the number of base pairs in the genome of interest. If the
target copy number/genome is unknown, the optimal starting
amount needs to be determined empirically using a dilution series
of each sample at the relevant range for the anticipated experiments.
2.1 Genomic DNA 1. We have routinely used DNA isolated from fresh frozen tissue
(gDNA) Isolation using the DNeasy® Blood and Tissue Kit (QIAGEN, catalog
#69504). If starting with FFPE tissues, we recommend using
the QIAamp® DNA FFPE Tissue Kit (QIAGEN, catalog
#56404) instead (see Note 1).
2. Water bath.
3. Xylene for deparaffinization of FFPE samples.
4. 100% ethanol.
5. Microcentifuge tubes, 1.5 mL (RNAse and DNAse free).
6. Microcentrifuge.
2.2 gDNA 1. QuantiT™ PicoGreen® dsDNA Assay Kit (Life Technologies,

Quantification (See catalog #p7589).
Note 2) 2. Falcon 96-well Black/Clear Flat-bottom Imaging Plate with
Lid (Fisher Scientific, catalog #353219).
3. Molecular biology grade water.
4. Microcentrifuge tubes, 1.5 mL (RNAse and DNAse free).
5. Fluorescence microplate reader, e.g., Fluoroskan Ascent from
ThermoLabsystems with Ascent software version 2.4.2.
2.3 Bisulfite 1. Molecular biology grade water.

Conversion of gDNA 2. EZ DNA Methylation Kit (Zymo Research, catalog #D-5001).
(See Note 3)
3. Water bath or other device that can be used to incubate the
samples at a constant temperature overnight.
4. Microcentifuge tubes, 1.5 mL (DNAse, RNAse free).
5. Microcentrifuge.
2.4 Assay 1. Control DNA: EpiTect Control DNA, 100% methylated (QIA-
Components GEN, catalog #59655) and EpiTect Control DNA, 100%
unmethylated (QIAGEN, catalog #59665, see Note 4).
2. Software for converting DNA sequence to bisulfite-converted

sequence, e.g., PyroMark Assay Design v2.0.1.15 (QIAGEN,
catalog #9019077).
3. Software for designing primers and probes, e.g., Primer
Express® Software v3.0.1 (Life Technologies, catalog
#4363991).
4. Probes and forward and reverse primers for both the reference
gene and gene of interest. To control for the total input of
methylated and unmethylated DNA, we use a reference gene
assay that is insensitive to CpG methylation status (referred to
as the C-less assay) [10].
5. Droplet Digital PCR Supermix for Probes (Bio-Rad, catalog
#1863010, see Note 5).
2.5 Bio-Rad ddPCR 1. Droplet Generation Oil for Probes (Bio-Rad, catalog
System Consumables #1863005).
tor (Bio-Rad, catalog #1864008). One is needed for every
eight reactions.
(Bio-Rad, catalog #1863009). One is needed for every
eightreactions.
4. ddPCR™ Droplet Reader Oil (Bio-Rad, catalog #1863004).
5. Pierceable Foil Heat Seals (Bio-Rad, catalog #1814040).
6. Microcentifuge tubes, 1.5 mL (DNAse, RNAse free).
7. Semiskirted 96-well PCR plate with notches in at least two
corners (e.g., Eppendorf, catalog #951020362).
2.6 Bio-Rad ddPCR 1. T100™ PCR Thermal Cycler (Bio-Rad, catalog #1861096).
System Equipment 2. QX200™ Droplet Generator (Bio-Rad, catalog #1864002).
3. PX1™ PCR Plate Sealer (Bio-Rad, catalog #1814000).
4. QX200™ Droplet Reader (Bio-Rad, catalog #1864003).
5. QuantaSoft™ Software, Regulatory Edition (Bio-Rad, catalog
#1864011).
3 Methods
Carry out all procedures at room temperature, unless otherwise

noted, and in a room guaranteed free of potentially contaminating
DNA, such as plasmids and amplicons. All steps after DNA isolation
and during setup of the PCR (prior to placing the samples in the
thermocycler) should be carried out in an environment kept
completely free of potentially contaminating DNA, such as in a
368 Ming Yu et al.
PCR workstation (see Note 6). All instruments used in the PCR
set-up steps should be decontaminated by exposure to UV light
(254 nm) for at least 15 min prior to starting this phase of Methy-
Light ddPCR.
3.1 Genomic DNA 1. Isolate gDNA from tissue using a kit such as the DNeasy®
(gDNA) Isolation and Blood and Tissue Kit or the QIAamp® DNA FFPE Tissue Kit
Quantification for FFPE tissues (see Note 1). Perform gDNA isolation accord-
ing to the manufacturer’s instructions for special considera-
tions when using FFPE tissues, see Note 7.
2. Measure gDNA concentration using the Quant-iT™ Pico-
Green® dsDNA Assay Kit according to manufacturer’s instruc-
tions scaled down to a 96-well format: Use a 96-well imaging
plate with black walls and a clear flat bottom. Add 100 μL of
DNA solution and 100 μL of Quant-iT PicoGreen reagent
working solution to each well for measurement (see Note 2).
3.2 Bisulfite 1. Perform bisulfite conversion of sample DNA using a kit such as
Conversion of gDNA the EZ DNA Methylation Kit from Zymo Research (D5002)
(See Note 3) following the manufacturer’s instructions with the following
optimizations.
2. Consistent with the manufacturer’s instructions in the EZ
DNA Methylation Kit, we recommend using less than 1 μg
DNA per column since too much DNA may result in incom-
plete conversion.
3. The overnight incubation (step 4 of the kit) may be done in a
water bath, thermocycler, or heat block. Any of these are
acceptable as long as the samples are in a DNA clean environ-
ment and protected from light. It is also important to avoid
excessive shaking of the samples during this process.
4. After the final wash step (step 11), empty the collection tube
and spin-dry the column for an additional 1 min in order to
ensure that there is no carryover of wash buffer into the elute.
5. To obtain the highest yield, repeat the elution step with both
rounds of eluate going into the same tube (10 μL eluate per
elution step, total volume ¼ 20 μL).
6. Bisulfite converted-DNA should be used as soon as possible
after conversion, or can be stored at 70 C for later use.
Freeze the samples immediately after conversion if they are to
be stored. Repeated freezing and thawing is strongly discour-
aged as greater than three freeze-thaw cycles can lead to DNA
fragmentation that impairs PCR DNA amplification. Therefore
we recommend creating aliquots of samples before freezing.
3.3 Primer and Probe The primer/probe design follows the principles of MethyLight
Design (See Note 8) assays in general, which has been discussed in detail previously,
with modifications noted below [11].
1. First virtually convert both the forward and reverse strand of a
500 bp sequence of DNA surrounding the CpG of interest to
its bisulfite-converted sequence using PyroMark Assay Design
v2.0 (QIAGEN).
2. Next, use ABI Primer Express software Version 3.0.1 (Applied
Biosystems, Life Technologies) to design the primers and
probe used in a MethyLight ddPCR assay (An example of the
use of ABI Primer Express designed primers and probes for a
methylated EVL assay (NP_057421) is described later and in
Yu et al. [9].) After virtual bisulfite conversion of the sequence,
input either the forward or reverse strand into Primer Express
set to the TaqMan® MGB Quantification setting.
3. Once the sequence is in Primer Express, follow the instructions
for use of the software to select primers and probes for the
locus of interest.
4. If an optimal set of primers and probes cannot be generated by
the software, then modify the sequence of interest repeatedly
by shortening or extending the sequence until a primer or
probe covers the CpG of interest.
5. In addition, if an acceptable primer and probe set is not found
despite modifying the target sequence, try using more lenient
parameters (e.g., alter melting temperature parameters) or
designing the assay for the complementary strand.
6. The selected probes are synthesized with a 50 fluorescent emit-
ter and a 30 nonfluorescent quencher with a minor groove
binding (MGB) domain. We design probes with 50 fluorescent
emitters that can be used in a duplex assay. So we typically
design the probes for the gene of interest with FAM and the
reference gene with VIC or HEX, which allows us to simulta-
neously detect and quantify the methylated target gene (e.g.,
EVL labeled with FAM) and the reference gene (e.g., C-LESS-
C1 [10] labeled with VIC).
The reference gene selected will be amplified regardless of
its methylation status and therefore is used to determine the
total amount of amplifiable DNA. Other investigators have
used beta-actin (ACTB) to normalize for input DNA, which
can also work as a reference gene but is less well studied
compared to C-LESS-C1 (Belshaw et al. [3]). A complete list
of the MethyLight primer and probe sequences for the EVL
and C-Less-C1 MethyLight ddPCR assays are provided in
Table 1. The location of CpGs in the MethyLight primers
and probes and the amplicons for methylated EVL and
C-LESS-C1 assays are provided in Fig. 2.
370 Ming Yu et al.
Table 1
Primer and probe sequences used in MethyLight ddPCR assays
Assay Primer/probe Primer/probe sequence (50 to 30 )

EVL Forward AACGACTCCGAATCCTCGAA
Reverse GCGAATAGTAACGCGCGTATT
Probe FAM-CGCGAACTAATCTCAACA-MGBNFQa
C-less-C1 Forward TTGTATGTATGTGAGTGTGGGAGAGAGA
Reverse TTTCTTCCACCCCTTCTCTTCC
Probe VIC-CCTCCCCCTCTAACTCTAT-MGBNFQa
MGBNFQ refers to a Minor Groove Binding nonfluorescent quencher in the 30 terminus of the probe
a
Fig. 2 Relative location of the CpG dinucleotides in the MethyLight primers and probes for the assay to detect
methylated EVL and for the C-LESS-C1 assays. The forward primers, probe and reverse primers are indicated
by the blue arrows, green lines and red arrows, respectively. The small vertical lines indicate the CpG sites.
The DNA strand shown is the top strand
3.4 Plate Lay-Out 1. Design the experimental plate layout using a 96-well plate in
Design columns (see Note 9).
2. Always include a NTC, a positive control using 100% methy-
lated Epitect DNA (Pos.), and a negative control using 100%
unmethylated EpiTect DNA (Neg.)
3.5 Droplet 1. Thaw all reagents and DNA on ice, and equilibrate to room
Generation and temperature for 3 min before generating droplets. Protect the
Transfer probe from exposure to light.
2. The “Master Mix” (MM) is a mixture of Supermix, primers,
probes, and water, in correct proportions, as described in
Table 2 below, which will be added to each sample. Calculate
the volumes of MM required based on 25 μL total volume per
sample (20 μL MM plus 5 μL DNA). The primer and probes
are used at final concentrations of 900 nmol/L and
~250 nmol/L, respectively according to manufacturer’s
recommendations. See Table 2 for an example of Master Mix
calculations. Reagents should be well mixed by pipetting or
gentle vortexing.
Table 2
Example master mix calculations
Component Starting Final 1

Water 5
2 ddPCR supermix for probes 2 1 12.5
20 FAMa 20 1 1.25
a
20 VIC 20 1 1.25
Bisulfite-converted DNA 2 ng/μL 5
Final volume in μL b
25
Final ng/μL in ddPCR 0.4
a
Recommend setting up 25 μL/reaction and loading 20 μL droplets in DG cartridge
b
20 primer-probe mix: 18 uM PCR primers (each), 5 uM probe
3. For each sample and control, prepare one “reaction tube”: a

1.5 mL microcentrifuge tube containing all of the Master Mix
and DNA for the replicates, with 20 μL MM and 5 μL DNA for
each replicate. Multiply MM and DNA volumes by the number
of replicates (e.g., for four replicates, mix 80 μL Master Mix
and 20 μL DNA in a tube). Store reaction tubes in the dark at
room temperature and start loading them one at a time onto
the generator cartridge immediately after the reagents are
combined.
4. Open the droplet generator cartridge holder by pressing
latches on each side. Slide the disposable cartridge into the
right side of the holder with the notch on the upper left corner
of the cartridge. Make sure it is in place before pressing the two
halves of the holder together.
5. Vortex the reaction tube at a moderate speed for at least 10 s
and transfer 20 μL of the reaction tube mixture into each well
in the middle row of the cartridge (see Note 10).
6. Pour droplet generator oil into a pipetting reservoir and use a
multichannel pipette to load 70 μL of oil into each well on the
bottom row of the cartridge. All eight wells must be filled or
the droplets will not be generated properly.
7. Hook a DG8™ Gasket for QX200™/QX100™ Droplet Gen-
erator over the cartridge holder.
8. Open the QX200 Droplet Generator by pressing the green
button on the top of the Generator and place the cartridge
holder in the instrument. Verify that the two left indicator
lights on the generator are green.
372 Ming Yu et al.
9. The run will begin automatically. When all three indicator

lights are green, the run is finished and the door may be
opened. Remove the cartridge holder from the instrument
and remove the gasket from the holder. Discard the gasket.
10. Using an electronic multichannel pipette set to a volume of
40 μL and speed of one (lowest), gently transfer the droplets
from the upper wells of the cartridge to one column of a
96-well plate (see Note 11).
11. Remove the cartridge from holder and discard.
3.6 Sealing the Plate 1. Configure the PX1 PCR plate sealer to seal for 5 s at 180 C.
2. Open the PX1 PCR plate sealer by pressing the arrow on the
instrument’s screen. Position your plate on the tray.
3. Position a pierceable foil seal over the wells of the plate with the
red stripe of the seal facing upward. Place the metal frame
included with the instrument over the PCR plate.
4. Close the door of the sealer and once the machine is up to
temperature, press “Seal.”
5. The PX1 will open automatically when it is finished sealing.
Turn the PCR plate horizontally 180 and repeat the steps
described above.
6. Thermal cycle the droplets within 30 min of pipetting the
reaction mixtures into the plate.
3.7 PCR 1. Run the plate in a thermocycler with the cycling conditions
recommended for the Supermix used. For cycling conditions
to use with ddPCR Supermix for Probes (Bio-Rad, catalog
#1863010), see Table 3.
2. After thermocycling, the plate should be read using the QX200
droplet reader immediately or kept at 4 C for no more than
1 day.
Table 3
Thermalcycling conditions for ddPCR Supermix for Probesa
Cycling step Temperature C Time Ramp rate Number of cycles

Enzyme activation 95 10 min 1

Denaturation 94 30 s 2 C/s 45
Annealing/extension 60b 1 min 2 C/s
Enzyme deactivation 98 10 min 1
Hold 4 Infinite 1

Use a heated lid set to 105 C and sample volume set to 40 μL
a
b
Annealing temperatures vary for each primer/probe set. Test for optimal temperature
3.8 Droplet Reading 1. Turn on the QX200 droplet reader and allow it to warm up for
30 min.
2. Verify that the left two lights are solid green. If the left light is
not solid, there is an issue with the power. If the second light is
not on then either the oil bottle is too low or the waste bottle is
too full. Correct these issues before proceeding with the plate
reading.
3. Open the plate holder by flipping the two black release tabs
into the up position and removing the cover. Place the plate on
the holder with well A1 in the upper left, replace the cover, and
flip the black tabs back down.
4. Open the reader by pressing the green button on top of the
door, place the plate holder in the instrument, and close the
door again using the same button. The plate indicator light
should now be solid green.
5. Connect the reader to your computer via a USB cable and open
the QuantaSoft software. Select Setup and define the experi-
ment design. Then select Run (see Note 13).
6. While the run is in progress, the right-most indicator light
flashes green. When the run is finished, all indicator lights
should be solid green. At this point, open the door and remove
the plate holder. Remove the PCR plate from the holder and
discard.
3.9 Interpreting 2D After the run is complete, we use QuantaSoft software to analyze
Plot Results (See the data in each well. If the plate was set up for absolute quantifica-
Table 4) tion (ABS) analysis. QuantaSoft sets the threshold automatically
and determines concentration in the data tables in the analysis
mode of the software. However, one should inspect each well and
manually adjust the threshold if necessary to ensure correct desig-
nation of the droplets in each well.
Figure 3 shows an example of a duplex MethyLight ddPCR
experiment in which the target gene EVL and the reference gene
C-LESS-C1 have been PCR amplified and viewed in a 2D plot of
droplet fluorescence. For detecting infrequently methylated alleles
in clinical samples, each sample is run in eight replicate wells and the
droplet counts (positive and negative) from all replicated wells are
combined to yield a “merged” well. The concentration and Poisson
confidence intervals for each “merged” well are computed using
the QuantaSoft software version 1.4.0.99 (Bio-Rad, Hercules,
CA). Troubleshooting advice can be found in Table 4.
374 Ming Yu et al.
Table 4
Troubleshooting table
Problem Possible reason Possible solution

No positive droplet counts from The ddPCR reaction mix has not Check the volumes and
positive control samples for the been assembled correctly concentrations of each
reference gene component in the PCR
reaction mix
No positive droplet counts of the The amount of input DNA is too Increase the amount of the input
reference gene from the low DNA
samples, but the positive
control works properly
The DNA concentration Bisulfite-converted DNA is Use recommended methods to
measurements by ddPCR are partially degraded as ddPCR prepare, handle and store
considerably lower than gives a concentration estimate bisulfite-converted DNA for
spectroscopic measurements of intact DNA targets methylation studies
Suboptimal assay design/ Use the recommended primer
efficiency and probe concentration.
Optimize ddPCR MethyLight
assay thermocycler conditions
by running a temperature
gradient experiment to
determine an optimal
annealing temperature
Only the reference gene gives Suboptimal assay design for the Redesign the assay
positive droplet counts. No gene of interest
signal from a newly designed The assay conditions are not Optimize assay conditions by
assay for the gene of interest optimal performing an annealing/
extension temperature
gradient
The ddPCR reaction mix is not Recheck the volumes and
assembled correctly concentrations of the
components in the PCR
reaction mix
The probe/primers sequences are Set up a conventional
not correct. This can happen if MethyLight assay to check if
an incorrect sequence is used the probe/primers will amplify
for synthesizing the primers or the DNA. Reorder probe/
probe primers if no PCR product
generated
Duplexing two assays together Duplex assays are interfering with Test the two assays separately in
does not give any positive each other singleplex assays
droplet counts
Positive droplet counts in NTC Contamination Decontaminate the PCR
workstation, PCR preparation
area, and equipment. Practice
good sample handling to avoid
contamination. Repeat the
experiment with new PCR
grade water and reagents.
(continued)
Table 4
(continued)
Problem Possible reason Possible solution

Separate the post-PCR area
from the PCR preparation
room
Significant number of mid-level The presence of PCR inhibitors in Dilute DNA samples further with
amplitude droplets (“Heavy DNA samples PCR grade water
rains” shown on the
fluorescence amplitude plot) in
DNA samples. Poor separation
of fluorescence intensity
between negative and positive
droplets
The number of total droplet The ddPCR reaction mix is not Use the recommended
counts per well is too low assembled correctly concentration of primers
(<10,000) (900 nM), probe (250 nM),
and 1 ddPCR Supermix for
probe
Suboptimal droplet generation Use a P-20 pipette and slowly
load the DG8 cartridge with
20 μL of samples into the
bottom of the sample well.
Then load 70 μL of oil to the
oil well. Begin droplet
generation immediately after
loading
Poor droplet handling Use the recommended electronic
multichannel P-50 pipette and
settings to ensure full-volume
transfer of droplets into
96-well plate. Make sure the
angle of the pipette tip near the
bottom of the well to avoid
shearing the droplet when
dispensing. Handle droplets in
a gentle and consistent way
Undersealed plates result in oil Properly seal the 96-well plate
evaporation and affect droplet
quality
Wide variation in technical Poorly mixed reaction mix When creating technical
replicates replicates, make sure the
reaction mixture is thoroughly
mixed
Poor well-to-well temperature Verify the thermal uniformity and
uniformity of the PCR thermal accuracy of the PCR thermal
cycler cycler
376 Ming Yu et al.
Fig. 3 Quantification of methylated EVL in colon mucosa biopsy samples by a duplex MethyLight ddPCR assay.
Genomic DNA samples were bisulfite converted and analyzed by MethyLight ddPCR. Each sample (8 ng or
2424.2 haploid genome equivalents of total input DNA) was partitioned into an average of 15,000 droplets per
well and replicated in 8 wells. The droplet counts (positive and negative) from all replicated wells were
combined to yield a “merged” well. The concentration and Poisson confidence intervals for each “merged”
well were computed using the QuantaSoft software version 1.4.0.99 (Bio-Rad, Hercules, CA). (a) 2D amplitude
plot of the duplex assay. For each fluorophore, a threshold was manually set to designate positive events.
Blue: methylated EVL-positive droplets (FAM); Green: C-Less-C1 positive droplets (VIC); Black: negative
droplets. (b) Measured concentration (copies/μL) of the methylated EVL (blue) in the positive control (EpiTect
100% methylated control DNA), NTC and a representative clinical sample. The total input DNA was estimated
based on a reference gene C-LESS-C1 assay (green). Error bars indicate the Poisson 95% confidence intervals
for each measurement. NTC: no template control
4 Special Considerations for Detection and Absolute Quantification of Rare

Methylated Alleles with ddPCR MethyLight
Rare methylation events occur in the context of field cancerization

effect, when one methylated allele of a candidate biomarker gene
exists in a background of (over 10,000) unmethylated alleles.
DDPCR MethyLight enables detection and absolute quantification
of rare methylation events at a level of sensitivity and precision
beyond the capabilities of conventional MethyLight [9]. This sec-
tion will discuss proper experimental setup to achieve a desirable
level of quantification (LOQ) and accurate quantification of rare
methylated alleles.
4.1 Sample DNA extracted from a biological sample will have variable integrity
Preparation and concentration, depending on the sample source. For example,
FFPE tissues, although used routinely in cancer research, provide
DNA that is highly fragmented and chemically cross-linked, which
may negatively impact the performance of a MethyLight ddPCR
assay. This should be kept in mind if poor quality results are gener-
ated from the test samples despite having good results from the
positive control samples.
Most methods in DNA methylation studies use bisulfite con-
version, which results in fragmented genomic DNA. One needs to
take into account the actual amount of amplifiable DNA in these
samples when calculating the amount of starting material needed.
Furthermore, bisulfite converted DNA may produce a “rain” pro-
file and a shift of baseline in a 1D amplitude plot (Fig. 4a). It is
possible to account for this artifact by simply setting the threshold
manually to obtain a good separation of positive droplets from
negative background and achieve accurate absolute quantification
(Figs. 3a and 4a). In addition, incomplete bisulfite-converted DNA
will result in misleading data and should be considered when an
unexpected amount of methylated DNA is detected or low PCR
amplification occurs.
4.2 Testing an Assay For development of methylation assays for assaying for a field
cancerization effect, one important consideration is the limit of
quantification (LOQ), which is the minimum concentration of
the methylated allele in a sample that can be reliably distinguished
from a background of a much higher proportion of unmethylated
alleles, within an acceptable (predetermined) coefficient of variation
(CV, measurement error). We use a multiplex assay, that includes an
assay for a reference gene (e.g., C-LESS-C1) to quantify the
amount of input DNA (over 99.9% of which are unmethylated),
and an assay for the methylated gene of interest (e.g., EVL) to
quantify the amount of methylated EVL allele. LOQ is typically
quoted as a ratio or a percentage: for example, 1 methylated allele in
378 Ming Yu et al.
Fig. 4 Analysis of limit of quantification (LOQ) for methylated EVL by MethyLight ddPCR. We determine the LOQ
by serial dilution of the 100% methylated EpiTect Methyl control DNA into a solution containing 100%
unmethylated EpiTect Unmethyl control DNA (8 ng of total input DNA in each well). Each sample is partitioned
into an average of 15,000 droplets per well and replicated in four wells. The droplet counts (see Subheading 4
with regard to protocol for selection of positive and negative droplets) from all four replicates are combined for
the final analysis. The concentration and Poisson confidence intervals for each “merged” well are computed
using the QuantaSoft software version 1.4.0.99 (Bio-Rad, Hercules, CA). (a) 1D plot shows the separation
between positive droplet (blue) and negative droplets (black). An event with fluorescence amplitude value
>2435 was considered a methylation-positive event (red line). Note the “rain” profile and a shift in baseline in
the 1D amplitude plot, possibly due to the bisulfite conversion treatment. (b) Representative example of results
of ddPCR of serial DNA dilution series. The concentration (copies/μL) of methylated EVL in serially diluted
samples (8 ng of total input DNA per well) are shown: S1 ¼ 80%, S2 ¼ 40%, S3 ¼ 20%, S4 ¼ 10%,
S5 ¼ 5%, S6 ¼ 2.5%, S7 ¼ 1.25%, S8 ¼ 0.625%, and S9 ¼ 0.313%. The error bars represent Poisson 95%
confidence intervals
1000 of unmethylated alleles, or 0.1%. In ddPCR, the LOQ is

primarily scored based on the number of unmethylated alleles that
are detected in the sample using the reference gene assay. For a
given assay, we determine the LOQ by serial diluting the 100%
methylated EpiTect Methyl control DNA with 100% unmethylated
EpiTect Unmethyl control DNA in 8 ng of total DNA in each well.
Figure 4 shows analysis of LOQ for methylated EVL by Methy-
Light ddPCR.
4.3 The Amount of One consideration for increasing the detection capability is to
Input DNA increase the amount of input DNA. For example, to reach a LOQ
of 0.001%, statistically at least 300,000 human haploid genome in a
well, or 1 μg of human DNA, must be screened within 20% CV. If
enough sample DNA is available, one can use about 1 μg of
bisulfite-converted DNA in the 20 μL reaction to effectively detect
more methylated alleles at a LOQ of 0.001%. It is important to
keep in my mind that the source of the DNA, such as a tissue biopsy
may provide very limited amount of DNA. Thus, a practical LOQ
of 0.1% is expected if the total input DNA is less than 10 ng
per well.
4.4 Experimental Random sample partition is the fundamental feature of ddPCR

Replications technology. It also has an important impact on improving an assay’s
LOQ by effectively diluting away the background DNA that can
interfere with detection of methylated alleles, resulting in greater
relative abundance of a rare methylated template in a given droplet.
Running the same samples in multiple wells will further increase the
number of positive events that can be detected by ddPCR, thus
improving the LOQ for low input DNA samples. To achieve a
lower LOQ than 1 in 25, 000 or 0.04% using human DNA, we
recommend replicates of four or higher and creating a merged well
for analysis for detecting rare methylated alleles.
4.5 Proper Controls A proper experimental setup for detecting rare methylated alleles
in Experimental Setup includes positive control wells (100% methylated control DNA),
negative control wells (100% unmethylated control DNA), and
non-template-control (NTC) wells. The number of positive dro-
plets in the negative control wells should be zero, which indicates
optimal specificity of the assay. In addition, NTC wells should also
produce zero positive droplets, which reflects optimal laboratory
practices. Refer to Table 4 if positive droplets are detected in NTC.
The false-positive rate must be considered when designing assays to
achieve accurate quantification.
4.6 Absolute We use the absolute quantification (ABS) experiment design to

Quantification Data quantify the concentration of the methylated alleles in a sample.
Analysis QuantaSoft software counts the number of positive and negative
droplets for each fluorophore in each sample, and fits the fraction of
380 Ming Yu et al.
positive droplets to a Poisson statistics to determine the concentra-

tion in copies/μL of the final ddPCR reaction. We then back-
calculate the original concentration in the starting material. This
type of ABS experiment does not require a standard curve as needed
in conventional MethyLight assays, thus simplifying the experimen-
tal planning and reducing the cost.
To analyze replications, select the merged button in Quanta-
Soft software to combine counts from all the wells with the same
sample name and analyze these data together in multiple wells.
4.7 Types of Errors For experiments with replicates, QuantaSoft software calculates
two types of errors: a theoretical replicate error (technical errors,
or Poisson error) and a standard error of the means (the total
error). It is recommended that total errors are reported for each
measurement. For samples with DNA concentration within the
range of dynamic range of ddPCR (3–106060.6 copies of target
molecules/well), the total error and the Poisson error should be
nearly identical.
However, for quantifying rare methylated alleles that are in the
extreme low end of the concentration range, both the Poisson error
and the total errors will tend to be larger. In addition, errors
introduced by pipetting during dilution or aliquoting are not neg-
ligible. Improving pipetting accuracy is helpful to reduce the total
error.
5 Notes
1. The methods described for isolating and quantifying genomic

DNA are based on protocols we have established in our labora-
tory and have yielded excellent results over many years. How-
ever, genomic DNA can be isolated from FFPE archived tissue
samples, fresh frozen tissues, blood, or stool samples by a
variety of methods, and other established methods that have
worked well for a laboratory should work well for this protocol
[12]. It is essential to include a proteinase K digestion step in
the DNA extraction step to eliminate chromatin structures that
might interfere with DNA dissociation, which is required for
the bisulfite treatment to convert the unmethylated cytosines
to uracil. For DNA extraction from larger FFPE samples or
fresh frozen samples, we prefer the DNeasy® Blood and Tissue
Kit (QIAGEN, catalog #69504). For isolating DNA from
FFPE tissues, we have used QIAamp® DNA FFPE Tissue Kit
(QIAGEN, catalog #56404), which is recommended for up to
eight FFPE tissue sections, each with a thickness of up to 10 μm
and a surface area of up to 250 mm2 in a single preparation. For
DNA extraction from larger FFPE samples or fresh frozen
samples, we use the DNeasy® Blood and Tissue Kit (QIAGEN,

catalog #69504).
2. Other established methods for DNA quantification that have
worked well for a laboratory should work well for this protocol.
However, we recommend using fluorometry based methods to
measure the DNA concentration (e.g., Quant-iT PicoGreen
dsDNA Assay Kit) rather than by absorbance using spectro-
photometry to avoid interference from other biomolecules and
dust particles and to obtain the most accurate estimates of
DNA concentration from FFPE tissue sources, which often
yields highly fragmented DNA.
3. For the majority of DNA methylation studies currently being
performed, the genomic DNA is subjected to a chemical mod-
ification induced by sodium bisulfite, which deaminates
unmethylated cytosine, followed by conversion to uracil and
ultimately to thymine after PCR amplification. Methylated
cytosine remains intact during this process. After bisulfite con-
version, the two strands of DNA are no longer complementary
to each other, which should be kept in mind during the design
of primers and probes as either strand of DNA can be used for
primer and probe selection. The resulting differential sequence
information between methylated and unmethylated alleles is
employed to design methylation-specific primers and probes.
Therefore, bisulfite conversion is a critical step for DNA meth-
ylation studies and bisulfite conversion efficiency of >95% is
necessary. It is essential that a robust sodium bisulfite conver-
sion method be used which is dependent on the starting mate-
rials. There is published literature on this topic that can be used
to guide the selection of a suitable method [13].
4. Control DNA: The EpiTect 100% methylated Control DNA
from QIAGEN are provided already bisulfite-converted and at
a concentration of 10 ng/mL. Store aliquots at 80 C until
use. Repeated freezing and thawing is strongly discouraged as
bisulfite-converted DNA is single stranded and less stable than
double stranded DNA. Other commercial sources such as
CpGenome Human Methylated & Unmethylated DNA stan-
dard set (Millipore, catalog #S8001) should work just as well.
5. ddPCR Supermix for Probes: Store at 20 C until use. Once
thawed, the Supermix can be stored at 4 C for up to 2 weeks.
Repeated freezing and thawing is not recommended, but we
have not observed a decrement in performance of the assays
with up to two freeze-thaw cycles.
6. For an environment completely free of potentially contaminat-
ing DNA, we have had good results using the dead air Air-
Clean600 Workstation from Air Clean Systems.
382 Ming Yu et al.
7. DNA can be isolated from FFPE tissues in the form of tissue

cores or sections. In either case, we use the QIAamp® DNA
FFPE Tissue Kit (QIAGEN, catalog #56404) with only slight
modifications: Cores of tissue will require longer incubations in
xylene, as well as a longer digestion period with a greater vol-
ume of Proteinase K solution (up to 4 days). For sections of
tissues on slides, we add a rehydration step after deparaffiniza-
tion by incubating the slides in a series of baths containing
decreasing concentrations of ethanol. The tissue is then
removed from the slide using a clean (i.e. DNA and RNA free,
RNAse free) razor blade and placed into a microcentrifuge tube
where it is then digested by Proteinase K as normal. We prefer to
perform the final elution in two steps in order to obtain the
greatest yield possible, as FFPE samples are often quite small.
8. Primer and probe design: Primers and probes that work on
conventional MethyLight should always work for MethyLight
ddPCR, but may require further optimization. A minor-groove
binding (MGB) domain on the quencher increases the melting
temperature of the probe, thus allowing for shorter, more
specific probes. The 30 emitters must be FAM and VIC/HEX
because these are the only emitters detected by the QX200
system droplet reader.
9. A 96-well plate is mandatory for use with the droplet reader as
opposed to PCR strip tubes or a 384-well plate. Designing
your plate in columns allows for easier transfer of samples
from the DG8 cartridge to the plate.
10. Droplet generation: One of the most crucial steps is the droplet
generation and transfer, during which a 20 μL PCR reaction is
randomly partitioned into on average 15,000 uniformly sized
droplets, which enable sensitive detection and precise quantifi-
cation. Handling the droplets in a gentle and consistent way is
the key to an optimal droplet count. Reaction mixes should be
well mixed prior to dispensing and the DG8 cartridge should
be preloaded in the DG8 cartridge holder. When loading
samples into the cartridge, the pipette tip should be held at a
~15 angle (from the vertical) and should rest on the ridge near
the bottom of the well while slowly dispensing to avoid intro-
duction of bubbles (which can disrupt droplet formation). Be
sure to fill all eight wells with reaction mixes; it is helpful to
have a second cartridge holder so that one can start preparing
the next set while the first set is running in the generator.
11. Transfer of droplets: After generating droplets in the DG8
cartridge, the droplet solution should look cloudy. We strongly
recommend transferring the droplets to a 96-well PCR plate
using an electronic multichannel pipette set at a very slow
speed. Gently dispense the droplets on the wall of the plate
wells near the bottom of the well. We also recommend

designing the experiment plate layout on a 96-well plate in
columns for ease of transfer from the eight-well droplet gener-
ator cartridges.
Acknowledgments
This work was supported by National Cancer Institute (NCI)

RO1CA194663, P30CA15704, UO1CA152756,
U54CA143862, and P01CA077852 (WMG); Burroughs Well-
come Fund Translational Research Award for Clinician Scientist
(WMG); NIH 2T32DK007742-16 (MY); the Lattner Foundation
(WMG), R.A.C.E. Charities (WMG).
References
1. Slaughter DP, Southwick HW, Smejkal W progression of esophageal adenocarcinoma.

(1953) Field cancerization in oral stratified Cancer Res 61:3410–3418
squamous epithelium; clinical implications of 8. Hayashi M, Guerrero-Preston R, Sidransky D,
multicentric origin. Cancer 6:963–968 Koch WM (2015) Paired box 5 methylation
2. Shen L, Kondo Y, Rosner GL, Xiao L, Hernan- detection by droplet digital PCR for ultra-
dez NS, Vilaythong J, Houlihan PS, Krouse sensitive deep surgical margins analysis of
RS, Prasad AR, Einspahr JG et al (2005) head and neck squamous cell carcinoma. Can-
MGMT promoter methylation and field defect cer Prev Res 8:1017–1026
in sporadic colorectal cancer. J Natl Cancer Inst 9. Yu M, Carter KT, Makar KW, Vickers K, Ulrich
97:1330–1338 CM, Schoen RE, Brenner D, Markowitz SD,
3. Belshaw NJ, Pal N, Tapp HS, Dainty JR, Lewis Grady WM (2015) MethyLight droplet digital
MP, Williams MR, Lund EK, Johnson IT PCR for detection and absolute quantification
(2010) Patterns of DNA methylation in indi- of infrequently methylated alleles. Epigenetics
vidual colonic crypts reveal aging and cancer- 10:803–809
related field defects in the morphologically nor- 10. Weisenberger DJ, Trinh BN, Campan M,
mal mucosa. Carcinogenesis 31:1158–1163 Sharma S, Long TI, Ananthnarayan S,
4. Grady WM, Parkin RK, Mitchell PS, Lee JH, Liang G, Esteva FJ, Hortobagyi GN, McCor-
Kim YH, Tsuchiya KD, Washington MK, mick F et al (2008) DNA methylation analysis
Paraskeva C, Willson JK, Kaz AM et al (2008) by digital bisulfite genomic sequencing and
Epigenetic silencing of the intronic microRNA digital MethyLight. Nucleic Acids Res
hsa-miR-342 and its host gene EVL in colorec- 36:4689–4698
tal cancer. Oncogene 27:3880–3888 11. Trinh BN, Long TI, Laird PW (2001) DNA
5. Zuccotti M, Grant M, Monk M (1993) Poly- methylation analysis by MethyLight technol-
merase chain reaction for the detection of ogy. Methods 25:456–462
methylation of a specific CpG site in the 12. Potluri K, Mahas A, Kent MN, Naik S, Markey
G6pd gene of mouse embryos. Methods Enzy- M (2015) Genomic DNA extraction methods
mol 225:557–567 using formalin-fixed paraffin-embedded tissue.
6. Herman JG, Graff JR, Myohanen S, Nelkin Anal Biochem 486:17–23
BD, Baylin SB (1996) Methylation-specific 13. Holmes EE, Jung M, Meller S, Leisse A,
PCR: a novel PCR assay for methylation status Sailer V, Zech J, Mengdehl M, Garbe LA,
of CpG islands. Proc Natl Acad Sci U S A Uhl B, Kristiansen G et al (2014) Performance
93:9821–9826 evaluation of kits for bisulfite-conversion of
7. Eads CA, Lord RV, Wickramasinghe K, Long DNA from tissues, cell lines, FFPE tissues,
TI, Kurumboor SK, Bernstein L, Peters JH, aspirates, lavages, effusions, plasma, serum,
DeMeester SR, DeMeester TR, Skinner KA and urine. PLoS One 9:e93933
et al (2001) Epigenetic patterns in the
Part V
Gene Expression and RNA Quantification

Chapter 22
Simultaneous Quantification of Multiple Alternatively

Spliced mRNA Transcripts Using Droplet Digital PCR
Bing Sun and Yun-Ling Zheng
Abstract
Currently there is no sensitive, precise, and reproducible method to quantitate alternative splicing of mRNA
transcripts. Droplet digital™ PCR (ddPCR™) analysis allows for accurate digital counting for quantifica-
tion of gene expression. Human telomerase reverse transcriptase (hTERT) is one of the essential compo-
nents required for telomerase activity and for the maintenance of telomeres. Several alternatively spliced
forms of hTERT mRNA in human primary and tumor cells have been reported in the literature. Using one
pair of primers and two probes for hTERT, four alternatively spliced forms of hTERT (α/β+, α+/β
single deletions, α/β double deletion, and nondeletion α+/β+) were accurately quantified through a
novel analysis method via data collected from a single ddPCR reaction. In this chapter, we describe this
ddPCR method that enables direct quantitative comparison of four alternatively spliced forms of the
hTERT messenger RNA without the need for internal standards or multiple pairs of primers specific for
each variant, eliminating the technical variation due to differential PCR amplification efficiency for different
amplicons and the challenges of quantification using standard curves. This simple and straightforward
method should have general utility for quantifying alternatively spliced gene transcripts.
Key words mRNA alternative splicing, mRNA quantification, hTERT, qPCR, Droplet digital PCR
1 Introduction
Humans produce around 150,000 different proteins from their

25,000–30,000 genes, by alternative splicing. More than 70% of
human protein-coding genes are alternatively spliced [1], allowing
the production of different proteins with diverse and even antago-
nistic functions from a single DNA sequence. As alternative splicing
affects numerous genes, it is not surprising that changes in alterna-
tive splicing are frequently associated with human diseases. A large
proportion of human genetic disorders result from splicing var-
iants. Tauopathies is an example of a disease caused by a change in
the ratio of protein isoforms generated by alternative splicing
[2]. Spinal muscular atrophy (SMA) is another example of a reces-
sive disease caused by a point mutation in an exonic regulatory
387
388 Bing Sun and Yun-Ling Zheng
element [3]. Abnormal splicing variants also contribute to the

development of cancer [4, 5]. Given the important role of alterna-
tive splicing in regulating cellular functions, accurate quantification
of multiple alternatively spliced transcripts will facilitate the discov-
ery of new biomarkers for clinical application and enhance our
understanding of the role of alternatively spliced transcripts in
health and disease.
Around 20 alternatively spliced variants of human telomere
reverse transcriptase (hTERT) have been reported; only four variants
(i.e., α/β+ [α deletion], α+/β [β deletion], α/β [α&β double
deletion], or α+/β+ [nondeletion]) are commonly present in most
tumor tissues and may serve as specific markers for cancer diagnosis,
prediction of clinical outcome, and drug targeting [6–12]. hTERT is
involved in the control of telomerase activity and plays significant
roles in cell replication and carcinogenesis. Only the full-length
hTERT (or α+/β+) has been shown to be associated with telomerase
activity. The α-deletion transcript has a 36-bp deletion within the
conserved reverse transcriptase motif and is a dominant negative
inhibitor of telomerase activity [5]. The β-deletion transcript has a
182-bp deletion leading to a truncated protein before the conserved
reverse transcriptase motifs, resulting in a catalytically inactive telo-
merase [8, 9]. Over 90% of tumor tissues have increased telomerase
activity when compared to adjacent normal tissues [10, 11]. How-
ever, the correlation between the level of telomerase activity and the
level of hTERT transcripts in the tumor tissues was low to moderate,
i.e., high level expression of full-length hTERT (or α+/β+) is not
always associated with high telomerase activity [12], which could be
the result of post-transcriptional processes leading to the formation
of several alternatively spliced variants that serve as an additional
mechanisms of telomerase regulation. The relative quantities of
each spliced variant may determine the overall telomerase activity.
For example, the co-presence of α-deletion hTERT will result in the
reduced telomerase activity depending on its expression level relative
to the expression level of full length hTERT (the ratio between α/
β+ and α+/β+).
Next generation sequencing is considered the gold standard for
profiling the expression pattern of alternatively spliced variants.
However its application is limited due to high cost and the large
amount of RNA required, and it does not typically render absolute
transcript copy numbers. Microarray assays (e.g., the Affymetrix
exon microarray, ExonHit) have been developed to detect alterna-
tively sliced variants for specific exons; however, these methods
often cannot distinguish between the different variants (i.e., con-
taining exon A only or exon B only from variants with both exons A
and B). Several real-time PCR (qPCR)-based assays for the enu-
meration of the common hTERT spliced variants have been devel-
oped to determine hTERT splicing patterns, but the correlation
Simultaneous Quantification of Multiple Alternatively Spliced mRNA. . . 389
between the levels of telomerase activity and hTERT transcript

levels is not consistent [12–16]. One caveat is that the qPCR-
based methods have limited sensitivity to precisely quantify some
of the alternatively spliced transcripts that are expressed at very low
levels. In addition, a separate qPCR reaction is required to measure
each alternatively spliced variant with different primer pairs. Varia-
tions in PCR amplification efficiency could compromise the quan-
tification accuracy, making comparisons less reliable. In the present
chapter, using hTERT expression as an example, we describe the
use of droplet digital PCR (ddPCR) for the quantification of the
four major alternatively spliced transcripts of hTERT, by using one
pair of primers in a single PCR reaction, and without the need for
internal standards. This simple method reliably quantitates four
distinguishable clusters of droplets and allows the direct compari-
son of expression levels among the four spliced variants in different
cell types.
2 Materials
2.1 Cell Lines All the cell lines (ovarian cancer OVCar-3, osteosarcoma SAOS-
Evaluated 2 and U2OS, breast cancer MCF-7, lung fibroblast WI-38, and
endometrium adenocarcinoma HEC-1-A) used in the study were
obtained from ATCC (Manassas, VA) or from Cell Culture Core
Facility of Lombardi Cancer Center, Georgetown University Med-
ical Center, and cultured with the media as recommended by the
providers. RNA can also be extracted from fresh tissues or frozen
fresh tissues, and used for transcript quantification with this
method.
2.2 Primer Design 1. All primers and probes were synthesized by Integrated DNA
and Sequences Technologies, Inc. (Coralville, IW).
2. The forward-primer for hTERT cDNA amplification (PCR-F
in Fig. 1) is located at upstream of α-deletion splicing site
(50 -GCCTGAGCTGTACTTTGTCA-30 ). The reverse-primer
(PCR-R in Fig. 1) is at downstream of β-deletion splicing site
with sequence of 50 -CAGCGTGGAGAGGATGGAG-30 . The
amplicon size are 376, 230, 194, and 412 bp for α/β+ (α
deletion), α+/β (β deletion), α/β (α&β double deletion),
or α+/β+ (full length), respectively.
3. Dual-labeled α-deletion probe was designed with FAM as the
fluorescent indicator and ZEN-Iowa Black as the quencher—
positioned as shown in Fig. 1. Sequence [50 -/FAM/
TACTTTGTC/ZEN/AAGGACAGGCTCACG/IABk/30 )]
partially overlaps the transcript both before and after the deleted
region, and specifically recognizes the α-deletion
transcripts only.
Fig. 1 Location of primers and probes for the detection of hTERT mRNA α-deletion (α/β+), β-deletion (α+/
β), and α/β double deletion. Deleted regions are designated by dashed lines
4. The dual-labeled β-deletion probe was designed in a similar

fashion [50 -/HEX/CTTCAAGAG/ZEN/CCACGTCC-
TACGTC/IABk/30 ], using HEX as the fluorescent
indicator.
5. The dual-labeled β probe was designed within the deleted β
region [50 /6-FAM/CTCTACCTT/ZEN/GACA-
GACCTCCAGC/IABkFQ/30 ], using FAM as the fluores-
cent indicator. This probe was used to identify the copresence
of single α-deletion and single β-deletion in the same droplet
using cloned hTERT variants.
6. GAPDH primers and probes: forward primer, 5-
0
-ATTCCACCCATGGCAAATTC-30 ; reverse primer 5-
0
-TGGGATTTCCATTGATGACAAG-30 with an amplicon of
72 bp; and probe, 50 -/HEX/CAAGCTTCCCGTTCT-
CAGCC/IABkFQ/30 was used as a normalization control
for the comparison between samples, assuming that GAPDH
expression is stable across all samples.
2.3 Positive Control To serve as positive controls for identification and quantification of
for hTERT Transcripts each variant, cloned hTERT DNA fragments of the four alterna-
tively spliced variants were amplified by PCR using the hTERT
PCR-F primer and another reverse primer (5-
0
-CAAACAGCTTGTTCTCCATGT-30 ) downstream of hTERT
PCR-R. These amplification products (419, 273, 237, and
455 bp for α/β+, α+/β, α/β, or α+/β+, respectively) were
previously cloned into pCR4-TOPO TA vector (Life Technologies,
Grand Island, NY), and their sequences were verified by Sanger
DNA sequencing (Genewiz, Germantown, MD). These controls
are available upon request.
2.4 ddPCR 1. 2 Droplet Digital™ PCR Supermix for probes (without

Related Materials dUTP; Bio-Rad, cat #186-3025).
and Reagents 2. Droplet generation oil for probes (Bio-Rad, cat #186–3005).
3. Eight-channel DG8™ droplet generation cartridges
(Bio-Rad).
4. Pierceable foil plate heat seal (Bio-Rad);
5. Heat-sealable semiskirted 96-well PCR plates (Eppendorf
North America, Inc., Hauppauge, NY).
6. Droplet Reader oil (Bio-Rad).
7. Pipet-Lite XLS (2–20 μL) and RT-L10F pipette tips (filtered,
Mettler Toledo), both from Rainin (Columbus, OH) for load-
ing ddPCR mix into droplet generation cartridge.
8. Pipet-Lite XLS (8 channels, 5–50 μL) and RT-L200F pipette
tips (filtered, Mettler Toledo), both from Rainin, for transfer-
ring droplets from droplet generation cartridge to Eppendorf
PCR plate.
9. Plate Sealer (Eppendorf).
10. Thermocycler T100 (Bio-Rad).
3 Methods
3.1 RNA Extraction 1. Total RNA from cultured tumor cell lines and fibroblasts was
and cDNA Synthesis extracted with TRIzol total RNA extraction kit (Life
Technologies).
2. RNA measurement: 1 μL of extracted total RNA was analyzed
by spectrometry on a NanoDrop-1000 to determine concen-
tration, with 260/280 ratio 2.0.
3. One μg total RNA was used for cDNA synthesis in a 10 μL
reaction, using the iScript Select cDNA synthesis kit (Bio-Rad)
and 19 nt oligo dT + A/G/C to ensure 1:1 reverse transcrip-
tion (see Subheading 4, Note 1).
3.2 Setting Up ddPCR 1. hTERT reaction mixtures were assembled by combining

Reaction 11.5 μL of 2 ddPCR Supermix (see Subheading 4, Note 2),
forward and reverse primers (250 nM final concentration),
probes (125 nM final concentration each), 1 μL cDNA (equiv-
alent to 50 ng of initial total RNA, depending on expression
levels and RT efficiency) and nuclease free water in a final
volume of 22.5 μL.
2. For GAPDH, the cDNA was diluted 1:2000 and mixed with
same reagents as above. Quantification was performed in a
separate ddPCR reaction.
3.3 ddPCR Reaction PCR reaction and detection were carried out according to manu-
Partioning facturer’s instructions [17].
and Amplification
1. Each assembled ddPCR reaction mixture (20 μL) was loaded
into the sample well of an eight-channel droplet generator
cartridge.
2. Seventy microliters of droplet generation oil was loaded into
the oil well for each channel.
3. The cartridge was placed into the droplet generator after secur-
ing a gasket over the loaded chip, for droplet generation (up to
20,000 droplets per reaction).
4. The droplets for each sample which collected in the cartridge
droplet wells were then manually transferred with a multichan-
nel pipette to a 96-well PCR Eppendorf plate.
5. The plate was heat-sealed with a foil seal and then placed on a
conventional thermal cycler (T100, Bio-Rad).
6. Thermal cycling conditions were 95 C for 10 min (1 cycle),
followed by 45 cycles of 95 C for 30 s and 60 C for 1 min,
then by 98 C for 10 min (1 cycle), and hold at 12 C until
detection. Longer extension time may be required for the
detection of long amplicon.
7. After PCR, the 96-well PCR plate was transferred to the drop-
let reader (QX100, Bio-Rad), which automatically reads the
droplets from each well of the plate using appropriate settings
predefined by the manufacturer (Bio-Rad).
3.4 Analysis To establish the transcript detection specificity of the multiplexed

of Positive Controls hTERT assay, each of the four cloned hTERT variants was amplified
by ddPCR in a separate reaction to determine the respective cluster
locations. Representative results are shown in a 2D plot in Fig. 2.
Droplets for each variant are clearly shown: α deletion (Fig. 2a), β
deletion (Fig. 2b), α&β double deletion (Fig. 2c, see Subheading 4,
Note 3), and nondeletion (Fig. 2d, see Subheading 4, Note 4). The
droplets with nondeletion hTERT variant appeared at slightly
higher fluorescence amplitude relative to the double negative dro-
plets (lacking any of the hTERT transcript isoforms) in the lower
left quadrant in Fig. 2a–c (see explanation in the Subheading 4,
Note 4). The ability of this multiplex assay to distinguish multiple
isoforms in the same well is shown in Fig. 3, in which 2, 3, or
4 cloned hTERT variants are combined in a single ddPCR reaction
to simulate samples from human cells or tissues. To identify the
location of copresence of α-deletion and β-deletion in the same
droplets, cloned α-deletion and β-deletion templates were titrated
and used at high concentrations with 25–50% droplets showing
positive signals (Fig. 3a, c, and d).
A: a-deletion only B: b-deletion only

4,000 4,000
a-/b+ a-/b- a-/b+ a-/b-
3,500 3,500
Droplet FAM intensity (a.u.)
3,000 3,000
2,500 2,500
2,000 2,000
1,500 1,500
1,000 a+/b+ a+/b- 1,000 a+/b+ a+/b-

500 500
0 0
0 500 1,000 1,500 2,000 2,500 0 500 1,000 1,500 2,000 2,500
C: a&b double deletion only D: Non-deletion only

4,000 4,000
a-/b+ a-/b- a-/b+ a-/b-
3,500 3,500
3,000 3,000
2,500 2,500
2,000 2,000
1,500 1,500
1,000 1,000 a+/b+ a+/b-
a+/b+ a+/b-
500 500
0 0
0 500 1,000 1,500 2,000 2,500 0 500 1,000 1,500 2,000 2,500
Fig. 2 2D display of droplet distribution after ddPCR amplification of cloned hTERT alternatively spiced
transcripts (a—α deletion only, b—β deletion only, c—α&β double deletion only, and d—Nondeletion
only) in a single droplet digital PCR reaction containing both α-deletion (FAM) and β-deletion (HEX) probes.
FAM-positive droplets (i.e., α/β+) are in the upper left quadrant. HEX-positive droplets (i.e., α+/β) are in
the lower right quadrant. Double-positive droplets (α/β in dashed rectangle) are in the upper right
quadrant. Weak FAM-positive (α+/β+ in dashed rectangle) or negative droplets are in the lower left quadrant
3.5 Verifying A separate ddPCR reaction with the same primer pair but including
the Detection only the nondeletion β probe (FAM-labeled, Fig. 1) and the
of Colocalized hTERT β-deletion probe (HEX) was performed to assess the ability to
Variants detect the copresence of single α-deletion and single β-deletion in
the same droplet. In the case of hTERT expression, the transcript
levels of each variant in all the cells evaluated are low, even in cancer
cells. Thus the likelihood of having colocalization of two or more
hTERT variants in one droplet is extremely low. It is generally
advised to titrate cDNA sample input for ddPCR to have less than
5000 positive droplets in a reaction in order to avoid the presence
of new clusters due to the colocalization of two or more variants in
one droplet (see Subheading 4, Note 5).
3.6 Data Analysis Instead of using autoanalysis after data acquisition with QuantaSoft
analysis software, manually select “+/”, “/+”, “/” counts
by using the “lasso” function in the 2D plots, relative to the positive
counts for each alternatively spliced transcripts, allowing the
A: a-deletion + b-deletion B: a-deletion + non-deletion

4,000 4,000
a-/b+ Co-presence a-/b+ a-/b+
3,500 3,500
3,000 3,000
2,500 2,500
2,000 2,000
1,500 1,500
1,000 a+/b+ a+/b- 1,000

a+/b+ a-/b+
500 500
0 0
0 500 1,000 1,500 2,000 2,500 0 500 1,000 1,500 2,000 2,500
C: a-deletion + b-deletion + non-deletion D: All 4 alternatively spliced variants

4,000 4,000
a-/b+ Co-presence a-/b+
3,500 3,500 a+/b-
3,000 3,000
2,500 2,500
2,000 2,000 Co-presence

1,500 1,500
1,000 a+/b+ a+/b- 1,000

a+/b+ a+/b-
500 500
0 0
0 500 1,000 1,500 2,000 2,500 0 500 1,000 1,500 2,000 2,500
Droplet HEX intensity (a.u.) Droplet HEX intensity (a.u.)
Fig. 3 2D display of droplet distribution after ddPCR amplification of cloned hTERT variants in different
combination (a—α-deletion + β-deletion, b—α-deletion + nondeletion, c—α-deletion + -
β-deletion + nondeletion, and d—all four alternatively spliced variants) in a single droplet digital PCR reaction
containing both α-deletion (FAM) and β-deletion (HEX) probes. FAM-positive droplets (i.e., α/β+) are in the
upper left quadrant. HEX-positive droplets (i.e., α+/β) are in the lower right quadrant. Double-positive
droplets (α/β in dashed rectangle, and the copresence of α- and β-deletion transcripts between dashed
lines) are in the upper right quadrant. Weak FAM-positive (α+/β+) or negative droplets are in the lower left
quadrant
calculation of the copy numbers of α-deletion (FAM-positive),

β-deletion (HEX-positive), α&β double deletion, or copresence of
single α- and β-deletions (FAM- and HEX-double positive) and
nondeletion forms (weak FAM-positive) in the four quadrants.
More tedious but accurate way to quantitate the four variants in a
given sample can be done by greying out all the droplets and
selecting FAM or HEX positive droplets that belong to a specific
group (for example, Fig. 3d cluster just above negative droplets
α+/β+ are selected as positive. For α/β+, high FAM cluster and
Copresence cluster would be selected, and all the others would
be grey).
4 Notes
1. It is important to make an effort to achieve 1:1 cDNA synthesis

from mRNA by using specific oligonucleotide for your tar-
geted gene transcripts. The oligo dT we used for reverse tran-
scription, not random primers, in combination with reverse
transcriptase with RNase H activity is preferred for cDNA
synthesis, to ensure transcripts were not used more than once
to generate cDNA.
2. To minimize “rain drops” (droplets with relatively much lower
fluorescent counts) when using cloned cDNA in a vector as
control, linearized template should be used. In addition,
primer design and PCR conditions need to be modified to
achieve high PCR amplification and specificity.
3. FAM- and HEX-double positive droplets were detected, as
shown in the upper right quadrant in Fig. 4a, c (dashed rec-
tangles). There are two possibilities for droplets shown in the
upper right quadrant: the presence of a transcript with α&β
double deletion or co-presence of α-deletion and β-deletion
transcripts in the same droplet. In the latter case, two indepen-
dent PCR amplifications using either α-deletion transcript or
β-deletion transcript as template, happened in the same drop-
let. In theory, the fluorescence intensity of double-positive
droplets would be stronger in droplets containing a α&β dou-
ble deletion variant than in droplets containing both α-deletion
and β-deletion variants, possibly due to the difference in length
of their amplicon, with shorter amplicon (α&β double deletion
variants) having higher PCR amplification efficiency. To dem-
onstrate this, we conducted experiments using cloned hTERT
variants with following combination: α-deletion and
β-deletion, α&β double deletion, and all four cloned fragments
together. As shown in Fig. 3a, c, the fluorescent intensity of the
droplets with copresence of single α-deletion and single
β-deletion in the upper-right quadrant (between the dashed
parallel lines) were noticeably weaker comparing to that with
α&β double deletion seen in the same quadrants of Figs. 2c and
3d (dashed rectangle). The large differences in signal intensity
between the copresence of α- and β-deletions and the α&β
double deletion allows the α&β double deletion droplets to
be counted. Accordingly, all the four cloned control templates
in the Fig. 3 were calculated and presented in Table 1. It was
shown that the counts for each variant are consistent among
the different combination measurements. As for the counts of
α&β double deletion and of co-presence, these two popula-
tions of double-positive droplets can be well separated and
A: U2OS Cells B: HEC-1-A Cells

4,000 4,000
a-/b+ a-/b- a-/b+ a-/b-
3,500 3,500
3,000 3,000
2,500 2,500
2,000 2,000
1,500 a+/b+ a-/b- 1,500 a+/b+ a-/b-
1,000 1,000
500 500
0 0
0 500 1,000 1,500 2,000 2,500 0 500 1,000 1,500 2,000 2,500
C: OVCar-3 Cells D: SAOS-2 Cells

4,000 5,000
a-/b+ a-/b+ b+ Co-presence
3,500
4,000
3,000
2,500 3,000
2,000
1,500 a+/b+ a+/b+
- 2,000
1,000 b-
1,000
500
0 0
0 500 1,000 1,500 2,000 2,500 1,000 1,500 2,000 2,500 3,000 3,500 4,000
Droplet HEX intensity (a.u.) Droplet HEX intensity (a.u.)
Fig. 4 2D display of droplet distribution after ddPCR amplification of mRNA from human cancer cell lines. Each
of the three cell line (a—U2OS, b—HEC-1-A, and c—OVCar-3) was amplified in a single droplet digital PCR
reaction containing both α-deletion (FAM) and β-deletion (HEX) probes. FAM-positive droplets (i.e., α/β+)
are in the upper left quadrant. HEX-positive droplets (i.e., α+/β) are in the lower right quadrant. Double-
positive droplets (i.e., α/β or the copresence of single α- and β-deletions) are in the upper right quadrant.
Weak FAM-positive (α+/β+) or negative droplets are in the lower left quadrant. d—cell line SAOS-2 was
amplified in a single droplet digital PCR reaction with nondeletion β (FAM) and β-deletion (HEX) probes.
FAM-positive droplets (i.e., β+) are in the dashed rectangle in upper half. HEX-positive droplets (i.e., β) are
in the dashed rectangle on the right. Double-positive droplets, i.e., copresence of non-β deletion (β+) and
β-deletion (β), are in the upper right quadrant
counted separately. The hTERT expression levels are generally

low in normal tissue or tumor samples, and thus co-presence of
α- and β-deletion variants in the same droplet is unlikely.
Figure 4d shows that the co-presence of β-deletion and non-β
deletion transcripts (the two most abundant spliced variants in
all the cells) in the same droplet is very rare (<0.1%). This
observation is true even in SAOS2 cells that have the highest
hTERT expression level among all the cells tested. Therefore,
the double-positive droplets in the upright corners of the upper
right quadrants in Fig. 4b, c represent the detection of α&β
double deletion transcripts.
Table 1
Calculated copies of each cloned hTERT constructs for the co-presence of α and βdeletion
constructs as shown in Fig. 3
Double-positive droplets
Non- Double- Co-

Mix of cloned variants α-deletion β-deletion deletion deletion presence
α/β+ and α+/β 1392 4580 36 2 202
α/β+ and α+/β+ 1392 3 2098 0 0
α/β+ and α+/β and α+/β+ 1258 4460 1774 0 192
α/β+ and α+/β and α+/β+ 1354 4140 1874 108 182
and α/β
Values were from single measurement
4. During our initial testing, it was observed that, in addition to

the quantification of α-deletion (α/β+, as in Fig. 4b, c) and
β-deletion (α+/β in the right lower quadrant, such as in
Fig. 4a–c), a group of weak FAM-positive droplets were
detected just above the negative droplets in the lower left
quadrant (as shown in dashed rectangles of Fig. 4a–c). We
suspected that these droplets contain nondeletion hTERT
transcripts (α+/β+) which were detected with weak FAM fluo-
rescent signals due to inefficient hydrolysis of α-deletion probe,
as α-deletion probe probably partially annealed to the nondele-
tion form at the annealing temperature during PCR. To con-
firm this, we carried out an experiment with cloned α-deletion
plasmids plus the cloned nondeletion plasmid as templates for
ddPCR. Compared to the results obtained with cloned
α-deletion plasmid only reaction as shown in Fig. 2a, a new
group of weak FAM-positive droplets in the lower left quadrant
(dashed rectangle) appeared after adding cloned nondeletion
hTERT fragments to the cloned α-deletion (Fig. 3b–d). The
quantification accuracy for the nondeletion form was also con-
firmed by the subtraction of α-deletion counts from non-β
deletion counts obtained from the experiments with nondele-
tion β probe and β-deletion probe. Although the detection is
coincident and may not apply to all situations, similar approach
with proper probe design plus optimization testing may work
for other gene transcript detection as well.
5. When the expression levels are high, precaution should be
taken to appropriately dilute template/sample if ddPCR will
be used. On the other hand, especially when the gene
expression level is very low, ddPCR amplification conditions

including the primer design should be optimized to avoid any
nonspecific amplification in order to obtain accurate counts for
each variant.
6. For normalization with a reference gene, titration of the input
template of the reference gene is required for accurate quanti-
fication because most of the reference genes such as GAPDH or
β-actin are expressed at much higher levels compared to the
targeted gene. After normalization by GAPDH expression, this
method permits the comparison of α-deletion, β-deletion, α&β
double deletions, nondeletion, and total hTERT expression
among different samples. Table 2 shows that SAOS-2 cells
expressed the highest level of total hTERT, double deletion
and β-deletion forms, whereas HEC-1-A cells expressed the
highest percentage of α-deletion among all the cells tested.
The U2OS is a cell line using an alternative lengthening of
telomeres (ALT) mechanism to maintain its telomere length
and have been shown lack of telomerase activity [18]. We
demonstrated the expression of extremely low levels of
β-deletion and nondeletion hTERT transcripts in this cell
line, indicating the high sensitivity of the method. The lack of
telomerase activity in the U2OS cells is likely due to the absence
of hTERC expression, the template component of telomerase
complex (unpublished observation with ddPCR). Lack of
hTERC expression is also observed in several other cell lines
depending on the ALT to maintain telomere length [19].
7. Most importantly, this method allows direct comparison of the
expression levels among the four alternatively spliced hTERT
variants. As shown in Fig. 4b, the α-deletion variant HEC-1-A
cells are expressed relatively higher in comparison to β-deletion
variant than that in OVCar-3 cells (Fig. 4c). The ratio of double
deletion variant to α-deletion variant in the OVCar-3 cells is
higher than that in HEC-1-A cells. The expression profiling of
MCF-7 cells obtained in the present study with α-deletion at
1.87%, β-deletion at 63.50%, double deletion at 2.79%, and
nondeletion at 31.85% (Table 2) are very similar to the data
reported previously with qPCR (estimated at 3.5%, 62.5%,
3.0%, and 31.5%, respectively) [15]. To the best of our knowl-
edge, the method presented here is the first to provide a simple,
precise and absolute quantification of four alternatively spliced
hTERT transcripts, in particular for the detection of single as
well as double deletions simultaneously for direct comparison.
8. The new method not only minimizes experimental errors that
may be introduced by pipetting, but also eliminates variations
from PCR amplification efficiency.
Table 2
The expresision of 4 alternatively spliced hTERT transcripts in cell lines
Expression relative to GAPDH (x 1000) Percentage of total hTERT expression
Cell line α-deletion β-deletion Double-deletion Non-deletion Total hTERT α-deletion β-deletion Double-deletion Non-deletion
SAOS-2 47.09 3668.28 253.03 1049.76 5018.16 0.94 73.10 5.04 20.92
U2OS 0.00 13.93 0.72 8.93 23.58 0.00 59.08 3.07 37.85
WI-38 0.00 1.44 0.00 0.47 1.91 0.00 75.44 0.00 24.56
HEC-1-A 37.61 162.44 14.35 622.07 836.46 4.50 19.42 1.72 74.37
OVCar-3 23.10 779.54 45.89 435.76 1284.28 1.80 60.70 3.57 33.93
MCF-7 20.83 708.88 31.12 355.50 1116.33 1.87 63.50 2.79 31.85
Values were the average of duplicate measurements
Simultaneous Quantification of Multiple Alternatively Spliced mRNA. . .
399
Acknowledgments
Research in YLZ’s laboratory is supported by grants from the

National Cancer Institute of the National Institutes of Health
(R01CA132996) and Susan G. Komen for the Cure (KG100283).
References
1. Matlin AJ, Clark F, Smith CWJ (2005) Under- 11. Fujiwara-Akita H, Maesawa C, Honda T et al
standing alternative splicing: towards a cellular (2005) Expression of human telomerase
code. Nat Rev Mol Cell Biol 6:386–398 reverse transcriptase splice variants is well cor-
2. Hong M, Zhukareva V, Vogelsberg-Ragaglia V related with low telomerase activity in osteosar-
et al (1998) Mutation-specific functional coma cell lines. Int J Oncol 26:1009–1016
impairments in distinct tau isoforms of heredi- 12. Zaffaroni N, Villa R, Pastorino U et al (2005)
tary FTDP-17. Science 282:1914–1917 Lack of telomerase activity in lung carcinoids is
3. Lefebvre S, Burglen L, Reboullet S et al (1995) dependent on human telomerase reverse tran-
Identification and characterization of a spinal scriptase transcription and alternative splicing
muscular atrophy-determining gene. Cell and is associated with long telomeres. Clin
80:155–165 Cancer Res 11:2832–2839
4. Faustino NA, Cooper TA (2003) Pre-mRNA 13. Yi X, Shay JW, Wright WE (2001) Quantitation
splicing and human disease. Genes Dev of telomerase components and hTERT mRNA
17:419–437 splicing patterns in immortal human cells.
5. Cáceres JF, Kornblihtt AR (2002) Alternative Nucleic Acids Res 29:4818–4825
splicing: multiple control mechanisms and 14. Ohyashiki JH, Hisatomi H, Nagao K et al
involvement in human disease. Trends Genet (2005) Quantitative relationship between func-
18:186–193 tionally active telomerase and major telomerase
6. Ulaner GA, JF H, TH V et al (1998) Telome- components (hTERT and hTR) in acute leu-
rase activity in human development is regulated kaemia cells. Br J Cancer 92:1942–1947
by human telomerase reverse transcriptase 15. Mavrogiannou E, Strati A, Stathopoulou A et al
(hTERT) transcription and by alternate splic- (2007) Real-time RT-PCR quantification of
ing of hTERT transcripts. Cancer Res human telomerase reverse transcriptase splice
58:4168–4172 variants in tumor cell lines and non-small cell
7. Colgin LM, Wilkinson C, Englezou A et al lung cancer. Clin Chem 53:53–61
(2000) The hTERTalpha splice variant is a 16. Hara T, Noma T, Yamashiro Y et al (2001)
dominant negative inhibitor of telomerase Quantitative analysis of telomerase activity
activity. Neoplasia 2:426–432 and telomerase reverse transcriptase expression
8. Listerman I, Sun J, Gazzaniga FS et al (2013) in renal cell carcinoma. Urol Res 29:1–6
The major reverse transcriptase-incompetent 17. Hindson BJ, Ness KD, Masquelier DA et al
splice variant of the human telomerase protein (2011) High-throughput droplet digital PCR
inhibits telomerase activity but protects from system for absolute quantitation of DNA copy
apoptosis. Cancer Res 73:2817–2828 number. Anal Chem 83:8604–8610
9. Lincz LF, Mudge LM, Scorgie FE et al (2008) 18. Ulaner GA, JF H, TH V et al (2000) Regula-
Quantification of hTERT splice variants in mel- tion of telomerase by alternate splicing of
anoma by SYBR green real-time polymerase human telomerase reverse transcriptase
chain reaction indicates a negative regulatory (hTERT) in normal and neoplastic ovary,
role for the beta deletion variant. Neoplasia endometrium and myometrium. Int J Cancer
10:1131–1137 85:330–335
10. Liu Y, BQ W, Zhong HH et al (2012) Quanti- 19. Hoare SF, Bryce LA, Wisman GBA et al (2001)
fication of alternative splicing variants of Lock of telomerase RNA gene hTERC expres-
human telomerase reverse transcriptase and sion in alternative lengthening of telomere cells
correlations with telomerase activity in lung is associated with methylation of the hTERC
cancer. PLoS One 7:e38868 promoter. Cancer Res 61:27–32
Chapter 23
Using Droplet Digital PCR to Analyze Allele-Specific

RNA Expression
Nolan Kamitaki, Christina L. Usher, and Steven A. McCarroll
Abstract
Genome-wide association studies have discovered thousands of common alleles that associate with human
phenotypes and disease. Many of these variants are in non-protein-coding (regulatory) regions and are
believed to affect phenotypes by modifying gene expression. In any organism with a diploid genome, such
as humans, measuring the expression of each allele of a gene provides a well-controlled way to identify allelic
influences on that gene’s expression. Here, we describe a protocol for precisely measuring the allele-specific
expression of individual genes. This method targets the nucleotide differences between the two alleles of a
gene within an individual and measures the “allelic skew,” the extent to which one allele is expressed more
than the other. We cover the design of effective assays, the optimization of reactions, and the interpretation
of the resulting data.
Key words Allele-specific expression, Allelic skew, Allelic imbalance, Droplet digital PCR, Digital
PCR, mRNA expression, Assay design
1 Introduction
Genome-wide association studies (GWAS) have mapped hundreds

of human traits to thousands of common variants [1]. Less than
30% of these variants are nonsynonymous or are in linkage disequi-
librium with a variant that is [2], suggesting that much of the
variation driving phenotypic differences does so not by altering
protein function, but by altering gene expression. Furthermore,
systematic substitution of DNA bases in known enhancers has
shown that the vast majority of these variants modestly affect
expression (less than twofold change) and very few abolish or
greatly increase expression [3]. Measuring these modest effects on
gene expression requires techniques that can both ascertain modest
differences in expression and can filter through the noise created by
environmental influences, genomic background, and other trans-
acting effects.
401
402 Nolan Kamitaki et al.
Once a candidate gene has been selected, usually by hypotheses

about the mechanisms and genetic architecture of a phenotype, we
can compare the expression of the two alleles within the same
biological sample, thus avoiding many sources of noise. In general,
a person’s two alleles are subject to the same environmental and
trans-acting genetic influences, which are potentially strong
enough to obscure modest genetic effects and can be difficult to
control for in study designs that assay total expression in genetically
diverse humans living in variable environments [4].
In the absence of allele-specific regulatory effects, the two
alleles will be expressed at a ratio of 1:1, but local cis-acting genetic
and epigenetic variation can cause one allele to be expressed more
than the other. Measuring this difference requires a technology
capable of measuring the two alleles precisely and with equal sensi-
tivity. Techniques that take this measurement after PCR amplifica-
tion—including RNA-seq and pyrosequencing—run the risk of
confusing amplification effects with real genetic effects. Even a
subtle difference of 1% in amplification efficiency will, after
30 PCR cycles, result in a 1.35-fold relative difference in the
abundance of the two alleles in the amplified material, an effect
size that would exceed that of most enhancer SNPs. In addition,
digital measurement, which counts the number of transcripts with
each allele in a sample, is preferable to analog measurement, which
estimates ratios, since digital measurement allows the statistical
significance of allelic skews to be estimated for each person.
The ability to digitally count the abundance of alleles in an
unamplified sample is therefore critical. 1-Step RT-droplet digital
PCR (RT-ddPCR) accomplishes this by partitioning the individual
RNA transcripts of a gene into separate reaction compartments
(droplets) before amplification/detection [5]. Within these
~20,000 droplets, the RNA is reverse-transcribed into cDNA and
amplified using a molecular assay that is composed of: a pair of
primers that will amplify both alleles, a fluorescence probe targeting
the reference allele, and a probe (with a different fluorophore)
targeting the alternate allele. The number of RNA molecules ori-
ginating from each allele is quantified by counting the number of
droplets that fluoresce with the corresponding probe colors.
Provided that the positive and negative droplets are clearly and
consistently distinguished, ddPCR is robust to differences in the
amplification kinetics of the alleles (a clear advantage over pyrose-
quencing and real-time PCR). Droplet digital PCR is also more
resistant to the effects of nonspecific binding of the probes to the
opposite allele, allowing the determination of the correct allele in
each droplet as long as each probe has greater affinity for its respec-
tive allelic target.
Here, we share our protocol for measuring the allelic skew of a
gene using ddPCR. We cover how to design a successful allele-
specific assay, how to screen and optimize that assay, and finally,
Using Droplet Digital PCR to Analyze Allele-Specific RNA Expression 403
how to use it in ddPCR. We focus on assay design in humans, but

the protocol is readily adaptable to any species with a diploid
genome and substantial heterozygosity. It is also readily adapted
to measure the expression levels of paralogous genes in a paralog-
specific manner.
However, this method has limitations in that it can only assay
those individuals heterozygous for the transcribed reporter SNP,
and historic recombination between the reporter SNP and the
functional variant may introduce uncertainty about the direction
of effect. Thus, the clearest overall picture may emerge when this
strategy is combined with another that assays total expression across
a large cohort [6].
2 Materials
2.1 Locus-Specific 1. 18 μM forward primer; 18 μM reverse primer; 5 μM 5’

Analysis Reagents FAM-labeled, 3’ Iowa Black ZEN-quenched probe (see Note
1) that targets the reference allele; and 5 μM 50 HEX-labeled, 30
Iowa Black ZEN-quenched probe that targets the alternate
allele. Store at 20 C away from light (see Note 2). Vortex
and centrifuge before use.
2.2 Biological 1. RNA samples (100 fg–100 ng of total RNA) from individuals
Sample heterozygous for the reporter SNP (see Note 3). They need not
be heterozygous for the candidate functional variant. Store at
80 C in RNase-free ddH2O. Vortex and centrifuge
before use.
2.3 ddPCR 1. 20 assay mix from Subheading 2.1.

Components 2. One-Step RT-ddPCR Advanced Kit for Probes (Bio-Rad).
and Equipment Store at 20 C. Invert to mix before use.
3. Droplet Generation Oil for Probes (Bio-Rad). Store at room
temperature.
4. Droplet Reader Oil (Bio-Rad). Store at room temperature.
5. DG8™ Cartridges for droplet generation (Bio-Rad).
6. DG8™ Gaskets for droplet generation (Bio-Rad).
7. QX200™ Droplet Digital PCR System: droplet generator and
cartridge holders, droplet reader, and QuantaSoft reader soft-
ware (Bio-Rad).
8. Rainin multichannel pipettors and corresponding tips for
pipetting 20 μL and 40 μL volumes (see Note 4).
9. Half-skirted Eppendorf 96-well PCR plates for droplet thermal
cycling and reading.
10. Pierceable, heat-sealable foil seals (e.g., Bio-Rad Pierceable

Foil Heat Seal).
11. Plate sealer capable of sealing for 5 s at 180 C (e.g., Bio-Rad
PX1™ Plate Sealer).
12. Thermal cycler, preferably one capable of a 4–12 C hold and a
heated lid.
2.4 Web Resources 1. UCSC genome browser (hg19): http://genome.ucsc.edu/cgi-

bin/hgGateway.
2. 1000 Genomes browser: http://browser.1000genomes.org/
index.html.
3. SNAP (SNP annotation and Proxy Search): https://www.bro
adinstitute.org/mpg/snap/.
4. Genotype-Tissue Expression (GTEx) Project: http://www.
gtexportal.org/home/.
5. IDT OligoAnalyzer: http://www.idtdna.com/calc/analyzer.
6. Primer3Plus primer design tool: http://www.bioinformatics.
nl/cgi-bin/primer3plus/primer3plus.cgi/ [7].
3 Methods
3.1 Selection The first step (see Fig. 1, Subheading 3.1) is to identify one or more
of Reporter SNPs reporter variants (transcribed heterozygous sites) in the gene of
for Assays interest, which is generally chosen through hypotheses about the
mechanisms of disease. These reporter variants are generally single
nucleotide polymorphisms (SNPs) within the target gene’s RNA
transcript. The chosen SNP will be used just as a reporter of a
transcript’s chromosome-of-origin; it need not be the variant caus-
ing the allelic skew, nor necessarily correlated via linkage disequilib-
rium with that variant, though it may be useful in downstream
analysis to know the relationship between the reporter and
causal SNP.
Using the UCSC genome browser, enter the gene name of
interest and select the assembly hg19 to focus on that locus. Under
the “Variation” header at the bottom of the page, change the
“1000G Ph1 Vars” track display setting to “full” and click
“refresh.” Any exonic variant appearing in this track, as well as
those in the 50 UTR and 30 UTR, can be used (see Fig. 2). In
addition, intronic SNPs can sometimes be utilized, especially if
RNA-seq data indicates that the corresponding sequence is
detected at a meaningful level in total RNA (see Note 5), reflecting
that sequence’s presence in cells for an appreciable period after
transcription and before splicing.
Fig. 1 A schematic following the steps of assay design for the GAPDH gene, as referenced sequentially in the
chapter text
Fig. 2 A screenshot of the UCSC genome browser
1. Find the parameters needed for evaluating potential reporter

SNPs. There are two main components to consider: the num-
ber of individuals that are heterozygous for the reporter SNP
and the abundance of transcription at that SNP. Note that due
to the evolutionary pressure on coding sequence variation,
many genes may have only one or two common, high read-
depth SNPs that are amenable for evaluating allelic skew.
(a) Because homozygotes cannot be evaluated, determine or
estimate the number of heterozygotes among the avail-
able RNA-sample donors. If genotype data are already
available, this can be measured directly. If not, heterozy-
gosity can be estimated from population-level data from a
1000 Genomes Project population sample whose ancestry
matches your study population. Obtain the minor allele
frequency (MAF) of the SNP using a source, such as the
1000 Genomes browser, and calculate the heterozygos-
ity—2(MAF)(1-MAF), which is derived from the
Hardy–Weinberg equilibrium value 2pq. The greater the
heterozygosity, the more samples will be usable.
(b) Determine or estimate the level of expression of the
potential reporter SNPs; this will be useful for determin-
ing the amount of RNA input and selecting a SNP that is
substantially expressed, despite alternative splicing or
potentially being in an intron (see Note 5). The average
number of sequencing reads, termed read depth and often
reported as RPKM or FPKM (reads or fragments per
kilobase per million reads), that contain the exon/intron
in which the reporter SNP resides can be used as a rough
proxy for expression. The read depth can be found for any
exon across a range of tissues at the GTEx website by
searching for a gene and positioning the mouse cursor
over the exon in the row corresponding to the desired
tissue type. The higher the read depth, the more the exon
containing the reporter SNP is expressed, and the more
informative that SNP will be given less input RNA.
(c) (Optional) If there is a prior hypothesis that a particular
functional variant is generating allele-specific expression
differences, then the linkage disequilibrium between this
variant and any potential reporter SNPs should be calcu-
lated using the 1000 Genomes browser or SNAP in the
population most genetically similar to that of the
RNA-sample donors (e.g., samples from donors of Cau-
casian ancestry should be evaluated using European-
ancestry populations such as CEU). A reporter SNP with
a high D’ (a measure of recombination) and an MAF
greater than the candidate functional variant’s MAF will
be most helpful for validating that candidate and deter-
mining its direction of effect (see Subheading 3.5, step
2 and Note 6).
2. Determine which potential reporter SNP(s) will be the most
informative. A general guideline for ranking these SNPs is
(heterozygosity)(expression_level), which assigns roughly
equal weighting to the number of heterozygotes, which influ-
ences sample size, and expression level, which influences mea-
surement precision. With ample RNA (>20 ng) and moderate
to high expression of the gene (>10 RPKM), the number of
heterozygous samples available will be more important than
the level of expression. Conversely, when a gene is expressed at
low levels or a smaller quantity of usable RNA is available for
analysis, it may be more important to select variants with higher
expression levels. Select the top reporter SNP(s) based on these
criteria. If there are multiple suitable reporter SNPs, it is useful
to select two or more with low D’ to each other, to increase the
likelihood of having a reporter SNP with high D’ to the func-
tional variant (see Subheading 3.1, step 1c).
3.2 Assay Design 1. Obtain the DNA sequence flanking the reporter SNPs for assay
design (see Fig. 1, Subheading 3.2, step 1). Enter the SNP
coordinates into the UCSC genome browser and navigate to
that genomic region. The browser should be zoomed in on
that variant.
(a) Obtain the surrounding DNA sequence by selecting View
> DNA from the tool bar.
(b) Under the options, add 100 extra bases upstream and
100 extra bases downstream.
(c) Check the “Mask Repeats” option and select “Mask to
N.” This blocks the primer-design software from using
sequences that are repeated many times in the human

genome.
(d) Hit the “extended case/color options” button and check
both the “Underline/Human mRNAs” option and the
“Bold/Common SNPs” option.
(e) Hit submit.
The resulting sequence will be 201 bp long, with the target
SNP in the middle, all SNPs in bold, and the exonic mRNA
sequence underlined. If not all of the sequence is underlined,
increase the “extra bases” parameter (possibly up to 1000s of
bases) until a total of 100 bases on either side of the SNP are
underlined. Repeat-masked sequence does not count in this
total.
2. Change the DNA sequence to exclude introns and SNPs (see
Fig. 1, Subheading 3.2, step 2). Unless designing a reporter
SNP assay to an intronic SNP, remove the intronic sequence
(the sequence that is not underlined), leaving the mRNA tran-
script. To prevent the primer-design software from designing
primers that bind to polymorphic bases, change all the bolded
SNPs (except the target SNP) to N (see Note 7).
3. Determine which strand the probes should bind (see Fig. 1,
Subheading 3.2, step 3). Selection of the best strand (plus or
minus strand) for an assay increases the extent to which the
correct allele-specific probe will outcompete the other probe
when amplifying the target allele and is done by balancing each
probe’s ΔTM (delta melting temperature ¼ correct match TM—
mismatch TM). Thus, the ΔTM’s for a pair of probes binding to
one strand of both allelic sequences will be compared to a pair
of probes binding to the other strand. To determine all possible
ΔTM’s, select the sequence from 10-bp upstream to 10-bp
downstream of the reporter SNP (Seq1). Copy/paste the
sequence and change the reporter SNP to its alternate allele
(Seq2). Obtain the reverse complement sequence for both
(cSeq1 and cSeq2). Use IDT OligoAnalyzer with these settings
for all following tests:
(a) Target type: DNA.
(b) Oligo Conc: 0.05 μM.
(c) Na+ Conc: 50 mM.
(d) Mg++ Conc: 3 mM.
(e) dNTPs Conc: 0 mM.
Enter Seq1 and select the Tm Mismatch tool. Change the
generated complementary sequence to cSeq2. Record the
DELTA TM. Repeat for Seq2, cSeq1, and cSeq2 with editing
the generated complementary sequence to cSeq1, Seq2, and
Seq1 respectively. If the DELTA TM values when entering Seq1
and Seq2 are more equal than those for cSeq1 and cSeq2,
proceed with the sequence already obtained. If cSeq1 and
cSeq2 are more equal, then proceed with the reverse comple-
ment of the 201-bp mRNA segment from the previous step. If
there were multiple potential reporter SNPs, prioritize on
those with a higher sum of DELTA TM for Seq1 and Seq2 or
cSeq1 and cSeq2.
4. Design the primers and probe to the reference allele using the
Primer3Plus primer design tool. Enter the 201-bp mRNA
sequence obtained in Steps 1–3 into the box at the top of the
webpage. Check “Pick hybridization probe (internal oligo)”
under the input sequence. Under “General Settings,” click the
“Mispriming/repeat library” pull-down menu above the
sequence box and choose “HUMAN.” The following condi-
tions are for an assay that will be run with an annealing temper-
ature around 60 C, but depending on local sequence
surrounding the SNP (for example, when this sequence is
AT-rich), it may be necessary to use a different annealing
temperature (see Note 8).
(a) Under “General Settings” set:
l The Primer Size as Min: 18, Opt: 20, Max: 27.
l The Primer Tm as Min: 57, Opt: 60, Max: 63.
l The Primer GC% as Min: 20, Opt: 50, Max: 80.
l The Concentration of divalent cations to 3.8.
l The Concentration of dNTPs to 0.8.
(b) Under “Advanced Settings” set:
l The Table of thermodynamic parameters to “SantaLu-
cia 1998”.
l The Salt correction formula to “SantaLucia 1998”.
(c) Under “Internal Oligo” set:
l The Hyb Oligo size as Min: 15, Opt: 20, Max: 27.
l The Hyb Oligo Tm as Min: 64, Opt: 65, Max: 70.
l The Hyb Oligo GC% as Min: 20, Opt: 50, Max: 80.
l The Hyb Oligo Mishyb Library to “HUMAN”.
(d) Force Primer3Plus to design the probe over the target
SNP by setting the Hyb Oligo Excluded Region to “1,80
121,80”.
(e) Click “Pick Primers” and select a probe and primer set in
which the probe does not start with a G (50 G) and the
probe overlaps the mutation of interest with at least four
bases on either side.
(f) If no assay designs pass these design restrictions, relaxing

the parameters can be attempted. However, designs
requiring significant changes may require trying a differ-
ent reporter SNP if one is available (see Note 9).
5. Design the probe to the alternate allele. Within the mRNA
sequence, change the target SNP to the alternate (nonrefer-
ence) allele. Copy and paste the left primer from Step 4 into the
box below “Pick left primer, or use left primer below:” and
copy and paste the right primer from Step 4 into the box below
“Pick right primer, or use right primer below.” Using the same
primer design settings as Step 4, select a probe that does not
start with a 50 G and overlaps the mutation of interest with at
least four bases on either side. Note, succeeding in the design
of the first allele does not guarantee success with the other
allele.
6. Ensure that the assay targets the correct region and is not
predicted to have off-target amplification. Use the UCSC
BLAT tool (under the “Tools” menu at the top of the page)
to check that the two primers and reference-allele probe match
only the region of interest perfectly (see Note 10). View the
region with the “Common SNPs” track displayed to ensure the
primers and probes do not bind over other SNPs. In addition,
check whether the primers and probes are likely to bind each
other by using the “Hetero-Dimer” option in the right menu
bar of IDT’s OligoAnalyzer; ΔGs lower than 7 should be
avoided.
7. Order the primers and probe from your usual oligonucleotide
supplier. For example, the 50 FAM and 50 HEX probe fluoro-
phores, both with the Iowa Black 30 ZEN quencher from
Integrated DNA Technologies. When making the 20 assay
mix, please note that the proportion of primers to probes is
different than in qPCR.
3.3 RT-ddPCR 1. (Optional) Before running the assay on samples of interest,

for Determining Allelic and particularly if the DELTA TM values from Subheading 3.2,
Imbalance step 3 were low, it may be useful to run trial reactions with
varying annealing temperatures using heterozygous genomic
DNA (see Fig. 3) and with varying RNA inputs to optimize
cluster separation (see Notes 16 and 17).
2. Thaw the RNA and the reagents of the One-Step RT-ddPCR
Advanced Kit for Probes on ice for 30 min. Vortex and centri-
fuge briefly.
3. For each reaction, assemble:
6.25 μL Supermix (One-Step RT-ddPCR Advanced Kit for
Probes).
12000
Channel 1 Amplitude
10000
8000
6000
4000
2000
0
0 1000 2000 3000 4000 5000 6000 7000
Channel 2 Amplitude
12000
Channel 1 Amplitude
10000
8000
6000
4000
2000
0
0 1000 2000 3000 4000 5000 6000 7000
Channel 2 Amplitude
12000
Channel 1 Amplitude
10000
8000
6000
4000
2000
0
0 1000 2000 3000 4000 5000 6000 7000
Channel 2 Amplitude
12000
Channel 1 Amplitude
10000
8000
6000
4000
2000
0
0 1000 2000 3000 4000 5000 6000 7000
Channel 2 Amplitude
12000
Channel 1 Amplitude
10000
8000
6000
4000
2000
0
0 1000 2000 3000 4000 5000 6000 7000
Channel 2 Amplitude
2.5 μL Reverse Transcriptase (One-Step RT-ddPCR Advanced

Kit for Probes).
1.25 μL 300 mM DTT (One-Step RT-ddPCR Advanced Kit
for Probes).
μL of 20 assay targeting the SNP.
variable input of RNA (100 fg–100 ng per reaction) (see Note 3).
<13.75 μL of DEPC-treated water.
25.0 μL total volume (see Note 11).
*Some may find it helpful to run a control sample with hetero-
zygous DNA to confirm where the droplets cluster. In addi-
tion, because the allelic imbalance statistics can be run using a
single reaction, duplicates of a sample are not necessary unless
the RNA is rare, necessitating combining several reactions to
have enough positive droplets to make a definitive call.
4. Vortex and spin the plate to collect the liquid at the bottom of
wells. Keep the plate protected from light until droplet genera-
tion, and allow the reactions to equilibrate to room tempera-
ture for 3 min prior to droplet generation (see Note 12).
5. Place a DG8 cartridge into the QX200 droplet generation
cartridge holder (Bio-Rad).
(a) Pipette 20 μL of the PCR mix into the middle row of the
cartridge. Only push down to the first stop when ejecting
liquid to ensure there are no air bubbles in the sample (see
Note 13). Using a Rainin multichannel pipettor with
Rainin tips is preferred at this stage (see Note 4).
(b) Pipette 70 μL of oil into the bottom row of the cartridge
using a new set of tips. Always be sure to pipette the oil
after the samples. The top row (where droplets are out-
putted) is left empty.
(c) Place a DG8 rubber gasket over the cartridge.
6. Place the cartridge holder with cartridge and gasket into the
QX200 droplet generator. Close the generator and droplets
will be formed.
7. When the triangles on the button on the droplet generator lid
return to being lit solid green, remove the cartridge. Carefully
remove its gasket and transfer the droplets in the top row to a
clean, half-skirted Eppendorf plate. It is important that the
pipetting at this stage is slow and careful with the pipette

Fig. 3 A temperature gradient run using the same sample and assay. Note how the clusters are intermingled in
(a), yet increase in separation with increasing annealing temperature in (b), (c), and (d) until an optimal
temperature is reached in (e). If this gradient were run with values below 50 C, we would see the clusters
start to intermingle again as we move away from the optimal annealing temperature
oriented at 45 degrees, otherwise the droplets may shear.

Afterward, discard the gasket and cartridge (see Note 14).
8. After all droplets are made, seal the droplet plate with a foil seal
by heating the top to 180 C for 5 s. Check to make sure that
the outlines of the wells are visible through the foil, indicating a
good seal.
9. Thermal cycle as follows:
42–50 C for 60 min (see Note 15).
95 C for 10 min.
40 cycles of 95 C for 30 s followed by 60 C (or identified
optimal annealing temperature; see Note 16) for 1 min.
98 C for 10 min.
4 C hold.
Use a 2.5 C per cycle ramp rate for all steps and set to 40 μL
volume. Droplets can be stored protected from light at 4 C
after cycling for up to 24 h before reading.
10. Set up a template on the QX200 droplet reader computer.
Open QuantaSoft. Under “Template” in the top left corner,
select “New” in order to fill in a new plate map. Double click
on the first well. In the “Sample” box, under “Experiment,”
select any of the “ABS” experiments, and then, under “Super-
mix,” select “One-Step RT-ddPCR Kit for Probes.” In the
“Target 1” box, enter the name of the FAM assay in the
“Name” field and select “Ch1 Unknown” from the “Type”
menu. In the “Target 2” box, enter the name of the HEX or
VIC assay in the “Name” field and select “Ch2 Target” from
the “Type” menu. Select all wells of the plate that will contain
samples that are using the same assays. Click the blue “Apply”
button in the top window. The name of each sample may also
be entered. Once finished, click “OK” and save the template.
11. Read the droplets on the QX200 droplet reader. Put the plate
into the plate holder of the QX200, then place the black plate
holder on top and click the silver tabs on either side down into
place, making sure the A1 well is in the top left corner. Close
the lid of the QX200. In QuantaSoft on the QX200 computer,
make sure the template created in Step 9 is loaded, then click
“Run.” On the popup menu that appears, select “FAM/HEX”
or “FAM/VIC,” depending on the pair of fluorophores used,
then click “OK.”
3.4 Data Finalization 1. After the ddPCR is complete, perform a well-by-well visual
and Quality Control inspection of droplet clusters in QuantaSoft by clicking “Ana-
lyze” on the leftmost menu followed by “2D Amplitude.”
Verify that there is clear separation between the positive and
negative clusters for both the reference and alternate allele
fluorophores. Some droplets between the positive and negative

clusters are acceptable, but if the clusters themselves overlap, it
will undermine the accuracy of the measurements. If all the
wells have bad cluster separation, first try optimizing the
annealing temperature and RNA concentration (see Notes 16
and 17, Fig. 4). If the assay is still not generating clean data,
redesign the assay to the same or a different reporter SNP (see
Subheading 3.1).
2. Evaluate and adjust the software’s automatic calling for each
droplet cluster in each well. Droplets in the top left corner of
the 2D amplitude plot are FAMþ, droplets in the bottom right
corner are VICþ (or HEXþ), droplets in the top right corner
are double positive, and droplets in the bottom left corner are
double negative. Though ddPCR is usually highly accurate for
assays targeting independent loci (as in CNV analysis, in which
the two probes interrogate different nucleic acid sequences),
for allele-specific expression, where the two sequences of inter-
est are highly similar, the signal in the two fluorescence chan-
nels is often not independent, due to the probes sometimes
binding the opposite allele. The lack of rectangularity in the
cluster locations (in two-dimensional fluorescence space) may
a. Balanced expression
12000
Channel 1 Amplitude
10000
8000
6000
4000
2000
0
0 1000 2000 3000 4000 5000 6000
Channel 2 Amplitude
b. Imbalanced expression
14000
12000
Channel 1 Amplitude
10000
8000
6000
4000
2000
0
0 1000 2000 3000 4000 5000
Channel 2 Amplitude
Fig. 4 (a) A sample where the alleles (blue and green) appear to have equal expression. (b) A sample where the
alleles have 90%:10% expression (blue:green)
complicate automated clustering and require interactive visual

clustering by the user. To adjust cluster assignment in an unbi-
ased fashion, highlight all the wells containing the same assay
on the plate view in upper left and demarcate cluster bound-
aries identically across all samples on the 2D amplitude plot by
using the “Threshold” or “Lasso” tools. If the “Status” col-
umn is set to “Check,” click anywhere in the amplitude plot to
get the software to recognize the data (see Note 18).
3. Export the data by selecting the wells and clicking “Export
CSV.” The columns “Ratio” and “FractionalAbundance” cor-
respond to different measures of allelic skew, (Allele A/Allele
B) and Allele A/(Allele A þ Allele B), respectively. The col-
umns following these with labels such as “PoissonFractionalA-
bundanceMax” and “PoissonFractionalAbundanceMin”
represent the upper and lower 95% confidence intervals for
Allele A/(Allele A þ Allele B), respectively (see Note 19).
3.5 Using 1. Using the “PoissonFractionalAbundanceMax” and “Poisson-

the Observed Allelic FractionalAbundanceMin” values, we can determine which
Imbalance to Map samples have significant allelic imbalance. Specifically, these
Functional Variants values define the 95% confidence interval around the estimate
of allelic imbalance recorded under “FractionalAbundance”
and thus identify any samples that do not overlap 0.5 (which
represents equal relative expression from each allele). The limit
of detection in terms of effect size is, of course, dependent on
the level of expression and RNA input, but for a moderately
expressed gene with 10–20 ng of RNA (roughly 3000 or more
positive droplets total), we can confidently determine effects
down to a 10% difference in expression. Ideally, all imbalanced
samples will display ratios with consistent absolute distances
from 0.5, suggesting a single common variant affecting expres-
sion across the population. A wide variance in the magnitude of
distance from 0.5 suggests multiple common influences acting
on this gene, making each effect difficult to identify in isolation.
2. Map the functional SNP by correlating the presence of imbal-
ance with each putative functional SNP’s heterozygosity. In
short, samples that have balanced expression of the reporter
SNP are homozygous for the functional variant(s), while sam-
ples that are imbalanced are heterozygous (see Fig. 5a). This can
be used to map the functional variant by finding the regional
SNPs that are always homozygous when there is balanced
expression and heterozygous when there is imbalanced expres-
sion (see Fig. 5b) (see Note 20). If all the imbalanced samples
are imbalanced in the same direction, this provides not only
increased power to find the functional variant, but also allows
the easy determination of which allele of the functional SNP is
associated with higher expression (see Note 21). However, this
a. Theory
Possible haplotypes (D’ = 0) Individuals in study
Reporter SNP Functional SNP

Individuals homozygous
A G for functional SNP will
have balanced
expression
A C
Individuals heterozygous
T G for functional SNP will
have imbalanced
expression
T C
0 0.5 1
Reference allele expression
for reporter SNP
Possible haplotypes (D’ = 1) Individuals in study
Reporter SNP Functional SNP
A G Allele G of the functional

SNP is always linked with
allele T of the reporter
A C SNP and always produces
higher expression of T.
Therefore, the direction of
T G
G is associated with higher
expressioin
0 0.5 1
Allele T expression
for reporter SNP
b. Finding the functional SNP
Is the putative functional SNP
heterozygous? ( yes)
SNP: 1 2 3 4 5
Separate samples into

balanced vs. imbalanced
Fig. 5 (a) The possible haplotypes for a reporter and functional variant with a D’ of 0 or 1, and the pattern of
imbalance those configurations would yield. (b) The technique used to map the functional variant by
correlating the presence of imbalance with heterozygosity
knowledge can also be obtained by determining the chromo-

somal phase molecularly in the DNA samples using methods
we have previously described [8]. A chi square test should be
used to determine the significance of the putative
functional SNP.
4 Notes
1. We have found that standard TaqMan probes generally per-

form well, but MGB or LNA probes may also be used. FAM
and HEX probes can be ordered through Integrated DNA
Technologies, and VIC probes can be ordered through Life
Technologies. HEX and VIC fluorescence is read in the same
channel in ddPCR, so either an HEX or VIC probe should be
used with a FAM probe.
2. For ease of use, 20 mixes can be stored at 4 C for short
periods (~1 month).
3. Because the level of expression across different genes varies by
orders of magnitude, there is not a single RNA input concen-
tration that will work for all genes. We generally start with an
input of 10 ng and adjust the concentration up or down until
we obtain 10–40% positive droplets in ddPCR, which limits the
likelihood of the clusters overlapping by limiting the size of
those clusters, and provides enough double-negative droplets
to confidently call clusters. This may require large amounts of
RNA for genes with low expression. Note that this sample
concentration is lower than most ddPCR assays.
4. Low-quality pipette tips can shred droplets during droplet
generation, likely because they shed tiny pieces of plastic into
the reaction. Though not strictly required, Rainin pipettors and
tips are preferred when working with droplets.
5. At minimum, the SNP has to be within the transcript of the
gene and preferably in an exon that is expressed in all isoforms
(unless only one isoform is of interest). We have found that
intronic SNPs can also be used, especially when they reside in
long introns and are far from the 30 ends of such introns,
ensuring slower degradation rates [9]. Assays on intronic
SNPs always require a total-RNA sample. Such assays also
tend to require more RNA input and are most feasible for
highly expressed genes (RPKM > 25). The GTEx website has
expression information for introns as a downloadable file under
the “Datasets” tab on “Download” page after registering for an
account. The current release is “GTEx_Analysis_V4_RNA-
seq_Flux1.6_intron_reads.txt.gz” at the time of writing. How-
ever, the data is given as read counts rather than equivalent
RPKM on each gene page and needs to be further manipulated
for direct comparison. Briefly, divide each row by the length of

the intron given in the first column, divide each column by the
number of unique mapped reads in that library (found in
“GTEx_Data_V4_Annotations_SampleAttributesDS.txt”
under “SMMPPDUN” column label), and multiply by one
billion (to correct for base/kilobase and read/(million read)
conversion).
6. When the reporter SNP is more common than the proposed
functional variant, particularly if D0 ¼ 1, it is possible to deter-
mine or validate the proposed functional variant by correlating
the heterozygosity of the other variants to the allelic imbalance
observed at the reporter. However, if the reporter SNP and
actual functional variant have D0 ¼ 1 and have similar allele
frequencies (i.e., when r2 ¼ 1), then there will be a limited set
of possible genotypes and most samples heterozygous for the
reporter SNP will have imbalance. Without any samples that
are balanced and heterozygous, mapping is nearly impossible
(because there is no balance/imbalance vs. heterozygosity to
correlate). If prioritizing on finding a reporter SNP with D0 ¼ 1
to the functional variant, then try to find a reporter SNP where
the allele linked to the minor allele of the functional variant is
twice as common as the functional variant’s minor allele fre-
quency. For example, if the functional variant has an
MAF ¼ 0.05, then a reporter SNP with D0 ¼ 1 and overlapping
reporter SNP allele frequency of 0.5 will yield 50% usable
reporter SNP heterozygotes, but with 45%:5% being bal-
anced:imbalanced. A reporter SNP with D0 ¼ 1 and overlap-
ping reporter SNP allele frequency of 0.1 will yield fewer
heterozygotes (18%), but more observations of imbalance
(9%:9% being balanced:imbalanced), thus increasing the
power of mapping.
7. SNPs in the primer or probe binding sites can prevent or hinder
amplification. The sites of common SNPs must be excluded
from the sequence used to design assays.
8. A temperature of 55 C is another annealing temperature to
consider for assay design at loci with lower nearby GC content.
Use the following settings, particularly if design selection at
60 C is difficult, while keeping the other settings constant:
(a) Under “General Primer Picking Conditions” set:
l The Primer Size as Min: 18, Opt: 20, Max: 27.
l The Primer Tm as Min: 55, Opt: 55, Max: 57.
l The Primer GC% as Min: 20, Opt: 50, Max: 80.
(b) Under “Internal Oligo (Hyb Oligo) General Conditions”

set:
l The Hyb Oligo size as Min: 15, Opt: 20, Max: 27.
l The Hyb Oligo Tm as Min: 56.5, Opt: 57.5, Max: 59.
l The Hyb Oligo GC% as Min: 20, Opt: 50, Max: 80.
9. Generally the hybridization probe will be the source of design
failure due to the constraint that the probe sequence must
overlap the variant. To identify the specific requirement
(s) that the surrounding sequence does not meet, enter the
21 bp sequence from Subheading 3.2, step 3 into the box
below “Pick hybridization probe (internal oligo), or use oligo
below” and click “Pick Primers.” The warning message at the
top of the design will help troubleshoot the design. If prob-
lematic, parameter relaxing might first begin with decreasing
“Hyb Oligo Max Mishyb” and increasing “Internal Oligo Max
Poly-X,” both under the “Internal Oligo” tab. If other settings
are leading to design failure, adjustments such increasing
annealing temperature range slightly can still lead to a usable
assay, but consider designing assays for other reporter SNPs.
10. If the primers are too short for BLAT, use the in silico PCR
function (“In-Silico PCR” under the UCSC genome browser
“Tools menu”). If the region of interest is in a segmental
duplication, one unique and perfect match may not be possi-
ble. Try choosing a different reporter SNP if the other duplica-
tions are also expressed.
11. Though the ddPCR instructions have 20 μL as the listed final
reaction volume before droplet generation, scaling reaction
volumes up by 25% ensures fewer air bubbles are introduced
in Subheading 3.2, step 4a.
12. ddPCR droplets are made in sets of eight samples at a time and
read in 96-well plate format, so it is easiest to set up the PCR in
a 96-well plate.
13. Pushing the pipette down to the final stop introduces air
bubbles, which compromise the number and quality of dro-
plets. If air bubbles are introduced into the sample chamber,
manually pop them with a clean pipette tip.
14. An automatic droplet generator (Bio-Rad QX200 AutoDG)
and associated consumables can be used instead of manual
droplet generation.
15. Reverse transcription temperature can be set to a higher tem-
perature (50 C) to denature any secondary structure that
may prevent conversion to cDNA, but using lower tempera-
tures (42 C) can help reduce RNA degradation in solution.
16. Though assays are designed to work best at a particular anneal-

ing temperature (60 C), many SNP assays yield cleaner data at
another temperature. Before running the assay on the samples
of interest, run a set of reactions containing identical reagents
and input sample on a gradient of melting temperatures (see
Fig. 4). Select the melting temperature with the best cluster
separation. An ideal plot will have tight clusters with ample
room to draw separating thresholds. The clusters need not lie
perfectly horizontal or vertical. In fact, it is expected that the
probes will cross-react with the other allele, causing the trajec-
tory of the single-positive droplets to form an acute angle,
rather than the normal 90 right angle, but this will not affect
the measurements as long as the four clusters are separate.
17. Because the correct identification of droplets containing one or
both alleles depends on the competitive binding of the probes
to sequences with which they are both similar, cluster separa-
tion can be more challenging. This is ameliorated by lower
RNA inputs, which yield fewer double-positive droplets. If
any two clusters overlap, run a range of reactions with decreas-
ing sample input until there is clear separation. Allele-specific
expression analysis benefits from using alleles as natural con-
trols for each other, and will often require less sample input
than total expression assays to obtain equivalent power in
detecting the same regulatory effect. An RNA sample can also
be assayed across multiple wells, with the resulting allele counts
then combined across wells to increase the resolution/preci-
sion of the analysis.
18. The software only calls wells with 10,000 or more droplets and
sets the “Status” column to “Check” for wells with fewer
droplets.
19. “FractionalAbundance” and the related confidence interval
defined by the min and max columns may better depict the
magnitude of deviation as compared to the nonlinear distances
between ratios. For example, if one allele generates 1 transcript
to the other allele’s 2 transcripts, the “Ratio” column would
read either 0.5 or 2, depending on the phase between the
reporter and functional variants. Whereas “FractionalAbun-
dance” would read either 33.3 or 66.6—depending on
phase—and thus more intuitively depict the same effect size.
20. Some samples could have a chromosome containing a rare
and/or recent recombination event that breaks the expected
LD pattern seen in most individuals of similar ancestry. Because
allele-specific imbalance is more resistant to variation arising
from genetic and environmental trans-acting factors (see 1), a
single sample can be confidently called as exhibiting allelic
imbalance or not. Based on the allelic imbalance of that sample
resembling that of other samples with equivalent haplotypes on

each side of the recombination event, this may suggest or
eliminate sets of variants normally segregating on the same
haplotype.
21. If all samples with significant allelic imbalance are imbalanced
in the same direction, then we automatically know which allele
is expressed more highly. This is because the reporter SNP and
the functional SNP have a D0 ¼ 1. Effectively, of four possible
combinations between two variants, a D0 ¼ 1 implies that one
or two of them are never seen. The reporter SNP heterozy-
gotes have only two possible genotypes (see Fig. 5a), resulting
in either balance or imbalance. This is different from the case
where D’ does not equal 1, where two different genotypes
result in imbalance and its not clear from genotypes alone
which allele is associated with higher expression.
Acknowledgments
Our understanding of allelic skew has greatly benefited from the

work of Tom Mullen and Jim Nemesh in our lab. This work was
supported by a grant from the National Human Genome Research
Institute (R01 HG006855, to SAM).
References
1. Hindorff LA, Sethupathy P, Junkins HA, Ramos Hariharan M, Dewey FE, Cheng Y, Clark MJ,
EM, Mehta JP, Collins FS, Manolio TA (2009) Im H, Habegger L, Balasubramanian S,
Potential etiologic and functional implications of O’Huallachain M, Dudley JT, Hillenmeyer S,
genome-wide association loci for human diseases Haraksingh R, Sharon D, Euskirchen G,
and traits. Proc Natl Acad Sci U S A Lacroute P, Bettinger K, Boyle AP,
106:9362–9367. https://doi.org/10.1073/ Kasowski M, Grubert F, Seki S, Garcia M,
pnas.0903103106 Whirl-Carrillo M, Gallardo M, Blasco MA,
2. Genomes Project C, Abecasis GR, Altshuler D, Greenberg PL, Snyder P, Klein TE, Altman RB,
Auton A, Brooks LD, Durbin RM, Gibbs RA, Butte AJ, Ashley EA, Gerstein M, Nadeau KC,
Hurles ME, McVean GA (2010) A map of Tang H, Snyder M (2012) Personal omics
human genome variation from population-scale profiling reveals dynamic molecular and medical
sequencing. Nature 467:1061–1073. https:// phenotypes. Cell 148:1293–1307. https://doi.
doi.org/10.1038/nature09534 org/10.1016/j.cell.2012.02.009
3. Patwardhan RP, Hiatt JB, Witten DM, Kim MJ, 6. Battle A, Mostafavi S, Zhu X, Potash JB, Weiss-
Smith RP, May D, Lee C, Andrie JM, Lee SI, man MM, McCormick C, Haudenschild CD,
Cooper GM, Ahituv N, Pennacchio LA, Shen- Beckman KB, Shi J, Mei R, Urban AE, Mon-
dure J (2012) Massively parallel functional dis- tgomery SB, Levinson DF, Koller D (2014)
section of mammalian enhancers in vivo. Nat Characterizing the genetic basis of transcrip-
Biotechnol 30:265–270. https://doi.org/10. tome diversity through RNA-sequencing of
1038/nbt.2136 922 individuals. Genome Res 24:14–24.
4. Cowles CR, Hirschhorn JN, Altshuler D, https://doi.org/10.1101/gr.155192.113
Lander ES (2002) Detection of regulatory vari- 7. Untergasser A, Nijveen H, Rao X, Bisseling T,
ation in mouse genes. Nat Genet 32:432–437. Geurts R, Leunissen JA (2007) Primer3Plus, an
https://doi.org/10.1038/ng992 enhanced web interface to Primer3. Nucleic
5. Chen R, Mias GI, Li-Pook-Than J, Jiang L, Lam Acids Res 35:W71–W74. https://doi.org/10.
HY, Chen R, Miriami E, Karczewski KJ, 1093/nar/gkm306
8. Regan JF, Kamitaki N, Legler T, Cooper S, 9. Gray JM, Harmin DA, Boswell SA, Cloonan N,
Klitgord N, Karlin-Neumann G, Wong C, Mullen TE, Ling JJ, Miller N, Kuersten S, Ma YC,
Hodges S, Koehler R, Tzonev S, McCarroll SA McCarroll SA, Grimmond SM, Springer M (2014)
(2015) A rapid molecular approach for chromo- SnapShot-Seq: a method for extracting genome-
somal phasing. PLoS One 10:e0118270. https:// wide, in vivo mRNA dynamics from a single total
doi.org/10.1371/journal.pone.0118270 RNA sample. PLoS One 9:e89673. https://doi.
Chapter 24
Very Low Abundance Single-Cell Transcript Quantification

with 5-Plex ddPCRTM Assays
George Karlin-Neumann, Bin Zhang, and Claudia Litterst
Abstract
Gene expression studies have provided one of the most accessible windows for understanding the molecular
basis of cell and tissue phenotypes and how these change in response to stimuli. Current PCR-based and
next generation sequencing methods offer great versatility in allowing the focused study of the roles of small
numbers of genes or comprehensive profiling of the entire transcriptome of a sample at one time. Marrying
of these approaches to various cell sorting technologies has recently enabled the profiling of expression in
single cells, thereby increasing the resolution and sensitivity and strengthening the inferences from
observed expression levels and changes. This chapter presents a quick and efficient 1-day workflow for
sorting single cells with a small laboratory cell-sorter followed by an ultrahigh sensitivity, multiplexed digital
PCR method for quantitative tracking of changes in 5–10 genes per single cell.
Key words Single-cell, Cell-sorting, Gene expression, Transcripts, RNA quantification, Digital PCR,
dPCR, Droplet digital PCR, ddPCR, Assay multiplexing, Radial multiplexing
1 Introduction
Gene expression studies give valuable insight into the properties of

cells and tissues and how these change in response to developmen-
tal signals and the environment. Methods to detect transcript levels
of both known and unknown genes have changed dramatically over
the past four decades becoming more efficient, more sensitive,
more quantitative and enabling the profiling of larger numbers of
gene transcripts at once. Historically, these studies have been done
with northern blots [1, 2], nuclease-protection assays [3, 4], or
PCR-based methods such as qRT-PCR [5, 6], targeting a limited
number of gene transcripts, or DNA microarrays targeting much
larger numbers [7–11]. Examples range across studies of bacteria,
e.g., sporulation in B. subtilis [12], humans, e.g., discrimination of
new breast cancer subtypes [13], and plants, e.g., light-induced
development of the photosynthetic apparatus [14]. More recently,
digital PCR (dPCR) and next-generation sequencing (NGS) have
423
424 George Karlin-Neumann et al.
been enlisted, in the former case to make absolute and highly

sensitive measurements of transcript levels [15–17] and in the
latter, to generate broad whole transcriptome or targeted RNA
transcript profiles [18–20]. Although prominently expressed
genes can be discerned in the complex mixture of cells comprising
a tissue or even a whole organism, transcripts from minority cell
populations such as stem cells are likely to be lost, especially for
transcripts present at low abundance per cell.
Furthermore, since a major goal of gene expression studies is to
identify genes cooperating to produce a phenotype of interest, the
mixing together of different cell types in a single sample, or even
transcriptional heterogeneity among apparently like cells, tends to
confound the effort as it is unclear which transcripts are present in
each individual cell (e.g., see Fig. 1, ref. 21). To address this chal-
lenge, various methods have been applied to isolate populations of
single cells (e.g., by fluorescence-activated cell sorting [FACS],
nanofluidic chambers, and most recently, droplets) and then to
analyze either selected transcripts or the entire transcriptome of
each cell by the methods mentioned above [22–25]. Each of the
Fig. 1 A 1-day workflow for single-cell sorting followed by RT-ddPCR™. Cultured cells (shown) or dissociated
cells are prepared as a suspension free of clumps and debris before using the S3e™ Cell Sorter to deliver a
single-cell into each PCR tube well which contains lysis buffer. cDNA generated from the unpurified RNA lysate
in each well is then divided into 2 ddPCR™ reactions per cell—one is assayed with 5-plex 1, the other with
5-plex 2. After converting all samples into droplets using the QX100™ or QX200™ Droplet Generator,
followed by thermocycling, the plate(s) of endpoint ddPCR™ reactions are read in the Droplet Reader to
quantify transcript levels for each of the 10 genes assayed
Single-Cell Transcript Quantification via Multiplexed ddPCRTM Assays 425
analysis methods has its strengths and drawbacks. Whereas NGS

enables the most complete profiling of the transcriptome, it is not
as sensitive as PCR-based methods—having capture efficiencies of
only ~10% [24, 25]—and thus only detects transcripts present at
>30–50 copies/cell. Unless unimolecular barcodes are used and
sequencing is done to a much greater depth (>tenfold more than
usual), it gives only relative quantification and is more costly and
laborious. PCR-based methods while only detecting a smaller sub-
set of genes per cell, can nonetheless detect transcripts present at
only a few copies/cell in the present method; can flexibly and
efficiently target any desired region in a transcript, no matter how
far 50 it is from the 30 mRNA end; can utilize customized assay
designs and thermal cycling conditions to more efficiently capture
difficult transcript sequences (e.g., GC-rich regions) and transcript
types (e.g., non-polyA-containing long noncoding RNAs); and is
much quicker and less costly. Digital PCR, in particular, can provide
the greatest sensitivity with absolute quantification of each gene
transcript species being targeted, without a standard curve. Fur-
thermore, because of the ability to multiplex and discriminate at
least five different gene transcripts in a single droplet digital PCR
reaction well (see ref. 26), it is possible to quantify 5–10 genes per
cell at very high sensitivity without requiring preamplification of
the single-cell cDNA as in all other methods. This both eliminates
potential distortion of transcript quantification that can occur dur-
ing preamplification and NGS library preparation, and provides a
quick and efficient method for profiling key signature genes in
populations of single cells.
Here we describe an ultrasensitive and rapid method for sorting
single cells into PCR wells using the Bio-Rad S3e Cell Sorter,
directly converting the unpurified RNA into cDNA, and utilizing
a strategy of “radial multiplexing” to quantify the absolute number
of transcripts per cell for ten genes via Bio-Rad’s QX200™ Droplet
Digital™ PCR (ddPCR™) System (see Figs. 1 and 2).
2 Materials
2.1 Single-Cell 1. For cultured cells, use appropriate growth medium, any neces-
Preparation sary growth supplements and sterile plasticware, along with
sterile hood and cell growth incubator. For P19 mouse embry-
onal carcinoma cells used here, normal-serum medium of
Monzo et al. was used [27].
2. Pipettes and pipette tips.
3. 50 mL Falcon centrifuge tubes.
4. 0.25% Trypsin–EDTA (1), phenol red solution for release of
adherent cells from culture dish surface (e.g., ThermoFisher
Scientific).
5. PBS (phosphate-buffered saline, pH 7.4).

6. BSA (20 mg/mL, Sigma).
7. 5 mL FACS tube with 40 μm cell-strainer (Corning).
8. Clinical centrifuge (providing 180 g-force).
9. Cell counter (e.g., Bio-Rad TC-20™ Automated Cell
Counter)
2.2 Single-Cell 1. S3e™ Cell Sorter (Bio-Rad, Cat. No. 145-1001).

Sorting 2. ProLine ™ Calibration Beads (Bio-Rad, Cat. No. 145-1081).
3. ProLine ™ Rainbow Beads (Bio-Rad, Cat. No. 145-1085).
4. 10% bleach solution.
5. Sterile distilled water.
6. PCR strip-tubes with caps (8 tubes/strip).
Fig. 2 Radial multiplexing with ddPCR™ assays. (a) A 2D droplet plot of a standard duplex assay, w/ cartoon of
a 3rd target along the 45 radius achieved by mixing together equal concentrations of FAM and HEX probes
specific for that target. (b) Real Triplex assay with high concentrations of all targets. This results not only in
primary clusters with droplets containing only a single type of transcript (1, 2 or 3), but also to secondary
(1 + 2, 1 + 3, 2 + 3) and tertiary (1 + 2 + 3) clusters with two and three target types present, respectively. In
practice, for single-cell and liquid biopsy applications, species concentrations are typically low and rarely
occupy these higher order clusters. (See Fig. 4 for examples of individual single-cell 2D droplet plots). (c)
Cartoon showing the concept of radial multiplexing extended to five quantifiable targets per ddPCR™ well
using a continuum of probe mix ratios, from 100% FAM to 100% HEX. (d) 5plex-2 (assaying a different set of
genes than 5plex-1) tested with purified bulk cell RNA from a mouse cell line, P19 embryonal carcinoma
(EC) cells. Note that only with this more abundant amount of input RNA are there a noticeable number of
droplets with more than a single transcript type. Compare with single cell transcript profiles in Fig. 4 which
rarely have any 2 clusters
Fig. 2 (continued)
7. DNA Suspension Buffer (10 mM Tris–HCl pH 8.0 with

0.1 mM EDTA);
8. 10% Triton X-100.
9. RNaseOUT (40; Thermo Fisher Scientific).
10. Vortexer.
2.3 cDNA Synthesis 1. Microcentrifuge (with 8-strip tube rotor).

2. iScript™ Advanced cDNA Synthesis Kit (Bio-Rad).
3. T100 (or C1000 Touch) Thermal cycler (Bio-Rad).
2.4 ddPCR™ 1. QX100™ or QX200™ Droplet Digital™ PCR System, with

Reactions Manual Droplet Generator and Automated Droplet Reader
(Bio-Rad).
2. Rainin 8-channel pipettors and pipette tips (see Note 1).
3. ddPCR™ Supermix for Probes (no dUTP; Bio-Rad).
4. ddPCR™ Taqman™ 50 nuclease assays with FAM- and/or
HEX-labeled probes.
6. DG8™ Cartridges and Gaskets (Bio-Rad).
7. Pierceable Foil Heat Seals (Bio-Rad).
8. 96-well semiskirted ddPCR™ plates (Bio-Rad, Cat.
No. 12001925).
9. PX1™ PCR Plate Sealer (Bio-Rad).
10. QuantaSoft™ Analysis Software 1.7.4.0917 or earlier versions,
or QuantaSoft™ Analysis Pro Software.
3 Methods
The present method involving probe multiplexing has been devel-

oped and tested with Bio-Rad’s QX100™ or QX200™ Droplet
Digital™ PCR system though, in principle, the approach should
also be deployable on other digital PCR platforms capable of
detecting at least two distinct fluorescence emission wavelengths
[28–32]. The Bio-Rad QX systems detect fluorescence in two
spectral regions well suited for FAM- and HEX-(or VIC-)labeled
Taqman probes, and in the simplest duplex assay format enables the
discrimination of two distinct DNA or RNA targets per reaction
well, where one probe is labeled in FAM and the other in HEX,
each hybridizing to different target sequences. When both targets
are abundantly present in a sample (as illustrated in the 2D droplet
plot in Fig. 2a), this results in discrete clusters of droplets repre-
senting the presence of a first target (e.g., cluster 1, containing
FAM signal only), a second target (e.g., cluster 2, containing
HEX signal only), both targets (cluster 3, containing both FAM

and HEX signals and their corresponding target molecules), or
neither target (cluster 4). By introducing a 50:50 mix of FAM
and HEX probes, both directed against a third target sequence, it
can be detected in a primary cluster located along the diagonal (or a
45 radius) as shown in Fig. 2a. An example of such a triplex assay is
shown in Fig. 2b where a sufficiently high concentration of the
3 targets leads to primary clusters with a single target type in each
(1, 2, or 3) closest to the origin, secondary clusters containing two
targets each (1 + 2, 1 + 3, and 2 + 3) further from the origin, and a
tertiary cluster containing all three targets (1 + 2 + 3) furthest out
of all. Fig. 2c cartoons the application of this probe mixing strategy
to five targets (a 5-plex) where ratios between 50:50 and 100%
(75:25 and 25:75, FAM:HEX) are also used to position the fourth
and fifth targets’ primary clusters along radii (hence, the term
“radial multiplexing”) between the vertical and diagonal (Assay 2)
and the diagonal and the horizontal (Assay 4). And finally, an
example of a 5-plex assay is shown in Fig. 2d.
3.1 Formulating A successful 5-plex ddPCR™ assay requires identification of five

and Optimizing a 5- individual assays, all with a similar annealing temperature, which do
Plex Droplet Digital not interfere with one another when combined into a single
PCR Assay ddPCR™ reaction. When formulating a 5-plex ddPCR™ assay,
follow these steps.
1. Obtain singleplex ddPCR™ Taqman 50 nuclease probe assays
with common annealing temperatures (e.g., 60 C). These may
either be from predesigned 20 assays optimized for ddPCR™
analysis (“https://www.bio-rad.com/digital-assays/#/”) or as
five sets of primers and their corresponding probes designed
against the transcript target sequences of interest according to
best practices (see Bio-Rad Bulletin #6407 and Note 2) and
synthesized by an oligo manufacturer (e.g., IDT, Coralville, IA
or BioSearch, Petaluma, CA). We recommend ordering them
first with a single probe fluorophore (e.g., four with FAM and
one with HEX) and confirming their proper singleplex perfor-
mance in ddPCR™ reactions, before ordering a second
HEX-labeled probe for three of the component assays. Ulti-
mately, eight probes will be necessary to detect the five tran-
scripts of interests (see Fig. 2c), where three of them require a
combination of both FAM- and HEX-labeled probes, one
requires only a FAM probe, and one only a HEX probe. If
formulating two 5-plex assays, design and test an additional five
gene assays as for the first 5-plex.
2. Run a ddPCR™ thermocycler temperature gradient (as shown
in Table 1) for each individual assay to test annealing/exten-
sion temperatures from ~5 C below to ~5 C above the
calculated Tm of the primers (e.g., 55–65 C) using cDNA
from an appropriate bulk total RNA target (e.g., from cell
430
Table 1
ddPCR plate set up
ddPCR plate setup
1 2 3 4 5 6 7 8 9 10 11 12

65 C A Gene Gene Gene Gene Gene Gene Gene Gene Gene Gene
1 assay 2 assay 3 assay 4 assay 5 assay> 6 assay 7 assay 8 assay 9 assay 10 assay
63.4 C B Gene Gene Gene Gene Gene Gene Gene Gene Gene Gene
1 assay 2 assay 3 assay 4 assay 5 assay 6 assay 7 assay 8 assay 9 assay 10 assay
George Karlin-Neumann et al.
62 C C Gene Gene Gene Gene Gene Gene Gene Gene Gene Gene
60.6 C D Gene Gene Gene Gene Gene Gene Gene Gene Gene Gene
59.2 C E Gene Gene Gene Gene Gene Gene Gene Gene Gene Gene
57.8 C F Gene Gene Gene Gene Gene Gene Gene Gene Gene Gene
56.4 C G Gene Gene Gene Gene Gene Gene Gene Gene Gene Gene
55 C H Gene Gene Gene Gene Gene Gene Gene Gene Gene Gene
lines or tissue of interest; Universal Mouse Reference RNA,

Agilent Catalog #740100; or Universal Human Reference
RNA, Agilent Catalog #740000). See Subheading 3.4 below
for cDNA synthesis procedure, scaling the reaction shown
there to 20 μL final volume, consisting of 12 μL of Reverse
Transcription Master Mix (Table 3) and 8 μL of total RNA and
water combined. See Subheading 3.5 below for set-up and
running of ddPCR™ reactions, however using temperature
gradient cycling conditions as illustrated in Table 1.
3. Select five assays that show optimal or near-optimal separation
of positive and negative droplets at a similar annealing temper-
ature for inclusion in the first 5-plex (see Note 3). If you are
creating two 5-plexes, it is recommended to include one assay
in common to the two multiplexes for more precise determina-
tion of transcript copies/cell from cDNA that is divided
between two ddPCR™ reactions per cell. For each set of
5 component assays in a multiplex, you will need to order
HEX probes with the same probe sequences as previously
ordered with FAM, for three of the assays (see Note 4). For-
mulate a 5-plex ddPCR™ assay (as shown in Table 4) by
adding assay 1 containing 100% FAM probe only, assay
2 with 75% FAM and 25% HEX probes, assay 3 with 50% of
both FAM and HEX probes, assay 4 with 75% FAM and 25%
HEX probes, and assay 5 with 100% HEX probe, each from a
20 stock with total probe concentration of 5 μM. If any of the
assays did not perform well at the desired temperature (or at
all), you may need to modestly redesign the assay for a higher
or lower temperature or select an alternative design for the
gene of interest.
4. Run ddPCR™ reactions for each 5-plex with cDNA from
control bulk RNA—just as was done for the singleplex assays
above—to check the separation of the clusters in 2D droplet
plots. You can change the total probe concentration for an
individual target to optimize the separation of the clusters (see
Note 4).
5. For quantification of each individual transcript in each 5-plex,
see Subheading 3.6 below. Compare the quantification of each
transcript level in the 5-plex to its quantification in singleplex
for the same control RNA target. Transcript copy numbers
should be very similar. However, if there is some unexpected
interference between any of the individual assays leading to
significant disagreement between the singleplex and multiplex
results (not commonly seen), it will be necessary to substitute
at least one of the interacting assays. After achieving good
cluster separation and accurate quantification in both multi-
plexes, the 5-plexes are now ready to use.
3.2 Preparation The 5-plex single-cell gene expression protocol described in this
of Single Cells chapter has been developed using adherent mammalian cells (P19
for FACS Sorting mouse embryonal carcinoma cells) grown according to Monzo
et al. [27]. Alternatively, sources of single cells may be derived
from suspension cell cultures; cells from blood, sputum or urine;
or cells dissociated from tissue (see Note 5). However, although
general guidance is given below for preparation of single cells ready
for sorting, these other cell sources have not specifically been tested
in this protocol.
1. First harvest cells by enzymatic release using 0.25% Trypsin-
EDTA at 37 C for 5 min, followed by quenching with media
containing serum.
2. Centrifuge cells at 180 g for 5 min at room temperature,
discard supernatant, resuspend in 10 mL PBS and count cells
with a cell counter (such as the Bio-Rad TC-20™ Automated
Cell Counter).
3. Centrifuge cells at 180 g for 5 min at room temperature,
discard supernatant and gently resuspend pellet to a concentra-
tion of 1–10 million cells/mL in PBS with 0.1% BSA.
4. Filter cells into a FACS tube via a 40 μM cell strainer to obtain a
single cell suspension (see Note 6). Keep on ice until sorting.
3.3 Flow-Sorting Filtered single cells are sorted with the Bio-Rad S3e™ Cell Sorter
of Single-Cells into into 8-well PCR strips containing Lysis buffer. Follow the manu-
Lysis Buffer facturer’s instructions for cell sorter operations, referring to
Bio-Rad Bulletin #10031105, if needed.
1. Prepare Lysis Buffer and Reverse Transcription Master Mix
according to the recipes in Tables 2 and 3, respectively.
2. Power on the S3e™ Cell Sorter and perform a QC run using
ProLine™ Calibration Beads, ensuring that it passes QC.
3. Perform a test sort with ProLine™ Rainbow Beads, making
sure the beads are sorted into the center of the PCR tubes.
4. Clean the system with 10% Bleach at low pressure; clean the
system again with sterile deionized water.
Table 2
Preparation of Lysis buffer
Lysis buffer 1000 μL

DNA suspension buffer 965 μL
(10 mM Tris–HCl, pH 8, 0.1 mM EDTA)
Triton X-100 (10%) 10 μL
RNAse OUT™ (40) 25 μL
Table 3
Preparation of Reverse transcription master mix
Reverse transcription master mix 1 130

5 iScript advanced reaction mix 2 μL 260 μL
iScript advanced reverse transcriptase 0.5 μL 65 μL
Nuclease-free water 3.5 μL 455 μL
Total 6 μL 780 μL
5. Add 4 μL of prechilled Lysis Buffer into each PCR tube on

8-well PCR strips, and keep on ice before use.
6. Load the sample tube containing the filtered cells onto the
sample station of the S3e™ sorter and collect data.
7. Set up gates for the target cells, making sure to exclude any cell
debris and doublets.
8. In the ProSort software, click on “Sort Logic”: select “8-well
strip”; choose “single sorting” mode, and set the sort limit to
“1” to sort one single cell into each well (see Note 7).
9. Put 2 8-well PCR strips containing Lysis Buffer onto the two
8-strip tube adaptors in the sorting chamber.
10. Click on “Sort” to sort single cells into each well.
11. Remove both strips out of the sorter, cap the tubes and store
them on ice.
12. Repeat steps 8–11, until single cells are sorted into as many as
12 strips (assuming 96 cells are to be collected).
13. Vortex the 8-well strips with sorted cells for 10 s and centrifuge
briefly.
14. Leave the PCR tubes on ice for 5 min for complete lysis (see
Note 8).
3.4 Preparation 1. Add 6 μL of Reverse Transcription master mix (see Table 3) into
of cDNA each well; then place the caps back onto the tube strips.
2. Centrifuge the strips for a few seconds, vortex and centrifuge
again.
3. Load PCR strips onto a thermal cycler* and incubate under the
following conditions:
(a) 42 C for 30 min.
(b) 85 C for 5 min.
(c) 4 C (infinite).
*Use heated lid set to 105 C.
If you wish to stop here, the cDNA may be frozen at 20 C at
this step.
Table 4
Preparation of ddPCRTM master mix
ddPCRTM master mix 1 well (1) 1 plate (120) 2 plates (240) per well (1.1)
Water 1 120 240 1.1
™
2 Droplet Digital PCR Supermix 10 1200 2400 11
for Probes (no dUTP)
20 assay 1 (100% FAM) 1 120 240 1.1
20 assay 2 (75% FAM:25% HEX) 1 120 240 1.1
20 assay 3 (50% FAM:50% HEX) 1 120 240 1.1
20 assay 4 (25% FAM:75% HEX) 1 120 240 1.1
20 assay 5 (100% HEX) 1 120 240 1.1
cDNA 4 480 960 4.4
Total 20 μL 2400 μL 4800 μL 22 μL
3.5 Setup 1. Prepare the ddPCR™ Master Mix according to Table 4 and
and Running Droplet store on ice.
Digital PCR Reactions l Prepare master mix for one or more plates, leaving out the
cDNA which is added separately.
l If you are using assays rather than 20 assay, the final concen-
tration is typically 900 nM for primers, 250 nM for probes.
2. Distribute 17.6 μL of the ddPCR™ Master Mix into each well
of a 96-well PCR plate.
3. Briefly centrifuge the PCR strips containing the RT reaction.
Open the caps carefully and transfer 4.4 μL of the cDNA into
each well on the 96-well plate containing the ddPCR™ Master
Mix. (Note: The remaining cDNA from each cell (also
~4.4 μL) may be frozen or prepared for running in a second
ddPCR™ reaction plate with a second 5-plex assay panel as
done for the first 5-plex).
4. Seal the plate. Centrifuge it for a few seconds, vortex and
centrifuge again.
5. To generate droplets for each ddPCR™ reaction, see Bio-Rad
Instruction Manual #10026322 (QX100™) or #10031907
(QX200™):
(a) Transfer 20 μL of each 22 μL sample—eight samples at a
time, into eight sample wells of a DG8™ droplet genera-
tion cartridge (previously secured in a DG8™ cartridge
holder).
(b) Add 70 μL Probes Droplet Generation Oil into the eight

companion oil wells.
(c) Seal with a gasket and place into the Droplet Generator.
(d) Follow manufacturer’s guidance in observing proper load-
ing of the DG8™ cartridge—adding sample first and
avoiding introduction of air bubbles into the
chamber well.
(e) Start droplet generation within 20 after loading of Droplet
Generation Oil using the QX100™ or QX200™ Droplet
Generator.
(f) Transfer droplets from the DG8™ cartridge output wells
to 8 wells of a 96-well PCR plate using an 8-channel P50
pipettor. Aspirate slowly at an ~70 degree angle.
(g) Repeat as needed until up to all 96 wells are loaded with
droplet reactions.
(h) Heat-seal the droplet plate with an adhesive foil seal using
the PX1™ PCR Plate Sealer.
6. Load the droplet plate onto a thermal cycler and cycle using the
protocol in Table 5.
7. After completion of thermal cycling, transfer the plate to the
QX100™/QX200™ Droplet Reader to read both FAM and
HEX fluorescence signals from all ddPCR™ wells, following
the manufacturer’s instructions (see Bulletin #10031906).
After entry of the experiment plate layout is completed in
QuantaSoft™ Software, select “Run” to begin the droplet
reading process. Select the appropriate dye set used and run
options when prompted.
Table 5
Thermal cycling protocol
Step Temperature Time Ramp # of cycles

1 95 C 10 min 2 C/s 1

2 94 C 30 sec 2 C/s
40
3 xx C* 1 min 2 C/s
4 98 C 10 min 2 C/s 1
5 4 C Hold 2 C/s 1
*The annealing temperature (Step 3) could vary depending on the optimal assay anneal-
ing temperature, though assays are typically designed for 60 oC
3.6 Quantification After data acquisition, select all samples in the QuantaSoft™ well
of ddPCRTM Reactions editor under “Analyze.”
1. Singleplex (or duplex) assays:
(a) To evaluate assay quality, examine the automatic thresh-
olding applied to the 1D (or 2D) amplitude data in the
“1D Amplitude” or “2D Amplitude” tabs, and if neces-
sary, set the thresholds or clusters manually. Negative and
positive clusters should be well-formed and well-separated
for assays that will be used to formulate the 5-plex(es)
(See, e.g., Fig. 2a). If they are not of adequate quality at
any of the annealing temperatures tested within the range
of the thermal cycler temperature gradient, the assays will
need to be redesigned and retested before proceeding.
(b) The concentration reported in the Data Window or
“Concentration” tab is “copies/μL” of the final 20 μL
ddPCR™ reaction. For selecting the assays to combine
into a 5-plex, be sure all of the assays perform well at a
common annealing temperature (i.e., good cluster sepa-
ration) and that the concentration value given for each
assay is consistent over several degrees higher and lower
than the annealing temperature selected for use in the
5-plex reactions. See Note 2.
2. 5-plex assays:
These higher multiplex reactions can be quantified by either of
two procedures. If using the newer QuantaSoft™ Analysis Pro
software version (QSAP™, available from the download tab
at: https://www.bio-rad.com/en-us/product/qx200-droplet-
digital-pcr-system?ID¼MPOQQE4VY), all genes of interest
(GOIs) in the multiplex can be quantified at once within the
program after defining the target clusters properly under the
“Advanced User Options.” This will provide concentrations
and confidence intervals for all targets defined in a single analy-
sis run. For further guidance, see Bio-Rad Bulletin #6827.
If using QuantaSoft™ 1.7.4.0917 or earlier software ver-
sions, it will be necessary to follow the iterative quantification
procedure, below, which identifies one gene of interest at a
time (in all wells selected on the plate) and provides the con-
centration (with 95% confidence interval) and number of tran-
script copies of that GOI in the 20 μL ddPCR™ reaction.
These values can be graphed within QuantaSoftTM and may
be exported as a “.csv” file for analysis in other software
applications.
(a) After 5-plex data acquisition, select all wells desired on the
plate using the “Well Editor” (see Fig. 3a).
(b) In the 2D droplet plot, using the cluster identification
tool, encircle the primary (i.e., single) positive cluster for
Fig. 3 Quantification of 5-plex transcript levels with QuantaSoft™ software. (a) A window from QuantaSoft™
software showing the Well Editor (for choosing wells being analyzed in the other graph and data windows), 2D
Droplet Plot (showing FAM and HEX fluorescent droplet intensities), and Data Table (showing various
computed values and sample and assay descriptors). (b) Iterative quantification of transcript levels. Note
that in each of the five 2D plots shown, only 1 primary cluster is labeled in blue and circled (i.e., the GOI being
quantified in that iteration), only 1 cluster labeled in grey (complete negatives), and the rest are in green and
are ignored for quantification purposes in the iteration shown in that plot
Gene 1 (defined in Fig. 2b) and designate it as “FAM

only” (shown as a blue cluster in the 2D plots of
Fig. 3b). Next, encircle the “all negative” cluster closest
to the origin as your “double-negative” cluster (shown as
grey droplets). Finally, use the freehand tool to encircle all
other droplets and droplet clusters as a single HEX cluster
(shown as green droplets). See Note 9.
(c) QuantaSoft™ will provide you with the concentration of

the Gene 1 transcript (per μL of the ddPCR™ reaction)
and the number of transcript copies per well (Nrxn) for all
selected wells (up to 96 wells per plate). To calculate the
copies per cell (Ncell) and for further analyses, export the
data for all wells as a first “.csv” file. After opening the .csv
file (e.g., in Microsoft EXCEL™), locate the column for
“CopiesPer20uLWell” (i.e., Nrxn) and divide each value
by the fraction of cDNA analyzed in the 20 μL ddPCR™
reaction. Thus, Ncell ¼ Nrxn/f, where f is the fraction of a
cell’s cDNA analyzed per ddPCR™ well (typically, f 0.4
[4 μL cDNA measured per well/~10 μL total cDNA per
cell]).
Note that typically, the ddPCR™ reaction is formu-
lated to contain 4.4 μL cDNA in a 22 μL sample volume,
of which only 20 μL is converted into droplets.
(d) Repeat steps (b) and (c) iteratively for each gene of inter-
est in all wells run with the same 5-plex assay (i.e., a total
of five times to analyze all five gene transcripts per
ddPCR™ reaction). See Note 10.
(e) After quantifying a bulk control RNA by both singleplex
and 5-plex assays, confirm that the 5-plex quantification of
your target transcripts is in agreement with your single-
plex quantifications run on the same sample(s). If there are
significant discrepancies, you will need to identify which
assay(s) is/are being compromised in the multiplex reac-
tion. You may then either modify the design for that
target gene region or chose another target sequence
within the transcript. Alternatively, if you are testing two
5-plexes, you may be able to switch an individual assay
between the two multiplexes to mitigate the interference.
(f) Once you have validated the performance of your multi-
plex(es), you are ready to use them for single-cell analyses
according to the workflow described in this chapter.
An example of data from a control RNA sample derived from a
mixture of both untreated and retinoic acid-treated P19 mouse
cells is illustrated in Fig. 2d. Additionally, Fig. 4 compares three
2D droplet plots from individual single cells vs. 2D droplet plots of
24 and 96 single-cell reactions superimposed. It is clear from this
example that most transcript levels per cell are low enough such
that very few if any transcripts partition into droplets containing
more than one gene transcript type (see Note 11). Figure 5 shows
the result of cell-to-cell variation in transcript levels for 3 genes
quantified by this workflow typifying low (B2M), intermediate
(SDH) and high (RPLP0) abundance transcripts analyzed in
24 untreated P19 mouse cells (see Note 12).
Fig. 4 Transcript levels for individual single-cells assayed with the 5-plex 1 ddPCR™ assay targeting GAP43,
Nanog, Mki67, Cadherin1, and RPLP0. Examples of 2D droplet plots for three P19 single cells are shown
assayed with 5-plex 1. Note the differences in transcript level between cells for the various GOIs by comparing
the number of droplets in the same encircled primary cluster in each of the typical cells shown. For
comparison, see the aggregate level of each transcript from 24 and 96 single cell results superimposed in
the adjacent 2D plots. Note some clusters are identified here by orange (rather than green) droplets merely for
better visualization of the five genes
4 Notes
1. Avoid low-quality pipette tips that may shed plastic particulates

into the DNA and Droplet Digital™ PCR solutions which can
then shred droplets. This results in poor quantification.
2. If primers and probes are ordered separately, it is possible to
formulate the 5-plex assay so that 1 μL contains 20 concen-
tration for all five assays (20 ¼ 18 μM each primer and 5 μM
each total probe concentration). This would potentially allow
the addition of the total amount of cDNA per cell to a single
ddPCR™ reaction (~8 μL cDNA/20 μL ddPCR™ reaction)
and thus double the sensitivity of the assay, at the expense of
only assaying for five genes per cell. However, it would need to
be established that the ddPCR™ reaction tolerated this larger
P19 single cell gene expression

10000
1000
copies/cells
100 B2M
SDH
RPLP0
10
1
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
cells
Fig. 5 Example of varying transcript abundances for three genes across 24 cells using the method described
here (but with two duplex assays run per cell). Transcript levels are shown for genes with low (beta-
2 microglobulin, B2M), intermediate (succinic dehydrogenase, SDH), and high (Ribosomal Protein Lateral
Stalk Subunit P0, RPLP0) abundance in untreated P19 mouse cells. Transcript levels for B2M vary by nearly
tenfold across these cells, whereas those for SDH and RPLP0 vary by only a fewfold in these same cells and
highlight the frequent noncorrelation of transcript levels for various genes across apparently like cells
volume of unpurified cDNA and Lysis Buffer and quantified

linearly with the increased sample amount.
3. Be sure the assays do not show signs of overquantification at
the selected annealing temperature, due to nonspecific target
amplification that may be seen at lower than optimal tempera-
tures; nor underquantification at higher temperatures where
the amplitude of positive droplets may still be distinguishable
from negative droplets but may give too low a concentration
value due to “molecular dropout” (i.e., where a droplet or
nano-chamber partition contains a target of interest, but for
various reasons, may not achieve sufficient fluorescence ampli-
tude at endpoint to be recognized as a positive partition).
Typically, the concentration value measured will remain con-
stant over a range of annealing temperatures and this is where
the optimal temperature should be chosen. Although the quan-
tification provided by digital PCR is in absolute copies per
transcript per volume of sample analyzed, it must be remem-
bered that there is generally some uncertainty about the effi-
ciency of reverse transcribing any given RNA target region into
cDNA. This can be influenced by quality of the RNA, sequence
of the target region and its context, the reverse transcriptase
used (e.g., RNase H or RNase H+) and the conditions of
cDNA synthesis. To determine the true quantity of a transcript

present in a sample requires a systematic effort to study these
factors for the transcript of interest and the use of multiple
assays designed against different regions of the same transcript.
Most often, it is sufficient to establish that the assay design
chosen gives unambiguous results, and is reproducible from
run-to-run and from sample to sample.
4. With regard to deciding which three assays are to be detected
with a combination of FAM and HEX probes, if some assays
have higher or lower fluorescence amplitude in singleplex than
others (potentially due to differences in amplicon lengths,
where longer amplicons tend to give lower amplitude droplets
with probes chemistry), it may make cluster identification and
thus quantification easier if lower amplitude assays are alter-
nated with higher amplitude ones to better spread out the
clusters from one another in 2D droplet space plots. Alterna-
tively, total probe concentration for a given target may be
increased or decreased to increase or decrease its droplet cluster
fluorescence amplitude, respectively. Note that assays yielding
short amplicons are generally preferred since they tend to give
clusters with higher fluorescence amplitude where neighboring
primary clusters will be better separated from one another than
if they are all at lower amplitudes.
5. Preparation of single cell suspensions from solid tissue requires
mechanical dissociation and/or enzymatic digestion for opti-
mal recovery of cells from the tissue. Conditions and enzyme
requirements for digestion of the tissue of interest will need to
be determined empirically.
6. It is necessary to filter the cell suspension to avoid clogging the
narrow bores of the sample injection needle and tubing on the
flow cytometer, which can be easily clogged by aggregated cells
and debris. The concentration also influences the rate of sort-
ing, which typically progresses at 2000–20000 cells/second;
thus, resuspension of cells at a density of 1–10 million cells/mL
in PBS/0.1% BSA is important.
7. If desired, it is also possible to sort more than 1 cell per well to
interrogate pools of cells of known number. We have done
pools of 10–10,000 cells per well to assess linearity and effi-
ciency of the lysis and overall protocol. Note, however, that
when sorting more than ~ 100 cells per well, the total volume
of crude lysate increases due to the sheath volume of sorting
buffer (~3.3 nL/sorted cell) and needs to be taken into con-
sideration for calculations of transcript levels per cell.
8. This lysis method efficiently extracts mRNA from single cells as
demonstrated by comparing average yields of transcripts per
cell (for multiple genes) determined on a population of sorted
single P19 cells, as compared to a bulk preparation of total

RNA extracted from 10,000 P19 cells sorted into a single well
using a guanidinium isothiocyanate lysis method (Aurum kit
from Bio-Rad; manuscript in preparation).
9. It is very important to be aware that in QS 1.7.4.0917 and
earlier software versions, all droplets are assigned as negative
droplets by default. If they are left as unassigned droplets, they
will be included in the total count of negative droplets (i.e., the
“All Negatives” cluster) resulting in under-quantification of
your gene of interest (GOI) when computed by the Poisson
equation. Thus, droplets for all other positive clusters besides
that for the gene of interest—including primary clusters for the
other gene targets and secondary and tertiary clusters, if pres-
ent—must be defined as another single-positive “meta-cluster”
(e.g., in HEX only, shown as green clusters in the Fig. 3b).
These green droplets will be ignored in calculation of the
GOI’s copy number per well by QuantaSoft™. If they are
defined as double-positive (i.e., orange) clusters, they will
incorrectly inflate the number of counted copies of the “FAM
only” GOI and give it an excessively high transcript copy
number per well.
10. If you were to analyze more than one 5-plex reaction per
96-well plate—for example, 48 cells analyzed per plate with
two 5-plexes per cell, each ddPCR™ reaction with ~half of a
cell’s cDNA—you would define clusters as described in Sub-
heading 3.6, step 2 (b) for the first 48 wells together (5-plex
1), then similarly for the second 48 wells together (5-plex 2),
then export the transcript concentrations and counts/well for
each first gene of interest (in both 5-plexes), followed itera-
tively by each of the other 4 genes in each 5-plex.
11. This method of radial multiplexing should also be readily
applicable to the analysis of a handful of tumor mutations in
circulating cell-free (cf)DNA, since such mutations are likely to
exist at low concentrations even if more than one of them is
present in a single plasma cfDNA sample. Hence, there too, as
in single cell analysis, it is unlikely that a significant number of
mutant target molecules would colocalize into the same drop-
let (i.e., in secondary and tertiary clusters) and therefore most
targets would be captured in analyzing primary clusters alone.
12. This method is capable of measuring as few as 3–5 transcripts
per cell, or approximately five- to tenfold more sensitive than
RNA-Seq methods.
Acknowledgments
We wish to thank our many colleagues who have assisted in the

development of the instrumentation and reagents necessary to
develop this protocol, especially Shenglong Wang who began ear-
lier single-cell work from which this was developed. Special thanks
also to Svilen Tzonev, Doug Hauge, Niels Klitgord, and Dimitri
Skvortsov for assistance with the quantification analysis, and to
Marcos Oquendo for assistance with the S3e cell sorter.
References
1. Thomas PS (1980) Hybridization of denatured peripheral blood T cells and leukemic Jurkat T
RNA and small DNA fragments transferred to cells. Lab Investig 83(6):765–776
nitrocellulose. Proc Natl Acad Sci U S A 77 11. Canales RD, Luo Y, Willey JC et al (2006)
(9):5201–5205 Evaluation of DNA microarray results with
2. Karlin-Neumann GA, Sun L, Tobin EM quantitative gene expression platforms. Nat
(1988) Expression of light-harvesting chloro- Biotechnol 24(9):1115–1122. https://doi.
phyll a/b-protein genes is phytochrome- org/10.1038/nbt1236
regulated in etiolated Arabidopsis thaliana 12. Losick R (2015) A love affair with Bacillus
seedlings. Plant Physiol 88:1323–1331 subtilis. J Biol Chem 290(5):2529–2538.
3. Berk AJ, Sharp PA (1977) Sizing and mapping https://doi.org/10.1074/jbc.X114.634808
of early adenovirus mRNAs by gel electropho- 13. Sorlie T, Tibshirani R, Parker J et al (2003)
resis of S1 endonuclease-digested hybrids. Cell Repeated observation of breast tumor subtypes
12(3):721–732 in independent gene expression data sets. Proc
4. Zinn K, DiMaio D, Maniatis T (1983) Identi- Natl Acad Sci USA 100(14):8418–8423.
fication of two distinct regulatory regions adja- https://doi.org/10.1073/pnas.0932692100
cent to the human beta-interferon gene. Cell 14. Tobin EM, Silverthorne J (1985) Light regula-
34(3):865–879 tion of gene expression in higher plants. Annu
5. Livak KJ, Schmittgen TD (2001) Analysis of Rev Plant Physiol 36:569–593. https://doi.
relative gene expression data using real- time org/10.1146/annurev.pp.36.060185.003033
quantitative PCR and the 2-ΔΔCT method. 15. Sanders R, Mason DJ, Foy CA et al (2013)
Methods 25:402–408. https://doi.org/10. Evaluation of digital PCR for absolute RNA
1006/meth.2001.1262 quantification. PLoS One 8(9):e75296.
6. Livak KJ, Wills QF, Tipping AJ et al (2012) https://doi.org/10.1371/journal.pone.
Methods for qPCR gene expression profiling 0075296
applied to 1440 lymphoblastoid single cells. 16. Hindson CM, Chevillet JR, Briggs HA et al
Methods 59(1):71–79. https://doi.org/10. (2013) Absolute quantification by droplet dig-
1016/j.ymeth.2012.10.004 ital PCR versus analog real-time PCR. Nat
7. Schena M, Shalon D, Davis RW et al (1995) Methods 10(10):1003–1005. https://doi.
Quantitative monitoring of gene expression org/10.1038/nmeth.2633
patterns with a complementary DNA microar- 17. Lemos DR, Babaeijandaghi F, Low M et al
ray. Science 270(5235):467–470 (2015) Nilotinib reduces muscle fibrosis in
8. DeRisi JL, Iyer VR, Brown PO (1997) Explor- chronic muscle injury by promoting
ing the metabolic and genetic control of gene TNF-mediated apoptosis of fibro/adipogenic
expression on a genomic scale. Science 278 progenitors. Nat Medicine 21(7):786–794.
(5338):680–686 https://doi.org/10.1038/nm.3869
9. Iyer VR, Eisen MB, Ross DT et al (1999) The 18. Nagalakshmi U, Wang Z, Waern K et al (2008)
transcriptional program in the response of The transcriptional landscape of the yeast
human fibroblasts to serum. Science genome defined by RNA sequencing. Science
283:83–87. https://doi.org/10.1126/sci 320(5881):1344–1349. https://doi.org/10.
ence.283.5398.83 1126/science.1158441
10. Lin Z, Fillmore GC, Um T-H et al (2003) 19. Wilhelm BT, Marguerat S, Watt S et al (2008)
Comparative microarray analysis of gene Dynamic repertoire of a eukaryotic transcrip-
expression during activation of human tome surveyed at single-nucleotide resolution.
Nature 453(7199):1239–1243. https://doi. Biomol Detect Quantif 10:15–23. https://

org/10.1038/nature07002 doi.org/10.1016/j.bdq.2016.05.002
20. Mortazavi A, Williams BA, McCue K et al 27. Monzo HJ, Park TIH, Montgomery JM et al
(2008) Mapping and quantifying mammalian (2012) A method for generating high-yield
transcriptomes by RNA-Seq. Nat Methods 5 enriched neuronal cultures from P19 embryo-
(7):621–628. https://doi.org/10.1038/ nal carcinoma cells. J Neurosci Methods
nmeth.1226 204:87–103. https://doi.org/10.1016/j.
21. Ståhlberg A, Bengtsson M (2010) Single-cell jneumeth.2011.11.008
gene expression profiling using reverse tran- 28. Zhong Q, Bhattacharya S, Kotsopoulos S et al
scription quantitative real-time PCR. Methods (2011) Multiplex digital PCR: breaking the
50:282–288. https://doi.org/10.1016/j. one target per color barrier of quantitative
ymeth.2010.01.002 PCR. Lab Chip 11(13):2167–2174. https://
22. Tang F, Barbacioru C, Wang Y et al (2009) doi.org/10.1039/c1lc20126c
mRNA-Seq whole-transcriptome analysis of a 29. Pender A, Garcia-Murillas I, Rana S et al
single cell. Nat Methods 6:377–382. https:// (2015) Efficient genotyping of KRAS mutant
doi.org/10.1038/nmeth.1315 non-small cell lung cancer using a multiplexed
23. Shalek A, Satija R, Shuga J et al (2014) Single- droplet digital PCR approach. PLoS One 10
cell RNA-seq reveals dynamic paracrine control (9):e0139074. https://doi.org/10.1371/jour
of cellular variation. Nature 510 nal.pone.0139074
(7505):363–369. https://doi.org/10.1038/ 30. Hughesman CB, XJD L, Liu KYP et al (2016)
nature13437 A robust protocol for using multiplexed drop-
24. Macosko EZ, Basu A, Satija R et al (2015) let digital PCR to quantify somatic copy num-
Highly parallel genome-wide expression ber alterations in clinical tissue specimens.
profiling of individual cells using nanoliter dro- PLoS One 11(8):e0161274. https://doi.org/
plets. Cell 161:1202–1214. https://doi.org/ 10.1371/journal.pone.0161274
10.1016/j.cell.2015.05.002 31. Kinz E, Leiherer A, Lang AH et al (2015)
25. Klein AM, Mazutis L, Akartuna I et al (2015) Accurate quantitation of JAK2 V617F allele
Droplet barcoding for single-cell transcrip- burden by array-based digital PCR. Int J Lab
tomics applied to embryonic stem cells. Cell Hematol 37(2):217–224. https://doi.org/10.
161:1187–1201. https://doi.org/10.1016/j. 1111/ijlh.12269
cell.2015.04.044 32. Madic J, Zocevic A, Senlis V et al (2016)
26. Whale A, Huggett J, Tzonev S (2016) Funda- Three-color crystal digital PCR. Biomol Detect
mentals of multiplexing with digital PCR. Quantif 10:34–46. https://doi.org/10.1016/
j.bdq.2016.10.002
Chapter 25
Quantification of Circulating MicroRNAs by Droplet

Digital PCR
Manuela Ferracin and Massimo Negrini
Abstract
MicroRNAs (miRNAs) are released in the blood as cell-free molecules either linked to Ago proteins and
LDL or enveloped inside exosomes and microvescicles. The amount of specific circulating microRNAs has
been discovered to change accordingly to a disease state and to be potentially used as a disease biomarker.
Sensitive and accurate methods for circulating microRNA quantification using probe-based or dye-based
digital PCR technology have been developed. With a digital PCR system it is possible to obtain the absolute
quantification of specific miRNAs, bypassing several issues related to low abundance targets and miRNA
normalization. This chapter addresses the workflow and methods for miRNA assessment in biological fluids
using EvaGreen-based droplet digital PCR as well as how to analyze and interpret results.
Key words MicroRNA, Serum, Plasma, Droplet digital PCR, Diagnostics, Cancer, Biomarkers
1 Introduction
The discovery of stable and assessable miRNA molecules in human

body fluids, including blood, saliva, urine, cerebrospinal fluid, and
synovial fluid, opened novel possibilities in their use as disease
biomarkers [1–3]. These miRNAs are detected outside the cells
and for this reason they are better known as cell-free or circulating
miRNAs; their amount can be extremely low but they present an
unexpected resistance to degradation. Despite the promising and
exciting features of circulating miRNAs, the low amount and lack of
known endogenous reference genes in body fluids provide a real
challenge for every reliable translational application [4, 5]. In order
to obtain trustworthy miRNA quantification—even for the less
abundant miRNA species—several protocols for the global circulat-
ing miRNA profiling and miRNA-specific validation have been
developed.
The methods applied to perform body fluid miRNA profiling
include probe-based or LNA-based quantitative PCR array, micro-
array, Nanostring nCounter technology, smallRNA sequencing [6].
445
446 Manuela Ferracin and Massimo Negrini
For each of these technologies, specific protocol variants have been

developed in order to manage these low-abundance RNA samples,
although satisfactory results are not always guaranteed [7].
In addition, the absence of true endogenous control miRNAs
outside cells generated novel issues to deal with during the valida-
tion step. Some laboratories decided to use the more stable circu-
lating miRNAs detected during the initial high-throughput
screening for sample normalization. Others adopted an approach
aimed at quantifying the amount of each miRNA normalized to
plasma, serum, or any other body fluid volume, using a sort of
absolute quantification [7, 8]. Digital PCR technology proved to
be very effective in the second approach, even exceeding the per-
formance of traditional quantitative PCR [8]. Indeed, digital PCR
target amount calculation was comparable to the use of raw CT
(cycle threshold) in qPCR experiments, with the advantage of
providing the exact count of miRNA molecules in the solution
and a better accuracy in low-abundance miRNA quantification.
The EvaGreen-based ddPCR efficiency has been previously deter-
mined [9]. The assay is precise, reproducible over a range of con-
centrations of four orders of magnitude, and sensitive, detecting a
target miRNA at levels down to 1 copy/μL.
Thus, this chapter focuses on how to perform miRNA quanti-
fication using EvaGreen-based droplet digital PCR technology with
QX200 system (Bio-Rad), specifically considering serum and
plasma clinical samples [9].
2 Materials
List of reagents, instrumentation, and support materials to run an

experiment.
2.1 Sample 1. EDTA Tubes (Vacuette or BD Vacutainer).

Preparation 2. Tabletop centrifuge.
3. 1.5 mL polypropylene tubes.
2.2 RNA Extraction 1. miRNeasy Mini Kit (Qiagen).

2. 100 nmole RNA oligo Cel-miR-39-3p (Integrated DNA Tech-
nologies). Sequence: UCACCGGGUGUAAAUCAGCUUG.
3. Tabletop centrifuge.
2.3 cDNA Synthesis 1. Universal cDNA synthesis kit II (Exiqon).

2. 200 μL PCR tubes or plates.
3. Thermal cycler.
Circulating miRNA Quantification 447
2.4 EvaGreen-Based 1. MicroRNA LNA PCR primer set (Exiqon).

Droplet Digital PCR 2. QX200 droplet generator (Bio-Rad).
3. QX200 droplet reader (Bio-Rad).
4. QuantaSoft software (Bio-Rad).
5. PX1 PCR plate sealer (Bio-Rad).
6. DG8 droplet generator cartridges and gaskets (BioRad).
7. QX200 ddPCR EvaGreen supermix (BioRad).
8. QX200 droplet generator oil for EvaGreen dye (BioRad).
9. 200 μL PCR plates (Eppendorf).
10. Pierceable foil plate seals.
11. Thermal cycler.
3 Methods
3.1 Sample Plasma and serum processing is a relevant step in circulating

Preparation miRNA quantification. There is no preferred procedure for plasma
and serum preparation. To be comparable, all the samples from the
same experiment must be processed using exactly the same
workflow.
1. Plasma preparation. Collect 5 mL blood in EDTA (ethylene-
diaminetetracetic acid) tubes (Vacuette or BD Vacutainer);
centrifuge samples at 1000 g for 10 min at room temperature
to remove blood cells, and dispense the supernatant plasma in
aliquots. Aliquots can be stored at 80 C until use.
2. Serum preparation. Collect 5 mL blood in serum tubes (Vacu-
ette or BD Vacutainer), keep at room temperature to clot for at
least 60 min, and then spin at 1000 g for 10 min; remove he
serum and dispense in aliquots. Aliquots can be stored at
80 C until use.
3.2 Protocol for Total We recommend starting from 200 μL serum or plasma. Total RNA
RNA (Including miRNA) can be isolated from serum or plasma using commercially available
Isolation kits. We suggest using miRNeasy Mini Kit (Qiagen) as described by
the supplier (see Note 1).
To monitor the extraction and reverse transcription reaction
occurrence, add 3 μL of a 4.16 nM solution of the synthetic
miRNA cel-miR-39-3p from C. elegans (e.g., custom synthesized
by Integrated DNA Technologies, sequence: UCACCGGGU-
GUAAAUCAGCUUG) to 1 mL QIAzol Lysis Reagent (Qiagen)
before adding the lysis regent to the sample (see Note 4).
This is the protocol for total RNA extraction using miRNeasy

Mini Kit (Qiagen).
1. Thaw serum/plasma samples on ice.
2. Add 1 mL of QIAzol Lysis Reagent (Qiagen) to 200 μL
serum/plasma.
3. Mix by vortexing and place the tube containing the homoge-
nate on the bench at room temperature for 5 min.
4. Add 3 μL of a 4.16 nM solution of the synthetic miRNA
cel-miR-39-3p from C. elegans.
5. Add 200 μL of chloroform. Shake the tube vigorously for 15 s.
Place the tube on the bench at room temperature for 2 min.
6. Centrifuge for 15 min at 12,000 g at 4 C to obtain phases
separation: the upper aqueous phase contains RNA.
7. Transfer the aqueous phase to a new tube, about 700 μL, avoid
transfer of any white interphase material.
8. Add 1 mL of 100% ethanol and mix by inversion.
9. Pipet up 700 μL of the sample into a Mini spin column placed
on 2 mL collection tube. Close the lid and centrifuge at
12,000 g for 15 s. Discard the flow-through.
10. Repeat this step using the remaining sample.
11. Add 700 μL Buffer RWT to the Mini spin column, close the lid
and centrifuge for 15 s at 12,000 g to wash the column.
Discard the flow-through.
12. Pipet 500 μL Buffer RPE into the Mini spin column. Close the
lid and centrifuge for 15 s at 12,000 g. Discard the flow-
through. Repeat this step again.
13. Centrifuge the Mini spin column at full speed for 2 min to dry
the spin column membrane from residual ethanol.
14. Transfer the Mini spin column to a new 1.5 mL collection tube
and pipet 35 μL RNase-free water on the membrane of the
column.
15. Centrifuge for 1 min at 12,000 g to elute the RNA.
Note: Since RNA concentration cannot be determined accurately,
we suggest using fixed volumes as a measure for input amount.
3.3 MicroRNA The conversion of RNA, including miRNA, to cDNA can be per-
Reverse Transcription formed using the Universal cDNA synthesis kit II (Exiqon) follow-
ing the company’s guidelines for miRNA profiling in serum and
plasma (see Note 2). The resulting cDNA has to be diluted at least
1:50 prior to amplification.
Table 1
Reverse transcription reagent mix components
Reagent Volume (μL)

5x reaction buffer 4
Nuclease-free water 11
Enzyme mix 2
Template total RNA 3
Total volume 20
Protocol for miRNA reverse transcription using the Universal

cDNA synthesis kit II (Exiqon).
1. Thaw the reverse transcription (RT) reagent mix components:
5 reaction buffer, nuclease-free water. Mix gently inverting
the tubes and place on ice. Immediately before use, remove the
enzyme mix from the freezer and place on ice. Spin down all
reagents.
2. Prepare the RT reagent mix on ice as described in Table 1.
Prepare at least 10% exceeding mix.
3. Mix the reaction by pipetting, and then spin down.
4. Incubate for 60 min at 42 C.
5. Heat-inactivate the reverse transcriptase for 5 min at 95 C.
6. Immediately cool to 4 C.
7. Store the cDNA at 20 C.
3.4 cDNA Dilution 1. Dilute the amount of cDNA template needed for the planned
ddPCR reactions in nuclease free water. We recommend dilut-
ing the cDNA reaction between 1:50 and 1:500 in water,
dependent on target miRNA abundance. Store the diluted
cDNA at 20 C. The diluted cDNA proved to be stable for
at least 3 months at 20 C.
3.5 Droplet Locked Nucleic Acid (LNA)-based miRNA-specific primers (Exi-

Generation and PCR qon) are used with a green fluorescent DNA-binding dye (Eva-
Green) in the QX200 droplet digital PCR system (Bio-Rad).
Droplet generation should be performed on 8 samples at a time.
Technical replicates are not required, due to the high reproducibil-
ity of this technology [8, 9]. A No Template Control (NTC)
sample should be always run in every plate for each different
ddPCR condition.
Table 2
ddPCR reagent mix components
Reagent Volume (μL)

2x EvaGreen Supermix 10
LNA primer set Variable (0.5–1)a
Nuclease-free water Variable
Diluted cDNA template 8
Total volume 20
a
See Note 5
For miRNA quantification, a 20 μL PCR mixture is prepared

containing 10 μL 2X EvaGreen supermix (Bio-Rad), 8 μL diluted
cDNA, and 0.25–1 μL of miRNA-specific miRCURY LNA PCR
primer set (Exiqon).
Protocol for miRNA quantification.
1. Thaw and equilibrate the EvaGreen Master mix (Bio-Rad),
microRNA primer set (Exiqon) and cDNA at room
temperature.
2. Mix thoroughly EvaGreen Master mix by inverting the tube
several times. Spin down all the reagents.
3. Prepare the ddPCR mix as described in Table 2. When multiple
ddPCR reactions are performed with the same microRNA
primer set, it is recommended to prepare a ddPCR mix
working-solution. Prepare at least 10% exceeding mix.
4. Once assembled, thoroughly mix and spin down the
ddPCR mix.
5. Dispense 12 μL of ddPCR mix into a 96-well PCR plate or
PCR tubes.
6. Add 8 μL of diluted cDNA template to each tube/well and mix
by pipetting.
7. Insert the droplet generator cartridge (DG8, Bio-Rad) into the
holder.
8. Transfer 20 μL of each prepared sample to the sample wells
(middle row) of the DG8 cartridge, carefully while avoiding air
bubbles at the bottom of the well.
9. Fill each oil well (bottom row) with 70 μL of EvaGreen droplet
generator oil (Bio-Rad).
10. Hook the gasket over the cartridge holder and insert into the
QX200 droplet generator (Bio-Rad).
Table 3
Thermal cycling conditions for droplet digital PCR
Cycling step Temperature ( C) Time Ramping rate Cycles

Enzyme activation 95 5 min ~2 C/s 1
Denaturation 95 30 s 40
Annealing/extension 56–60a 1 min
Signal stabilization 4 5 min 1
90 5 min 1
Hold (optional) 4 Infinite 1
Optimal temperature should be determined for each primer set. See Note 5
a
11. Close the lid to start droplet generation. When droplet gener-
ation has finished, open the lid, remove the disposable gasket.
The top wells of the cartridge contain the droplets.
12. Pipet slowly and smoothly 40 μL of the contents of the eight
top wells (the droplets) into a single column of a 96-well PCR
plate.
13. Seal the PCR plate with foil immediately after transferring
droplets to avoid evaporation. Use pierceable foil plate seals
that are compatible with the needles in the QX200 droplet
reader.
14. Begin thermal cycling (PCR) within 30 min of sealing the
plate.
15. Perform the cycling protocol according to Table 3.
3.6 Droplet Reading 1. Power on the QX200 droplet reader (Bio-Rad).

and Data Analysis 2. Move the plate from the thermal cycler to the droplet reader.
3. Place the 96-well PCR plate containing the post-PCR droplets
into the base of the plate holder.
4. Place the top of the plate holder on the PCR plate. Firmly press
both release tabs down to secure the PCR plate in the holder.
5. Start the QuantaSoft software from the system PC.
6. In QuantaSoft software, click Setup to define your experiment
then click Run to start the droplet reading.
7. When droplet reading is complete, click “Analyze” button to
open and analyze the data.
8. Use the 2D amplitude plot in QuantaLife software (Bio-Rad)
to select the positive droplets (lasso tool) (Fig. 1a).
9. Use the Events tab to check the number of positive and total
droplets. A total number of 18,000–21,000 droplets is usually
achieved with EvaGreen ddPCR (Fig. 1b, c).
Fig. 1 Positive droplets selection. (a) Positive droplet selection from the 2D Plot using the lasso function and
manually drawing a circle around the proper cloud. (b) Total number of positive (blue) and negative (green)
droplets visualized in the QuantaLife software for each sample. (c) 1D Plot representation of positive and
negative droplets. (d) Some LNA primer set (miR-125a-5p in the panel) displays a nonspecific product
formation, which is evident also in NTC well. A miRNA-specific concentration can still be obtained using the
2D plot and excluding the off-target amplification product (see Note 3)
10. Once the positive droplets are selected, export the miRNA
concentration values using the export .csv option from the
Concentration tab.
11. Assuming that each miRNA molecule is reverse-transcribed in
a cDNA molecule, we can calculate the absolute copies in 1 μL
of plasma or serum multiplying the obtained concentration in
ddPCR reaction value for a dilution factor (DF ¼ 145.83 for
1:50 cDNA dilution).
4 Notes
1. Plasma can be prepared using different protocols. The most

common protocol suggests to use two centrifugations at
1200 g either at room temperature or at þ4 C. The
increasing of the spin rate reduces the number of exosomes,

microvescicles, and platelets in the sample and therefore
changes the miRNA composition [10]. It is therefore manda-
tory to compare samples that have been prepared following the
same protocol.
2. When a certain number of circulating miRNAs needs to be
assessed, a universal cDNA system that reverse-transcribes all
the miRNAs in one reaction (e.g., the systems proposed by
Exiqon, Qiagen, and now also by Applied Biosystems/Thermo
Fisher) is preferable to the still largely used system developed
by Applied Biosystems/Thermo Fisher consisting of a miRNA-
specific reverse transcription with stem-loop primers. Indeed, a
universal reverse transcription system provides the flexibility to
subsequently select the subset of miRNAs to test starting from
the same RT, therefore reducing the amount of RNA required
for multiple miRNAs assessment.
3. The NTC (No Template Control) sample needs to be run in
every plate to verify the expected fluorescent signal of the
individual droplets. There is always the possibility that due, to
the short length of the miRNAs, nonspecific products are
amplified. Since the size of targeted miRNA amplicons and
that of primer dimers are relatively close in size, it may be
difficult to discriminate them using fluorescence amplitude
only. Analysis using 2D plots adds a second fluorescent dimen-
sion to the dropltes and sometimes allows separation of two
distince populations. If the “cloud” of nonspecific target ampli-
fication is not overlapping with the true signal (Fig. 1d), the
miRNA can still be quantified selecting only the true positive
droplets. In this case, the NTC sample is essential for droplet
selection. In the case of overlapping true- and false-positive
droplet distributions, the assay cannot be used with the
ddPCR system. A different assay, from a different supplier,
should be considered.
4. Working in every step with fixed input volumes it is possible to
obtain the absolute quantification of miRNA copies, without
the need of additional normalization steps. Therefore, results
obtained using droplet digital PCR can be easily used to calcu-
late the number of each miRNA molecule that are present in
1 mL of plasma or serum. Spiked-in exogenous miRNAs (e.g.,
cel-miR-39) can be used to monitor the occurrence of RNA
extraction or RT reaction but not to normalize the miRNA
levels, because their recovery is more variable than that of
endogenous miRNAs [9].
Fig. 2 Droplet distribution for two different miRNA assays. miR-320a (a) and miR-21-5p positive and negative
droplet amplitudes (upper panels) and concentration (lower panels) in seven representative plasma samples
and NTCs. The fluorescence amplitude can change according to the assay. Results are presented as copies
per microliter in the amplification reaction. Error bars represent the Poisson 95% confidence interval
5. The amount of each miRNA in the blood is very different and

some miRNAs result to be more abundant than others. Using a
1:50 diluted cDNA in a ddPCR reaction is usually sufficient to
obtain an adequate number of positive and negative droplets.
In case of positive droplet saturation (i.e., no negative droplets)
a further dilution of the cDNA sample is necessary.
6. Different EvaGreen-based miRNA assays can generate differ-
ent positive and negative droplet amplitudes (Fig. 2). Every
LNA primer set should be optimized before running all the
samples. The result of miR-125a-5p and miR-425-5p optimi-
zation is represented in Figs. 3 and 4 respectively. Specifically,
by changing the amount of primer (usually in the range of
0.25–1 μL per ddPCR reaction) and the annealing temperature
(usually in the range of 56–60 C), it is possible to identify the
combination that determines a better separation between posi-
tive and negative droplets.
Fig. 3 Influence of primer concentration and annealing temperature on the fluorescence amplitude of droplets
in EvaGreen-based ddPCR. 1D Plots and concentration measurements from the miR-125a-5p obtained at
annealing temperatures of 56, 58, and 60 C and using 0.5 or 1 μL primer for two different plasma samples
(S1, S2) and NTC. This assay does not work at 60 C in EvaGreen master mix. The best positive-negative
droplet separation can be obtained performing the PCR at 56 and using either 0.5 or 1 μL primer. The
nonspecific product is still visible in 1D Plots (see Note 6)
Fig. 4 Influence of primer concentration and annealing temperature on the fluorescence amplitude of droplets
in EvaGreen-based ddPCR. 1D Plots and concentration measurements from the miR-425-5p obtained at
annealing temperatures of 56, 58, and 60 C and using 0.5 or 1 μL primer for two different plasma samples
(S1, S2) and NTC. This assay work properly at all temperatures in EvaGreen master mix. The best positive-
negative droplet separation can be obtained performing the PCR at 56 and using 1 μL primer
Acknowledgments
The work was supported by funding from the Italian Association

for Cancer Research (AIRC) to MF (MFAG 11676) and to MN
(Special Program Molecular Clinical Oncology - 5 per mille
n. 9980, 2010/15) and from the Italian Ministry of Instruction,
University and Research FIRB 2011 (Project RBAPIIBYNP) and
University of Ferrara (FAR 2012-14) to MN.
References
1. Chen X, Ba Y, Ma L, Cai X, Yin Y, Wang K et al of cancer and other diseases. Cell Res 18
(2008) Characterization of microRNAs in (10):997–1006. PubMed PMID: 18766170.
serum: a novel class of biomarkers for diagnosis Epub 2008/09/04. eng.
2. Lawrie CH, Gal S, Dunlop HM, Pushkaran B, PubMed PMID: 24533657. Pubmed Central
Liggins AP, Pulford K et al (2008) Detection of PMCID: 3943687. Epub 2014/02/19. eng.
elevated levels of tumour-associated micro- 7. Ferracin M, Lupini L, Salamon I, Saccenti E,
RNAs in serum of patients with diffuse large Zanzi MV, Rocchi A et al (2015) Absolute
B-cell lymphoma. Br J Haematol 141 quantification of cell-free microRNAs in cancer
(5):672–675. PubMed PMID: 18318758. patients. Oncotarget 6(16):14545–14555.
Epub 2008/03/06. eng. PubMed PMID: 26036630. Epub 2015/06/
3. Mitchell PS, Parkin RK, Kroh EM, Fritz BR, 04. Eng.
Wyman SK, Pogosova-Agadjanyan EL et al 8. Hindson CM, Chevillet JR, Briggs HA, Galli-
(2008) Circulating microRNAs as stable chotte EN, Ruf IK, Hindson BJ et al (2013)
blood-based markers for cancer detection. Absolute quantification by droplet digital PCR
Proc Natl Acad Sci U S A 105 versus analog real-time PCR. Nat Methods 10
(30):10513–10518. PubMed PMID: (10):1003–1005. PubMed PMID: 23995387.
18663219. Pubmed Central PMCID: Epub 2013/09/03. eng.
2492472. Epub 2008/07/30. eng. 9. Miotto E, Saccenti E, Lupini L, Callegari E,
4. Jarry J, Schadendorf D, Greenwood C, Negrini M, Ferracin M (2014) Quantification
Spatz A, van Kempen LC (2014) The validity of circulating miRNAs by droplet digital PCR:
of circulating microRNAs in oncology: five comparison of EvaGreen- and TaqMan-based
years of challenges and contradictions. Mol chemistries. Cancer Epidemiol Biomark Prev
Oncol 8(4):819–829. PubMed PMID: 23(12):2638–2642. PubMed PMID:
24656978. Epub 2014/03/25. eng. 25472671. Epub 2014/12/05. eng.
5. Witwer KW (2015) Circulating MicroRNA 10. Cheng HH, Yi HS, Kim Y, Kroh EM, Chien
biomarker studies: pitfalls and potential solu- JW, Eaton KD et al (2013) Plasma processing
tions. Clin Chem 61(1):56–63. PubMed conditions substantially influence circulating
PMID: 25391989. Epub 2014/11/14. Eng. microRNA biomarker levels. PLoS One 8(6):
6. Moldovan L, Batte KE, Trgovcich J, Wisler J, e64795. PubMed PMID: 23762257. Pubmed
Marsh CB, Piper M (2014) Methodological Central PMCID: 3676411. Epub 2013/06/
challenges in utilizing miRNAs as circulating 14. eng.
biomarkers. J Cell Mol Med 18(3):371–390.
Chapter 26
Droplet Digital PCR for Absolute Quantification

of Extracellular MicroRNAs in Plasma and Serum:
Quantification of the Cancer Biomarker hsa-miR-141
Maria D. Giraldez, John R. Chevillet, and Muneesh Tewari
Abstract
Droplet-based digital PCR provides high-precision, absolute quantification of nucleic acid target sequences
with wide-ranging applications for both research and clinical diagnostic applications. Droplet-based digital
PCR enables absolute quantification by counting nucleic acid molecules encapsulated in discrete, volumet-
rically defined water-in-oil droplet partitions. The current available systems overcome the previous lack of
scalable and practical technologies for digital PCR implementation. Extracellular microRNAs in biofluids
(plasma, serum, urine, cerebrospinal fluid, etc.) are promising noninvasive biomarkers in multiple diseases
and different clinical settings (e.g., diagnosis, early diagnosis, prediction of recurrence, and prognosis).
Here we describe a protocol that enables highly precise and reproducible absolute quantification of
extracellular microRNAs using droplet digital PCR.
Key words Digital PCR, MicroRNA, Biofluids, Plasma, Serum, Absolute quantification, qPCR,
Reproducibility
1 Introduction
Digital PCR is a method that provides high-precision, absolute

quantification of nucleic acid target sequences based on multiple
partitioning of individual analyte molecules at limiting dilution, in
its original form described as resulting in one or zero molecules in
most reactions [1]. After endpoint PCR, the starting concentration
of template is determined by Poisson statistical analysis of the
number of positive (containing amplified target) and negative
(no amplified target detected reactions). The digital PCR concept
[2] has many potential advantages over real-time PCR, including
the capability to obtain absolute quantification without external
references, robustness to variations in PCR efficiency [3] and
potentially higher resilience to PCR inhibitors [4, 5]. Importantly,
technology has become commercially available that overcomes the
previous lack of scalable and practical technologies for digital PCR
459
460 Maria D. Giraldez et al.
implementation [6]. In this sense, the current commercial systems

allow reactions to be partitioned into nanoliter-size (Bio-Rad)
[7, 8] or picoliter-size (RainDance) [9], aqueous droplets in oil
rather than multiwell plates. Rapid microfluidic analysis of
thousands or millions of droplets per sample (Bio-Rad and Rain-
Dance, respectively) makes droplet-based digital PCR feasible for
routine use. In addition, the practical dynamic range of the system
is substantially improved with highly uniform droplets (e.g.,
Bio-Rad QX100/200), which (with Poisson correction for multi-
ple target molecules per droplet) enable the precise calculation of
concentrations even above conditions of limiting dilution [7, 8]
(e.g., up to and sometimes even beyond an average of five copies
per droplet).
We have developed a protocol for highly precise and reproduc-
ible absolute quantification of extracellular microRNAs using the
Bio-Rad Droplet Digital™ PCR (ddPCR™) system and compared
its performance against RT-qPCR [10]. Extracellular microRNAs
in biofluids (plasma, serum, urine, cerebrospinal fluid, etc.) are
being actively studied as potential noninvasive biomarkers in multi-
ple diseases and different clinical settings (e.g., diagnosis, early
diagnosis, prediction of recurrence, and prognosis) [11]. For exam-
ple, abundance of circulating, cell-free hsa-miR-141 in plasma or
serum has been found to be a biomarker for several epithelial cancer
types, including prostate, colorectal, and breast cancers [12–19]. It
is being studied as a biomarker for prognosis, as well as for mini-
mally invasive (i.e., via blood sample) diagnosis and potentially even
early detection of cancer. In our hands, the use of ddPCR for
microRNA quantification provided higher precision and less day-
to-day variation than RT-qPCR [10]. This greater precision was
associated with a somewhat greater sensitivity of ddPCR than
qRT-PCR for some of the microRNA assays evaluated, often
when sensitivity was assessed as limit of quantification, but also in
some cases when assessed as limit of detection [10]. These features
along with the possibility of obtaining absolute quantification with-
out any standard reference (i.e., without requiring a standard
curve) make ddPCR a very attractive technique for microRNA
research. In any case, the better reproducibility provided by
ddPCR along with the possibility of direct comparison of results
intra-lab (not only day-to-day results from the same project, but
also potentially current results to archival data from older projects)
and potentially across different labs, is expected to enable increased
reproducibility in research, multi-institutional data comparisons
and collaboration, and ultimately faster research progress.
Droplet Digital PCR for Absolute Quantification of Extracellular MicroRNAs. . . 461
2 Materials
2.1 Instruments 1. QX100™ or QX200™ Droplet Digital™ PCR (ddPCR™)

systems (each consists of two instruments, the Droplet Gener-
ator and the Droplet Reader) (Bio-Rad).
2. C1000 Touch™ thermal cycler with a gradient-enabled
96-deep well module (Bio-Rad).
3. PCR Plate Heat Sealer (Eppendorf) or PX1™ PCR Plate Sealer
(Bio-Rad).
4. Single channel manual pipettes, 2, 20, 200, and 1000 μL (Pipet
Lite LTS, Rainin).
5. Eight-channel manual pipettes 10, 20, 50, and 200 μL (Pipet
Lite LTS, Rainin).
2.2 Reagents 1. miRNA synthetic template/s (IDT). See Note 1.

for Creating Dilution 2. Nuclease-free water.
Series (Optional)
3. MS2 carrier RNA 0.8 μg/μL (Roche, PN: 10165948001).
4. Eppendorf twin.tec PCR plate 96 (Eppendorf, PN:
951020362).
5. 15 mL conical tubes (Corning, PN: 352196).
6. 1.7 mL low adhesion microcentrifuge tubes (GeneMate, PN:
C-3302-1).
2.3 Reagents 1. TaqMan® MicroRNA Reverse Transcription Kit (Applied Bio-

for Reverse systems, PN: 4366596 or 4366597) Kit Contents: 10 RT
Transcription Buffer, dNTP mix, RNase Inhibitor, Multiscribe™ reverse
transcriptase (RT) enzyme.
2. TaqMan® MicroRNA Assays 5X Reverse Transcription primer
for your specific target (Applied Biosystems, PN: 4427975).
3. Eppendorf twin.tec PCR plate 96 (Eppendorf, PN:
951020362).
4. Adhesive PCR foils (Thermo Scientific, PN: AB-0626).
5. 1.7 mL low adhesion microcentrifuge tubes (GeneMate, PN
C-3302-1).
2.4 Reagents 1. TaqMan® MicroRNA 20X PCR assay (Applied Biosystems,

for Preparation of PCR- PN: 4427975).
Ready Samples, 2. 2 ddPCR™ Supermix for Probes No dUTP (Bio-Rad, PN:
Droplet Generation 186-3024).
and PCR Amplification 3. Droplet Generation Oil for Probes (Bio-Rad, PN: 186-3005).
of Droplets
4. DG8™ Cartridges for QX100/QX200 Droplet Generator
(Bio-Rad, 24 cartridges PN: 186-4008).
5. DG8™ Gaskets for QX100/QX200 Droplet Generator

(Bio-Rad, 24 gaskets PN: 186-3009).
6. Eppendorf twin.tec semiskirted PCR plate 96 (Eppendorf, PN:
951020362).
7. Pierceable foil heat seal (Bio-Rad, PN:1814040).
8. 1.7 mL low adhesion microcentrifuge tubes (GeneMate, PN:
C-3302-1).
9. Sterile disposable reagent reservoir (Costar, PN: 07-200-127).
2.5 Reagents 1. ddPCR Droplet Reader Oil (Bio-Rad, PN:186-3004).

for Droplet Reading
3 Methods
3.1 Blood Specimen Methods for sample processing and RNA isolation have been
Processing for Plasma described by us elsewhere [12, 20, 21]. Briefly, in the latest itera-
and RNA Isolation tion of the plasma collection protocol, we collect whole blood
samples in K2EDTA tubes and separate the plasma within 2 h
using a two-stage centrifugation protocol to ensure that it is free
not only of cells but also of significant numbers of contaminating
platelets [21]. We discard plasma samples that show visual evidence
of hemolysis. Total RNA containing microRNAs is purified from
200 μL of plasma or serum using the miRNeasy kit (Qiagen) with
slight modification as previously described [20] although we have
also successfully employed an alternative protocol using the miR-
Vana PARIS RNA isolation kit (Ambion) [12]. We find that it is not
uncommon for the concentration of purified RNA to be too low for
quantification using absorption spectrophotometry. We use a fixed
volume of RNA eluate rather than a fixed mass of RNA as input into
the subsequent reverse transcription reaction.
3.2 Create Dilution All the following steps should be performed on ice unless otherwise
Series (Optional specified. See Note 3.
Note 2)
1. Thaw synthetic miRNA template/s (10 μM) and MS2
carrier RNA.
2. Label four low adhesion 1.7 mL microcentrifuge tubes I–IV for
each synthetic template. See Note 4.
3. Prepare MS2 diluent in a 15 mL conical tube as shown in
Table 1.
4. Mix gently by inversion.
5. Pipet the volumes of MS2 diluent shown in column 3 of
Table 2 into the microcentrifuge tubes prelabeled as I–IV for
each microRNA dilution series.
Table 1
MS2 diluent
1
Nuclease-free water 9.585 mL
MS2 carrier RNA (0.8 μg/μL) 415 μL
Total volume 10 mL
Table 2
microRNA synthetic template dilutions
Copies/μL in RT Copies/μL in ddPCR

assay when 2 μL when 1 μL of
of dilution is synthesized cDNA
Dilution # Transferred Diluent H2O/ miRNA added to a 10 μL is added to 20 μL
(Tube Label) miRNA (μL) MS2 (μL) copies/μL RT assay ddPCR assay
Tube or well Supplied 6.02 1012
I 10 990 6.02 1010
II 10 990 6.02 10 8
III 10 990 6.02 10 6
IV 10 990 6.02 10 4
1 (96-well 41 59 24,691 4936 250
plate starts)
2 50 50 12,345 2468 125
3 50 50 6173 1234 62.5
4 50 50 3086 617 31.3
5 50 50 1543 309 15.6
6 50 50 772 154 7.8
7 50 50 386 77 3.9
8 50 50 193 39 1.95
9 50 50 96 19 0.98
10 50 50 48 10 0.49
11 50 50 24 5 0.25
12 0 50 0 0 0
6. Pipet 59 μL of MS2 diluent in the first well of a 96-well plate

and 50 μL for the 2nd–12th consecutive wells.
7. Spin briefly the synthetic miRNA template/s (10 μM), mix well
by pipetting up and down several times, and briefly spin down
again.
8. Transfer 10 μL of the corresponding synthetic miRNA into the
microcentrifuge tube prelabeled as I.
9. Mix this solution 5 with the full volume by pipetting up and
down to ensure complete mixing and spin down briefly.
10. Transfer 10 μL of the volume contained in the microcentrifuge
tube I into the microcentrifuge tube prelabeled as II.
11. Repeat steps 9–10, changing the volume transferred as indi-
cated in Table 2, until the entire dilution series has been
completed in tubes II–IV and then in wells 1 through 12 of
the microplate. See Note 5.
3.3 Reverse All the following steps should be performed on ice unless otherwise
Transcription specified. See Note 3.
1. Thaw 10X RT buffer, dNTP mix, and 5X RT primer for the
miRNA of interest. Keep RNAse inhibitor and RT enzyme cold
on a portable cooler. See Notes 6 and 7.
2. Vortex RT primer, 10X RT buffer, and dNTP mix to ensure
adequate mixing and spin down briefly. Spin down briefly
RNAse inhibitor and RT enzyme.
3. Add the RT Master Mix components listed in Table 3 to a low
adhesion microcentrifuge tube. Mix well by pipetting up and
down several times before adding RNAse inhibitor and reverse
transcription enzyme. Then mix again by pipetting up and
down five times.
Table 3
Reverse transcription master mix
1
10x RT buffer 1
100 mM dNTPs 0.1
RT primer (5) 2
RNAse inhibitor 0.12
Multiscribe RT 0.66
RNA 2
Total volume 10 μL
Table 4
Reverse transcription thermal conditions
16 C 30 min

42 C 30 min
85 C 5 min
4 C Hold
4. Add 8 μL of RT master mix into each relevant well of a 96-well

plate using a multichannel pipette.
5. Pipet 2 μL of the synthetic microRNA template dilutions or
extracellular RNA sample into the reverse transcription mix.
Mix with full solution volume (10 μL) using a multichannel
pipette. Please note that we have tested this protocol for plasma
and serum. It may also be useful for analysis of extracellular
RNAs from other sources such as urine, sputum, and CSF but
would need to be validated for each source biofluid type.
6. Cover the 96-well plate with foil.
7. Spin down briefly the 96-well plate.
8. Place the 96-well plate onto a compatible thermal cycler. Close
the lid and perform reverse transcription using the thermal
conditions specified in Table 4.
3.4 Preparation All the following steps should be performed on ice unless otherwise
of PCR-Ready Samples specified.
1. Thaw 2X Bio-Rad supermix and the 20X primer/probe mix
that is specific for your target miRNA. See Notes 6 and 7.
2. Spin down briefly 2X Bio-Rad PCR supermix and mix by
pipetting up and down several times (a concentration gradient
may form during 20 C storage). See Note 8.
3. Spin down primers and vortex to ensure an appropriate mixing.
4. Prepare the ddPCR Master Mix by pipetting the reagents
shown in Table 5 into a low adhesion microcentrifuge tube
and mix thoroughly by pipetting up and down several times.
Table 4 provides volumes of components needed for one reac-
tion of Master Mix. In practice, the volumes should be mul-
tipled by the number of reactions needed and increased by
another 10% each to have an excess of master mix to compen-
sate for pipetting losses which are inevitable when subsequently
drawing multiple aliquots from the Master Mix tube.
5. Add 21 μL of the master mix into each relevant well of a new
96-well plate on ice.
6. Dilute the cDNA solution 1:1 with nuclease-free water.
Table 5
ddPCR master mix
1
20 primers/probe TM 1.1
2 Bio-Rad MM v1.2 11.1
Total master mix volume for 1 reaction 21 μL
Note that the total volume of master mix and cDNA for each reaction is 23 μL, even
though only 20 μL is loaded onto the DG8 chip. The excess is included to cover for any
pipetting losses.
7. Pipet 2 μL of diluted cDNA into the master mix.

8. Seal the 96-well plate and spin down briefly (e.g., 1 min at
600 g)
3.5 Droplet 1. Dispense droplet-generating oil into a reservoir. See Note 9.

Generation 2. Turn plate thermo sealer on to warm up at least 10 min
and Amplification before use.
of Droplets
3. Let reaction mixes stand at room temperature for 3 min before
transferring into cartridge.
4. Load a DG8 cartridge in a DG8 cartridge holder placing the
notch in the cartridge at the upper left of the holder.
5. Transfer the reaction mixes to the middle rows of the DG8
cartridge using a multichannel pipette and pipetting slowly to
avoid the generation of air bubbles. See Notes 10 and 11.
6. Add 70 μL of the droplet generator oil into each of the bottom
wells of the DG8 cartridge using a multichannel pipette.
7. Attach a gasket on top of the DG8 cartridge by hooking it on
both sides of the cartridge holder.
8. Open the droplet generator by pressing the button on the
green top and place the cartridge holder into the instrument.
9. Close the droplet generator by pressing the same button. The
instrument will start generating droplets.
10. After droplet generation is completed (all three indicator lights
on the droplet generator are solid green), open the droplet
generator and take out the cartridge holder with the cartridge
in place. See Note 12.
11. Remove the disposable gasket from the holder and discard
it. Keep the cartridge in the holder.
12. Transfer 40 μL droplets from the top wells of the cartridge to a
96-well PCR plate (see Note 13) by pipetting gently with a
Table 6
ddPCR cycling conditions
95 C 10 min
40 cycles of
94 C 30 s
60 C 60 s

98 C 10 min
Then
4 C Hold
multichannel pipette. See Notes 14 and 15. Cover the plate

during the process to prevent evaporation and contamination.
13. Seal the plate immediately after transferring the droplets with a
pierceable foil seal to avoid evaporation. See Notes 16 and 17.
14. Place the sealed 96-well plate on a thermal cycler and perform
the cycling conditions specified on Table 6. See Notes 18
and 19.
3.6 Droplet Reading 1. When the PCR amplification is completed remove the 96-well
plate from the thermal cycler and load it onto a droplet reader.
See Note 20. Press the button on the green lid of the reader to
open the instrument and place the 96-well plate in the plate
holder (well A1 of the PCR plate must be in the top left
position).
2. Confirm the three indicator lights of the droplet reader are
solid green. See Note 21.
3. Open QuantaSoft software and click Setup in the left naviga-
tion bar to define your experiment. Use the well selector to
define your settings.
Sample:
l Name. Enter the sample ID corresponding to each well.
l Experiment. Select Abs Quant from the drop-down menu.
l Mastermix. ddPCR supermix for probes (no dUTP).
Target 1:
l Name. Enter the name of your target microRNA.
l Type. Select Ch1 unknown or NTC (non template control)
for each well.
Target 2:
l As we are analyzing only one target for experiment, you do
not need to fill out any information here.
Click Apply or OK to save your settings (by clicking Apply you

save settings without exiting the well editor. In contrast by
clicking OK you save changes and close the well editor).
4. Once you have defined your experiment settings click Run on
the left navigation bar to start the run. In the Run Options
window that will appear select the detection chemistry (under
set probe dye, click FAM/VIC from the drop-down menu).
Click OK. After that, a green circle will appear on the left
navigation bar and flash periodically to indicate the run is in
progress. See Note 22.
5. When droplet reading is complete (all four indicator lights are
solid green) open the door by pressing the button on the lid
and remove the plate holder from the unit. See Note 23.
Remove the 96-well PCR plate from the holder and discard it.
3.7 Analyze Results 1. Visualize the data collected from each channel clicking the 1-D
amplitude tab in the QuantaSoft software. In our case only one
channel (channel 1, FAM) is used. The software plots each
droplet in a sample on a graph of fluorescence intensity
vs. droplet number (events). See Fig. 1a. You can also visualize
the data as a histogram of events vs. amplitude. This tab pro-
vides options for adjusting the thresholds used in assigning
positives and negatives for the channel.
2. Select the same threshold for all your samples (a threshold of
4000 has typically performed well for the plasma/serum miR-
NAs we have analyzed, although the appropriate threshold
could vary depending on the intensity of positive and negative
events) using the multisample threshold tool. See Note 24.
3. Click the tab Events to review the number of droplet events
(you can see positive, negative, or total droplet counts, or any
combination of these) counted for each well/sample.
Figure 1b. Consider excluding from the analysis samples with
a total number of droplets <10,000 to achieve more narrow
95% confidence intervals.
4. Click the tab Concentration to see the estimated copies/μL of
your target for each sample as well as the Poisson 95% confi-
dence limits. Figure 1c.
5. Export your results to an Excel file by clicking on Export CSV.
3.8 Interpretation The interpretation of data with respect to biomarker analysis for
of Results—Biomarker plasma or serum microRNAs is still an active research area, and it
Analysis depends not only on the specific biomarker but also on the disease
context. Thus, a universal, specific “threshold” at which miR-141
level is indicative of disease has not been established. In the research
setting where two groups of individuals are compared (e.g., cases
which have a disease vs. healthy controls), both parametric and
Fig. 1 (a) Quantification of serial dilution of synthetic miR-141 using ddPCR. (b) and (c) Quantification of
hsa-miR-141 in plasma samples from two healthy controls (K and W) and two patients with advanced prostate
cancer (J and R). Three technical replicates per sample were measured. In ddPCR, the QuantaSoft software
measures the numbers of droplets that are positive and negative for each fluorophore. Each droplet in a
nonparametric statistical tests are frequently used. In addition,

Receiver Operating Characteristic (ROC) curve analysis is an
important biomarker analysis approach, which can characterize
the relationship between sensitivity and specificity of the biomarker
with respect to distinguishing case specimens from controls.
4 Notes
1. Prepare individual single-use 10 μM aliquots of the synthetic

microRNAs and keep them at 80 C until use.
2. Creation of a synthetic miRNA dilution series is not mandatory
for ddPCR, as quantification with this technology does not rely
on a standard curve. However, the performance of a synthetic
miRNA dilution series is advisable to evaluate the efficiency of
your assays. Such a dilution series provides a quality control to
confirm that the assays and protocol are working well, and it is
advisable to include this regularly in runs for such quality
control purposes. Using synthetic miRNA dilution series
experiments, it is possible to detect assays with low efficiency
that would benefit from further optimization before their use
in ddPCR experiments. Assay optimization can sometimes be
needed to find proper annealing conditions, and in some other
cases to discriminate between similar targets. Some factors that
could contribute to low apparent assay efficiency include low
target accessibility (e.g., primers not being able to anneal due
to nonseparation of the target DNA strands), and degraded
DNA (i.e., if the DNA is fragmented in the region where
primer(s) would bind). It is important to note, however, that
for most assays once optimized, normalization for assay effi-
ciency is typically not required (except in extreme cases) even
when the efficiency of the assays that you plan to use for an
experiment varies somewhat, as the effect of this variation is
typically minimal on ddPCR quantification in contrast to
qPCR.
Fig. 1 (continued) sample is plotted on a 1D intensity graph (a) of fluorescence intensity versus droplet
number (events). Vertical yellow lines in the plot show where droplet data from each well start and end, and
the assigned threshold appears as a horizontal pink line. All positive droplets (those above the threshold
intensity indicated by the horizontal pink line) are scored as positive, and each is assigned a value of one. All
negative droplets (those below the threshold) are scored as negative, and each is assigned a value of 0 (zero).
The number of positive or negative droplets vs. total number of droplets can be visualized in a bar graph (b).
This counting technique provides a digital signal from which to calculate the starting target concentration by a
statistical analysis of the numbers of positive and negative droplets in a given sample. The software fits the
fraction of positive droplets to a Poisson distribution to determine the absolute starting copy number in units of
copies/μL input sample, and then plots the target concentration of each sample as copies per μL (c). Error bars
in the concentration plot reflect total error or Poisson 95% confidence limits
3. Working with RNA requires special precautions because of the

chemical instability of the RNA and the ubiquitous presence of
RNases. Ensure good laboratory practice for handling and
storage of RNA to achieve optimal performance of the
protocol:
l Always wear disposable gloves, and work in a nuclease-free
environment.
l Treat surfaces of benches and glassware with commercially
available RNase inactivating agents such as RNaseZap (Life
Technologies).
l Use nuclease-free, low nucleic acid binding plasticware and
filter barrier pipette tips.
l Always ensure that all reagents and chemicals purchased
commercially are guaranteed to be RNase-free.
l Keep tubes capped when possible, always spin tubes before
opening.
l Work on ice.
l For long-time storage, RNA may be stored at 80 C.
l Avoid repeated freeze–thaw cycles.
4. Create an individual dilution series for each microRNA that
you plan to evaluate in your experiment.
5. Change the micropipette tip between every dilution step to
avoid adsorptive carryover.
6. It is advisable to prepare the master mix in a pre-PCR space
working in a PCR hood if possible, to minimize chances of
contamination during PCR setup.
7. Prepare an individual specific master mix for each microRNA to
be tested.
8. After thawing, store the leftover 2X Bio-Rad PCR supermix at
4 C for 1 to 2 weeks, avoid refreezing.
9. The oil requirement depends on the number of samples that
you are processing. If you are not processing a whole 96-well
plate, instead of dispensing the total volume contained in the
oil generating bottle (7 mL), adjust the volume to the number
of samples that you have as listed in Table 7.
10. Air bubbles can cover the bottom of the well and result in
2500–7000 fewer droplets and poor data quality.
11. All eight sample wells in the DG8 droplet generator cartridge
must contain sample (create nontemplate control reactions or
use Bio-Rad 1 ddPCR buffer control if you are not proces-
sing enough samples to occupy the full chip), and all eight oil
wells must contain droplet generation oil.
Table 7
Droplet generator oil requirements
# Wells Volume of oil (μL)a

8 700
24 1820
48 3500
96 6860
a
Note that the volume of 700 μL includes excess to compensate for losses when pipetting
using a multichannel pipettor
12. The time required for droplet generation is typically 2 min for
eight samples.
13. As this 96-well plate will be later placed in the droplet reader, it
is important to use a plate compatible with this instrument.
(Semiskirted plates such as twin.tec semiskirted plates from
Eppendorf not only fit properly in the instrument but also
provide an optimal rigidity required during the reading
process.)
14. Use a 50 μL manual multichannel pipette and pipet gently to
avoid damaging the droplets.
15. Do not aspirate >40 μL, as this causes air to percolate through
the droplet solution which can shear and/or cause coalescence
of droplets prior to thermocycling.
16. Make sure that you are using pierceable foil heat seal compati-
ble with the sample needle in the droplet reader. (The instru-
ment will pierce the plate seal to aspirate each sample during
the droplet reading.) Do not use self-adhesive PCR films.
17. Begin thermal cycling (PCR) within 30 min of sealing the
plate, or store the plate at 4 C for up to 4 h prior to thermal
cycling.
18. Use a 2.5 C/s ramp rate to ensure each droplet reaches the
correct temperature for each step during the cycling.
19. 40 cycles of PCR is enough for an optimized ddPCR assay. Do
not exceed 50 cycles.
20. Once droplets have undergone PCR-amplification the product
is quite stable. You can leave the plate in the thermal cycler
overnight at 10 C or store at 4 C for up to 3–4 days before
proceeding to read the droplets.
21. The first green solid light indicates that the droplet reader is on,
the second one that the level of the reader oil and waste bottle
are adequate to perform the run, and the third one that there is
a plate in place.
22. It might take up to 1 min for the green circle to appear after
clicking OK.
23. The QX100 droplet reader can process 32 wells/h.
24. Although changing the threshold typically has a minimal effect
on ddPCR results (in contrast to qPCR), we recommend
setting the same threshold for each sample in all your experi-
ments to make the results as comparable as possible.
Acknowledgements
M.D.G acknowledges initial support from a Rio Hortega Fellow-

ship and later from a Martin Escudero Fellowship.
M.T. acknowledges support from the Department of Defense
Peer-Reviewed Cancer Research Program Award CA100606 and
NIH US National Institutes of Health Transformative R01 grant
R01DK085714.
References
1. Vogelstein B, Kinzler KW (1999) Digital PCR. for absolute quantitation of DNA copy num-
Proc Natl Acad Sci U S A 96(16):9236–9241 ber. Anal Chem 83(22):8604–8610
2. Sykes PJ, Neoh SH, Brisco MJ, Hughes E, 8. Pinheiro LB, Coleman VA, Hindson CM,
Condon J, Morley AA (1992) Quantitation of Herrmann J, Hindson BJ, Bhat S, Emslie KR
targets for PCR by use of limiting dilution. (2012) Evaluation of a droplet digital polymer-
BioTechniques 13(3):444–449 ase chain reaction format for DNA copy num-
3. Bustin SA, Nolan T (2004) Pitfalls of quantita- ber quantification. Anal Chem 84
tive real-time reverse-transcription polymerase (2):1003–1011
chain reaction. J Biomol Tech 15(3):155–166 9. Kiss MM, Ortoleva-Donnelly L, Beer NR,
4. Rački N, Dreo T, Gutierrez-Aguirre I, Warner J, Bailey CG, Colston BW, Rothberg
Blejec A, Ravnikar M (2014) Reverse transcrip- JM, Link DR, Leamon JH (2008) High-
tase droplet digital PCR shows high resilience throughput quantitative polymerase chain
to PCR inhibitors from plant, soil and water reaction in picoliter droplets. Anal Chem 80
samples. Plant Methods 10(1):42 (23):8975–8981
5. Dingle TC, Sedlak RH, Cook L, Jerome KR 10. Hindson CM, Chevillet JR, Briggs H, Galli-
(2013) Tolerance of droplet-digital PCR vs chotte EN, Ruf IK, Hindson BJ, Vessella RL,
real-time quantitative PCR to inhibitory sub- Tewari M (2013) Absolute quantification by
stances. Clin Chem 59:1670–1672 droplet digital PCR versus analog real-time
6. Baker M (2012) Digital PCR hits its stride. Nat PCR. Nat Methods 10:1003–1005
Methods 9:541–544 11. Weiland M, Gao XH, Zhou L, Mi QS (2012)
7. Hindson BJ, Ness KD, Masquelier DA, Small RNAs have a large impact: circulating
Belgrader P, Heredia NJ, Makarewicz AJ, microRNAs as biomarkers for human diseases.
Bright IJ, Lucero MY, Hiddessen AL, Legler RNA Biol 9(6):850–859
TC, Kitano TK, Hodel MR, Petersen JF, Wyatt 12. Mitchell PS, Parkin RK, Kroh EM, Fritz BR,
PW, Steenblock ER, Shah PH, Bousse LJ, Wyman SK, Pogosova-Agadjanyan EL,
Troup CB, Mellen JC, Wittmann DK, Erndt Peterson A, Noteboom J, O’Briant KC,
NG, Cauley TH, Koehler RT, So AP, Dube S, Allen A, Lin DW, Urban N, Drescher CW,
Rose KA, Montesclaros L, Wang S, Stumbo Knudsen BS, Stirewalt DL, Gentleman R, Ves-
DP, Hodges SP, Romine S, Milanovich FP, sella RL, Nelson PS, Martin DB, Tewari M
White HE, Regan JF, Karlin-Neumann GA, (2008) Circulating microRNAs as stable
Hindson CM, Saxonov S, Colston BW (2011) blood-based markers for cancer detection.
High-throughput droplet digital PCR system Proc Natl Acad Sci U S A 105
(30):10513–10518
13. Cheng H, Zhang L, Cogdell DE, Zheng H, circulating MicroRNA signature as a biomarker
Schetter AJ, Nykter M, Harris CC, Chen K, for prostate cancer in a high risk group. J Clin
Hamilton SR, Zhang W (2011) Circulating Med 4(7):1369–1379
plasma MiR-141 is a novel biomarker for met- 18. Madhavan D, Peng C, Wallwiener M,
astatic colon cancer and predicts poor progno- Zucknick M, Nees J, Schott S, Rudolph A,
sis. PLoS One 6(3):e17745 Riethdorf S, Trumpp A, Pantel K, Sohn C,
14. Bryant RJ, Pawlowski T, Catto JW, Marsden G, Chang-Claude J, Schneeweiss A, Burwinkel B
Vessella RL, Rhees B, Kuslich C, Visakorpi T, (2016) Circulating miRNAs with prognostic
Hamdy FC (2012) Changes in circulating value in metastatic breast cancer and for early
microRNA levels associated with prostate can- detection of metastasis. Carcinogenesis 37
cer. Br J Cancer 106(4):768–774 (5):461–470
15. Madhavan D, Zucknick M, Wallwiener M, 19. Sun Y, Liu Y, Cogdell D, Calin GA, Sun B,
Cuk K, Modugno C, Scharpff M, Schott S, Kopetz S, Hamilton SR, Zhang W (2016)
Heil J, Turchinovich A, Yang R, Benner A, Examining plasma microRNA markers for
Riethdorf S, Trumpp A, Sohn C, Pantel K, colorectal cancer at different stages. Oncotar-
Schneeweiss A, Burwinkel B (2012) Circulat- get 7(10):11434–11449
ing miRNAs as surrogate markers for circulat- 20. Kroh EM, Parkin RK, Mitchell PS, Tewari M
ing tumor cells and prognostic markers in (2010) Analysis of circulating microRNA bio-
metastatic breast cancer. Clin Cancer Res 18 markers in plasma and serum using quantitative
(21):5972–5982 reverse transcription-PCR (qRT-PCR). Meth-
16. Nadal E, Truini A, Nakata A, Lin J, Reddy RM, ods 50(4):298–301
Chang AC, Ramnath N, Gotoh N, Beer DG, 21. Cheng HH, Yi HS, Kim Y, Kroh EM, Chien
Chen G (2015) A novel serum 4-microRNA JW, Eaton KD, Goodman MT, Tait JF,
signature for lung cancer detection. Sci Rep 5 Tewari M, Pritchard CC (2013) Plasma proces-
(12464) sing conditions substantially influence circulat-
17. Kelly BD, Miller N, Sweeney KJ, Durkan GC, ing microRNA biomarker levels. PLoS One 8
Rogers E, Walsh K, Kerin MJ (2015) A (6):e64795
Part VI
Other Uses of Partitioning

Chapter 27
Droplet Digital™ PCR Next-Generation Sequencing

Library QC Assay
Nicholas J. Heredia
Abstract
Digital PCR is a valuable tool to quantify next-generation sequencing (NGS) libraries precisely and
accurately. Accurately quantifying NGS libraries enable accurate loading of the libraries on to the sequencer
and thus improve sequencing performance by reducing under and overloading error. Accurate quantifica-
tion also benefits users by enabling uniform loading of indexed/barcoded libraries which in turn greatly
improves sequencing uniformity of the indexed/barcoded samples. The advantages gained by employing
the Droplet Digital PCR (ddPCR™) library QC assay includes the precise and accurate quantification in
addition to size quality assessment, enabling users to QC their sequencing libraries with confidence.
Key words Droplet, Digital PCR, NGS, Sequencing library quantification, Sequencing library quali-
fication, Library balancing, Library loading
1 Introduction
Next-generation sequencing (NGS) has unlocked vast genetic

information and accelerated discoveries in the life sciences and
medicine which were unobtainable before. In order to utilize
NGS technology, sequencing libraries are constructed typically
with adapter sequences which bind to sequencing surfaces and are
clonally amplified to boost signal. An important feature of such
libraries is to ensure that the adapted library fragments are at a
concentration which allows for efficient clonal amplification to
occur without mixing with another adapted library fragment.
This is achieved in various forms such as creating an emulsion
where droplets of the emulsion contain a sequencing surface such
as a bead particle and a single adapted library fragment which can
bind to oligonucleotides complementary to the adapter sequence
and thus be clonally PCR amplified in the droplets from the single
starting template. Examples of the emulsion methodology include
Ion Torrent™ and Roche 454® sequencing technologies [1, 2]. If
the emulsions contain two or more adapted library molecules, then
477
478 Nicholas J. Heredia
the resulting bead sequencing surfaces will generate mixed reads

which cannot be differentiated. Thus, quantification of the input of
adapted library molecules in to the emulsion is a critical component
of the sequencing workflow.
Another example of achieving clonal amplification is to create
adapted library fragments and flow them on a flow cell which has
complementary oligonucleotide adapter sequences arrayed on the
surface which can bind the adapted molecules [3]. An important
aspect of this method is that the adapted library molecules are at a
concentration where they are diffuse enough on the flow cell sur-
face such that when they amplify on the surface they can remain
distinct from other adapted library molecules and not merge
together during cluster generation. An Example of this method of
clonal amplification is Illumina sequencing technology. If clusters
merge together, the instrument is not capable of distinguishing and
deconvoluting the signal generated during sequencing by synthesis,
where fluorescently labeled nucleotides are incorporated into the
growing strands on the sequencing clusters. Thus, it is important to
load the sequencing flow cell surface at a concentration where most
adapted library molecules will be spatially distinct from neighbor-
ing adapted library molecules during cluster generation, without
cluster merging. Again, this is an example of a critical point in the
workflow to have accurate quantification of the adapted library
molecules to prevent cluster merging if the concentration were
too high, but also not to underload the sequencing surface and
lose valuable potential sequencing reads. This protocol will focus
on using droplet digital PCR and the Illumina sequencing work-
flow for NGS library quantification to achieve optimal loading of
the sequencing flow cell and to gain equal representation of
sequencing libraries by precisely balancing indexed/barcoded
libraries when they are combined in sequencing runs.
The Droplet Digital PCR Next-Generation Sequencing
Library QC Assay enables the user get highly accurate library
quantification, and also provides information on the relative library
size and amount of adapter dimers generated in the library con-
struction process. These additional metrics give the user confi-
dence in the quality of their library prior to sequencing which
typically required separate qPCR and gel analysis in the past.
Another improvement of digital PCR is that the absolute quantifi-
cation is not dependent on a set of standards with specific sizes and
sequences which may or may not be relevant to the PCR efficien-
cies of the user’s actual samples. Digital PCR provides better
accuracy, confidence, and reproducibility in the user’s
measurements [4].
Droplet Digital™ PCR Next-Generation Sequencing Library QC Assay 479
2 Materials
Set up all ddPCR reactions at room temperature after thawing the

various components of the ddPCR™ Library Quantification Kit for
Illumina TruSeq, from Bio-Rad part number 1863040. Accurate
library quantification is achieved by assaying the NGS libraries at
between 100 and 5000 copies/μL.
Kit components include:
ddPCR Supermix for Probes (no dUTP) at 2 concentration,
ddPCR library quantification assay at 20 concentration.
Other materials required:
Nuclease-free water, TE buffer (20 mM Tris–HCl, 0.1 mM
EDTA, pH 8), 1.5 mL LoBind eppendorf tubes, Twin-Tec® semi-
skirted 96-well PCR plates (Eppendorf 951020362), PX1™ PCR
Plate Sealer (Bio-Rad 1814000), Pierceable Foil Heat Seal
(Bio-Rad #1814040), QX200™ Droplet Digital™ PCR System
(Bio-Rad 1864001), Droplet Generation Oil for Probes (Bio-Rad
1863005), ddPCR™ Droplet Reader Oil (Bio-Rad 1863004),
sterile reagent trough, LTS Rainin pipette tips with filters (Rainin
GP-L200F, GP-L10F), benchtop microcentrifuge, Microseal® ‘B’
Adhesive Seals (Bio-Rad #MSB1001).
3 Methods
1. Thaw kit components at room temperature. Vortex the super-

mix and assay vials thoroughly and spin briefly.
2. Generate a dilution series of Illumina library from the user’s
unknown stock TruSeq library to be quantified by digital PCR.
Dilutions to achieve quantification in the dynamic range of
digital PCR will depend on factors such as which library prepa-
ration kit or method was used, starting concentration of nucleic
acids, and the number of preamplifications cycles used, if any.
Suggestions for achieving a working range of dilutions based
on the original illumine TruSeq LT library preparation are
shown in Table 1. In general, a good starting dilution point
would be 107. Repeat dilution series for additional samples.
Note: additional dilution points may be necessary for different
library construction methods in order to achieve the dynamic
range of ddPCR which currently spans about four orders mag-
nitude. Samples should be diluted in low adsorbing plastic
tubes or plates, such as Eppendorf lobind 1.5 mL. Surfactants
should not be added to samples to prevent adsorption as they
can interfere with droplet making.
3. Pipet 16 μL of 2 mastermix generated in Table 2 into each
well of the 96-well plate.
Table 1
Library dilution scheme
Dilution
Step Tube 1 Tube 2 Tube 3 Tube 4 Tube 5
Sample 5 μL of stock 5 μL of Tube 5 μL of Tube 10 μL of Tube 10 μL of Tube
into into 495 μL 1 into 495 μL 2 into 495 μL 3 into 90 μL 4 into 90 μL
TE TE TE TE TE
TE buffer
Dilution 100 100 100 10 10
2 4 6 7
Final 10 10 10 10 108
dilution
Table 2
Reaction setup
Component Volume/Reaction Final Concentration

2 ddPCR supermix for probes (no dUTP) 10 μL 1
20 ddPCR library quantification assay 1 μL 1
RNase-/DNase-free water 5 μL —
Total volume 16 μL 1
4. Pipet 4 μL of the appropriate dilution series generated from

Table 1 into the corresponding wells according to a plate layout
design such as Table 3. Include no template controls (NTCs)
by adding 4 μL of nuclease-free water in place of actual sample.
5. Seal the plate with adhesive seal, such as Bio-Rad Microseal
“B,” and vortex the plate and briefly centrifuge.
6. Remove the adhesive seal and transfer 20 μL of the mastermix
into the sample inlet wells on a droplet generation cartridge,
add 70 μL of droplet generation oil for probes to the oil inlets
wells on the cartridge, place a cartridge gasket on the cartridge
and insert the cartridge in to the droplet generator instrument.
7. After droplet generation, carefully transfer the newly formed
droplets to a new 96-well plate, repeat droplet generation for
remaining samples and transfer to the plate, place a foil seal on
the plate and heat seal the plate using the PX1 heat sealing
instrument.
8. Transfer the sealed plate to a thermal cycler and run a thermal
cycling protocol according to Table 4.
Table 3
ddPCR 96 well plate map
1 2 3 4 5 6 7 8 9 10 11 12
A NTC NTC NTC NTC NTC NTC NTC NTC NTC NTC NTC NTC
B Lib Lib Lib Lib Lib Lib Lib Lib Lib Lib Lib Lib
1 (108) 1 (108) 1 (107) 1 (107) 1 (106) 1 (106) 8 (108) 8 (108) 8 (107) 8 (107) 8 (106) 8 (106)
C Lib Lib Lib Lib Lib Lib Lib Lib Lib Lib Lib Lib
2 (108) 2 (108) 2 (107) 2 (107) 2 (106) 2 (106) 9 (108) 9 (108) 9 (107) 9 (107) 9 (106) 9 (106)
D Lib Lib Lib Lib Lib Lib Lib Lib Lib Lib Lib Lib
3 (108) 3 (108) 3 (107) 3 (107) 3 (106) 3 (106) 10 (108) 10 (108) 10 (107) 10 (107) 10 (106) 10 (106)
E Lib Lib Lib Lib Lib Lib Lib Lib Lib Lib Lib Lib
4 (108) 4 (108) 4 (107) 4 (107) 4 (106) 4 (106) 11 (108) 11 (108) 11 (107) 11 (107) 11 (106) 11 (106)
F Lib Lib Lib Lib Lib Lib Lib Lib Lib Lib Lib Lib
5 (108) 5 (108) 5 (107) 5 (107) 5 (106) 5 (106) 12 (108) 12 (108) 12 (107) 12 (107) 12 (106) 12 (106)
G Lib Lib Lib Lib Lib Lib Lib Lib Lib Lib Lib Lib
6 (108) 6 (108) 6 (107) 6 (107) 6 (106) 6 (106) 13 (108) 13 (108) 13 (107) 13 (107) 13 (106) 13 (106)
H Lib Lib Lib Lib Lib Lib Lib Lib Lib Lib Lib Lib
7 (108) 7 (108) 7 (107) 7 (107) 7 (106) 7 (106) 14 (108) 14 (108) 14 (107) 14 (107) 14 (106) 14 (106)
Lib library, NTC no template control
Table 4
Cycling protocol for Bio-Rad’s C1000 Touch™ Thermal Cyclera
Cycling Step Temperature, C Time Ramp Rate Number of Cycles

Enzyme activation 95 10 min 1
Denaturation 94 30 s
Annealing/extension 60 1 min Approximately 2.0 C/s 40
Enzyme deactivation 98 10 min 1
Hold (optional) 12 or 4 1 1

Use a heated lid set to 105 C and set the sample volume to 40 μL
a
9. Transfer the thermal cycled plate to the Droplet Reader instru-

ment, create a plate setup for the run using QuantaSoft soft-
ware, and read the droplets.
10. Verify the correct automatic thresholding for each well data set
which was read and adjust manually if necessary.
11. Calculate the concentration of the NGS library stock by the
following method: multiply the ddPCR concentration value of
copies/μL by the factor of the dilution used for that value.
Then multiply this new value by the factor to account for the
dilution in the ddPCR master mix reaction, for example a
factor of 5 if 4 μL of diluted library was placed into 16 μL of
supermix and assay. The stock concentration of each library can
be backcalculated from this method if the ddPCR concentra-
tion measurements falls between 100 and 5000 copies/μL. If
quantification is outside this range, it will be necessary to assay
additional dilution points.
Example:
A value of 253 copies/μL was obtained by ddPCR for a library
sample that was assayed at the 1 106 dilution point. To
backcalculate, multiply 253 copies/μL by 1 106 to account
for the dilution point, and then multiply by a factor of 5 for the
ddPCR reaction dilution. This equals 1.265 109 copies/μL.
To obtain nM concentration, take 1.265 109 copies/μL and
multiply by 1 106μL/L and divide this by 6.023 1023
copies per mole ¼ 2.1 109 M or 2.1 nM
12. Access NGS library quality by quantitating adapter dimers. It is
possible to assess adapter dimer concentrations if there are two
or more distinct clusters and the user is confident, based on
expected sizes of the library that there is an additional unac-
counted for cluster which should be assigned to adapter
dimers. The TaqMan assay employed in this library QC
would yield a cluster, if adapter dimers are present, as the
highest fluorescent cluster on the plot. Figure 5a demonstrates
a fluorescent two-dimensional plot of such an adapter dimer

cluster. By using the lasso tool in QuantaSoft, the user should
assign the negative droplet cluster as double negative (gray),
the intermediate cluster or clusters FAM positive only (Blue),
and the highest cluster as double positive (orange) such as in
Fig. 5a. This enables the user to get a total concentration call of
the adapter dimers and the library using the concentration in
the FAM channel, and the HEX channel concentration will be
the adapter dimer concentration. The user can then take ratios
of adapter dimer concentration to the total to get a percentage
of the library that is adapter dimer.
4 Data Analysis
The TruSeq library quantification kit contains a 50 -nuclease assays

designed to detect and quantify both the P5 and P7 adaptor arms
(Fig. 1a, b). The combined signal from each assay is thus used to
confirm the formation of bona fide library fragments, which display
themselves as double-positive clusters within the 2-channel plot
(Fig. 2). Additionally, by exploiting the ability of the QX-100
platform to detect subpopulations of templates with differing
amplification efficiencies and with different combinations of P5
and P7 moieties a qualitative measure of library purity can be
obtained.
Once data has been acquired on a Droplet Reader instrument,
it is important to view the 2-channel plot of the data. This plot
enables the user to visualize fluorescent intensity from the two
probes in the library quantification kit, the FAM intensity is plotted
on the Y-Axis, and the HEX/VIC intensity is plotted on the X-Axis.
The 2-channel plot also gives a qualitative sense of the library due
to the position of the fluorescent intensity of the clusters. Since the
assay is based on two hydrolysis probes, library molecules having
both a P5 and P7 adapter will be double positive. If there is no
insert in the library molecule and hence it only has adapter-dimers,
this product is the most efficient at PCR amplification in the drop-
let. Thus, droplets containing only adapter-dimers will have the
property of being the highest fluorescent amplitude on a
2-channel plot, and they typically form a secondary cluster to the
cluster which is comprised of droplets with library molecules having
the P5 and the P7 adapters, as well as, library insert. Likewise, the
longer the library insert between the two adapters, the less efficient
that PCR reaction will be, and thus these droplets will have a lower
fluorescent intensity or clustering on the 2-channel plot. Due to
these fluorescent intensity differences, the user is able to get a
qualitative sense of the library insert size distribution which is
inversely related to the fluorescent intensity of the droplets, the
larger the insert, the lower the intensity. This effect is shown in
A P5
Rd1 SP P DNA insert
T A
Index A
P P
Rd2 SP
P7
P5 P7
Rd1 SP DNA insert Rd2 SP
Index
Index
Rd2 SP Rd1 SP
P7 P5
Ligate index adapter
P5 Rd1 SP DNA insert Index

5'
3'
Rd2 SP P7
B Denature and amplify
for final product
H Probe 1 Q F Probe 2 Q
FP
RP
Fig. 1 (a–b) Diagrams representing the illumina library architecture of the adapter structures (a), and the
hydrolysis probes and primers for assaying library molecules (b)
Fig. 2 A 2-channel plot showing well-formed library molecules in droplets (orange cluster), and negative
droplets (Gray cluster)
12,000
10,000 Adapter dimers
FAM amplitude
8,000
6,000
4,000 Decreasing size of library insert with

adapters as part of construct
2,000
0
0 1,000 2,000 3,000 4,000 5,000 6,000 7,000 8,000
HEX amplitude
Fig. 3 A two-channel plot representing the inverse relationship between library insert size and fluorescent
intensity. The highest fluorescent intensity droplets are highlighted within the outlined red oval which
correspond to library molecules with no inserts and are adapter dimers
Fig. 4 Highlighted in blue oval, are ill-formed library molecules in droplets which potentially have multiple P7
adapters attributing a more FAM fluorescent intensity to those droplets compared to the well-formed library
molecules in the orange, double positive, cluster
Fig. 3. Figure 4 demonstrates poor library formation of a few

droplets with potentially multiple P7 adapters added to these
library molecules in these droplets as they are not along the double
positive cluster, but rather possess more FAM fluorescence com-
pared to the vast majority of well-formed library molecules in the
double positive cluster. Figures 5 and 6 represent three ddPCR
well’s worth of data for a dilution series of library molecules span-
ning 106–108dilutions. Figure 5a is the 2-channel plot which has
been manually thresholded using the lasso tool in QuantSoft soft-
ware. The main cluster of library molecules which have insert are
intentionally selected as FAM positive only (blue), and the empty
library molecules which are adapter dimers, are intentionally
selected as double positive for FAM and HEX (orange). These
intentional assignments for the library clusters enable the user to
Fig. 5 (a) Coloring scheme of library clusters which enable quantification of well-formed library molecules
(blue), and adapter-dimers (orange). (b) Quantification of concentration of a dilution series from the data
shown in Fig. 5 which demonstrates good linearity over the tenfold dilution series. The blue points represent
well-formed library plus adapter dimers, and the green points represent only the adapter dimers quantifica-
tion. (c) QuantaSoft can take the ratio of the HEX concentration to FAM concentration for the data in Fig. 5b.
This essentially calculates what percent of the adapter-dimers constitute the total library being assayed
quantify the total number of library molecules which could occupy

space on sequencing flow cell, specifically those library molecules
which contain either insert or empty adapter-dimers and these are
quantified in the FAM channel data in QuantaSoft. By assigning the
adapter dimers cluster as double positive, it is possible to simulta-
neously quantify what percent of the total library molecules are
empty adapter dimers, which can be quantified in the HEX channel
c
100
10
Ratio
0.233 0.234 0.256
0.1
0.01
Lib8 10E-6 Lib8 10E-7 Lib8 10E-8
Sample
Fig. 5 (continued)
800
700
600
PF reads (x 1,000)
500
400
300
200
100
0
0
1
2
y1
y2
y3
y4
y5
y6
y7
y8
y9
ge
y1
y1
y1
era
rar
rar
rar
rar
rar
rar
rar
rar
rar
rar
rar
rar
Lib
Lib
Lib
Lib
Lib
Lib
Lib
Lib
Av
Lib
Lib
Lib
Lib
Fig. 6 Twelve TruSeq LT DNA libraries were quantified by ddPCR and subse-
quently pooled together at equimolar concentrations based on the ddPCR
results. The combined libraries were then loaded on a single sequencing run
and exhibit excellent balancing within less than 15% difference (total number of
reads passing filter for each library). PF passing filter
on QuantaSoft. Figure 5b plots the concentration values across the

dilutions series in the FAM and HEX channels according to the
assignments in Fig. 5a. The data obtained in Fig. 5c allows the user
to compute the percent of adapter-dimers (HEX) compared to the
total library molecules (FAM) values. Figure 6 demonstrates the
power of ddPCR library QC by quantitating 12 different sequenc-
ing libraries, pooling them together based on the results at equi-
molar concentrations, and sequencing them to show excellent
concordance between ddPCR balancing and the number of reads
obtained from each library on the illumina sequencer.
5 Conclusion
Digital PCR offers exceptional sensitivity for NGS library quantifi-

cation and can improve sequencing results by providing precise
measurements of library loaded on to the sequencing surface.
Digital PCR also offers unique capabilities such as library QC in
the same measurement which give additional valuable information
such as a relative library insert sizing visualization and library qual-
ity. By using digital PCR to evaluate NGS libraries, it is possible to
measure such quality metrics as adapter dimers in a library and
overall library construction quality. Undoubtedly as digital PCR
applications expand, the synergy between digital PCR and NGS will
likewise grow, especially in the areas of rare species quantification
and verification, and absolute counting.
References
1. Rothberg JM et al (2011) An integrated semi- 3. Bentley DR et al (2008) Accurate whole human
conductor device enabling non-optical genome genome sequencing using reversible terminator
sequencing. Nature 475:348–352 chemistry. Nature 456(7218):53–59
2. Margulies M (2005) Genome sequencing in 4. Hindson BJ (2011) High-throughput droplet
microfabricated high-density picolitre reactors. digital PCR system for absolute quantitation of
Nature 437(7057):376–380 DNA copy number. Anal Chem 83
(22):8604–8610
Further Reading
Laurie MT, Bertout JA, Taylor SD, Burton JN, Quail MA, Kozarewa I, Smith F, Scally A, Stephens
Shendure JA, Bielas JH (2013) Simultaneous PJ, Durbin R, Swerdlow H, Turner DJ (2008)
digital quantification and fluorescence-based A large genome center’s improvements to the
size characterization of massively parallel Illumina sequencing system. Nat Methods
sequencing libraries. BioTechniques 55 5:1005–1010
(2):61–67 Sequencing Library qPCR Quantification Guide
Buehler B, Hogrefe HH, Scott G, Ravi H, Pabón- Catalog # SY-930-1010, Part #11322363 Rev.
Peña C, O’Brien S, Formosa R, Happe S (2010) C (2011). Illumina
Rapid quantification of DNA libraries for next- White RA III, Blainey PC, Fan HC, Quake SR
generation sequencing. Methods 50:S15–S18 (2009) Digital PCR provides sensitive and abso-
QPCR NGS Library Quantification Kit (illumina lute calibration for high throughput sequenc-
GA), Part Number G4880A, Publication num- ing. BMC Genomics 10:116–128
ber 5990–3065. (2012). Agilent
Chapter 28
Phasing DNA Markers Using Digital PCR

John Regan and George Karlin-Neumann
Abstract
Besides quantifying the absolute number of copies of known DNA targets, digital PCR can also be used to
assess whether two nonpolymorphic gene sequences or two heterozygous markers reside on the same DNA
molecule (i.e., are physically linked). Some useful linkage applications include: phasing variants to define a
haplotype; genotyping of inversions; determining the presence of multimarker pathogenic bacteria in a
metagenomic sample; and assessing DNA integrity. This chapter describes an efficient and cost-effective
method for analyzing linkage of any two genetic sequences up to at least 200 Kb apart, including phasing of
heterozygous markers such as that which occur abundantly in the cystic fibrosis transmembrane conduc-
tance regulator (CFTR) gene.
Key words Phasing, Linkage, Haplotype, Compound heterozygote, Complex alleles, Co-localiza-
tion, Genotyping inversions, DNA integrity, DNA sizing, Large DNA isolation, Digital PCR, Droplet
Digital PCR
1 Introduction
A number of biological and clinical questions require not only

knowing whether given sequences are present in a sample—some-
thing capably done by digital PCR—but also whether these
sequences (or markers) exist on the same chromosome or homo-
logue, or how close they are to one another in a chromosomal
region. Some examples include: in clinical microbiology, determin-
ing whether multimarker pathogenic bacteria such as Shiga toxin-
producing Escherichia coli or methicillin-resistant Staphylococcus
aureus are present [1, 2]; for personalized medicine, deciding
whether an advanced-stage non-small-cell lung carcinoma harbor-
ing both T790M and C797S mutations in the EGFR oncogene, is
likely to be responsive (if they are in trans) or insensitive (if they are
in cis) to a combination of tyrosine kinase inhibitors (TKIs) [3]; and
for diagnosing cystic fibrosis in a patient with two sequenced muta-
tions in their CFTR gene, determining whether the patient has two
pathogenic CFTR alleles (i.e., a compound heterozygote, where
they are in trans) and thus will have the disease, or whether they
489
490 John Regan and George Karlin-Neumann
have one pathogenic allele harboring both mutations (a complex

allele, where they are in cis) and one wild-type allele, and will likely
not have the disease [4]. Besides the last two examples, assessment
of phasing also facilitates the investigation of allelic imbalance for
understanding the transcriptional basis of various phenotypes (See
chapter 23 entitled “Using droplet digital PCR to analyze allele-
specific RNA expression” by Kamataki et al.), and enables the
verification and genotyping of chromosomal inversions which
may underlie both disease and nondisease phenotypes [5]. Lastly,
digital PCR has been used to assess the intactness of a DNA sample
to determine its fitness for a subsequent analysis, such as phasing
itself [4] or genotyping [6].
An underappreciated feature of digital PCR is that the very act
of partitioning a sample into many small microreactions to make
digital copy number measurements, also enables the determination
of whether two targets being queried are physically linked (i.e.,
found on the same molecule or chromosome). For diploid organ-
isms, markers (or targets) can be either homozygous or heterozy-
gous in nature. In both cases, physical linkage can be established if
the DNA sample(s) in question has sufficiently large molecules to
contain both marker sequences. If both markers are heterozygous
(e.g., Aa and Bb), then in principle it is possible to phase the
markers using digital PCR, that is, to determine whether the sample
is AB/ab or Ab/aB (the underline denotes markers are on the same
chromosome). Prior to digital PCR, obtaining linkage or phase
information was extremely difficult. Commonly used approaches
including single-molecule dilution (SMD) followed by amplifica-
tion with either gel analysis or sequencing [7–9], long-range PCR
[10], linking-emulsion PCR [11, 12], and cloning strategies [13]
are either very laborious and/or greatly limited in their ability to
phase distant markers. As a result, studies that required long-range
phase information (5–200 kb) were very hard to carry out.
The advent of two-color commercial digital PCR instruments
with partitions created by either fixed nanochambers or droplets
has greatly simplified carrying out linkage experiments. By labeling
each of the two markers in question with distinguishable fluoro-
phores, digital PCR systems are able to determine linkage status
since linked markers increase the number of double-positive partitions
that contain both markers of interest above what would be expected
from random co-localization of two unlinked markers. Figure 1
provides an overview of how digital PCR quantifies the presence
of linked markers. This inference can be tested by appropriately
digesting the DNA sample with a restriction enzyme to separate the
two markers into now unlinked DNA molecules.
The first utilization of a commercial digital PCR system to
determine haplotype was recently published [4]. The authors
focused on phasing common mutations in the cystic fibrosis gene
(CFTR). Their work demonstrated phasing of markers between
Phasing Markers with Digital PCR 491
a Unlinked alleles distribute and co-localize randomly across droplets
Emulsify PCR
G A
FAM
C T
HEX
b Statistical excess of double-positive droplets when the alleles are physically linked
Emulsify PCR
G A
FAM
C T
HEX
Fig. 1 Measuring linked species. Marker-specific fluorescent probes (FAM: blue, HEX: green) are used to
detect the markers of interest in undigested genomic DNA. Following endpoint PCR, droplets are positive for
one fluorophore (blue or green), both fluorophores (orange), or neither (black). (a) Markers on DNA from
different chromosomes (or DNA molecules) partition independently into droplets (including some that
randomly colocalize into the same droplets and become double positive droplets). (b) Markers on the same
unbroken strand systematically colocalize into the same droplets because they are physically linked, resulting
in an overabundance of double-positive droplets above what would be expected from chance co-localization
of unlinked markers
110 kb and 116 kb apart with 100% concordance to inheritance

information. The method, called “Drop-Phase,” requires less than
4 h, is scalable to hundreds of samples, is effective at genomic
distances up to ~200 kb, and is low cost.
It is important to stress that the distance over which linkage
measurements can be made is a function of the intactness of DNA
being used. Published work has shown there to be a significant
difference in the ability of some DNA extraction kits to preserve the
intactness of DNA during the purification process [4]. Currently, it
is possible to establish linkage from cultured cells out to ~200 kb.
Other extraction kits or new advances in sample preparation could
extend the distance at which linkage between two markers can be
established. However, at this time, it is unclear if mega-based length
DNA will require larger partition sizes to enable very long-distance
linkage measurements.
This chapter presents two protocols to assess linkage. The first,
the standard linkage protocol, is used to determine whether two
nonpolymorphic sequences are linked. This protocol is independent
of zygosity. The second, the phasing protocol, is particularly aimed
at diploid organisms and is used to phase heterozygous markers.
2 Materials
The protocols below specifically reference Bio-Rad’s Droplet Digi-

tal™ (ddPCR™) system. However, the principles presented are
likely applicable to other commercial and noncommercial digital
PCR systems.
2.1 Equipment 1. Droplet Digital PCR System composed of a droplet generator

(manual or AutoDG) and droplet reader (QX100 or QX200)
(Bio-Rad).
2. PX1™ PCR Plate Sealer (Bio-Rad).
3. C1000 Touch™ Thermal Cycler (Bio-Rad).
2.2 DNA Samples 1. DNA samples should be extracted with a sample preparation
chemistry that is known to sufficiently preserve the intactness
of the DNA and used as soon as possible, preferably before
freezing the DNA (see Notes 1–3).
2. Optional: High molecular weight intact genomic control DNA
(see Note 4).
2.3 Reagents 1. ddPCR SuperMix for Probes (no dUTP) (Bio-Rad).

and Consumables 2. DNA Suspension Buffer/10 mM Tris, 0.1mM EDTA, pH 8.0
(Teknova).
3. Nuclease-free water (Ambion).
4. 20 FAM- and HEX-labeled TaqMan probe assays (Final 1
concentration of each primer is 900 nM, and of each probe is
250 nM).
6. Appropriate restriction enzyme for confirming linkage (choice
depends on target sequences and cuts the sample DNA
between the two markers being evaluated but not within the
assay amplicons themselves).
7. DG8™ Droplet Generatior Cartridges and Gaskets (for man-
ual droplet generation) or DG32™ Automated Droplet Gen-
erator Cartridges (both from Bio-Rad).
8. Rainin normal-bore P20 and P200 filter pipette tips (Note:
Lower quality pipette tips may result in shredding of droplets).
9. ddPCR 96-well semiskirted plates (Bio-Rad, Cat.
#12001925). These plates are specially formulated to maxi-
mize stability and recovery of droplets and their use is strongly
recommended.
10. PCR Plate Heat Seal, foil, pierceable (Bio-Rad, #1814040).
3 Methods
Two protocols are provided: a standard linkage protocol where

phase is not applicable and a phasing protocol for determining the
cis–trans relationship between two heterozygous markers (Fig. 2).
The standard linkage protocol is used to establish whether two
DNA sequences are linked. It can be applied to any markers regard-
less of ploidy or zygosity. In contrast, the phasing protocol, is most
frequently applied to diploid organisms to determining whether
linkage exists between two heterozygous sites, but in principle can
be applied to organisms with higher order zygosities.
3.1 Preparation Choosing the correct extraction methodology and sample type is
of Genomic DNA critical (see Note 5). When starting from cultured cells or cells from
whole blood, silica membrane methods generally only allow for
linkage measurements out to 30 kb. In contrast, Life Technologies’
PrepFiler Forensic DNA Extraction Kit regularly yields DNA that is
sufficiently intact to permit linkage measurements out to 200 kb.
Others have reported success with a phenol-chloroform protocol
(personal communication).
For the PrepFiler kit:
1. Suspend the cells at ~7 106 cells/mL in 1 PBS and remove
40 μL for extraction.
2. Follow the manufacturer’s recommendations, except reduce
the lysis step down to 10 min at 70 C with no shaking.
3. In addition, use gentle end-over-end inversion rather than vor-
texing, except after the addition of 180 μL isopropyl alcohol,
when the lowest rpm setting of a vortex mixer should be used.
4. Generally, minimize the speed and duration of the centrifuga-
tion steps.
5. Elute the extracted DNA in 50 μL of kit-provided elution
buffer.
a b
Standard Linkage Protocol for Phasing Protocol for
Homozygous Markers Heterozygous Alleles
Paternal A B A B
chromosome
Maternal A B a b
chromosome
Phase information Phase information is

not applicable present
Fig. 2 Use cases for linkage studies. (a) Standard linkage protocol is used when querying two loci that are
homozygous in nature. (b) Phasing protocol is used when querying two loci that are heterozygous in nature
3.2 Standard The standard linkage protocol should be used when verifying
Linkage Protocol whether two genes are present in the same organism (i.e., pathogen
virulence testing), to genotype inversions [5], and to test whether
the DNA sample is intact enough (see Note 1) to permit successful
phasing determination (Fig. 3).
3.2.1 Design/Selection For standard linkage experiments, a pair of conventional

and Optimization FAM-labeled and HEX-labeled TaqMan probe assays are combined
of Standard Linkage in a single tube. Suitable wet-lab validated assays for this purpose
Assays can be obtained through the Bio-Rad assay design website
(https://www.bio-rad.com/digital-assays/#/), or custom designs
can be provided from raw sequence at both this site and that of IDT
(http://www.idtdna.com/calc/analyzer).(See Subheading 3.3.1
below for additional design information.) These assays must
a FAM-Assay HEX-Assay
A B C
Paternal
chromosome
A B C
Maternal
chromosome
Chr 7 Chr15
b
FAM-Assay HEX-Assay
A B C
Paternal
chromosome
A B C
Maternal
chromosome
Chr 7 Chr15
Fig. 3 Duplex assays for standard linkage protocol. (a) The test duplex includes an FAM-labeled assay
targeting the “A” locus and a HEX-labeled assay targeting the “B” locus, which is presumed to be on the same
chromosome as the “A” locus and close enough such that sufficient intact DNA remains for a positive linkage
measurement. (b) The control duplex includes an FAM-labeled assay targeting the “A” locus and a
HEX-labeled assay targeting the “C” locus, which is known to be on a different chromosome than the “A”
locus. This control duplex provides a realistic background estimate of linkage caused by nonrandomly
partitioned DNA due to entanglement of large DNA molecules (as occurs in the test reaction). This approach
is preferred over use of a restriction enzyme to separate “A” from “B” in the test duplex reaction because
digested DNA more easily distributes randomly into droplets than undigested DNA and may underestimate the
amount of background linkage
produce sufficient fluorescence amplitude separation between posi-

tive and negative droplet clusters for unambiguous assignment of
droplets into clusters and provide specific target detection. Gener-
ally, standard ddPCR thermal cycling parameters are used with
annealing temperatures in the 55–60 C range. Optimal cycling
conditions for the assays should be determined in advance by
running a thermal gradient experiment from ~5 C below to
~5 C above the predicted Tm (melting temperature) of the assay
primers.
3.2.2 Assembling 1. Create 3 technical replicates for both test and control ddPCR
Standard Linkage reactions, by assembling the triplicate reaction mixtures
Reactions detailed in Tables 1a and 1b, respectively.
2. Vortex both 60 μL reaction mixtures, and then briefly centri-
fuge the contents.
Table 1a
Assay components for test reactions
Test reaction component Volume (μL)

ddPCR SuperMix for Probes (no dUTP) 37.5
20 FAM Assay for Marker A 3.75
20 HEX Assay for Marker B 3.75
Nuclease Free Water 15
Total 60
Table 1b
Assay components for control reactions (markers known not to be linked)
Control reaction component Volume (μL)

20 FAM Assay for Marker A 3.75
*
20 HEX Assay for Marker C 3.75
Total 60
*Marker C should reside on a different chromosome from Marker A. If the organism is a haploid organism (e.g.,
bacterium), instead of assembling a different duplex assay for the control, maintain the original A and B duplex, but
instead add a restriction enzyme at 10 U/60 μL reaction to specifically cut the DNA between A and B (see Note 6).
3. Dispense 18 μL of each reaction mixture into 3 wells of a

96-well plate (6 wells total).
4. Individually add 4.5 μL of undigested DNA (5 ng/μL) to each
well using a P-20 pipette tip (see Note 7). Do not pipet up and
down with the P-20 pipette tip (see Note 8).
5. Using a normal-bore pipette tip and a P-200 with the volume
set to 18 μL, slowly pipet up and down 20 times, making sure
not to create bubbles (see Note 9). If bubbles are created, spin
down the plate at 1500 RCF for 1 min.
6. Generate droplets by either:
(a) Adding 20 μL of the ddPCR reaction mixture into a
droplet generation cartridge using a multichannel pipet-
tor and manually generate droplets according to the
manufacturer’s instructions, followed by their careful
transfer to a 96-well PCR plate, using a P-50 multichannel
pipettor. Repeat until all samples have been partitioned
into droplets and transferred to the PCR plate, 8 samples
at a time.
(b) Or by leaving the entire 22.5 μL reaction in the plate,
sealing the plate, and transferring the sample plate to an
AutoDG for automated droplet generation.
(c) Note that regardless of whether droplets are generated by
the manual or automated systems, all 8 wells of a DG chip
must be filled with either sample reaction mix or control
buffer, along with DG oil in the corresponding DG8
wells.
7. Seal the droplet plate using a heat-seal foil and the PX1 heat
sealer set to 180 C for 5 s.
8. Thermal-cycle the droplets using conditions previously deter-
mined to be optimal for all assays on the plate. Standard ddPCR
cycling conditions (95 C for 10 min; 40 cycles of 94 C for
30 s followed by 60 C for 1 min; 98 C for 10 min; 12 C
hold; 2.5 C per cycle ramp rate for all steps) should be fol-
lowed, except for using the optimal anneal/extension temper-
ature if it differs from 60 C.
9. Read the droplets on the QX100 or QX200 Droplet Reader.
3.2.3 Analyzing the Data 1. In QuantaSoft Analysis software, manually examine each well
for Linkage of Markers for droplet quality in 2D fluorescent amplitude droplet plots.
Discard any wells that have an abundance of droplets (i.e.,
>5%) that align on a diagonal running through the double
negative cluster, which is indicative of sheared droplets.
(If this occurs excessively, pay attention to pipetting the dro-
plets more slowly when transferring them to the PCR plate and
use high quality pipette tips which do not shed plastic
Fig. 4 Screen captures from a QuantaSoft table showing the total concentration of A, B, and C markers,
regardless of linkage status, and the “Linkage” column, which displays only the concentration of linked
molecules (AB and AC) in copies/μL
particulates into DNA preps and ddPCR reactions.) Select the

remaining wells and classify the droplets while viewing the 2D
fluorescent amplitude plot.
2. Using the data available in the QX100/200 QuantaSoft Anal-
ysis table (Fig. 4), calculate the percentage of linked AB mole-
cules for the wells containing assays for both A and B using
Eq. 1. Handle each well separately, rather than merging the
wells, to protect against the possibility that one poorly mixed
well skews the analysis. If the replicates are not within 20% of
each other for the percentage of linked molecules, then the
mixing of the DNA sample into the reactions was not sufficient
and caution should be exercised when interpreting the
findings.

2λAB
% Linked AB ¼ 100 ð1Þ
ðλA þ λB Þ
l Where λAB is the value found in the “Linkage” column,

which is expressed in copies/μL.
l Where λA is the value taken from the “Conc (copies/μL)”
column for the concentration of Marker A. λA is a summa-
tion of the linked and unlinked A molecules.
l Where λB is the value taken from the “Conc (copies/μL)”
column for the concentration of Marker B. λB is a summa-
tion of the linked and unlinked B molecules.
Using the numbers shown for Duplex 4, 56:3% of
#1 in Fig.
l
A and B molecules are linked : AB ¼ 2∗58:8

ð105þ104Þ 100.
3. Using the data available in the QuantaSoft table, calculate the

percentage of linked AC molecules for the AC duplexes using
Eq. 2.

2λAC
% Linked AC ¼ 100 ð2Þ
ðλA þ λC Þ
l Where λAC is the value found in the “Linkage” column,

l Where λA is the value taken from the “Conc(copies/μL)”
column for the concentration of marker A. λA is a summa-
tion of the linked and unlinked A molecules.
l Where λC is the value taken from the “Conc (copies/μL)”
column for the concentration of marker C, here representing
the unlinked control locus. λC is a summation of the linked
and unlinked C molecules.
4, 1:3% of
Using the numbers shown for Duplex#2 in Fig.
l
A and C molecules are linked : AC ¼ 2∗1:5

ð116þ108Þ 100:
In this case, the 1.3% of linked AC molecules is artificial
linkage due to incomplete mixing (see Notes 8 and 9).
4. For the three test and three control reactions, calculate the
combined mean and standard deviation for each.
5. Multiply each standard deviation by 2 and compare the lower
confidence bar of the test samples with the upper confidence
bar of the control samples. If these bars overlap, linkage cannot
be claimed. If they do not overlap, then linkage can be claimed
at >95% confidence.
3.3 Phasing Protocol This phasing protocol is designed to determine if alleles “a” and
“b” are on the same chromosome. Across two heterozygous sites,
there are four unique FAM-HEX duplexes that can be assembled
to determine linkage (Fig. 5). A minimum of two of these
duplexes needs to be assembled, tested, and compared as one
serves as a test duplex and the other a control duplex for artificial
linkage. If the percentage of linked molecules is expected to be less
than 25%, all four duplexes should be assembled, tested and
compared.
3.3.1 Design/Selection Assay design and data interpretation for phasing experiments is
and Optimization slightly more complicated than for the Standard Linkage Protocol
of Phasing Assays due to the heterozygous nature of the targeted loci. This can result
in cross-reactivity of an assay with the nontargeted allele as well as
competition for reagents in droplets containing both alleles, and
can produce up to 16 different clusters on a 2D fluorescent ampli-
tude plot (Fig. 6).
The droplets in the plot shown in Fig. 6 are easily classified—as
indicated in the figure, but it is not uncommon to observe plots
where the clusters are not as orthogonal due to cross-reactivity, and
even experts in digital PCR can easily misclassify the clusters.
a c
FAM-Assay HEX-Assay
A B A B
Paternal Paternal
chromosome chromosome
FAM-Assay HEX-Assay
a b a b
Maternal Maternal
Chr 7 Chr 7
b d
FAM-Assay HEX-Assay
A B A B
Paternal Paternal
HEX-Assay FAM-Assay
a b a b
Maternal Maternal
Chr 7 Chr 7
Fig. 5 Duplexes assembled in phasing protocol. (a) Assays targeting the “A” and “B” alleles are combined.
(b) Assays targeting the “A” and “b” alleles are combined. (c) Assays targeting the “a” and “b” alleles are
combined. (d) Assays targeting the “a” and “B” alleles are combined
Accordingly, phasing assays must not only have sufficient amplitude

for easy positive droplet classification, but steps should be taken to
minimize cross-reactivity. At a minimum, the probes should either
have a Tm ~1–2 C above the 55 C melting temperature of the
assay primers or should be designed with a Tm enhancer, such as
linked nucleic acids (LNAs from Exiqon) or a minor groove binder
(MGB from ThermoFisher Scientific). In addition, adding a dark
competitor probe at 250 nM final concentration (equivalent to the
concentration of the probe with which it is competing) is strongly
recommended (see Figs. 7 and 8, and Note 10).
For common clinically relevant variants, Bio-Rad offers many
wet-lab validated and in silico designed assays for digital PCR. In
addition, their website allows for users to input COSMIC ID, gene
symbol and/or amino acid change, and/or nucleotide change. Alter-
natively, raw sequence can be submitted to their website for custom
assay design (https://www.bio-rad.com/digital-assays/#/). If order-
ing from this site, it is still necessary to design a dark competitor probe
for the nontargeted alleles. To do so:
1. Go to IDT’s oligo analyzer (http://www.idtdna.com/calc/
analyzer) and input ~27 bases of the nontargeted sequence
with the base of interest centrally located.
2. Adjust the default parameters such that Mg++ is 3.8 mM and
the dNTPs concentration is 0.8 mM.
Fig. 6 Droplet cluster identification and classification in the presence of assay competition and cross-reacting
fluorescent probes (i.e., NO dark competitor probes present). Major clusters (designated as “✚” in the figure)
are comprised of single- and double-positive droplets that contain the FAM (A)- and/or HEX (B)-targeted alleles
(without the nontargeted “a” and “b” alleles), as well as the double-negative clusters (,). Additionally, at
high enough DNA loads per reaction, some droplets contain both targeted and nontargeted alleles (A and a or
B and b). When both alleles are present in the same droplet (alleles shown in bold on the plot), the two
templates compete for the same primer pair resulting in droplets with reduced fluorescent amplitude since
only half of the end-product is specific to the included probe (and this is not mitigated by the presence of dark
competitor
J probes). Droplets that lack the targeted allele, but contain the nontargeted allele (denoted by
the symbol), experience probe cross-reactivity that generates as many as 7 additional clusters that are
offset from the four major clusters (✚). If dark competitor probes had been added to the reaction these offset
clusters are often driven toward the axes to overlap with the main FAM and HEX clusters, including the
competition clusters (see Fig. 7)
3. Click “Analyze” and examine the Melt Temperature. Ideally,

design the dark competitor to have a melting temperature
which is equivalent to the Tm of the fluorescent probe with
which it is competing (~57 C; see above and Note 10).
4. If the melting temperature is too high, remove bases equally
from the ends such that the allele of interest remains centrally
located within the competitor probe.
5. After settling on the design, the probe should be ordered with a
30 BHQ quencher (/3BHQ_1/), but no fluorophore.
Fig. 7 Influence of dark competitor probes (see Note 10). Phasing experiments without competitor probes
(as in this 2D plot) can have as many as 16 clusters both because the probes for each targeted sequence will
often cross-react with the nontargeted allele when present and because the presence of both alleles in a
single droplet will result in competition that lowers the targeted alleles endpoint yield and thus fluorescence
amplitude creating additional subclusters. Adding dark competitor probes into the reaction drives the cross-
reacting clusters surrounded by a blue box, (those clusters having targets cross-reacting with the “A” probe)
and a green box (those clusters having targets cross-reacting with the “B” probe) into the conventional
clusters that lie closest to the HEX and FAM axes, respectively. The seven clusters denoted with an asterisk
(*) are present due to PCR competition, since a single primer pair amplifies both targeted and nontargeted
alleles and thus results in only about ½ of the final PCR product in these droplets with commensurately
reduced hydrolysis of the corresponding fluorescent probe. To reduce the number of droplets in these
competition-based clusters, load less DNA into the reaction
6. Spike this dark competitor probe into the fluorescent assay

targeting the allele of interest such that the final concentration
in the ddPCR reaction mixture is 250 nM (see Note 10).
3.3.2 Assembling 1. Plan to create three technical replicates as in the Standard

Phasing Protocol Reactions Linkage Protocol above.
2. If a high degree of linkage is expected for the alleles of interest
(e.g., >25%), only two of the four reactions (e.g., Tables 2a
and 2b) are required. However, if <25% of the alleles are
expected to be linked, then all four duplexes should be run
(see Tables 2a, 2b, 2c & 2d), as indicated in Fig. 5a–d, to
mitigate the risk that artificial linkage confounds the results
(see Note 9). In such cases, two of the duplexes should have
measurable linkage for the alleles in question and the other two
should not.
Fig. 8 Influence of adding dark competitor probes to phasing assay. (a) Phasing assay without the addition of
dark competitor probes. (b) Phasing assay with the addition of dark competitor probes. Notice that the dark
competitor probes do not eliminate the additional minor clusters resulting from the simultaneous presence of
both targeted and nontargeted alleles in droplets (i.e., the A,a; B,b; A,B,b; A,a,B and A,a,B,b clusters seen in
Fig. 6 still remain)
Table 2a
Assay components for AB reactions
AB test reaction component Volume (μL)

20 FAM Assay for Allele A 3.75
20 HEX Assay for Allele B 3.75
Total 60
Table 2b
Assay components for Ab reactions
Ab test reaction component Volume (μL)

20 FAM Assay for Allele A 3.75
20 HEX Assay for Allele b 3.75
Total 60
Table 2c
Assay components for ab reactions (only required if low % of linkage
expected)
ab test reaction component Volume (μL)

20 FAM Assay for Allele a 3.75
20 HEX Assay for Allele b 3.75
Total 60
Table 2d
Assay components for aB reactions (only required if low % of linkage
expected)
aB test reaction component Volume (μL)

20 FAM Assay for Allele a 3.75
20 HEX Assay for Allele B 3.75
Total 60
3. Vortex the 60 μL reaction mixtures, and then briefly centrifuge

the contents.
4. Dispense 18 μL of each reaction mixture into 3 wells of a
96-well plate (12 wells total, if running all 4 duplexes,
6 wells, if running only 2 of the duplexes).
5. Individually add 4.5 μL of undigested DNA (10 ng/μL) to

each well using a P-20 pipette tip (see Note 7). Do not pipet up
and down with the P-20 pipette tip (see Note 8).
6. Using a normal-bore pipette tip and a P-200 with the volume
set to 18 μL, slowly pipet up and down 20 times, making sure
not to create bubbles (see Note 9). If bubbles are created, spin
down the plate at 1500 RCF for 1 min.
7. Generate droplets by either:
(a) Transferring 20 μL of the reaction mixture into a droplet
generation cartridge and manually generating droplets
according to the manufacturer’s instructions, followed
by their careful transfer to a 96-well PCR plate, using a
P-50 multichannel pipettor. Repeat until all samples have
been partitioned into droplets and transferred to the PCR
plate, 8 samples at a time.
(b) Or by leaving the entire 22.5 μL reaction in the plate,
sealing the plate, and transferring it to an AutoDG for
automated droplet generation.
8. Seal the droplet plate using a heat-seal foil and the PX1 heat
sealer set to 180 C for 5 s.
9. Thermal-cycle the droplets using conditions previously deter-
mined to be optimal for all assays on the plate. Standard ddPCR
cycling conditions (95 C for 10 min; 40 cycles of 94 C for
30 s followed by 55 C for 1 min; 98 C for 10 min; 12 C
hold; 2.5 C per cycle ramp rate for all steps) should be
followed, except for using the optimal anneal/extension tem-
perature if it differs from 55 C.
10. Read the droplets on the QX100 or QX200 Droplet Reader.
3.3.3 Analyzing the Data 1. Barring significant differences between the three wells per
for Linkage of Markers duplex (e.g., obvious shearing of the droplets in one well),
select all three wells and classify the droplets while viewing
the 2D fluorescent amplitude plot. (See Subheading 3.2.3,
step 1, for further guidance on well analysis). In phasing
experiments, PCR competition between alleles and probe
cross-reactivity can make classifying droplets difficult. For guid-
ance on understanding these principles and their influence on
droplet cluster location within the 2D fluorescent plot see
Figs. 7 and 9.
2. If running the AB, Ab, ab, and aB duplexes, use the data
available in the QX100/200 QuantaSoft Analysis table
(as seen in Fig. 10), to calculate the percentage of linked
molecules.
a
(detects A allele strongly, a allele weakly) b
(detects A allele strongly, a allele weakly)

(AB, A+B)
(Ab, A+b)
FAM Fluorescence
FAM Fluorescence
(ab, a+b) (aB, a+B)
HEX Fluorescence HEX Fluorescence

(detects B allele strongly, b allele weakly) (detects B allele strongly, b allele weakly)
Fig. 9 Shifting cluster location. For linked targets, the location of the cluster(s) with an excess of droplets shifts
depending on whether the included duplex is a perfect match for the linked alleles or if one of the assays only
cross-reacts with the linked allele. (a) When the sample (AB, ab) contains linked alleles that are specifically
targeted by the included assays (FAM-labeled assay detects “A” and HEX-labeled assay detects “B”), then
there is an overabundance of the major clusters denoted with arrowheads. (b) When a different sample of the
opposite genotype (Ab, aB) is queried using the same duplex without dark competitor probes, the cluster with
excess double-positive droplets shifts to the ones where the nontargeted allele is detected through cross-
reactivity (denoted with arrowheads). If dark competitor probes were used, these cross-reactive droplets
would ideally merge with the conventional major FAM (A)-only and HEX (B)-only clusters in Fig. 6, effectively
causing an overabundance of single-positive droplets and an underabundance of double-positive droplets and
the expected number of double-positive droplets containing the two targeted alleles (all of the A,B-containing
droplets in the upper right of Fig. 6) would be no greater than by chance alone
Fig. 10 Screen captures from a QuantaSoft table showing the total concentration of A, B, a, and b alleles,
regardless of linkage status, and the “Linkage” column, which displays only the concentration of linked
molecules (AB, Ab, ab, and aB) in copies/μL
3. For calculating the percentage of linked AB molecules use Eq. 3.

2λAB
% Linked AB ¼ 100 ð3Þ
ðλA þ λB Þ
l Where λAb is the value found in the “Linkage” column,
l Where λA is the value taken from the “Conc(copies/μL)”
column for the concentration of allele A. λA is a summation
of the linked and unlinked A molecules.
l Where λB is the value taken from the “Conc(copies/μL)”
column for the concentration of allele B. λB is a summation
of the linked and unlinked B molecules.
l Using the numbers shown for Duplex #1 in Fig. 10,
57:8%
of A and B molecules are linked : AB ¼ 2∗60:7
ð104þ106Þ 100:
4. For calculating the percentage of linked Ab molecules use Eq. 4.

2λAb
% Linked Ab ¼ 100 ð4Þ
ðλA þ λb Þ
l Where λAb is the value found in the “Linkage” column,

l Where λA is the value taken from the “Conc (copies/μL)”
column for the concentration of allele A. λA is a summation
of the linked and unlinked A molecules.
l Where λb is the value taken from the “Conc (copies/μL)”
column for the concentration of allele b. λb is a summation of
the linked and unlinked b molecules.
0% of A and b molecules are linked : Ab ¼ ð104þ107
2∗0
Þ 100:
In this case, no linkage is found between the A and b
markers (see Notes 8 and 9).
5. For calculating the percentage of linked ab molecules use Eq. 5.

2λab
% Linked ab ¼ 100 ð5Þ
ðλa þ λb Þ
l Where λab is the value found in the “Linkage” column,
l Where λa is the value taken from the “Conc (copies/μL)”
column for the concentration of allele a. λa is a summation
of the linked and unlinked a molecules.
l Where λb is the value taken from the “Conc (copies/μL)”
column for the concentration of allele b. λb is a summation
of the linked and unlinked b molecules.

55:2% of a and b molecules are linked : ab ¼ ð109þ106Þ 100.
2∗59:3
6. For calculating the percentage of linked aB molecules use Eq. 6.

2λaB
%Linked aB ¼ 100 ð6Þ
ðλa þ λB Þ
l Where λaB is the value found in the “Linkage” column,
l Where λa is the value taken from the “Conc (copies/μL)”
column for the concentration of allele a. λa is a summation
of the linked and unlinked a molecules.
l Where λB is the value taken from the “Conc (copies/μL)”
column for the concentration of allele B. λB is a summation
of the linked and unlinked B molecules.
0:87% of a and B molecules are linked aB ¼ ð108þ107Þ 100:
2∗0:93
In this case, the 0.87% of linked aB molecules is artificial

linkage due to incomplete mixing (see Notes 8 and 9).
7. For the AB, Ab, ab, and aB reactions, calculate the mean and
standard deviation across triplicate reactions.
8. Compare the % of linked markers and using the criteria in steps
9 and 10 below, determine which of the following three
options is most likely:
(a) Alleles AB and ab are linked.
(b) Alleles Ab and aB are linked.
(c) The Aa and Bb loci are either on different chromosomes or
the DNA linking the two loci is too fragmented to allow
for phasing of the alleles.
9. In the case where AB and ab are linked, these duplexes should
yield similar percentages of linked alleles. Similarly, if Ab and aB
are linked, these duplexes should yield similar percentages of
linked alleles. In contrast, the other two duplexes serve as
controls and any linkage observed with them is artificial linkage
due to insufficient mixing. It is acceptable to have artificial
linkage in the control wells, so long as sufficient technical
replicates were performed and the test wells have statistically
more linked molecules than the control wells.
10. If the difference in the percent linked between the four
duplexes is small, determine the mean and standard deviation
for the percentage of linked molecules for each of the four
duplexes. For the two most linked (which should both be for
the phased alleles) and two least linked duplexes (which should
both be for the unphased alleles), recalculate the mean and

standard deviation to include the six wells across each of these
higher and lower linkage duplexes, respectively. Next, subtract
two standard deviations from the mean of the higher linkage
duplexes and add two standard deviations to the mean of the
lower linkage duplexes and check to see if the error bars overlap
(i.e., if the first value is greater than the second). If the error
bars overlap, either the alleles are not on the same chromosome
or the DNA sample is too fragmented for making a linkage
determination.
4 Notes
1. When starting from cultured cells, DNA extracted using silica-

based columns (e.g., Qiagen’s DNeasy Blood & Tissue Kit)
and polysaccharide precipitation–based chemistry (e.g., Ther-
moFisher Scientific’s PrepFiler Forensic DNA Extraction Kit)
permit linkage assessment out to ~ 25 kb and ~200 kb, respec-
tively [4]. If working with an untested extraction chemistry,
first perform a “mile marker series” as described in Regan et al.
[4], where a series of duplex assays are assembled, each having
the same HEX-labeled assay specific to an anchor sequence,
and each with a unique FAM-labeled assay specific to a con-
served region of the genome at increasing distances away from
the anchor sequence. For each distance (e.g., 1, 10, 33, 60,
100, 150, 200 kb) separating the mile markers and the anchor
sequence calculate the percentage of linked molecules and fit
these data with an exponential regression line to learn the
theoretical limit in kilobases at which a successful linkage mea-
surement can be made. This information can then be used to
assess whether linkage status between other markers of interest
can be determined with this DNA prep.
2. There may be an upper limit to the length of DNA strands that
are compatible with reliable droplet formation from Bio-Rad’s
QX200 digital PCR system. The QX200 generates droplets
that are ~ 0.85 nL in size; and we believe the upper limit for
DNA strand length in these droplets is ~500 kb. Longer DNA
strands become entangled and cause localized viscosity
changes, which increase droplet polydispersity with a bias
toward larger droplet formation. Excessive DNA input levels
(e.g., >66 ng/well) of highly intact DNA, such as Promega’s
human DNA (Cat # G1471 or G1521), will completely pre-
vent droplet formation. In light of this, when using undigested
DNA, always confirm average droplet counts are in excess of
15,000/well if using QuantaSoft 1.7 or in excess of 12,000/
well if using an older version of QuantaSoft. If working with
DNA that regularly fails to form droplets, mix the sample more
vigorously than would otherwise be recommended to slightly
shear the DNA before going into droplet generation.
3. After the DNA is extracted, immediately perform the linkage
tests, as storing the DNA for extended periods will cause
additional DNA fragmentation. If the experiment is to be
performed within a week’s time, leave the sample at 4 C. If
the sample is likely not to be processed for more than a week,
freezing is recommended, but avoid repetitive freeze-thaw
cycles. Always store DNA for linkage analysis in a buffered
solution to minimize hyrdrolysis (e.g., Teknova’s DNA suspen-
sion buffer: 10 mM Tris, 0.1 mM EDTA, pH 8.0).
4. In evaluating different sample preparation chemistries for their
ability to purify intact DNA, as of this writing, Promega’s
human DNA (https://www.promega.com/products/
biochemicals-and-labware/nucleic-acids/genomic-dna/) may
serve as a good control DNA. Promega DNA is highly intact
and accordingly very viscous and enables linkage of two mar-
kers separated by at least 200 Kb to be detected by the proce-
dure described in this chapter. In the absence of restriction
enzyme digestion or aggressive pipetting, it does not allow
droplet formation when included in ddPCR reactions at
>66 ng/well (>1 copy/droplet).
5. DNA extracted from plasma (cfDNA) or FFPE is generally too
fragmented to justify a linkage experiment that includes two
different assays. Instead, a single primer-pair assay (<500 bp in
length) that contains FAM- and HEX- labeled probes for the
markers of interest can often be used to establish linkage using
the protocol described in this chapter.
6. A temptation in linkage experiments is to include a restriction
enzyme in the control wells, where the restriction enzyme
digests the DNA linking the two markers. However, digested
DNA mixes more easily to homogeneity than undigested DNA
where the DNA is entangled. To avoid incorrectly genotyping
samples due to artificial linkage, when possible, use undigested
DNA for the controls in linkage experiments. This requires
varying the assays used between the test and control wells. An
exception to this recommendation is for haploid genomes and
for triplex inversion genotyping experiments [5] where the test
and control assays are included in the same well.
7. When performing a linkage experiment, the amount of DNA
put into the reaction to minimize statistical error is different
from the amount of DNA (120 ng/well or ~1.6 copies/drop-
let (cpd)) needed for the most precise quantification given
Bio-Rad’s QX200 partition volume of 0.85 nL. Instead, for
linkage applications, it is a balance between loading sufficient
DNA to maximize the number of linked molecules scored in a

well (appearing as double-positive droplets), while minimizing
the number of double-positive droplets due to randomly colo-
calized targets. Our modeling suggests that the optimal DNA
load is ~0.2 target copies per droplet (cpd). For human geno-
mic DNA, we typically add ~22.5 ng of highly intact human
gDNA into a 22.5 μL reaction, allowing for as much as 50%
overestimation of the amplifiable DNA amount by UV mea-
surement. If performing a phasing experiment, where only a
single chromosome is targeted, this number should be doubled
to ~40 ng/reaction. If the sample DNA is of generally smaller
size distribution but still has some molecules large enough to
contain both target sequences, it may be necessary to increase
the number of wells run to achieve statistical confidence in a
linkage result for that sample.
8. Undermixing the reaction mixtures can result in significant
“artificial” linkage measurements. To protect against this risk,
for the technical replicates, individually add and mix the DNA
into wells which already contain Supermix and assay. This
approach is preferable to splitting a large reaction mixture
already containing DNA between replicate wells, as this may
mask incomplete mixing and give inaccurate linkage
measurements.
9. Insufficiently mixed samples result in underquantification of the
sample, as the DNA is not randomly distributed into the dro-
plets, resulting in a deficit of droplets containing the target
DNA. Another way to think about this is that undermixed
have more double-negative and more double-positive droplets
than in a properly mixed sample. The higher number of double-
negative droplets translates into a lower concentration measure-
ment for both markers, while the higher number of double-
positive droplets translates into a statistically significant linkage
measurement, even though the markers may not be linked (i.e.,
artificial linkage through entanglement of DNA molecules).
Unpublished data suggests that when using 80% stroke volume,
it is necessary to pipet up and down 10 times for digested and
20 times for undigested samples to mix them to homogeneity.
These same data found that wide-bore pipette tips require even
more pipette strokes to achieve homogeneity. As a result, wide-
bore pipette tips are not recommended unless the sample is
pipetted up and down at least 40 times to produce sufficient
mixing of DNA in the reaction; failure to mix sufficiently can
lead to artificial linkage measurements as high as 40%.
10. Standard assays for phasing contain 900 nM primers, 250 nM
each fluorescent probe, and 250 nM dark competitor probe.
Dark competitor probes are designed to bind to the nontar-
geted allele to reduce probe cross-reactivity (e.g., dark
competitor is specific to Marker a where the fluorescent probe

is specific to Marker A). Dark competitor probes have no 50
fluorophore but instead a 30 quencher which should match the
quencher being used for the corresponding fluorescent probe
(e.g., an Iowa Black quencher (IABkFQ) from IDT (Coralville,
Iowa, USA)). The melting temperature of the dark competitor
probe should the same as the Tm of the fluorescent probe with
which it is competing.
Acknowledgments
Steven McCarroll, Nolan Kamitaki, Tina Legler, Samuel Maars,

Mario Caceres, and Svilen Tzonev, Niels Klitgord, and Samantha
Cooper contributed to the knowledge presented in these protocols.
References
1. Wolk DM, Struelens MJ, Pancholi P, Davis T, 6. Didelot A, Kotsopoulos SK, Lupo A, Pekin D,
Della-Latta P, Fuller D, Picton E, Li X, Atochin I, Srinivasan P, Zhong Q,
Dickenson R, Denis O, Johnson D, Chapin K Olson J, Link D, Laurent-Puig P, Blons H,
(2009) Rapid detection of Staphylococcus Hutchison JB, Taly V (2013) Multiplex
Aureus and methicillin-resistant S. Aureus picoliter-droplet digital PCR for quantitative
(MRSA) in wound specimens and blood cul- assessment of DNA integrity in clinical sam-
tures: multicenter preclinical evaluation of the ples. Clin Chem 59(5):815–823. https://doi.
Cepheid Xpert MRSA/SA skin and soft tissue org/10.1373/clinchem.2012.193409
and blood culture assays. J Clin Microbiol 47 7. Chen N, Schrijver I (2011) Allelic discrimina-
(3):823–826. https://doi.org/10.1128/ tion of cis-trans relationships by digital poly-
JCM.01884-08 merase chain reaction: GJB2 (p.V27I/p.
2. Muniesa M, Hammerl JA, Hertwig S, Appel B, E114G) and CFTR (p.R117H/5T). Genetics
Brussow H (2012) Shiga toxin-producing in medicine: official journal of the American
Escherichia Coli O104:H4: a new challenge College of Medical Genetics 13
for microbiology. Appl Environ Microbiol 78 (12):1025–1031. https://doi.org/10.1097/
(12):4065–4073. https://doi.org/10.1128/ GIM.0b013e3182272e0b
AEM.00217-12 8. Paul P, Apgar J (2005) Single-molecule dilu-
3. Rosell R, Karachaliou N (2016) Lung cancer: tion and multiple displacement amplification
using ctDNA to track EGFR and KRAS muta- for molecular haplotyping. BioTechniques 38
tions in advanced-stage disease. Nat Rev Clin (4):553–554, 556, 558–559
Oncol 13(7):401–402. https://doi.org/10. 9. Dear PH, Cook PR (1993) Happy mapping:
1038/nrclinonc.2016.83 linkage mapping using a physical analogue of
4. Regan JF, Kamitaki N, Legler T, Cooper S, meiosis. Nucleic Acids Res 21(1):13–20
Klitgord N, Karlin-Neumann G, Wong C, 10. McDonald OG, Krynetski EY, Evans WE
Hodges S, Koehler R, Tzonev S, McCarroll (2002) Molecular haplotyping of genomic
SA (2015) A rapid molecular approach for DNA for multiple single-nucleotide poly-
chromosomal phasing. PLoS One 10(3): morphisms located kilobases apart using long-
e0118270. https://doi.org/10.1371/journal. range polymerase chain reaction and intramo-
pone.0118270 lecular ligation. Pharmacogenetics 12(2):93–99
5. Puig M, Pacheco S, Izquierdo D, Regan JF, 11. Wetmur JG, Kumar M, Zhang L,
Karlin-Neumann G, Cáceres M (2016) Drop- Palomeque C, Wallenstein S, Chen J (2005)
let digital PCR (ddPCR)-based validation and Molecular haplotyping by linking emulsion
genotyping of human inversions. Eur J Hum PCR: analysis of paraoxonase 1 haplotypes
Genet 24 E(Suppl. 1):353 and phenotypes. Nucleic Acids Res 33
(8):2615–2619. https://doi.org/10.1093/ 13. Kitzman JO, Mackenzie AP, Adey A, Hiatt JB,
nar/gki556 Patwardhan RP, Sudmant PH, Ng SB, Alkan C,
12. Wetmur JG, Chen J (2008) An emulsion poly- Qiu R, Eichler EE, Shendure J (2011)
merase chain reaction-based method for Haplotype-resolved genome sequencing of a
molecular haplotyping. Methods Mol Biol Gujarati Indian individual. Nat Biotechnol 29
410:351–361 (1):59–63. https://doi.org/10.1038/nbt.1740
Chapter 29
ddTRAP: A Method for Sensitive and Precise

Quantification of Telomerase Activity
Andrew T. Ludlow, Dawne Shelton, Woodring E. Wright,
and Jerry W. Shay
Abstract
Telomerase is a cellular RNA template-dependent reverse transcriptase that adds telomere repeats to the 30
ends of chromosomes. Telomerase is expressed almost universally in tumor cells (>85%) to maintain
telomere length, thus providing the ability of tumor cells to avoid senescence and to have unlimited
replication ability, one of the key hallmarks of cancer. ddTRAP (droplet digital Telomere Repeat Amplifica-
tion Protocol) is a two-step assay with whole cell lysates that utilizes a telomerase-mediated primer
extension followed by droplet digital PCR (ddPCR) detection of extended products. The adoptation of
the TRAP assay to ddPCR has resulted in improved throughput, increased sensitivity and better repeatabil-
ity of the TRAP assay. The protocol described below details our procedures for ddTRAP.
Key words Telomerase activity, Droplet digital PCR
1 Introduction
Telomere maintenance and telomerase enzyme activity are almost

universal features of transformed cancer cells [1–4]. Telomeres are
repeat DNA sequences (50 -TTAGGGn-30 ) found at the ends of
chromosomes [5, 6]. Telomeres function the in completion DNA
replication each cell cycle and also provide a “capping” function by
preventing recombination, end-to-end fusions, and degradation at
the chromosome ends [7, 8]. Telomeres, therefore prevent
unwanted DNA damage signaling at chromosome ends and may
be a potent initial tumor suppressor mechanism when telomeres
become critically short and cells enter senescence [9, 10]. In normal
diploid cells telomeres shorten with each cell division due to the
end replication problem (i.e., due to the placement of the terminal
RNA primer, the inability of DNA polymerase to fully replicate the
lagging strand of DNA, and the lack of the enzyme telomerase
there is a loss of about 60 nucleotides per cell division/DNA
replication cycle). Almost all tumor cells adapt a telomere
513
514 Andrew T. Ludlow et al.
maintenance mechanism that utilizes telomerase to maintain telo-

meres [1]. Telomerase is a ribonucleoprotein complex that utilizes a
protein component with reverse transcriptase activity (hTERT) and
a RNA template (hTR or hTERC) to add telomere repeats to
chromosome ends [11]. The two major molecular properties that
are commonly assayed are telomerase activity (ability of the enzyme
to extend a substrate by adding nucleotides) and repeat addition
processivity (the number of TTAGGG repeats added to a substrate)
[12]. Since the vast majority of cancer cells utilize telomerase to
maintain telomeres, telomerase activity an important transforma-
tion biomarker, and thus accurate detection is critical. The discov-
ery that telomerase synthesizes telomeric repeats onto 30 ends of
chromosomes led to the development of an assay for the detection
and measurement of its activity in cells and tissues and was estab-
lished over 20 years ago [1].
To measure the activity of telomerase the most common assay is
the telomere-repeat amplification procedure or TRAP, a two-step
procedure involving telomerase mediated primer extension and
PCR-based detection of extended products. Briefly, cells are lysed
in a buffer containing a detergent, then a portion of the lysate is
mixed with an extension reaction containing the telomerase sub-
strate (a DNA oligonucleotide) and dNTPs which are used by
telomerase (if present in the sample) to add hexameric telomere
repeats. Finally, the extended substrates are PCR amplified and
detected (See Fig. 1 for a graphical representation of the ddTRAP
assay). We have recently adapted the original TRAP assay that
provided only relative quantification (quantitated to an internal
standard DNA) to be detected with intercalating DNA dye (i.e.,
Evagreen®) on the Bio-Rad droplet digital PCR QX150/200 Eva-
green® compatible droplet reader, we call this assay the “ddTRAP”
[13]. The adaptation of the TRAP assay to droplet digital PCR
improved the sensitivity, repeatability, and throughput of the TRAP
assay [13]. These properties of ddTRAP have enabled single cell
telomerase assays and the screening of genes and small molecules
aimed at manipulating telomerase activity. The detailed procedure
below describes how to perform a ddTRAP to assay and quantitate
telomerase activity in adherent or suspension transformed cells and
primary human cells in culture.
2 Materials
Prepare all solutions and set up all reactions using typical PCR
precautions see Note 1.
2.1 Whole Cell 1. Use of a well-characterized telomerase positive cell line such as
Lysate Preparation HeLa, HEK293, H1299, MDA-231, HCT116, or HT1080 as
a positive control.
ddTRAP: A Method for Sensitive and Precise Quantification of Telomerase Activity 515
Fig. 1 TRAP assay theory. Tissue culture cells are plated on a 10 cm culture dish and grown to about 90%
confluence. Cells are then counted and pelleted in 1 million cell aliquots. Whole cell lysates are prepared and
then added to the extension mix containing the telomerase substrate primer (TS primer) and dNTPs. The lysate
containing active telomerase enzymes will extend the TS primer using the RNA template of the enzyme. The
extension reaction is heat-inactivated, and the products are then PCR amplified in the presence of the reverse
primer ACX and the forward primer TS to amplify the telomerase-extended substrates. The PCR products are
double stranded and can be detected using Evagreen® dsDNA binding dye in the QX100/200 droplet digital
PCR reader. Note—the bolded nucleotides in the ACX primer are mismatches with the telomerase-extended
products
2. A cell line lacking telomerase activity such as U2OS (human

osteosarcoma) or primary (nontransformed) BJ fibroblasts
should be used as a negative control.
3. Scale up the cell line you plan to use as positive and negative
controls and freeze aliquots for future use to avoid uninten-
tional changes that could occur in long term cultured cell lines.
4. NP-40 lysis buffer (RNase/DNase-free): 10 mM Tris–HCl,
pH 8.0; 1 mM MgCl2; 1 mM EDTA; 1% (vol/vol) NP-40;
0.25 mM sodium deoxycholate; 10% (vol/vol) glycerol;
150 mM NaCl; 5 mM β-mercaptoethanol; 0.1 mM AEBSF
(4-(2-Aminoethyl)benzenesulfonyl fluoride hydrochloride).

(See Note 2 for cell extract preparation.)
5. Ca2+- and Mg2+-free PBS: 137 mM sodium chloride, 1.5 mM
potassium phosphate, 7.2 mM sodium phosphate, 2.7 mM
potassium chloride, pH 7.4.
6. RNase/DNase-free DEPC-treated dH2O (Ambion).
7. Cell counting device such as a hematocytometer, Coulter®
counter, or automated cell counter (e.g., TC20, Bio-Rad).
Alternatively, protein quantification can be determined prior
to extension reaction with a BCA protein assay for cell extracts
(see Note 2).
8. Bicinchoninic Acid protein assay kit—such as BCA protein
assay kit (23225, ThermoFisher).
9. Spectrophotometer that can measure absorbance at 562 nm
(for BCA assay).
2.2 Telomerase 1. Ultrapure BSA (50 mg/mL, Ambion).

Substrate Primer 2. 50 dNTP mix (2.5 mM of each dATP, dCTP, dGTP, dTTP).
Extension Reaction
3. 10 μM telomerase substrate, “TS” primer (telomerase sub-
strate/primer for extension; 50 -AATCCGTCGAGCAGAGTT
HPLC purified).
4. 10 TRAP buffer (RNase/DNase-free): 200 mM Tris–HCl,
pH 8.3; 15 mM MgCl2; 630 mM KCl; 0.5% Tween 20;
10 mM EGTA.
5. Thin walled (250 μL volume) PCR grade tubes/stripes/plates.
6. Thermocycler.
2.3 ddPCR Detection 1. 10 μM ACX primer (reverse amplification primer for detec-
of Telomerase tion—50 -GCGCGGCTTACCCTTACCCTTACCCTAACC
Extended Substrates HPLC purified—see Note 9 about primers).
2. 10 μM telomerase substrate, “TS” primer (telomerase substrate
for extension, also forward primer for detection—5-
0
-AATCCGTCGAGCAGAGTT HPLC purified).
3. 2 Evagreen® super mix for ddPCR (Bio-Rad).
5. Twin-Tec 96-well plate—Eppendorf 951020362 (Fisher).
6. Pierceable foil heat seal—(Bio-Rad #1814040).
7. Droplet generator cartridges (DG8) (Bio-Rad 186-3008).
8. Droplet generator oil (Bio-Rad 186-3005).
9. Droplet cartridge gaskets (DG8) (Bio-Rad 186-3009).
10. QX200 evagreen ddPCR supermix (Bio-Rad).
11. Thermocycler capable of fitting the 96-well skirted plates and

adjusting the temperature ramp rate (i.e., Bio-Rad T100).
12. Droplet reader oil.
13. Droplet reader (Bio-Rad QX150/200) capable of reading Eva-
green® DNA binding dye.
14. PX1 PCR plate sealer (Bio-Rad).
3 Method/Procedures
3.1 Preparation 1. Culture cells to a desired density, typically to a density of 90%

of Tissue Culture Cells confluence (one needs to be consistent from experiment to
for ddTRAP Assay experiment since in some cases cell density can affect telome-
rase activity). We typically grow cells on 10 cm dishes for TRAP
assay analysis.
2. Trypsinize and count cells see Note 1.
3. Pellet cells. Centrifuge cells at 2000 g at room temperature
and aspirate media (see Note 2). Cells can be washed with cold
1 PBS following initial pelleting to remove carry over tissue
culture media but this is not necessary.
4. If cell counting is not performed, ddTRAP assays can be per-
formed using protein concentration (see Note 13).
5. Pellets can be used immediately by placing cells on ice in lysis
buffer.
6. Pellets can be also be stored for later use by flash freezing in
liquid nitrogen and stored at 80 C (see Note 2).
3.2 Lysis of Cells 1. Generate a whole cell lysate by adding 40 μL of NP40 lysis
buffer to pelleted cells (per 1 million cells) and incubate on ice
for 40 min (see Notes 12 and 14).
2. During the lysis period periodically vortex or mix vigorously at
least twice (e.g., at 15 min and 30 min of lysis). This is essential
to ensure the cells are lysed and homogenous (see Note 14).
3.3 Telomerase 1. While the cells are in lysis buffer, set up the extension reaction
Primer Extension mix as shown below in Table 1 and store on ice.
Reaction Setup 2. Set up tubes for dilution of the lysates on ice.
3. Lysates need to be diluted to a cell equivalent of 1250 cells per
microliter (2 μL of lysate in 18 μL of NP40 lysis buffer) on ice.
4. Prepare all diluted lysates before mixing with the extension
reaction. Please see Note 3 below for details on extension
reactions.
Table 1
Extension reaction setup
Reagent X1 (volumes in μL) X10 (volumes in μL)

10 TRAP buffer 5 50
50 mg/mL BSA 0.4 4
10 μM TS primer 1 10
2.5 mM dNTPs 1 10
1250 cell equivalents lysate 1 –
Water (dH2O) 40.6 406
Table 2
PCR setup
Reagent Volume (μL) 1 Volume (μL) 10 Volume (μL) 20

2 Evagreen ddPCR super mix 11 110 220
10 μM TS 0.11 1.1 2.2
10 μM ACX 0.11 1.1 2.2
25 cells/μL extension product 2 –
dH2O 8.8 88 176
5. In an ice old tube containing 49 μL of extension reaction mix

add 1 μL of diluted lysate to thin walled (250 μL) PCR tubes.
Reactions should be mixed by pipetting.
6. When all lysates are added to the extension reaction mix, the
PCR plate/tubes should be immediately moved to a thermo-
cycler set with the following reaction conditions: 25 C for
40 min, 95 C for 5 min, and 12 C hold. Please see Notes
15–17 for further information on the lysis step and storage of
lysates.
3.4 Digital PCR Setup 1. Using the PCR precautions outlined in see Notes 4–11 prepare
a master mix containing a final concentration of; 1 Evagreen
ddPCR Super Mix v2.0 (Bio-Rad), 50 nM TS primer, 50 nM
ACX primer, 50 cell equivalents or less of extension product
and dH2O to 20 μL per sample, with 10% extra. See Table 2 for
example volumes.
2. Allow reaction mix to reach room temperature (see Notes 5–8).
3. Set up the droplet generation (DG) cartridge. Load 20 μL of
PCR mixture into the sample well in the cartridge. Then add
70 μL of droplet generation oil for Evagreen® into the oil well.

Secure a gasket on the DG cartridge.
4. Place assembled droplet cartridge into the DG machine.
5. Remove cartridge from the DG machine once droplet genera-
tion cycle is completed (about 90 s).
6. Transfer droplets to PCR plate. Remove gasket gently. Using
an eight-channel pipet remove ~42 μL of emulsion (droplets)
from the droplet wells in the cartridge and place into a 96-well
plate (see Note 9).
7. Seal plate. Once all samples are loaded into the 96-well PCR
plate, the plate must be foil sealed to prevent evaporation and
light exposure of the emulsion (droplets).
8. Run PCR. Load the 96-well plate into the thermocycler and
close the lid. In these experiments, we use a Bio-Rad T100.
PCR reaction conditions—All ramp rates between temperature
steps must be set to 2.5 C/sec in order to achieve even heating
of the reaction mixture.
Thermocycling:
95 C for 5 min (activation of hot-start polymerase).
40 cycles of:
95 C for 30 s.
54 C for 30 s.
72 C for 30 s.
22 C hold.
Timing ¼ 1 h 45 min.
3.5 Detection 1. Read the droplets. Load the 96-well plate into the droplet
of the Telomerase reader plate holder matching well “A1” in the proper
Extended PCR orientation.
Amplified Substrates 2. Set up the reader. Open® software on the desktop of the laptop
computer synchronized with the droplet reader. Double click
well “A1” in the plate template. This will change the upper part
of the screen to display “sample name,” “experiment,” “assay
name,” and assay channels (fluorescence dyes fam or Vic®/hex.
See Note 10.
3. Define the wells. Click “Experiment” and a pull-down menu
with choices for experiment types will be displayed. Choose
“absolute quantification” for the experiment type. Select wells
(highlight wells) and click apply or press enter. With the same
wells highlighted, Select assay channels as unknown channel
1 (6FAM/Evagreen®). Click “apply” or press enter. Name the
assay—As an example for ddTRAP put the word ddTRAP and
indicate the extension time used (i.e., ddTRAP 40 min ext.)
and press apply. Name the samples.
4. Run the plate. Click run. A screen will prompt you to save the
template. Give your plate a logical name (record this in your lab
note book or experiment log in Excel® etc.). Click “save.” A
screen will prompt you to pick the dye types (Fam/Vic or
Fam/Hex) for ddTRAP pick FAM/Vic and if you want to
read the wells in columns or rows.
3.6 Data Analysis 1. Determine if the sample is valid for further analysis. The data
generated from the droplet reader is given in several formats.
The most important component that should be addressed
when analyzing ddTRAP data is the number of droplets gen-
erated. This can be queried several ways in the QuantaSoft®
software. By clicking “events,” the total number of droplets per
well can be visualized in a histogram. Alternatively under the
“table” display the column titled “accepted droplets” will give
the investigator the same information. We typically consider a
sample valid for further analysis with more than 10,000
accepted droplets; more stringent criteria of 12,000 accepted
droplets can also be applied, but the same criteria should be
applied across all samples in a given experiment and all experi-
ments in a particular manuscript unless noted in the figure
legend.
2. Set the threshold. Setting the thresholds between positive and
negative droplets can be a subjective task when analyzing digi-
tal PCR data and thus certain standard criteria should be
applied for each new assay. For the ddTRAP assay, positive
droplets typically fall between 6000 and 10,000 fluorescent
units, however since telomerase generates amplicons of various
sizes and is a “GC”-rich template, longer molecules (i.e., those
with more repeats added) appear lower on the amplification
plot. We use the following guides to set thresholds for ddTRAP
assays: (1) A “no-template lysis buffer control” (NTC-LB)
sample should be analyzed first, set threshold at 2000 fluores-
cence units above population of negative droplets (typically
around 4000 fluorescence units see Fig. 1). There may be
some background in this sample, however given the quantita-
tive nature of ddTRAP, this background can and should be
subtracted from all positive wells. See Fig. 1 for an example of
background in NTC-LB controls. (2) Set thresholds for all
wells. We typically use the same threshold for all wells as was
set for the NTC-LB. Occasionally, the population of negative
droplets will vary between wells and in this case a well specific
background needs to be set to analyze the data. We recom-
mend in this case that the threshold be set at a minimum of
2000 fluorescent units above the “negative population.”
(3) Dealing with “rain.” “Rain” or intermediate droplets
between the negative and positive populations are common
Table 3a
Data from select columns from a QuantaSoft spreadsheet
Concentration
Sample type Experiment Accepted droplets (molecules per microliter)
NTC-LB Abs quant 14,267 1.64
Positive control cancer line Abs quant 12,587 43.8
with the ddTRAP assay and can have an impact on the quanti-
tative results. Users must be cautious when setting thresholds
as to not skew data one way or another. To avoid potential
problematic “rain” we recommend using as few cell equivalents
as possible and repeating analysis with samples that produce a
large amount of “rain.” When this is not possible the user must
set thresholds equivalently for all samples to be compared.
Once the thresholds are set quantitative information is now
available (see Notes 18 and 19).
3. Extracting quantitative data. In terms of the digital TRAP assay
we refer to the data as “the number of telomerase extended
substrates/products per number of cell equivalents analyzed.”
This information is obtained after correction for background
signal and number of cell equivalents analyzed. From the
QuantaSoft® software refer to the “concentration in molecules
per microliter” column in the tabular view of the software. This
table can be exported as a .csv file and analyzed in Excel®. First
we generate background corrected data by subtracting the
NTC-LB control concentration (molecules per microliter)
from each unknown sample and positive control. This back-
ground corrected value is then converted to total telomerase-
extended products by multiplying by 20 (20 μL PCR volume),
this number is then divided by the number of cell equivalents
added to the extension reaction (typically 50). See example data
below.
4. Determining telomerase-extended products per cell. If
1,000,000 cells are harvested, lysed in 40 μL, the lysate diluted
1:20 to give 1250 cells per microliter, 1 μL of the diluted lysate
added to a 50 μL telomerase extension (25 cell equivalents per
microliter) and then 2 μL of the extension (50 cell equivalents)
added to a 20 μL PCR the For example calculation see Note 20
(Table 3a and 3b).
Table 3b
Example calculations for telomerase-extended products per cell
Total telomerase
Data NTC-LB background extension Telomerase extension products
quantification correction’ products per cell
Description of Subtract “NTC-LB” Multiply Divide total telomerase extension
step concentration from background products by the number of cell
unknowns to generate corrected equivalents added to the
background corrected data concentration extension reaction. In this
by 20 example 50 cell equivalents.
Positive 42.16 843.2 16.864
control
cancer line
4 Notes
1. Cell counts are critical to determine cell equivalents added to

the extension reactions and for calculating the number of
telomerase-extended products per cell.
2. Pellets are typically stable under ideal conditions for 1 year at
80 C. If you freeze cells it is best to remove all extra liquid
before placing in the 80 C.
3. Following the extension reaction small white particulate matter
may be observed in the extension reaction, to avoid carrying
this material over to the ddPCR setup briefly centrifuge the
PCR tubes and avoid pipetting from the bottom of the exten-
sion reactions. The amount of detergent from the lysate/
extension can negatively impact droplet formation. We found
that 0.008% of NP40 did not negatively impact droplet forma-
tion. We suggest testing concentrations of detergent higher
than 0.008% on droplet formation prior to performing assays
if higher concentration lysates are used.
4. Ten percent (10%) extra volume of each reagent should be used
so that the final reaction volume is 22 μL to help prevent
volume shortage when pipetting into the droplet generator
cartridge.
5. You must make samples in intervals of 8.
6. The polymerase in the 2 ddPCR Evagreen® Super Mix is
hot-start so all steps should be performed at RT.
7. Performing PCR set up on ice may increase solution viscosity
(2 PCR master mix) which will affect droplet formation and is
not recommended.
8. Order of loading the droplet generation cartridge essential.

The oil is “heavy” which if loaded first causes the oil to flood
the microfluidics and create a situation of poor droplet forma-
tion. Thus it is essential that you load your PCR sample first
then the oil for proper droplet formation. After the PCR and
oil are added to the cartridge wells attach a droplet generator
cartridge gasket. The droplet generation machine works only
with a full cartridge of 8 wells and a gasket properly attached to
the cartridge. The machine stops when air (i.e., an empty well)
is encountered or when a gasket is misplaced on the cartridge.
If you only have 12 samples you must still set up 16 reactions,
just four of the reactions will be blanks (mixing 11 μL
2 ddPCR supermix Evagreen® control with 11 μL of water
per open well) or no template controls (NTC).
9. Transferring the droplets to the PCR plate is critical. Move
slowly and consistently. Do not repeat pipet or completely
aspirate the droplets from the pipet, as this will enhance droplet
breaking/blending.
10. Evagreen® is read on the same channel as 6-fam.
11. Due to the sensitivity of the PCR-based TRAP assay, avoiding
contamination is of utmost importance for assay success. We
suggest setting up a dedicated set of pipettes and an area for
ddTRAP assays only. Always using barrier tips. Do not use ART
tips. Since telomerase is a reverse transcriptase with an RNA
component, special precaution must be taken to ensure an
RNase-free environment. All solutions should be made with
DEPC-treated water, and benches, pipettes and labware
decontaminated with an RNase inactivator such as
RNaseZap®.
12. An alternative lysis buffer using CHAPS (0.5–1.0%) can be
used in place of NP40 to generate cell extracts instead of
whole cell lysates. Whole cell lysates give maximal activity
detection, which is important for comparing telomerase-
modulating compounds quantitatively [14, 15]. If desired,
cell extracts can be made (lyse as above but spin down debris
at 10,000 g for 10 min at 4 C), and supernatant removed
for analysis. Cell extracts will underestimate the telomerase
activity but may be useful in analyzing primary samples that
may contain PCR inhibitors (such as primary tumor samples,
[14]). CHAPS lysis buffer and cell extracts should be used
when analyzing tissue samples. Mechanical homogenization
(avoid heat treatment) or a mortar and pestle can be used to
help disrupt the tissue. Centrifuge as above and determination
of protein concentration as in note.
13. Determination of equivalent loading input for telomerase
assays can be done in two ways: Cell counting or protein
concentration. We prefer cell counting when using whole cell

lysates. We advise pelleting large numbers of cells so that the
impact of loss of cells due to aspiration techniques is minimal.
We typically use greater than 300,000 cell pellets and prefer
1 106 cell pellets for optimal data. Protein concentration
should be determined when cell extracts are used in ddTRAP.
For ddTRAP 1–6 μg of protein is sufficient to detect telome-
rase activity from HeLa cells. Also, protein concentrations are
needed for tissue samples but caution should be noted that
telomerase positive cells will be comixed with telomerase silent
stromal cells.
14. Cell pellets must be thoroughly lysed in NP40 buffer for a
minimum of 40 min and a maximum of 1 h on ice. For pellets
up to 1 million cells, typically 40 μL of NP40 lysis buffer is
sufficient for lysis (1 million cells lysed in 40 μL of buffer results
in 25,000 cell equivalents per microliter of lysate). We do not
recommend using large volumes of lysis buffer to avoid dilu-
tions; this may cause loss of telomerase activity or data that is
not quantitative and repeatable. Do not lyse more than 45 min.
15. Lysates can also be stored at 80 C but telomerase enzyme
activity decreases overtime in the freezer, thus we recommend
the use of fresh lysates and that lysates are aliquoted to avoid
freeze–thaw cycles.
16. We have found that adding 1250 cell equivalents to a 50 μL
extension reaction is the most reproducible in the ddTRAP
assay (this results in a final extension reaction cell equivalent of
25 cells per microliter). Since we typically use 1 million cell
count pellets, a 1:20 dilution in NP40 lysis buffer is necessary
(2 μL of lysate in 18 μL of NP40 lysis buffer) and will generate
a diluted lysate with 1250 cell equivalents per microliter. Once
the lysates are diluted, 2 μL are added to the extension reaction
mixture in thin wall PCR tubes/plates on ice. Tip—make sure
that all lysates and dilutions are homogenous prior to pipet-
ting. Important control reactions should be set up to confirm
assay integrity. Controls such as treatment with RNase A to
digest the RNA component of telomerase and or heat treat-
ment (95 C for 10 min) prior to extension can be performed
to ensure specific detection of telomerase in the ddTRAP assay
(Fig. 2). These controls are important for analysis of new
tumor lines (i.e., lines with unknown telomere maintenance
strategies) and for laboratories unfamiliar with the telomerase
and TRAP assays (Fig. 3).
17. An extension time of 40 min is satisfactory for quantitative
detection of telomerase (see Fig. 4), longer extension times
(up to 2 h) can be used to detect maximal telomerase activity
however longer extension times risk degradation of the enzyme
Fig. 2 Workflow and optimization of droplet digital TRAP. (a) The ddTRAP workflow. Cells are lysed and then
diluted to a concentration of 1250 cells per μl, telomerase extension products generated at a concentration of
25 cells/μl, then telomerase is heat inactivated and extension products dispersed into droplets. PCR
thermocycling is done for 40 cycles and droplets analyzed for the presence or absence of fluorescence by
the droplet reader (QX150/200 Evagreen® compatible machine). (b) ddTRAP output showing BJ fibroblasts
(input of 100 cell equivalents, telomerase negative), H1299 cells (input of 100 cell equivalents, telomerase
positive), a lysis buffer only control, and a control with no primers and input of 100 cell equivalents of H1299
lysate to test for specificity of amplification. Only very low background signals are seen in these controls. Each
well or sample of the ddPCR analyzes about 17,000 droplets. Event number at the bottom of the output
represents the number of droplets counted in the wells overtime. Each dot on the ddPCR output represents a
unique droplet that is either positive or negative for fluorescent signal. Fluorescence amplitude is a measure of
the fluorescence detected for each droplet in the assay. Fluorescence amplitude is used to separate the
positive and negative droplets. Since the droplets are detected with EvaGreen® double stranded DNA binding
dye there will be inherent background fluorescence of DNA molecules not amplified during PCR. The heat map
scale represents the density of droplets at given fluorescent amplitudes. NTC-LB ¼ no-template control-lysis
buffer ([13]; Figure is reproduced with permissions and slight modification from NAR and Oxford press -
Ludlow, A.T., Robin, J.D., Sayed, M., Litterst, C.M., Shelton, D.N., Shay, J.W. and Wright, W.E. (2014)
Quantitative telomerase enzyme activity determination using droplet digital PCR with single cell resolution.
Nucleic Acids Res, 42, e104)
Fig. 3 ddTRAP control reactions and linearity. (a) Heat and RNase A inactivation of telomerase resulted in virtually
no ddTRAP signal. This permitted a small background correction to be included, and established the specific
measurement of telomerase activity from lysates. Background signal is highlighted by the red box. (b) HeLa cell
dilution series from 50 to 1 cell equivalents produced a linear relationship (R2 ¼ 0.99) between input and
detection of telomerase enzyme activity as indicated by total product generated in the ddPCR. Error bars are the
standard deviation of the replicates. Although the extract was not made from a single cell, ddTRAP was able to
reproducibly detect telomerase activity above background at the dilution equivalent to one cell input [13]
at 25 C and thus have not been tested currently in the TRAP

assay. Linear dilution analysis should be performed by input-
ting different amounts of cell equivalents into the extension
reactions. Two types of linearity can be performed here:
(1) Inputting different numbers of cell equivalents into the
extension reaction (Fig. 2) or 2. Making an extension reaction
and then diluting it prior to PCR to test the PCR linearity
[13]. The extension products are single stranded DNA and
should be used as soon as possible in the digital PCR reaction.
We have found that storage of extension products at 4–12 C
overnight does not result in loss of detectable extension pro-
ducts (Fig. 5).
Fig. 4 Comparison and correlation of ddTRAP to gel based TRAP. ddTRAP quantification is less variable than
gel based TRAP and allowed accurate determination of the IC50 of Imetelstat in HeLa (0.2 μM). HeLa cells
were incubated with Imetelstat (0, 0.125, 0.25, 0.5, 1, and 3 μM) for 72 h. Cells were pelleted and triplicate
extracts prepared from three separate tissue culture experiments (nine total extracts and extensions per dose).
(a) ddTRAP quantification with 50 cell equivalents added to the PCR. (b) Gel quantification (representative gel
image of two experiments in Fig. 3c. (c) Gel based TRAP was performed with 125 cell equivalents. Data are
expressed as relative telomerase activity compared to control (untreated HeLa) and standard error of the
mean. (d) Correlation analysis of gel-based TRAP to ddTRAP. The P < 0.0001 positive relationship indicates
that the methods are measuring the same phenomenon. IC ¼ Internal competitive telomerase activity
substrate (also known as ITAS) [13]
18. Setting digital PCR thresholds with “rainy” data. Since telo-
merase uses whole cell lysates and generates a “GC”-rich
amplicon of various sizes there can be samples that produce
“rain” or droplets of intermediate fluorescence intensity
between the major populations of negative and positive dro-
plets (see Fig. 4). This can cause problems in the analysis. For
this reason controls are essential in the ddTRAP. We also always
try to use 50 cell equivalents when possible. If samples seem to
produce a lot of “rain” at this cell input we dilute the sample
until better separation is achieved. Alternatively, the extension
can be diluted and corrected for cell equivalents in the analysis.
19. We found that purchasing HPLC purified primers was neces-
sary for the success of this assay, it will not work otherwise.
Fig. 5 Dealing with “rainy” data. ddTRAP fluorescent droplet outputs are shown on the right, and quantified
outputs graphed on the left. Lysates were incubated with TS substrate for various amounts of time prior to
heat inactivation, using 100 cell equivalents from H1299 cells. Data are presented as means of background
corrected total telomerase products generated standard error of the mean. The horizontal black line
represents the threshold used in these samples. It is essential when setting threshold for “rainy” data that
controls are used as a guide and the same threshold is used for all samples [13]
We suggest diluting stock primers at a concentration of

100 μM and making working concentration aliquots of no
more than 100 μL and avoiding freeze–thaw cycles.
20. Example calculation of cell equivalents: 1,000,000/40 μL lysis
buffer ¼ 25,000 cells/μL lysate; diluted 1:20 ¼ 1250 cells/μL
lysate; 1 μL of 1250 cell equivalents/50 μL extension reac-
tion ¼ 25 cell equivalents/μL extension reaction *2 μL into
20 μL ddPCR ¼ 50 cell equivalents analyzed.
Acknowledgments
We would like to acknowledge Viresh Patel and George Karlin-

Neumann at Bio-Rad for materials support in the development of
this technique and critical reading of the manuscript. We also
acknowledge Oxford University Press for permissions to use
Figure published in our article in Nucleic Acids Research. Ludlow,
A.T., Robin, J.D., Sayed, M., Litterst, C.M., Shelton, D.N., Shay, J.
W., and Wright, W.E. (2014) Quantitative telomerase enzyme
activity determination using droplet digital PCR with single cell
resolution. Nucleic Acids Res, 42, e104.
Funding
National Institute of Health [AG01228]; National Cancer Institute
[CA154805, P50CA70907, T32-CA124334-07, 5P30
CA142543-03]. Funding for open access charge: National Insti-
tute of Health Grant [AG01228]. This work was performed in
laboratories constructed with support from National Institutes of

Health grant C06 RR30414.
Conflict of interest statement
Dawne N. Shelton is a employee of Bio-Rad Laboratories. Bio-Rad
Laboratories provided materials and technical support for the
development of this technique.
References
1. Kim NW, Piatyszek MA, Prowse KR, Harley 9. Shay JW, Wright WE (2005) Senescence and
CB, West MD, Ho PL, Coviello GM, Wright immortalization: role of telomeres and telome-
WE, Weinrich SL, Shay JW (1994) Specific rase. Carcinogenesis 26:867–874
association of human telomerase activity with 10. Wright WE, Shay JW (1992) The two-stage
immortal cells and cancer. Science mechanism controlling cellular senescence and
266:2011–2015 immortalization. Exp Gerontol 27:383–389
2. Shay JW, Wright WE (1996) Telomerase activ- 11. Shay JW, Wright WE (2010) Telomeres and
ity in human cancer. Curr Opin Oncol 8:66–71 telomerase in normal and cancer stem cells.
3. Shay JW, Wright WE (1996) The reactivation FEBS Lett 584:3819–3825
of telomerase activity in cancer progression. 12. Nandakumar J, Cech TR (2013) Finding the
Trends Genet 12:129–131 end: recruitment of telomerase to telomeres.
4. Shay JW, Wright WE (2011) Role of telomeres Nat Rev Mol Cell Biol 14:69–82
and telomerase in cancer. Semin Cancer Biol 13. Ludlow AT, Robin JD, Sayed M, Litterst CM,
21:349–353 Shelton DN, Shay JW, Wright WE (2014)
5. Blackburn EH (1991) Telomeres. Trends Bio- Quantitative telomerase enzyme activity deter-
chem Sci 16:378–381 mination using droplet digital PCR with single
6. Blackburn EH (2000) Telomeres and telome- cell resolution. Nucleic Acids Res 42:e104
rase. Keio J Med 49:59–65 14. Herbert BS, Hochreiter AE, Wright WE, Shay
7. de Lange T (2009) How telomeres solve the JW (2006) Nonradioactive detection of telo-
end-protection problem. Science merase activity using the telomeric repeat
326:948–952 amplification protocol. Nat Protoc
8. de Lange T (2010) How shelterin solves the 1:1583–1590
telomere end-protection problem. Cold Spring 15. Norton JC, Holt SE, Wright WE, Shay JW
Harb Symp Quant Biol 75:167–177 (1998) Enhanced detection of human telome-
rase activity. DNA Cell Biol 17:217–219
Chapter 30
Highly Efficient and Reliable DNA Aptamer Selection Using

the Partitioning Capabilities of ddPCR: The Hi-Fi SELEX
Method
Aaron Ang, Eric Ouellet, Karen C. Cheung, and Charles Haynes
Abstract
In addition to its growing use in detecting and quantifying genes and larger genomic events, the partition-
ing used in digital PCR can serve as a powerful tool for high-fidelity amplification of synthetic combinatorial
libraries of single-stranded DNA. Sequence-diverse libraries of this type are used as a basis for selecting
tight-binding aptamers against a specific target. Here we provide a detailed description of the Hi-Fi SELEX
protocol for rapid and efficient DNA aptamer selection. As part of that methodology, we describe how
Hi-Fi SELEX gains advantages over other aptamer selection methods in part through the use of the massive
partitioning capability of digital PCR.
Key words Digital PCR, Aptamers, SELEX, Biological affinity reagents
1 Introduction
Biological reagents that recognize target molecules with both high

affinity and specificity are routinely required in the life sciences and
clinics, where in the latter case they often serve as capture, diagnos-
tic, or therapeutic agents. Among the available classes of affinity
reagents, antibodies currently are the gold standard. Antibodies,
including monoclonal antibodies (mAbs), bind their antigen with
high affinity; nM or tighter equilibrium dissociation constants
(Kds) are frequently reported [1]. Moreover, binding is generally
highly selective, as demonstrated by the ability of mAbs to discrim-
inate posttranslational modifications to proteins, as well as among
more subtle protein isoforms [2, 3]. These exquisite binding prop-
erties can serve to maximize therapeutic mAb potency and mini-
mize off-target effects, while the relatively large size of mAbs
enables long circulation half-lives [4, 5].
However, antibodies are limited in certain important aspects.
As large complex multisubunit proteins, they are sensitive to
531
532 Aaron Ang et al.
environmental conditions, and can rapidly be inactivated under

acidic conditions or at elevated temperatures. Though significant
advances have been made toward their manufacture, mAbs remain
relatively expensive to produce and purify at larger scales. Most
mAbs are incapable of permeating cells efficiently, which effectively
restricts their application to extracellular antigens. Moreover,
though humanized-antibody technology has greatly reduced
immune responses, therapeutic mAbs often do not escape immune
surveillance completely, further challenging their long-term
efficacy [6].
As a result, the development of mAb alternatives for use as
research and diagnostic affinity reagents, as well as therapeutic
agents, has gained considerable interest in recent years [7]. A num-
ber of simplified antibody forms including nanobodies, VH and VL
antibody domains, and single-chain variable fragments have proven
effective as mAb surrogates [8]. In addition, a particular class of
nucleic acids, known as aptamers, has emerged as a potent option
[9]. Each composed of a short single-stranded (ss) oligonucleotide,
aptamers can be produced relatively inexpensively at large scale with
high precision. The discovery of useful aptamers is likewise facili-
tated by the ability to easily synthesize large semicombinatorial
libraries of ssDNA or ssRNA that can be subjected to in vitro
Darwinian-type selection strategies to enrich and select sequences
that exhibit high affinity and specificity for a target as a result of
their unique folds. However, standard aptamer selection methods,
which typically employ a recursive selection process known as sys-
tematic evolution of ligands by exponential enrichment (SELEX),
are not sufficiently robust to ensure the timely and cost-effective
discovery of aptamers suitable for further development [10–13].
We therefore recently reported on a new method, the high-
fidelity SELEX (Hi-Fi SELEX) platform [14, 15], that greatly
improves the speed and robustness of aptamer selection, in part
by exploiting the partitioning capabilities of droplet digital PCR
(ddPCR) to preserve library integrity during regeneration. The
method requires either a commercial droplet generation system or
the ability to create emulsions suitable for PCR. It does not require
droplet-reading capabilities, making the method cost-effective
since the remaining elements of Hi-Fi SELEX use consumables
and equipment that are available in almost all standard molecular
biology laboratories. Here, we describe the Hi-Fi SELEX method,
for which the basic processing scheme is shown in Fig. 1, in
sufficient detail to enable a user to reliably apply it to discover
useful DNA aptamers against a target of interest. A number of
“Notes” are also provided to give a better understanding of the
methods we employ to maximize the speed and overall perfor-
mance of the protocol.
Highly Efficient and Reliable DNA Aptamer Selection Using the Partitioning. . . 533
Fig. 1 Schematic showing the sequence of operations comprising a single round of Hi-Fi SELEX. Adapted from
[15] with journal permission
2 Materials
2.1 Oligonucleotides The Hi-Fi SELEX protocol selects aptamers from a semicombina-
torial library of members in which each member is composed of
three oligonucleotides hybridized together. As shown in Fig. 2,
these oligonucleotides include an 80-nt ssDNA “library” sequence,
a 20-nt ssDNA DNA sequence (50 -Comp) complementary to the
50 -end of each 80-nt member of the library, and a 20-nt ssDNA
sequence (30 -Comp) that is 50 -phosphorylated and complementary
to the 30 -end of each library member. Each of the three required
components (80-nt ssDNA library, 30 -Comp, 50 -Comp) is synthe-
sized chemically by a reputable vendor (we use IDT Inc.; Coralville,
IA) capable of delivering it in the quantities (typically a few μmoles),
uniformities (ssDNA library and Comp sequences are indepen-
dently processed by HPLC (strong anion exchange column) to
eliminate truncated members and isolate a tight band of the desired
molecular weight), and purities needed to satisfy the DNA quality
and diversity requirements of the Hi-Fi SELEX process. Each com-
ponent is then reconstituted in 1 AF buffer to create the required
stock solutions (see Note 1). A ssDNA library structure and
corresponding set of 50 -Comp and 30 -Comp sequences appropriate
for conducting Hi-Fi SELEX are provided below.
Fig. 2 Basic structures of the semicombinatorial DNA libraries used in standard SELEX (standard library) and
Hi-Fi SELEX (competent library): Hi-Fi SELEX libraries differ from standard SELEX libraries through their use of
novel fixed-region complements (blocking elements) to improve the functional diversity of the starting
semicombinatorial library. Each member of the Hi-Fi SELEX ssDNA library is therefore composed of an
80 nt library sequence containing a 40-mer random core region (N40) flanked by a 50 universal 20-mer flanking
sequence and a 30 universal 20-mer primer binding sequence, with each flanking sequence hybridized to its
complement, which are hereafter denoted as 50 -Comp and 30 -Comp, respectively. By eliminating single-
strand structures within the fixed regions, the competent Hi-Fi SELEX library isolates aptamer fold and function
to within the variable core region of the library, while reducing artifacts that might compromise or eliminate
the discovery of a tight-binding sequence within that region. Unblocked, the fixed-region sequences within a
given selection library can interfere with aptamer fold and function through their potential to adopt stable
secondary structures created through either (1) self-association, or (2) association with complementary
nucleotides within the variable core region, the opposing fixed region, or both. These types of unwanted
structures can occur either within an individual library member or between complementary regions of different
members of the library [14], and we have previously shown that the net effect is to significantly reduce the
total functional diversity of the library [15]
2.2 Hi-Fi SELEX 1. Aptamer Folding (AF) buffer: 20 mM Tris–HCl (pH 7.4),
Reagents 140 mM NaCl, 5 mM KCl, 1 mM MgCl2, 1 mM CaCl2.
and Consumables 2. ssDNA Library Stock Solution: 1 μmole purified 80-nt ssDNA
in AF buffer (10 μl of a 100 μM stock required for complete
(three rounds) Hi-Fi SELEX selection). The general structure
of each library member is 50 -flanking sequence (20 nt) –N40
–flanking sequence (20 nt)-30 , with one suitable library
sequence being library [50 -TCGCACATTCCGCTTCTACC
–N40 –CGTAAGTCCGTGTGTGCGAA-30 ].
3. 50 Complementary Blocker (50 -Comp): for the library
sequence shown above, 50 -GGTAGAAGCGGAATGTGCGA-
30 resuspended at 100 μM stock in AF buffer (10 μl each total
needed for complete Hi-Fi SELEX run).
4. 30 Complementary Blocker (30 -Comp): for the library
sequence shown above, 50 -TTCGCACACACGGACTTACG-
30 resuspended at 100 μM stock in AF buffer (10 μl each total
needed for complete Hi-Fi SELEX run).
5. Stringent Wash (SW) buffer: 20 mM Tris–HCl (pH 7.4), 1 M

NaCl, 5 mM KCl, 1 mM MgCl2, 1 mM CaCl2, 0.005%
Tween 20.
6. Denaturing Elution (DE) buffer: 50 mM NaOH (followed by
neutralization with 50 mM HCl).
7. Tris-EDTA (TE) buffer: 10 mM Tris, 1 mM EDTA, adjusted
to pH 8.0 with HCl.
8. Target Immobilization (TI) buffer: 100 μl of 100 mM sodium
phosphate buffer (pH 7.5).
9. Surface Passivation (SP) buffer: AF buffer supplemented with
0.005% Tween 20.
10. λ-exonuclease; lambda exonuclease (New England Biolabs).
11. Nunc Immobilizer Amino Plates (ThermoFisher Scientific).
12. Thermomixer fitted with a plate adapter (e.g., Eppendorf
Thermomixer C).
13. NanoSep® Omega 10K MWCO centrifugal filter unit (Pall
Corporation).
14. Standard PCR thermal cycler (e.g., Bio-Rad C1000 or Eppen-
dorf Mastercycler Thermal Cycler).
2.3 ddPCR and qPCR 1. Forward Primer (FP): 50 -TCGCACATTCCGCTTCTACC-30

Reagents resuspended in AF at 100 μM stock concentration (for library
sequence shown).
2. Phosphorylated Reverse Primer (RP): 50 -p-TTCGCACA-
CACGGACTTACG-30 resuspended in AF at 100 μM stock
concentration (library sequence shown).
3. SYBR Green qPCR master mix: 2 iQ SYBR Supermix
(Bio-Rad).
4. ddPCR master mix: ddPCR™ Supermix for probes without
dUTP (Bio-Rad).
5. Thermal Cycler capable of real-time detection (e.g., Bio-Rad
CFX96).
6. DG8™ Droplet-Generation Cartridge (Bio-Rad).
7. Droplet Digital™ PCR (ddPCR™) Droplet Generator
(Bio-Rad QX100™ or QX200™).
3 Methods: The Hi-Fi SELEX Protocol
Prior to the initiation of any aptamer selection work, it is essential

that proper steps be taken to prevent cross-contamination of mate-
rial recovered between selection rounds. Regular decontamination
of all work surfaces and regular changes of gloves during the
process is necessary to prevent such occurrences. Moreover, stock

reagents used for PCR amplification must be kept separate from the
initial library as well as from any recovered pools of library members
during selection. Aliquots of each stock reagent should be made in
designated nucleic acid template-free areas. If aptamer selection is
to become routine in the lab, a dedicated separate environment for
conducting Hi-Fi SELEX must be created, with regular mainte-
nance and decontamination of all surfaces and instruments prior to
and at the completion of each round of selection.
3.1 Library Design In its standard format [10–12], SELEX enriches a subset of short
and Synthesis single-stranded RNA or DNA sequences from a synthetic library
composed of a semicombinatorial population whose total diversity
is similar to the body’s own antibody repertoire. The process is
typically performed in vitro, creating the capacity to modulate the
nature and stringency of the screening conditions to more effi-
ciently select library members showing high affinity or specificity
to a target.
SELEX can operate on either a ssDNA or ssRNA library, with
the choice typically made by carefully examining the intended
application of the aptamer. RNA-based aptamers have a more flexi-
ble backbone that allows them to adopt a wider range of secondary
and tertiary structures. Though built on a less flexible backbone,
ssDNA aptamers have higher chemical stability, and screening of
DNA libraries requires fewer processing steps. In this chapter, the
technological advances offered by Hi-Fi SELEX, including its use
of the partitioning step of ddPCR, are specifically applied to ssDNA
aptamer selections.
DNA libraries used for SELEX typically contain ~1014 unique
members of equivalent length, each composed of a random oligo-
nucleotide sequence within a variable core region that is flanked by
universal fixed sequences at the 50 and 30 ends. Library diversity is
largely encoded in the variable core region. This region can be
created utilizing different randomization strategies and nucleotide
chemistries. Most often, chemical syntheses that equally weight the
frequency of each naturally occurring nucleotide are employed, but
variable core regions composed of partially randomized sequences
[16], genomic DNA inserts [17, 18], and various chemically mod-
ified nucleotides [19–21] have also been used with success.
Hi-Fi SELEX typically employs a variable core region that is
40 nt in length (Fig. 2) so as to generate reasonable diversity while
keeping the mass of the starting library suitable for the screening
process. However, smaller or longer randomizations can be used.
The sequences of the universal fixed regions (typically 20 nt each)
flanking the 50 and 30 ends of each variable core sequence are
engineered to achieve high-fidelity amplification of retained library
members by eliminating, or at least minimizing, self-association
and primer-dimer pairing reactions that can promote formation of
unwanted by-products during PCR. In our lab, this is typically

achieved by designing both the flanking sequences and the asso-
ciated forward (FP) and reverse (RP) primers using Primer3 soft-
ware (http://biotools.umassmed.edu/bioapps/primer3_www.
cgi), and then scoring them for self-complimentary and primer-
dimer formation using the OligoAnalyzer tool (https://www.
exiqon.com/ls/Pages/ExiqonOligoOptimizerTool.aspx) of Exi-
qon, Inc. Several pairs of 20-mer flanking sequences designed
using the same or a similar strategy have been reported and suc-
cessfully used [22–24].
The random core region of the 80-mer ssDNA library is gener-
ally created combinatorially by dosing the required nucleotides at
an A:C:G:T molar ratio of 3:3:2:2.4 in order to achieve equal
probability incorporation of each nucleotide within the variable
core region [25]. The 50 -phosphorylated 30 -Comp sequence is
used as the RP for PCR amplification.
3.2 Target Levels of nonspecifically retained library members are known to

Immobilization dictate the degree of enrichment of useful aptamers achieved in a
given round of selection [26]. Practitioners of SELEX have
attempted to quantify the overall quality of each round of selection
through a parameter PE known as the “partition efficiency”:
½A
PE ¼ ð1Þ
½AP
where
X
n
½A ¼ ½A i ð2Þ
i¼1
Here, [A] is the total molar concentration of all unbound

library members, given by the sum of the concentration [Ai] of
each unbound library sequence i; likewise [AP] is the molar con-
centration of all bound library members recovered in the fraction
eluted in DE buffer. In SELEX, per round PE values generally
range from 10 to 1000, due in part to nonspecific retention of
library members. Reducing the starting population of ~1014 mem-
bers to a manageable number of candidate aptamers, say ~104, then
requires a relatively large number of successful selection rounds
(generally identified by an improvement in the mean Kd of the
retained pool).
In contrast, Hi-Fi SELEX generates per round PE values
greater than 105 (usually close to 106) by displaying the target on
a tailored substrate and in conditioned solvents that together
greatly inhibit nonspecific retention. Nunc Immobilizer plates,
which present on the surface of each well a layer of end-grafted
hyperbranched polyglycol chains at densities near or above those
needed for soft brush formation, display good passivation against
Fig. 3 Schematic of chemistries used to immobilize target protein, neutralize reactive groups, and passivate
nonspecific binding sites on the surface of a Nunc Immobilizer Amino plate
nonspecific adsorption of proteins [27] or oligonucleotides

[28, 29], even in cases where the zeta potentials of the underlying
base surface and sorbate are opposite in sign [27, 30]. This capabil-
ity is integral to the Target Immobilization procedure itemized
below:
1. Immobilize the target protein (50–100 nM target in TI buffer)
to the Nunc Immobilizer Amino C8 strips by incubating the
filled well overnight at 4 C (see Fig. 3 and Notes 2 and 3 for
information on the coupling chemistry).
2. Wash the wells three times with 300 μl of SP buffer (see Note 4).
3. Fill the wells once more with 300 μl of SP buffer.
4. Incubate for 1 h under gentle agitation (we set a thermomixer
fitted with a plate adapter at 500 rpm) at room temperature to
complete the neutralization reaction.
5. Aspirate the SP buffer out of each well at the end of the
neutralization/passivation process.
6. Perform a final wash (3 of 300 μl of SP buffer) before sealing
the wells with optical film.

7. Store the processed and sealed C8 strip at 4 C until
further use.
3.3 Partitioning Prior to Hi-Fi SELEX screening, the library is subjected to a

and Retained Fraction thermal denaturation–renaturation cycle designed to promote
Recovery annealing of the 50 -Comp and 30 -Comp strands to each 80-mer
library member, as well as folding of each member into its thermo-
dynamically favourable conformation at screening conditions [31]
(see Note 5). Note that the success of Hi-Fi SELEX in reliably and
efficiently discovering useful aptamers requires strict adherence to
this protocol during each selection round. The combined use of an
aptamer-to-target ratio of 100:1 to 1000:1 (see Note 6),
conditioned solvents, and high-salt wash steps creates the stringent
partitioning conditions (see Note 7) required to reduce the
sequence diversity 105–106 orders of magnitude in each of the
first few selection rounds. Approximately 108 unique library
sequences are therefore retained and recovered in the first partition-
ing step—a number optimal for the remaining steps of the selection
cycle.
The library partitioning and retained fraction recovery protocol
is as follow:
1. Mix 1 nmole (~1014 sequences) of the ssDNA library from the
prepared stock solution with equimolar amounts of 50 -Comp
and 30 -Comp in 100 μl of SP buffer.
2. Heat that mixture to 95 C for 5 min (denaturation) before
slowly cooling it down to 25 C at a rate of 0.5 C min1
(renaturation) in a standard thermal cycler.
3. Add the resulting 10 μM of competent library to a neutralized
and passivated Nunc Immobilizer well displaying the target.
4. Equilibrate that system for 1 h at 25 C under gentle agitation
(500 rpm) in a thermomixer equipped with a plate reader (see
Note 6).
5. Remove unbound and weakly bound members by washing
three times with 300 μl of SP buffer.
6. Apply a second more stringent pair of 3 washes with 300 μl of
SW buffer. These washing steps do not require agitation (see
Notes 7 and 8).
7. Elute the retained pool of specifically bound library members
from the Nunc wells by incubating with 100 μl of DE buffer for
10 min at 70 C and an agitation rate of 600 rpm.
8. After agitation, neutralize the pH by adding 100 μl of
50 mM HCl.
9. Repeat steps 7 and 8.
10. Combine the eluted aptamer pools.
11. Desalt that pool by exchange into AF buffer and concentrating
to 25 μl (final) using a centrifugal filter unit operating at
14,000 g for 15 min (see Note 9). Begin that centrifugation
without addition of AF buffer so as to remove as much DE

buffer as possible. Then add AF buffer and begin buffer
exchange.
3.4 Retained Pool Following each selection round, the polymerase chain reaction
Amplification by (PCR) is generally used to amplify the pool of retained library
ddPCR sequences. In SELEX, this step is conducted as a conventional
bulk PCR reaction using a standard thermal cycler and universal
forward and reverse primers targeting the fixed flanking sequences
of each library member.
This approach tends to be problematic, as illustrated in Fig. 4
for a relatively simple pool of 105 unique library members. Though
the desired 80-bp dsDNA amplicon is created, a maximum in its
total abundance is typically reached after a limited number of cycles.
Beyond that cycle, various artifacts, including formation of 80-mer
heteroduplexes hybridized together through only their common
flanking sequences, oligonucleotide stretches within the universal
primer regions mispriming certain variable core region sequences,
and improperly extended products acting as spurious primers on
heterologous sequences, promote conversion of the library to
increasingly aberrant high molecular weight (HMW) by-products.
To avoid these complications, Hi-Fi SELEX replaces traditional
PCR with the partitioning capabilities of ddPCR. When as few as
105 retained library members are recovered in a given Hi-Fi SELEX
round and then amplified by the ddPCR-based protocol described,
an 80-bp amplicon concentration of greater than 1 μM is generally
realized in the final ~25 μl sample. Moreover, all amplicons pro-
duced are in their fully complementary (homoduplexed) dsDNA
state. As a result, a regenerated ssDNA library can be created from
as little as 105 retained members in quantities sufficient to not only
proceed to the next selection round, but also to determine the
mean binding affinity of the enriched pool after each selection
round, providing a metric of how the overall selection is
proceeding.
The amplification protocol used in Hi-Fi SELEX is conducted
in two steps as follows:
Determine the concentration of the retained library CLibrary:
1. Dilute 0.5 μl of the desalted concentrated pool of retained
80-nt library members in 9.5 μl nanopure water.
2. To a qPCR well, add 5 μl of the solution created in step 1,
along with SYBR green mastermix (to 1 final), FP and RP
(250 nM final each). Top up the final volume in the well to
20 μl by adding nanopure water.
3. Begin cycling with an initial activation step at 95 C for 3 min
followed by 39 cycles of amplification, each composed of dena-
turation at 95 C for 30 s and annealing/extension at 60 C for
Fig. 4 Comparison of amplification of a retained pool (105) of library members by conventional PCR (standard
SELEX; upper panel) and by ddPCR (Hi-Fi SELEX; lower panel): Although the desired 80-bp dsDNA amplicon is
created in conventional PCR, a maximum in its total abundance is typically reached after a limited number of
cycles. Beyond that cycle, various artifacts, including formation of 80-mer heteroduplexes hybridized together
through only their common flanking sequences, oligonucleotide stretches within the universal primer regions
mispriming certain variable core region sequences, and improperly extended products acting as spurious
primers on heterologous sequences, promote conversion of the library to increasingly aberrant high molecular
weight (HMW) by-products. A small-scale “pilot” PCR reaction [32, 33] is therefore generally performed to
determine the maximum number of PCR cycles that can be conducted before accumulating unacceptable
amounts of by-products, which are known to adversely affect selection and must therefore be removed by gel
electrophoresis or other means [34]. That pilot reaction typically shows that standard PCR amplification of the
retained pool must be stopped at ca. 22–25 cycles since HMW by-products generally start accumulating when
the amplicon concentration reaches ca. 20–50 nM. Termination of amplification at this relatively low cycle
number generally yields ~1010–1012 80-bp amplicons, or 1% of that needed to initiate the next selection
round. As a result, in SELEX, the PCR step must be multiplexed across 100 or more parallel reactions, each
amplifying between ~ 105–106 library members to create 1011–1012 amplicons per well. The products of the
parallel reactions are then pooled and concentrated to reach the concentration (i.e., 1014 amplicons in 100 μl)
required for downstream processing and the next round of SELEX. The use of emulsions in ddPCR to isolate
and amplify single templates by PCR is well established, and it is known that the resulting partitioning of single
templates into individual droplets reduces formation of unwanted by-products when coamplifying mixtures of
templates (e.g., multiple genes) [35, 36]. Spurious priming events are greatly reduced within each droplet, in
part because competition between different templates and biases resulting from differences in amplification
efficiencies are avoided [37]. Moreover, post amplification, the emulsions can be broken to recover the full set
of amplicons in an aqueous phase suitable for downstream processing. In Hi-Fi SELEX, the pool (~108) of
competent 80-nt ssDNA library members retained after a selection round is therefore partitioned among a
similar number of nL-sized droplets. ddPCR partitions ca. 20,000 droplets per well, which means 100 wells
30 s. Set the heating and cooling rates at 2.5 C s1 for even
heat distribution in the well.
4. Determine CLibrary by comparing the recorded quantitation
cycle Cq to corresponding Cq data from a standard curve
derived from a dilution series of the initial 100 μM library
stock with sequence diversity from 108 down to 102 unique
library members.
5. From CLibrary and the retained library volume (~25 μl) the
number of ddPCR wells required to conduct the retained
library amplification is computed by assuming 17,000 readable
drops having a CPD of 50 are created per ddPCR well. If, for
example, the retained library contained 108 members (typical
for 1st and 2nd rounds of Hi-Fi SELEX), 120 wells are
required (120 wells 17000 droplets 50 copies/
droplet ¼ 108).
Droplet based amplification of the retained pool (see Note 10)
6. From the remaining 24–25 μl of desalted retained library,
prepare an appropriate volume (¼ 20 μl # of required wells
(¼ ~2–2.4 ml)) of ddPCR sample mixture by adding ddPCR
master mix (1 final), 900 nM each of FP and phosphorylated
RP (final) and nanopure water.
7. Load a 20 μl aliquot of this sample mixture into each well of a
droplet-generation cartridge.
8. Add 70 μl of fluorinated oil to each of the corresponding oil
wells of the cartridge.
9. Insert and process the cartridge in the droplet generator.
10. Transfer the stable emulsion (containing ca. 17,000 readable
droplets) formed in each well into a well of a standard 96-well
PCR plate for thermal cycling.
11. Repeat steps 7 to 10 until the sample is fully processed.
12. Begin cycling with an initial activation step at 95 C for 5 min,
followed by 35 amplification cycles (note that this is fewer
cycles than typically used for ddPCR quantitation), each
composed of denaturation at 95 C for 30 s then annealing/
extension at 64 C for 30 s. Set the heating and cooling rate-
s at 2.5 C s1 to ensure even heating of all droplets in
each well.
Fig. 4 (continued) are required to accommodate 50 templates into each droplet (CPD ¼ 50). As a result of the
low sequence heterogeneity per droplet, minimal HMW by-products formation is observed over 40 or more
ddPCR cycles, permitting high-fidelity end-point amplification of all retained library members into more than
1014 total copies of the desired 80-bp dsDNA amplicon products. Adapted from [15] with journal permission
13. Immediately after amplification, pool all of the wells and spin at
5000 g to separate the reacted droplets from the continuous
oil (bottom) phase.
14. Discard the continuous oil phase.
15. Recover the double-stranded DNA by subjecting the droplet
phase to a freeze (80 C for 15 min)/thaw cycle.
16. Immediately spin down the frozen droplets at 14,000 g for
5 min to create sufficient force to burst them.
17. Repeat steps 15 and 16 two more times to generate and
recover a clear aqueous (top) phase containing the soluble
amplified material. Set aside 5 μl of the clarified aqueous phase.
18. Process the remaining clarified aqueous phase in a centrifugal
filter unit to concentrate the amplified library to ~25 μl.
19. Analyze the 5 μl aliquot from step 18 on a 1.5% agarose gel
alongside a properly sized molecular weight ladder to confirm
proper amplification of the library.
3.5 Measuring In general, a proper balancing of retained library sequence diversity

the Sequence Diversity and mean binding affinity must be maintained across selection
of the Amplified rounds for a Hi-Fi SELEX screening to prove successful in discov-
Retained Pool ering a useful set of candidate aptamers. Maintaining that balance
requires methods to quantify both the mean Kd and the total
sequence diversity of the retained pool after each selection round.
Retained library diversities can be determined using next-
generation sequencing, but that approach is neither fast nor inex-
pensive [38, 39].
In Hi-Fi SELEX, the sequence diversity of a retained pool of
library members is estimated using a novel qPCR assay that is
simple and inexpensive, requiring equipment common to all molec-
ular biology labs. The area of the melting peak for each population
is recorded and the two peak areas are used to compute the fraction
fhetDNA of amplicons in the heteroduplexed state:
A 67 C
f hetDNA ¼ ð3Þ
ðA 67 C þ A 81 C Þ

This melt analysis is repeated for serial reductions in the

sequence diversity of the dsDNA-amplicon representations of the
starting library (see above) to generate data as shown in Fig. 5. The
corresponding peak areas for the set of serial reductions are used to
create a standard curve relating fhetDNA to the known sequence
diversity (Fig. 6). For those standards, the area of the Tm ¼ 67 C
melting peak decreases with decreasing sequence diversity, reflect-
ing the higher probability of forming fully complementary homo-
duplexes (which melt at Tm ¼ 81 C) in such systems. The value of
fhetDNA measured for a retained pool thereby permits estimation of
its sequence diversity (see Note 11).
Fig. 5 Fluorescence (SYBR green) based melt analysis for serial reductions in the sequence diversity of
dsDNA-amplicon representations of an 80-nt Hi-Fi SELEX library: The double stranded amplicons of retained
library members are homogeneous in terms of their two flanking sequences, while presenting a highly diverse
ensemble of variable core-region sequences. Denaturation of the amplified library followed by cooling to 55 C
therefore results in two distinct dsDNA populations: fully homoduplexed amplicons characterized by a
Gaussian melting envelope centered at a relatively high melting temperature (Tm ~ 81 C), and heteroduplexes
exhibiting only partial complementarity (typically through only their common flanking sequences). The
heteroduplexed pool of amplicons collectively exhibits a Gaussian melting peak characterized by a much
lower Tm (~67 C).
Fig. 6 Standard curve for qPCR-based sequence diversity determination: normalized melt peak areas (a) and
fhetDNA values (b) as a function library diversity
The protocol used to measure sequence diversity is as follows:

1. Amplify the remaining 5 μl volume of the diluted working
aliquot prepared in step 1 from Section 3.4 by qPCR using
SYBR Green master mix (1 final) and qPCR thermal cycling
conditions composed of initial activation at 95 C for 5 min
followed by 13–15 amplification cycles: denaturation at 95 C
for 30 s, annealing at 64 C for 30 s and extension at 72 C for
30 s, with the heating and cooling ramp rate set at 3 C s1 (see
Note 12).
2. Following cycling, cool the SYBR dye-bearing amplicons to

55 C.
3. Then, melt by heating from 55 C to 95 C in 0.5 C incre-
ments in a real-time PCR thermal cycler and record the
required A67 C and A81 C values.
4. Determine the sequence diversity from the standard curve
relating fhetDNA to the known sequence diversity (e.g., Fig. 6).
3.6 Enzymatic The ddPCR amplification step generates 80-bp dsDNA products
Regeneration from which the sense strands must be recovered in order to con-
of the Single-Stranded tinue the Hi-Fi SELEX process. A number of purification methods
Library have been established for stoichiometric removal of the antisense
strand from the sense (aptamer library) strand of amplicons, includ-
ing alkaline denaturation followed by streptavidin capture of bioti-
nylated antisense strands (generated by chemically modifying the
RP) [40] or electrophoretic separation of poly-T-labeled antisense
strands [41]. The relative performance of these various methods
has been studied by Civit et al. [42] and others [43, 44], who have
collectively shown that enzymatic digestion of 50 -phosphorylated
(PO4) antisense strands with λ-exonuclease generally works best.
However, λ-exonuclease hydrolysis activity is appreciably lower
on ssDNA than on complementary dsDNA [45]. As a result, we
have shown that enzymatic regeneration of a ssDNA library stalls
when acting on heteroduplexed amplicons produced in the bulk
PCR amplification step employed in standard SELEX [15]. A par-
tially regenerated library composed of a mixture of desired 80-nt
ssDNA and partially processed dsDNA material is therefore created
(Fig. 7, right half), compromising the next round of selection.
In Hi-Fi SELEX, in addition to mitigating formation of
unwanted HMW products, the ddPCR based amplification of
retained members described above provides a means to eliminate
formation of heteroduplexes during amplification and thereby yield
a pool of fully homoduplexed amplicons (as evidenced by melt
analyses of those pools (Fig. 7)). Complete stoichiometric regener-
ation of the required 80-nt ssDNA library by λ-exonuclease proces-
sing can then be achieved (Fig. 7, left half) using the following
protocol:
1. Generate the sense-strand library from the 25 μl pool of con-
centrated amplicons prepared in step 18 from Section 3.4 by
reacting with 5 U of λ-exonuclease at 37 C for 1 h, followed by
heat inactivation at 75 C for 10 min.
2. Purify the digested product using standard phenol chloroform
extraction by adding to the sense-strand library a 2 volume
(i.e., ~50 μl) of phenol:chloroform:isoamyl alcohol (25:24:1).
3. Vortex to create an emulsion and centrifuge at 14,000 g for
5 min at room temperature.
Fig. 7 Efficient stoichiometric regeneration of the ssDNA library in Hi-Fi SELEX: (a) ddPCR amplification of
retained members results in formation of fully complimentary homoduplexed amplicons, while conventional
bulk PCR used in SELEX yields a mixture of homoduplexed and heteroduplexed amplicons; (b) λ-exonuclease
processing of the homoduplexed amplicons produced by Hi-Fi SELEX results in stoichiometric recovery of the
ssDNA library in 30 min, while relatively little ssDNA product is recovered from the bulk PCR product. Adapted
from [15] with journal permission
4. Recover the top aqueous phase and repeat steps 2 and

3 once more.
5. Mix the recovered top aqueous phase with 2.5 volumes of
ice-cold 100% ethanol and incubate at 80 C for 2 h to
precipitate the nucleic acids.
6. Following this incubation, centrifuge the tube at 14,000 g
for 30 min in a refrigerated centrifuge set at 4 C.
7. Wash the resulting pellet, which should be clearly visible, twice
with 500 μl of 70% ethanol by centrifugation at 14,000 g and
4 C for 2 min.
8. Air-dry the recovered nucleic acid pellet.
9. Resuspend the pellet in 25 μl AF buffer. Following regenera-
tion, the retained library is typically at a concentration of
ca. 1 μM, which is suitable for the next round of selection.
3.7 Measuring In conventional SELEX, the selection process is typically monitored

the Mean Kd by measuring the mean Kd of retained pools after blindly conduct-
of the Retained ing a few selection rounds to enrich sufficient amounts of high-
and Regenerated affinity binders to make Kd determination possible (typically by
Library surface plasmon resonance (SPR) or fluorescence spectroscopy).
Parallel PCR amplifications followed by antisense strand removal

are required to produce the required quantities of regenerated
ssDNA library. If the mean Kd is to be determined by spectroscopy,
that process must include fluorescent labeling or biotinylation of
the regenerated library; no modification is required if SPR is to be
used. Either process is time consuming, fails to provide information
in the critical early rounds of selection, and is costly if SPR is used to
measure mean Kd values. Moreover, a risk of altered binding prop-
erties that bias selection is created if modification of library mem-
bers is required.
To reduce costs and eliminate need for specialized SPR or
fluorescence spectroscopy equipment, Hi-Fi SELEX incorporates
a novel qPCR-based method to quantify the mean Kd. The method
is based on creating a standard relation to quantify the concentra-
tion of library CLibrary in a given sample before and after incubation
with an immobilized target, and then using that knowledge to
measure the adsorption isotherm for binding of the regenerated
library to protein target presented on Nunc Immobilizer Amino
plates as described above. This simple qPCR method has been used
to show that Hi-Fi SELEX typically delivers a retained pool offering
a mean Kd μM after the first round of selection, and a mean Kd of
order nM after three rounds of selection [15]. It may be used to
monitor the mean Kd after each round of selection and to terminate
selection when the mean bulk affinity of retained members either
remains unchanged for consecutive rounds or has reached a suitable
value (often a mean Kd of order nM) for a particular application. It
requires accurate determination of CLibrary values, necessitating a
statistically more robust measurement protocol.
The protocol is as follow:
1. 2 dilute in SP the retained 25 μl of 80-nt ssDNA library.
2. Take 4 μl of that diluted library and dilute it 1000-fold in SP.
3. From step 2, prepare a serial dilution set (2, 4, 8, 16,
32, 64, 128, 256) to a final volume of 500 μl each.
4. Subject two 5 μl aliquots of each dilution to qPCR-based
determination using steps 2 and 3 from Section 3.4.
5. Record the mean Cq value (for a given RFU threshold value T)
and standard error computed from the duplicates for each
dilution.
6. In a base-10 semi-logarithmic graph, plot the threshold cycle
versus the dilution factor and fit the data to a straight line and
record the slope and error in it.
7. Compute the amplification efficiency E from that slope as
10–1/slope 1.
8. Determine CLibrary from the Cq value for each dilution using
the standard relation (see Note 13)

log C Library ¼ logðT Þ C q logðE Þ ð4Þ
9. In duplicate, incubate each of the 8 serial dilutions (200 μl

each) in a target-presenting Nunc well for 1 h at 25 C with
gentle mixing at 300 rpm.
10. Remove the solution and then successively wash the retained
pools in the equilibrated wells 3 with 300 μl of AF.
11. Remove the final wash from each well.
12. Elute the retained members in 100 μl of 50 mM NaOH at
70 C.
13. Immediately neutralize the eluted pool from each well with
100 μl of 50 mM HCl.
14. Add 100 μl of 10 mM TE buffer.
15. Dilute 5 μl of each neutralized eluted pool by 50-fold to make
the library concentration appropriate for analysis.
16. In duplicate, quantify the CLibrary in each neutralized eluent by
qPCR according to steps 2 and 3 from Section 3.4.
17. For each dilution, determine the fraction θ of the library that is
bound to the target as the ratio of the dilution-corrected values
of CLibrary for the eluted and initial samples. The concentration
of library remaining in the solution phase (Cfree) after equili-
bration with the immobilized target is given by the difference
in dilution-corrected CLibrary values for the initial and eluted
libraries, respectively.
18. From the θ and Cfree values, construct the binding isotherm as
illustrated in Fig. 8. Nonlinear fitting of the Langmuir iso-
therm relation
C free
θ¼ ð5Þ
K d þ C free
to the binding isotherm data yields a value for the mean Kd.
4 Notes
1. The initial library must be synthesized at a 1 μmole scale in

order to generate enough material for conducting Hi-Fi
SELEX. Following synthesis, the library is purified as described
and then lyophilized. When required, the lyophilized product
is reconstituted at a 100 μM concentration in 1 AF buffer and
stored at 20 C away from all other reagents used in the Hi-
Fi-SELEX protocol.
2. Covalent immobilization of the target onto the well surface
proceeds through reaction of amino groups (or other nucleo-
philes) on the protein with the electrophilic coupling agents
Fig. 8 Representative adsorption isotherm and mean Kd value determined using the qPCR based binding
analysis method: data shown are for a retained pool of tight binding library members with fixed regions against
human α-thrombin
displayed on polyglycol chain ends of the Nunc Immobilizer

well. The manufacturer recommends conducting this reaction
at pH 9.6, but the use of a basic pH raises possibilities for
chemical modifications (e.g., de-amidation) that alter target
protein structure, chemistry and/or activity. Hi-Fi SELEX
therefore uses the alternate but equally efficient coupling con-
ditions described in steps 1–7 of the protocol. Under these
milder conditions, the reaction proceeds more slowly and, in
coordination with selection of an appropriate target solution
concentration, permits facile control of the surface density of
immobilized target.
3. The hydrodynamic diameter of a random 80-mer ssDNA apta-
mer is ca. 8 nm. We therefore set the solution concentration
(50–100 nM) of the target protein in the coupling reaction so
as to achieve a mean distance of separation of immobilized
target molecules that is a bit greater than 8 nm (i.e.,
9–10 nm). For thrombin (MW ¼ 37 kDa), this is achieved
using a solution concentration of 80 nM. A lower concentra-
tion would be required for targets of higher molecular weight,
and vice versa. To determine the optimal solution concentra-
tion, the bound moles of target are measured as a function of
solution concentration using a mass balance and A280 nm values
for the starting and final solution. The contact area of the well
(0.95 cm2) is then used to compute the average distance of
separation of immobilized target as a function of the starting
solution concentration used.
4. During this wash sequence, the amines (Tris (hydroxyl-methyl-
aminomethane)) present in SP buffer neutralize unreacted
electrophiles displayed on the end-grafted polyglycol surface.

Supplementing the SP buffer used in both the 3X wash and 1X
incubation steps with 0.005% Tween 20 is absolutely essential
to fully passivate the surface against nonspecific retention.
5. In conventional SELEX, one round of screening in the absence
of the molecular target is often first performed in an attempt to
reduce the number of members of the library retained through
mechanisms unrelated to the target. This step is not required in
Hi-Fi SELEX, as the unique surface passivation methods used
have been shown to reduce nonspecific retention to undetect-
able levels [15].
6. The competent library concentration used (10 μM), in combi-
nation with setting the surface density of immobilized target to
achieve a mean separation distance of ~9 nm, results in library
screening within the theoretically preferred 100:1 to 1000:1
aptamer-to-target range [46, 47], while also eliminating the
possibility of bridging of bound library members between
proximal targets, an effect that has been shown to confound
aptamer selection [48].
7. The polyanionic form of aptamers naturally creates the
unwanted potential to over-select for sequences whose binding
is dominated by coulombic (electrostatic) interactions. The
second set of high-salt washes employed in Hi-Fi SELEX are
designed to remove library members that bind the target
through ion-exchange type mechanisms that generally lack
sufficient specificity for the target.
8. It is important to ensure that the immobilized target and
aptamers are always kept hydrated. When done properly, the
passivation step performed after target immobilization will
significantly increase the hydrophilicity of the Nunc Immobi-
lizer Amino well surface, in part due to the presence of Tween-
20. The surface of the wells thereby remains hydrated during
various required solution exchanges provided vacuum aspira-
tors are not used. A single or multichannel pipette should
instead be used to add and remove solutions.
9. The molecular weight of the 80-nt aptamer is 24 kDa. We
therefore use a 10 kDa molecular weight cut-off membrane
for this step to ensure good aptamer recovery yields. Buffer
exchange should proceed until the final concentration of salt in
the aptamer pool falls to within the mM range.
10. The number of reaction wells required (and thus total droplets
generated) is defined by setting the average copies per droplet
(CPD ¼ ln(empty droplets/total read droplets)) at 50 library
members. We have found that at a CPD of 50 (or lower),
end-point amplification of the library is achieved with minimal
accumulation of HMW by-products [15]. However, ddPCR at
a CPD > 50 results in by-product formation and a concomi-

tant loss of desired 80-bp amplicon product (and product
quality) in a manner similar to that observed in standard PCR
of retained library members. The CPD limit applies to libraries
retained in the first two rounds of selection, where each of the
50 templates within each droplet is almost certain to harbour a
unique sequence within its variable core region. In later rounds
of selection, duplicate library sequences will increasingly parti-
tion into any given drop, diminishing template heterogeneity
per drop and the probability of forming by-products during
amplification. Higher CPDs may then be used, up to a CPD of
no greater than 1000.
11. As indicated in Fig. 6, the qPCR-based diversity assay provides
reliable estimates of sequence diversity for retained libraries
comprising 1000 unique variable-region sequences. Hi-Fi
SELEX screening is typically stopped before the retained
library diversity falls to that limit, as the target mean Kd of
the retained population has generally been met. But in a case
where characterization of less diverse libraries is required, one
may employ an alternative PCR-based method for characteriz-
ing pool diversities that relies on fluorometric quantitation of
dsDNA after amplification. The protocol for that method is
provided in a previous publication [15].
12. The number of cycles is set so as to generate sufficient amplified
material from the amount of retained library present in the
previously quantified starting aliquot. Typically, ca. 13 to
15 cycles is sufficient to generate the 10 ng of 80-bp ampli-
con required for the diversity assay. Note that cycling should be
stopped (~16–20 cycles) before production of HMW amplifi-
cation by-products is observed in gel-based visualization of the
reaction product. The pilot PCR reaction is performed first to
determine the optimal cycle number, where a small amount of
the pool is amplified using standard 50 μl PCR. A sample is
taken out every 3 cycles and a gel-based visualization is per-
formed similar to that in Fig. 4. The optimal cycle number
should yield enough material to proceed to the next step. The
amount of product can be quantified using NanoDrop™
ND2000 spectrophotometer.
13. Determination of library concentrations can be done using
other methods. In our experience, the oft-used NanoDrop™
ND2000 spectrophotometer is not sufficiently accurate in
measuring dsDNA at concentrations below ca. 50 μg/ml,
typically overestimating the true concentration. The Qubit™
Fluorescence Monitor is generally accurate at these concentra-
tions, and one can use it as a secondary check of CLibrary values.
References
1. Edwards BM, Barash SC, Main SH, Choi GH, speed, and success of SELEX methods and
Minter R, Ullrich S, Williams E, Du Fou L, aptamer characterization. Mol Ther Nucleic
Wilton J, Albert VR, Ruben SM, Vaughan TJ Acids 3:e183
(2003) The remarkable flexibility of the human 14. Ouellet E, Lagally ET, Cheung KC, Haynes C
antibody repertoire–isolation of over one thou- (2014) A simple method for eliminating fixed-
sand different antibodies to a single protein, region interference of aptamer binding during
BLyS. J Mol Biol 334:103–118 SELEX. Biotechnol Bioeng 111
2. Xing PX, Russell S, Prenzoska J, McKenzie JF (11):2265–2279
(1994) Discrimination between alternatively 15. Ouellet E, Foley JH, Conway EM, Haynes C
spliced STP-A and -B isoforms of CD46. (2015) Hi-fi SELEX: a high-fidelity digital
Immunology 83:122–127 PCR based therapeutic aptamer discovery plat-
3. Raska M, Czernekova L, Moldoveanu Z, form. Biotechnol Bioeng 112(8):1506–1522
Zachova K, Elliott M, Novak Z, Hall S, 16. Bartel DP, Zapp ML, Green MR, Szostak JW
Hoelscher M, Maboko L, Brown R, Smith P, (1991) HIV-1 rev regulation involves recogni-
Mestecky J, Novak J (2014) Differential glyco- tion of non-Watson-Crick base pairs in viral
sylation of envelope gp120 is associated with RNA. Cell 67:529–536
differential recognition of HIV-1 by virus- 17. Lorenz C, von Pelchrzim F, Schroeder R
specific antibodies and cell infection. AIDS (2006) Genomic systematic evolution of
Res Ther 11:1–16 ligands by exponential enrichment (genomic
4. Mould DR, Sweeney KR (2007) The pharma- SELEX) for the identification of protein-
cokinetics and pharmacodynamics of monoclo- binding RNAs independent of their expression
nal antibodies–mechanistic modeling applied levels. Nat Protoc 1:2204–2212
to drug development. Curr Opin Drug Discov 18. Zimmermann B, Bilusic I, Lorenz C, Schroe-
Devel 10:84–96 der R (2010) Genomic SELEX: a discovery
5. Chapman AP, Antoniw P, Spitali M, West S, tool for genomic aptamers. Methods
Stephens S, King DJ (1999) Therapeutic anti- 52:125–132
body fragments with prolonged in vivo half- 19. Keefe AD, Cload ST (2008) SELEX with mod-
lives. Nat Biotech 17:780–783 ified nucleotides. Curr Opin Chem Biol
6. Chames P, Van Regenmortel M, Weiss E, Baty 12:448–456
D (2009) Therapeutic antibodies: successes, 20. Kuwahara M, Obika S (2013) In vitro selection
limitations and hopes for the future. Br J Phar- of BNA (LNA) aptamers. Artif DNA PNA
macol 157:220–233 XNA 4:39–48
7. Ruigrok VJ, Levisson M, Eppink MH, 21. Klussmann S, Nolte A, Bald R, Erdmann VA,
Smidt H, van der Oost J (2011) Alternative Furste JP (1996) Mirror-image RNA that
affinity tools: more attractive than antibodies? binds D-adenosine. Nat Biotech
Biochem J 436:1–13 14:1112–1115
8. Skerra A (2007) Alternative non-antibody scaf- 22. Hall B, Micheletti JM, Satya P, Ogle K,
folds for molecular recognition. Curr Opin Pollard J, Ellington AD (2001) Design, syn-
Biotechnol 18:295–304 thesis, and amplification of DNA pools for
9. Jayasena SD (1999) Aptamers: an emerging in vitro selection. In: Current protocols in
class of molecules that rival antibodies in diag- molecular biology. John Wiley & Sons, Inc.,
nostics. Clin Chem 45:1628–1650 Hoboken, NJ, pp 24.2.1–24.227
10. Tuerk C, Gold L (1990) Systematic evolution 23. Jiménez E, Sefah K, López-Colón D, Van
of ligands by exponential enrichment: RNA Simaeys D, Chen HW, Tockman MS, Tan W
ligands to bacteriophage T4 DNA polymerase. (2012) Generation of lung adenocarcinoma
Science 249(4968):505–510 DNA aptamers for cancer studies. PLoS One
11. Ellington AD, Szostak JW (1990) In vitro 7(10):e46222
selection of RNA molecules that bind specific 24. Sefah K, Shangguan D, Xiong X, O’Donoghue
ligands. Nature 346(6287):818–822 MB, Tan W (2010) Development of DNA
12. Robertson DL, Joyce GF (1990) Selection aptamers using cell-SELEX. Nat Protoc
in vitro of an RNA enzyme that specifically 5:1169–1185
cleaves single-stranded DNA. Nature 344 25. Gopinath SC, Sakamaki Y, Kawasaki K, Kumar
(6265):467–468 PK (2006) An efficient RNA aptamer against
13. Ozer A, Pagano JM, Lis JT (2014) New tech- human influenza B virus hemagglutinin. J Bio-
nologies provide quantum changes in the scale, chem 139:837–846
26. Weng C-H, Huang C-J, Lee G-B (2012) Lu H, Makhijani VB, McDade KE, McKenna
Screening of aptamers on microfluidic systems MP, Myers EW, Nickerson E, Nobile JR,
for clinical applications. Sensors 12 Plant R, Puc BP, Ronan MT, Roth GT, Sarkis
(7):9514–9529 GJ, Simons JF, Simpson JW, Srinivasan M, Tar-
27. Gong P, Grainger DW (2007) Nonfouling sur- taro KR, Tomasz A, Vogt KA, Volkmer GA,
faces: a review of principles and applications for Wang SH, Wang Y, Weiner MP, Yu P, Begley
microarray capture assay designs. Methods Mol RF, Rothberg JM (2005) Genome sequencing
Biol 381:59–92 in microfabricated high-density picolitre reac-
28. Cattani-Scholz A, Pedone D, Blobner F, tors. Nature 437:376–380
Abstreiter G, Schwartz J, Tornow M, Andruzzi 38. Schütze T, Wilhelm B, Greiner N, Braun H,
L (2009) PNA-PEG modified silicon platforms Peter F, Mörl M, Erdmann VA, Lehrach H,
as functional bio-interfaces for applications in Konthur Z, Menger M, Arndt PF, Glökler J
DNA microarrays and biosensors. Biomacro- (2011) Probing the SELEX process with next-
molecules 10:489–496 generation sequencing. PLoS One 6:e29604
29. Schlapak R, Pammer P, Armitage D, Zhu R, 39. Hoon S, Zhou B, Janda KD, Brenner S, Scol-
Hinterdorfer P, Vaupel M, Frühwirth T, nick J (2011) Aptamer selection by high-
Howorka S (2005) Glass surfaces grafted with throughput sequencing and informatic analy-
high-density poly(ethylene glycol) as substrates sis. BioTechniques 51:413–416
for DNA oligonucleotide microarrays. Lang- 40. Kai E, Sumikura K, Ikebukuro K, Karube I
muir 22:277–285 (1998) Purification of single stranded DNA
30. Poncin-Epaillard F, Vrlinic T, Debarnot D, from asymmetric PCR product using the
Mozetic M, Coudreuse A, Legeay G, El SMART system. Biotechnol Tech 12:935–939
Moualij B, Zorzi W (2012) Surface treatment 41. Pagratis NC (1996) Rapid preparation of single
of polymeric materials controlling the adhesion stranded DNA from PCR products by strepta-
of biomolecules. J Funct Biomater 3:528–543 vidin induced electrophoretic mobility shift.
31. SantaLucia J Jr, Hicks D (2004) The thermo- Nucleic Acids Res 24:3645–3646
dynamics of DNA structural motifs. Annu Rev 42. Civit L, Fragoso A, O’Sullivan CK (2012)
Biophys Biomol Struct 33:415–440 Evaluation of techniques for generation of
32. Lou X, Qian J, Xiao Y, Viel L, Gerdon AE, single-stranded DNA for quantitative detec-
Lagally ET, Atzberger P, Tarasow TM, Heeger tion. Anal Biochem 431:132–138
AJ, Soh HT (2009) Micromagnetic selection of 43. Svobodova M, Pinto A, Nadal P, O’Sullivan
aptamers in microfluidic channels. Proc Natl CK (2012) Comparison of different methods
Acad Sci 106:2989–2994 for generation of single-stranded DNA for
33. Nieuwlandt D (2000) In vitro selection of SELEX processes. Anal Bioanal Chem
functional nucleic acid sequences. Curr Issues 404:835–842
Mol Biol 2:9–16 44. Avci-Adali M, Paul A, Wilhelm N, Ziemer G,
34. Musheev MU, Krylov SN (2006) Selection of Wendel HP (2009) Upgrading SELEX tech-
aptamers by systematic evolution of ligands by nology by using lambda exonuclease digestion
exponential enrichment: addressing the poly- for single-stranded DNA generation. Mole-
merase chain reaction issue. Anal Chim Acta cules 15:1–11
564:91–96 45. Lee G, Yoo J, Leslie BJ, Ha T (2011) Single-
35. Nakano M, Komatsu J, Matsuura S-i, molecule analysis reveals three phases of DNA
Takashima K, Katsura S, Mizuno A (2003) degradation by an exonuclease. Nat Chem Biol
Single-molecule PCR using water-in-oil emul- 7:367–374
sion. J Biotechnol 102:117–124 46. Irvine D, Tuerk C, Gold L (2001) Selexion:
36. Williams R, Peisajovich SG, Miller OJ, systematic evolution of ligands by exponential
Magdassi S, Tawfik DS, Griffiths AD (2006) enrichment with integrated optimization by
Amplification of complex gene libraries by non-linear analysis. J Mol Biol 222
emulsion PCR. Nat Methods 3:545–550 (3):739–761
37. Margulies M, Egholm M, Altman WE, 47. Vant-Hull B, Payano-Baez A, Davis RH, Gold
Attiya S, Bader JS, Bemben LA, Berka J, L (1998) The mathematics of SELEX against
Braverman MS, Chen Y-J, Chen Z, Dewell complex targets. J Mol Biol 278(3):579–597
SB, Du L, Fierro JM, Gomes XV, Godwin 48. Ozer A, White BS, Lis JT, Shalloway D (2013)
BC, He W, Helgesen S, Ho CH, Irzyk GP, Density-dependent cooperative non-specific
Jando SC, Alenquer MLI, Jarvie TP, Jirage binding in solid-phase SELEX affinity selec-
KB, Kim J-B, Knight JR, Lanza JR, Leamon tion. Nucleic Acids Res 41:7167–7175
JH, Lefkowitz SM, Lei M, Li J, Lohman KL,
INDEX
A Cell-free DNA (cfDNA) ..................................... 5, 46–50,

52–62, 161, 164, 166, 167, 194, 196–199, 202,
Absolute quantification...................................4, 6, 18, 19, 211, 214, 223, 278, 279, 298, 303–305, 310,
42, 70, 99, 113, 120, 168, 230, 258, 265, 276, 312, 318, 319, 335–337, 340–343, 345–347,
283, 316, 323, 350, 364, 373, 377–380, 398,
365, 442, 509
425, 446, 453, 459–470, 478, 519 Cell-sorting........................................................... 424, 426
Accuracy............................................................... 7, 15, 16, Cerebrospinal fluid (CSF) ..............................4, 101, 102,
18, 20, 21, 78, 115, 119, 123, 138, 144, 174,
107, 111, 114, 116, 120, 445, 460, 465
183, 230, 258, 296, 375, 380, 389, 397, 414, Circulating cell-free DNA (ccfDNA)............45, 324–327
446, 478 Clonal evolution............................................................ 275
Allele burden ........................................... 5, 257, 267–269
Co-localization ............................................ 393, 490, 491
Allele-specific expression............................. 407, 414, 420 Colorectal cancer (CRC) ....................................... 13, 363
Allelic imbalance......................... 408–414, 418, 420, 490 Complex alleles ............................................................. 490
Allelic skew ..........................................402, 404, 406, 415
Compound heterozygote ............................................. 489
Aptamers........................................................... 7, 531–552 Control ................................................................ 7, 32, 46,
Assay design..........................................................v, 4, 5, 7, 74, 100, 115, 136, 144, 162, 178, 195, 213, 234,
32, 109, 144, 146–148, 151, 152, 174, 194, 195, 261, 278, 313, 325, 343, 351, 366, 402, 431,
201, 242, 277, 279–280, 289, 290, 312, 313,
446, 480, 492, 514, 550
351, 352, 354, 357–359, 367, 369, 374, 403, Control materials ............................................... 19, 20, 45
405–410, 425, 441, 494, 497, 499 Copy number variants (CNVs) .................. 174, 176, 177
Assay multiplexing ....................... 79, 283–286, 327, 428
Copy number variation (CNV) .................................4, 16,
45, 143–159, 326
B
Counting ....................................................................8, 19,
BCL1/IGH ......................................................... 230–237, 25–43, 332, 335, 352, 402, 470, 488, 516, 517,
240–244, 250, 251 523
BCL2/IGH .......................................................... 241, 251 CRISPR/Cas9............................................................... 350
Biofluids......................................................................... 465
Biological affinity reagents............................................ 531 D
Biomarkers........................................................... 4, 12, 41, Deoxyribo nucleic acid (DNA) .................................3, 12,
111–125, 303, 304, 320, 335, 364, 388, 445,
45, 70, 100, 112, 144, 161, 173, 194, 210, 231,
459, 514 258, 275, 323, 335, 349, 363, 365, 387, 423,
Blood .......................................................... 5, 26, 45, 100, 446, 490, 513, 532
111, 127, 162, 178, 194, 210, 232, 259, 279,
Diagnostics .............................................................. 12, 13,
303, 323, 337, 366, 432, 445, 460, 493 21, 22, 41, 45, 46, 48, 70, 111, 112, 128, 235,
Breast cancer........................................................ 4, 5, 169, 237, 245, 248, 257, 258, 304, 531
170, 209–226, 234, 304, 315, 318, 321, 389,
Digital PCR (dPCR)..................................................3, 11,
423, 460 25, 46, 70, 162, 423
DNA extraction............................................................. 128
C
DNA integrity ......................................................... 74, 92,
Calibration............................................................. 12, 112, 137, 377
123, 426, 432 DNA methylation ........................................296, 363–383
Cancer................................................................4, 12, 143, Droplet ................................................................ 5, 14, 61,
161, 174, 193, 209, 234, 275, 303, 323, 363, 70, 99, 122, 128, 144, 163, 174, 198, 210, 230,
388, 460, 513 279, 304, 324, 336, 350, 364, 389, 402, 425,
Cell-free circulating tumor DNA ................275, 303–321 446, 460, 478, 490, 514, 532
vol. 1768, https://doi.org/10.1007/978-1-4939-7778-9, © Springer Science+Business Media, LLC, part of Springer Nature 2018
555
DIGITAL PCR: METHODS AND PROTOCOLS
556 Index
Droplet digital PCR (ddPCR) ..................................5, 51, I
70, 99, 113, 128, 144, 163, 173, 194, 210, 230,
277, 304, 323, 336, 350, 363, 389, 402, 425, Immunoglobulin heavy-chain gene (IGH) ....... 230–240,
446, 460, 479, 492, 515, 532 242–244, 248, 251
Duplex digital PCR.............................................. 128, 276 Inhibitor tolerance ........................................................ 111
E J
Efficiency ............................................................. 4, 12, 16, JAK2 V617F..................................................... 5, 257–272

19–21, 46, 47, 52, 54–56, 58, 59, 61, 70, 92, 123,
L
155, 239, 249, 253, 280, 319, 374, 381, 389,
391, 395, 398, 402, 440, 441, 446, 459, 470, Large DNA isolation............................................ 240, 367
537, 548 Library balancing ................................................. 478, 487
EGFR .......................................... 193–196, 199–202, 489 Library loading............................................ 305, 307, 478
Enterococcus ......................................................... 127, 128, Limit of blank (LOB) ........................................ 21, 34, 35
130, 133, 136, 139 Limit of detection (LOD) .........................................7, 21,
Extraction ..................................7, 45, 74, 101, 112, 128, 27, 33, 34, 36, 58, 117, 355, 414, 460
162, 211, 232, 305, 324, 337, 391, 446, 491, 545 Linkage .................................................................v, 6, 401,
404, 407, 490–499, 503, 505–510
F Liquid biopsy.........................................4, 5, 26, 210, 426
False negative .......................................................... 20, 21,
M
26, 31, 32, 38–40, 79, 92, 139, 315
False positive........................................................ 8, 21, 26, Mantle cell lymphoma (MCL) ....................................230,
31, 32, 35, 36, 38–41, 62, 70, 79, 94, 95, 119, 235–238, 240, 245, 246, 248, 250
202, 205, 297, 309, 315, 320, 321, 453 MethyLight ......................................................... 364, 365,
Fecal contamination......................................... 4, 127–139 369, 370, 373, 374, 376–380, 382
Follicular lymphoma (FL) .................................. 230, 235, MethyLight ddPCR ............................................ 364, 365,
236, 241, 244, 252 368–370, 373, 374, 376–379, 382
Microbial source tracking ............................................. 128
G MicroRNA (miRNA) .............................................. 6, 445,
Gene expression ...................................................... 6, 143, 454, 459
363, 397, 401, 423, 424, 432 Minimal residual disease (MRD).................................229,
Genetic testing .................................................6, 275, 335 230, 233–236, 238, 242–246, 248, 250–252, 254
Genetically modified organism (GMO).......................v, 4, Mitochondrial DNA (mtDNA)....................................... 4,
69, 70, 74, 77, 78 111, 114, 116, 120, 303
Genome editing ..................................................v, 12, 349 Mobile elements ............................................................ 174
Genomic structural variation.............................................v mRNA alternative splicing................................... 387–400
Genotyping.......................................................... 193, 194, mRNA expression ................................................ 408, 409
202, 205, 267, 276, 323, 337, 343, 344, 490, 509 Multiple myeloma (MM).............................................230,
Genotyping assay design.....................194, 267, 326, 327 235, 236, 238, 239, 246
Genotyping inversions ......................................... 490, 494 Multiplex ....................................................................5, 16,
GM Maize .........................................................69, 71, 72, 48, 69–95, 100, 113, 117, 119, 128, 277, 294,
75, 76, 80, 83, 85, 87, 88, 90, 91, 95 327, 377, 392, 425, 431, 436, 438
Mutation detection .............................................. 258, 292
H Myeloproliferative neoplasms (MPNs) ........................ 257
Haplotype ............................................144, 416, 421, 490 N

HER2................................................................ 4, 161–172
HF183 .................................................128, 130, 133, 139 Nanofluidic chip.............................................................. 14
HHV_6A ........................................................99, 103, 104 Next generation sequencing (NGS) ...........................3, 5,
HHV_6B ........................................................99, 103, 104 6, 230, 275, 298, 423, 425, 477–479, 482, 488
Homology-directed repair (HDR) ................. 6, 349–361 Nonhomologous end joining (NHEJ) ........................... 6,
Human herpesvirus 6..............................................99–109 349, 351, 352, 356–360
Human telomerase reverse transcriptase Noninvasive diagnosis................................................... 303
(hTERT) ..................................388–396, 398, 514 Noninvasive genetic profiling ....................................... 275
DIGITAL PCR: METHODS AND PROTOCOLS
Index 557
Non-small cell lung cancer (NSCLC).................... 5, 193, S
194, 202
Nucleic acids quantification.........................................4, 6, Sensitivity.................................................................. v, 5, 8,
13, 16, 18, 479 14, 33–36, 38–40, 42, 56, 69, 78, 94, 102, 107,
109, 112, 113, 115, 119, 128, 161, 166, 174,
P 176, 177, 183, 202, 210, 213, 214, 220, 229,
230, 245, 254, 258, 259, 261, 275, 279, 287,
Partitioning ..................................................... v, 6, 15, 16, 290, 297–300, 305, 307, 309, 324, 350, 361,
19, 25, 28–31, 41, 99, 113, 144, 175, 258, 402, 364, 377, 389, 398, 402, 425, 439, 460, 470,
459, 490, 531, 541 488, 514, 523
Pathogen detection ....................................................... 275 Sequencing library qualification .......................... 347, 478
Performance characteristics ............................... 18, 37, 38 Sequencing library quantification ....................... 478, 488
Personalized medicine .................................................. 489 Serum..........................................................................6, 49,
Phasing ................................................................. 489–511 101, 102, 107, 210, 303, 319, 365, 432,
PIK3CA ......................................................................... 210 446–448, 452, 453, 459
Plasma ................................................................5, 45, 107, Single cell.............................................................. 423–442
194, 210, 279, 303, 323, 335, 365, 446, 460 Single color........................................................... 323–333
Poisson distribution .................................................27–30, Single nucleotide variations (SNVs) .................. 174–176,
35, 41, 70, 169, 222, 276, 332, 470 181, 186, 326
Polymerase chain reaction (PCR) ........................ 74, 173, Somatic mosaicism ................................................v, 4, 174
230, 258, 540 Specificity ................................................... v, 8, 20, 33–35,
Precision ...............................................................v, 4, 6, 7, 38–40, 61, 62, 78, 94, 104, 109, 113, 128, 175,
11, 14–16, 18, 20, 56, 58, 61, 78, 113, 144, 157, 179, 186, 202, 205, 210, 213, 220, 244, 251,
159, 174, 186, 187, 195, 216, 261, 265–269, 275, 276, 290, 297, 299, 326, 379, 392, 395,
272, 287, 297, 364, 377, 407, 420, 459, 460, 532 470, 525, 531, 532, 536, 551
Standardization ..................................................... 46, 111,
Q
230, 245, 258
Quantification ..................................................... 3, 12, 25, Statistics ...................................................... 15, 16, 18, 19,
46, 69, 99, 112, 128, 162, 174, 230, 245, 276, 25, 27, 28, 30, 31, 34, 94, 157, 158, 203, 225,
303, 323, 335, 350, 364, 388, 425, 445, 459, 258, 264, 315, 380, 412
478, 514 Subsampling ............................................................ 16, 26,
Quantitative PCR (qPCR).........................................4, 12, 27, 30, 31, 36, 40–42
16, 19, 20, 69, 70, 113–115, 128, 144, 148, 175, Systematic evolution of ligands by exponential
230, 245, 251, 254, 276, 280, 343, 344, 389, enrichment (SELEX)................................... 7, 531,
398, 410, 446, 470, 473, 478, 535, 540, 542, 533, 534, 541, 544, 547, 551
545, 548–550, 552
T
R
t(11;14) translocation................................................... 230
Radial multiplexing ................................................. 6, 425, t(14;18) translocation................................................... 230
426, 429, 442 TALEN ........................................................ 350, 360, 361
Rare allele detection.........................................5, 265, 277 Telomerase activity.................6, 388, 389, 398, 513–528
Rare mutation detection............................................... 292 Transcripts ..................................... 6, 387, 393, 394, 399,
Real-time PCR ...........................................................3, 74, 402, 420, 423–425, 429, 438, 441, 442
78, 92, 144, 336, 337, 341–343, 346, 347, 388, Transplantation ............................................100, 335–347
402, 459
Rejection monitoring.................................................... 335 V
Reproducibility......................................................... v, 4, 7, Validation............................................11–22, 38, 78, 111,
13, 20, 22, 112, 113, 123, 230, 258, 449, 460, 144, 174, 177, 186, 245, 283–286, 310, 313,
478 316, 354, 355, 445, 446
Resistance mutations........................................5, 205, 304 Viral detection ............................................................... 100
Ribo nucleic acid (RNA) ...........................................3, 74,
129, 164, 304, 342, 388, 401–421, 424, 446, W
461, 490, 513
Risk biomarkers ............................................................. 364 Water quality ............................................. 4, 11, 128, 129
RNA quantification ........................................................... 6 Whole-genome sequencing ........................ 176, 304, 310

(Methods in Molecular Biology) George Karlin-Neumann, Francisco Bizouarn - Digital PCR - Methods and Protocols-Humana Press (2018)

Uploaded by

Copyright:

Available Formats

(Methods in Molecular Biology) George Karlin-Neumann, Francisco Bizouarn - Digital PCR - Methods and Protocols-Humana Press (2018)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

(Methods in Molecular Biology) George Karlin-Neumann, Francisco Bizouarn - Digital PCR - Methods and Protocols-Humana Press (2018)

Uploaded by

Copyright:

Available Formats

Methods in

Molecular Biology 1768

For further volumes:

Methods and Protocols

George Karlin-Neumann and Francisco Bizouarn

ISSN 1064-3745 ISSN 1940-6029 (electronic)

© Springer Science+Business Media, LLC, part of Springer Nature 2018

Printed on acid-free paper

Pleasanton, CA, USA George Karlin-Neumann

PART I BACKGROUND CONCEPTS

PART II ABSOLUTE QUANTIFICATION

5 Multiplex Droplet Digital PCR Protocols for Quantification

PART III COPY NUMBER VARIATION

11 Detection and Quantification of Mosaic Genomic DNA

PART IV RARE MUTATION AND RARE ALLELE DETECTION

PART V GENE EXPRESSION AND RNA QUANTIFICATION

22 Simultaneous Quantification of Multiple Alternatively

23 Using Droplet Digital PCR to Analyze Allele-Specific

PART VI OTHER USES OF PARTITIONING

27 Droplet Digital™ PCR Next-Generation Sequencing Library

ALEXEJ ABYZOV  Department of Health Sciences Research, Center for Individualized

EMILY C. LEIBOVITCH  Neuroimmunology Branch, Viral Immunology Section, National

PETAR PODLESNIY  Neurobiology Unit, Institut d’Investigacions Biomèdiques de Barcelona,

CHRISTINA M. WOOD-BOUWENS  Division of Oncology, Department of Medicine, Stanford

Entering the Pantheon of 21st Century Molecular Biology

After a 25 year gestation period, digital PCR (dPCR) has finally

RNA from diverse biological sample types. Although its earlier

phenomenon that was little recognized before the availability of

of gene editing experiments where they are able to simultaneously

smaller reactions can greatly improve the signal-to-noise ratio over

of our assay and whether or not we are confident we have detected a

7. Mellert H, Foreman T, Jackson L, Maar D, (6):991–994. https://doi.org/10.1373/

Basic Concepts and Validation of Digital PCR Measurements

Key words Digital PCR, Nucleic acids quantification, Validation

Use of dPCR technology is undoubtedly on the rise considering the

been used in clinical research including development and detection

containing no target DNA through their fluorescence signal. The

2.2 Commercial In 2006, Fluidigm Corporation launched a commercial dPCR

3 Basic Concepts of dPCR

4 Validation of dPCR Assays

4.1 Validation, The term verification is defined as “provision of objective evidence

quantity values being attributed to a measurand, based on the

Target Reaction Endpoint

sampling pipetting partition number of poor or non-specific proportion of (+)

conditions optimized using qPCR are not always optimal for

dPCR has increasingly been proved as a “ground-breaking” tech-

evident by the published literature from both academia and the

1. Strain MC, Lada SM, Luong T et al (2013) https://doi.org/10.1016/j.foodchem.2014.

Fundamentals of Counting Statistics in Digital PCR: I Just

Digital PCR (dPCR) technology promises to change how genomic

In digital PCR the PCR reaction is first partitioned into indepen-

Let us first consider a test that detects a single species of target

5000 expect 10 copies

Fig. 1 Subsampling process. A patient has 5 L of blood in which 5000 molecules

10 actual target molecules. Our ability to make confident state-

0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14

80 90 100 110 120

4 Multiple Occupancy and Partitioning

The original concept of dPCR assumed a limiting dilution regime,

ALEXEJ ABYZOV Department of Health Sciences Research, Center for Individualized

EMILY C. LEIBOVITCH Neuroimmunology Branch, Viral Immunology Section, National

PETAR PODLESNIY Neurobiology Unit, Institut d’Investigacions Biomèdiques de Barcelona,

CHRISTINA M. WOOD-BOUWENS Division of Oncology, Department of Medicine, Stanford