Methods in

Molecular Biology 1730

Martin Giera Editor

Clinical
Metabolomics
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
 Clinical Metabolomics

Methods and Protocols

Edited by

Martin Giera
Center for Proteomics and Metabolomics, Leiden University Medical Center
Leiden, The Netherlands
Editor
Martin Giera
Center for Proteomics and Metabolomics
Leiden University Medical Center
Leiden, The Netherlands

ISSN 1064-3745 ISSN 1940-6029 (electronic)

Methods in Molecular Biology
ISBN 978-1-4939-7591-4 ISBN 978-1-4939-7592-1 (eBook)
https://doi.org/10.1007/978-1-4939-7592-1

Library of Congress Control Number: 2017963765

© Springer Science+Business Media, LLC 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.

Cover illustration: Cover art design by Dr. Magnus Palmblad

Printed on acid-free paper

This Humana Press imprint is published by Springer Nature

The registered company is Springer Science+Business Media, LLC
The registered company address is: 233 Spring Street, New York, NY 10013, U.S.A.
Preface

In recent years, metabolomics has become an inevitable tool in several clinical research fields,
helping to discover new diagnostic markers and molecules and furthering our understand-
ing of pathophysiological processes. Unlike the field of clinical chemistry which today is
integrated into many clinical processes, clinical metabolomics is a much more “juvenile”
discipline still on its way to become fully integrated into modern health care. Nevertheless,
metabolomics is at the core of several very promising initiatives evolving around persona-
lized health care and precision medicine.
Ideally, clinical metabolomics should be seen as a complimentary discipline to clinical
chemistry. The much more hypothesis-driven exploratory nature of clinical metabolomics
allows it to fill the pipelines of clinical chemistry with novel disease markers and diagnostic
patterns. Besides this, clinical metabolomics is very well suited to help clinicians and
biologists understand pathophysiological processes in detail, hopefully allowing us to design
novel treatment strategies and therapies. Its multidisciplinary nature covering (analytical)
chemistry, biology, bioinformatics, and pathology necessitates that scientists from various
fields understand each other. Hence, a common fundament for communication is a manda-
tory prerequisite for the successful embedding of clinical metabolomics into modern
disease-related research. When communicating with colleagues from various disciplines, it
is of utmost importance to the planning of joint studies that every partner understands the
needs and limitations of one another. In multidisciplinary projects, this particularly applies
to the fact that each partner should be aware of practical requirements and limitations of the
different methods and technologies used. Therefore, exchanging experimental protocols
and making colleagues aware of critical practical considerations is of vital importance for a
successful study outcome.
With this book, we hope to present a comprehensive compendium of clinical metabo-
lomics protocols covering LC-MS-, GC-MS-, CE-MS-, and NMR-based clinical metabo-
lomics as well as bioinformatics and study design considerations. We hope that this book will
serve as the basis for the successful (practical) communication between scientists from
several fields, including chemists, biologists, bioinformaticians, and clinicians, ultimately
leading to successful study design and completion.

Leiden, The Netherlands Martin Giera

v
Contents

Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v
Contributors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi

PART I CLINICAL METABOLOMICS AND LIPIDOMICS

1 Metabolomics as a Tool to Understand Pathophysiological Processes. . . . . . . . . . 3
Julijana Ivanisevic and Aurelien Thomas
2 Metabolomics in Immunology Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
Bart Everts

PART II LC-MS-BASED METABOLOMICS

3 LC-MS-Based Metabolomics of Biofluids Using All-Ion

Fragmentation (AIF) Acquisition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
Romanas Chaleckis, Shama Naz, Isabel Meister, and Craig E. Wheelock
4 Lipid Mediator Metabolomics Via LC-MS/MS Profiling and Analysis . . . . . . . . . 59
Jesmond Dalli, Romain A. Colas, Mary E. Walker, and Charles N. Serhan
5 UHPSFC/ESI-MS Analysis of Lipids. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
Miroslav Lı́sa and Michal Holčapek
6 LC-MS/MS Analysis of Lipid Oxidation Products in Blood
and Tissue Samples. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
Yiu Yiu Lee and Jetty Chung-Yung Lee
7 Serum Testosterone by Liquid Chromatography Tandem Mass
Spectrometry for Routine Clinical Diagnostics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
Lennart J. van Winden, Olaf van Tellingen, and Huub H. van Rossum
8 LC-MS/MS Analysis of Bile Acids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
Sabrina Krautbauer and Gerhard Liebisch
9 LC-MS/MS Analysis of Triglycerides in Blood-Derived Samples. . . . . . . . . . . . . . 111
Madlen Reinicke, Susen Becker, and Uta Ceglarek
10 LC-MS/MS Analysis of the Epoxides and Diols Derived
from the Endocannabinoid Arachidonoyl Ethanolamide . . . . . . . . . . . . . . . . . . . . . 123
Amy A. Rand, Patrick O. Helmer, Bora Inceoglu,
Bruce D. Hammock, and Christophe Morisseau
11 Sphingolipid Analysis in Clinical Research. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
Bo Burla, Sneha Muralidharan, Markus R. Wenk,
and Federico Torta
12 Shotgun Lipidomics Approach for Clinical Samples . . . . . . . . . . . . . . . . . . . . . . . . . 163
Lars F. Eggers and Dominik Schwudke

vii
viii Contents

13 Establishing and Performing Targeted Multi-residue Analysis

for Lipid Mediators and Fatty Acids in Small Clinical Plasma Samples . . . . . . . . . 175
Theresa L. Pedersen and John W. Newman
14 Chemical Isotope Labeling LC-MS for Human Blood
Metabolome Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
Wei Han and Liang Li
15 Direct Infusion-Tandem Mass Spectrometry (DI-MS/MS)
Analysis of Complex Lipids in Human Plasma and Serum
Using the Lipidyzer™ Platform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
Baljit K. Ubhi

PART III GC-MS-BASED METABOLOMICS

16 Exploratory GC/MS-Based Metabolomics of Body Fluids . . . . . . . . . . . . . . . . . . . 239

Carole Migné, Stéphanie Durand, and Estelle Pujos-Guillot
17 GC-MS Analysis of Short-Chain Fatty Acids in Feces, Cecum
Content, and Blood Samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
Lisa R. Hoving, Marieke Heijink, Vanessa van Harmelen,
Ko Willems van Dijk, and Martin Giera
18 GC-MS Analysis of Medium- and Long-Chain Fatty Acids
in Blood Samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
Lisa R. Hoving, Marieke Heijink, Vanessa van Harmelen,
Ko Willems van Dijk, and Martin Giera
19 Analysis of Oxysterols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
Fabien Riols and Justine Bertrand-Michel
20 Analysis of Metabolites from the Tricarboxylic Acid Cycle
for Yeast and Bacteria Samples Using Gas Chromatography
Mass Spectrometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277
Reza Maleki Seifar, Angela ten Pierick, and Patricia T.N. van Dam
21 GC-MS Analysis of Lipid Oxidation Products in Blood, Urine,
and Tissue Samples. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
Anne Barden and Trevor A. Mori

PART IV CE-MS-BASED METABOLOMICS

22 Metabolic Profiling of Urine by Capillary Electrophoresis-Mass

Spectrometry Using Non-covalently Coated Capillaries. . . . . . . . . . . . . . . . . . . . . . 295
Rawi Ramautar
23 CE-MS for the Analysis of Amino Acids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
Karina Trevisan Rodrigues, Marina Franco Maggi Tavares,
and Ann Van Schepdael

PART V NMR-BASED METABOLOMICS

24 NMR Analysis of Fecal Samples. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 317

Hye Kyong Kim, Sarantos Kostidis, and Young Hae Choi
 Contents ix

25 Quantitative Analysis of Central Energy Metabolism in Cell

Culture Samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 329
Sarantos Kostidis

PART VI MALDI-BASED TECHNIQUES AND MASS SPECTROMETRY

IMAGING OF CLINICAL SAMPLES

26 Mass Spectrometry Imaging of Metabolites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 345

Benjamin Balluff and Liam A. McDonnell

PART VII STUDY DESIGN, DATA ANALYSIS, AND BIOINFORMATICS

27 Quality-Assured Biobanking: The Leiden University Medical

Center Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 361
Rianne Haumann and Hein W. Verspaget
28 Extracting Knowledge from MS Clinical Metabolomic Data:
Processing and Analysis Strategies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 371
Julien Boccard and Serge Rudaz

Perspectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 385

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 387
Contributors

BENJAMIN BALLUFF  Maastricht MultiModal Molecular Imaging Institute (M4I),

Maastricht University, Maastricht, The Netherlands
ANNE BARDEN  Medical School, Royal Perth Hospital Unit, University of Western Australia,
Perth, WA, Australia
SUSEN BECKER  Institute of Laboratory Medicine, Clinical Chemistry and Molecular
Diagnostics, University Hospital Leipzig, Leipzig, Germany; LIFE-Leipzig Research Center
for Civilization Diseases, Leipzig University, Leipzig, Germany
JUSTINE BERTRAND-MICHEL  MetaToul-Lipidomic MetaboHUB Core Facility, INSERM
U1048, Toulouse Cedex 4, France
JULIEN BOCCARD  School of Pharmaceutical Sciences, University of Geneva, University of
Lausanne, Geneva, Switzerland; Swiss Centre for Applied Human Toxicology (SCAHT),
Universities of Basel and Geneva, Basel, Switzerland
BO BURLA  Singapore Lipidomics Incubator (SLING), Life Sciences Institute, National
University of Singapore, Singapore, Singapore
UTA CEGLAREK  Institute of Laboratory Medicine, Clinical Chemistry and Molecular
Diagnostics, University Hospital Leipzig, Leipzig, Germany; LIFE-Leipzig Research Center
for Civilization Diseases, Leipzig University, Leipzig, Germany
ROMANAS CHALECKIS  Division of Physiological Chemistry 2, Department of Medical
Biochemistry and Biophysics, Karolinska Institutet, Stockholm, Sweden; Gunma University
Initiative for Advanced Research (GIAR), Gunma University, Gunma, Japan
YOUNG HAE CHOI  Natural Product Laboratory, Institute of Biology, Leiden University,
Leiden, The Netherlands
ROMAIN A. COLAS  Lipid Mediator Unit, Biochemical Pharmacology, William Harvey
Research Institute, Barts and the London School of Medicine, Queen Mary University of
London, London, UK
JESMOND DALLI  Lipid Mediator Unit, Biochemical Pharmacology, William Harvey
Research Institute, Barts and the London School of Medicine, Queen Mary University of
London, London, UK
PATRICIA T.N. VAN DAM  Department of Biotechnology, Faculty of Applied Sciences, Delft
University of Technology, Delft, The Netherlands
KO WILLEMS VAN DIJK  Einthoven Laboratory for Experimental Vascular Medicine,
Department of Human Genetics, Leiden University Medical Center (LUMC), Leiden, The
Netherlands; Division of Endocrinology, Department of Medicine, Leiden University
Medical Center (LUMC), Leiden, The Netherlands
STÉPHANIE DURAND  Université Clermont Auvergne, INRA, UNH, Plateforme
d’Exploration du Métabolisme, MetaboHUB Clermont, Clermont-Ferrand, France
LARS F. EGGERS  Division of Bioanalytical Chemistry, Research Center Borstel, Borstel,
Germany
BART EVERTS  Department of Parasitology, Leiden University Medical Center (LUMC),
Leiden, The Netherlands
MARTIN GIERA  Center for Proteomics and Metabolomics, Leiden University Medical Center
(LUMC), Leiden, The Netherlands

xi
xii Contributors

BRUCE D. HAMMOCK  Department of Entomology and Nematology, UC Davis

Comprehensive Cancer Center, University of California, Davis, Davis, CA, USA
WEI HAN  Department of Chemistry, University of Alberta, Edmonton, AB, Canada
VANESSA VAN HARMELEN  Einthoven Laboratory for Experimental Vascular Medicine,
Department of Human Genetics, Leiden University Medical Center (LUMC), Leiden,
The Netherlands
RIANNE HAUMANN  Department of Biobanking, Leiden University Medical Center, Leiden,
The Netherlands
MARIEKE HEIJINK  Center for Proteomics and Metabolomics, Leiden University Medical
Center (LUMC), Leiden, The Netherlands
PATRICK O. HELMER  Institute of Inorganic and Analytical Chemistry, University of
M€ unster, M€
unster, Germany
MICHAL HOLČAPEK  Department of Analytical Chemistry, Faculty of Chemical Technology,
University of Pardubice, Pardubice, Czech Republic
LISA R. HOVING  Einthoven Laboratory for Experimental Vascular Medicine, Department of
Human Genetics, Leiden University Medical Center (LUMC), Leiden, The Netherlands
BORA INCEOGLU  Department of Entomology and Nematology, UC Davis Comprehensive
Cancer Center, University of California, Davis, Davis, CA, USA
JULIJANA IVANISEVIC  Metabolomics Platform, Faculty of Biology and Medicine, University of
Lausanne, Lausanne, Switzerland
HYE KYONG KIM  Natural Product Laboratory, Institute of Biology, Leiden University,
Leiden, The Netherlands
SARANTOS KOSTIDIS  Center for Proteomics and Metabolomics, Leiden University Medical
Center (LUMC), Leiden, The Netherlands
SABRINA KRAUTBAUER  Institute of Clinical Chemistry and Laboratory Medicine, University
of Regensburg, Regensburg, Germany
JETTY CHUNG-YUNG LEE  School of Biological Sciences, The University of Hong Kong, Hong
Kong, SAR, China
YIU YIU LEE  School of Biological Sciences, The University of Hong Kong, Hong Kong, SAR,
China
LIANG LI  Department of Chemistry, University of Alberta, Edmonton, AB, Canada
MIROSLAV LÍSA  Department of Analytical Chemistry, Faculty of Chemical Technology,
University of Pardubice, Pardubice, Czech Republic
GERHARD LIEBISCH  Institute of Clinical Chemistry and Laboratory Medicine, University of
Regensburg, Regensburg, Germany
LIAM A. MCDONNELL  Fondazione Pisana per la Scienza ONLUS, Pisa, Italy
ISABEL MEISTER  Division of Physiological Chemistry 2, Department of Medical Biochemistry
and Biophysics, Karolinska Institutet, Stockholm, Sweden; Gunma University Initiative for
Advanced Research (GIAR), Gunma University, Gunma, Japan
CAROLE MIGNÉ  Université Clermont Auvergne, INRA, UNH, Plateforme d’Exploration
du Métabolisme, MetaboHUB Clermont, Clermont-Ferrand, France
TREVOR A. MORI  Medical School, Royal Perth Hospital Unit, University of Western
Australia, Perth, WA, Australia
CHRISTOPHE MORISSEAU  Department of Entomology and Nematology, UC Davis
Comprehensive Cancer Center, University of California, Davis, Davis, CA, USA
SNEHA MURALIDHARAN  Singapore Lipidomics Incubator (SLING), Department of
Biological Sciences, National University of Singapore, Singapore, Singapore
 Contributors xiii

SHAMA NAZ  Division of Physiological Chemistry 2, Department of Medical Biochemistry and

Biophysics, Karolinska Institutet, Stockholm, Sweden
JOHN W. NEWMAN  Obesity and Metabolism Research Unit, United States Department of
Agriculture (USDA), Agricultural Research Service (ARS), Western Human Nutrition
Research Center, University of California, Davis, Davis, CA, USA; Department of
Nutrition, University of California, Davis, Davis, CA, USA; NIH West Coast
Metabolomics Center, University of California, Davis, Davis, CA, USA
THERESA L. PEDERSEN  Advanced Analytics, Woodland, CA, USA
ANGELA TEN PIERICK  BioAnalytical Chemistry, Department of Chemistry and
Pharmaceutical Sciences, Vrije Universiteit Amsterdam, Amsterdam, The Netherlands
ESTELLE PUJOS-GUILLOT  Université Clermont Auvergne, INRA, UNH, Plateforme
d’Exploration du Métabolisme, MetaboHUB Clermont, Clermont-Ferrand, France
RAWI RAMAUTAR  Division of Systems Biomedicine and Pharmacology, Leiden Academic
Center for Drug Research, Leiden University, Leiden, The Netherlands
AMY A. RAND  Department of Entomology and Nematology, UC Davis Comprehensive
Cancer Center, University of California, Davis, Davis, CA, USA
MADLEN REINICKE  Institute of Laboratory Medicine, Clinical Chemistry and Molecular
Diagnostics, University Hospital Leipzig, Leipzig, Germany; LIFE-Leipzig Research Center
for Civilization Diseases, Leipzig University, Leipzig, Germany
FABIEN RIOLS  MetaToul-Lipidomic MetaboHUB Core Facility, INSERM U1048, Toulouse
Cedex 4, France
KARINA TREVISAN RODRIGUES  Institute of Chemistry, University of Sao Paulo (USP), Sao
Paulo, SP, Brazil; Department of Pharmaceutical and Pharmacological Sciences,
Pharmaceutical Analysis, KU Leuven-University of Leuven, Leuven, Belgium
HUUB H. VAN ROSSUM  Laboratory of Clinical Chemistry and Hematology, The Netherlands
Cancer Institute, Amsterdam, The Netherlands
SERGE RUDAZ  School of Pharmaceutical Sciences, University of Geneva, University of
Lausanne, Geneva, Switzerland; Swiss Centre for Applied Human Toxicology (SCAHT),
Universities of Basel and Geneva, Basel, Switzerland
ANN VAN SCHEPDAEL  Department of Pharmaceutical and Pharmacological Sciences,
Pharmaceutical Analysis, KU Leuven-University of Leuven, Leuven, Belgium
DOMINIK SCHWUDKE  Division of Bioanalytical Chemistry, Research Center Borstel, Borstel,
Germany
REZA MALEKI SEIFAR  Department of Biotechnology, Faculty of Applied Sciences, Delft
University of Technology, Delft, The Netherlands; DSM Food Specialties B.V., Delft, The
Netherlands
CHARLES N. SERHAN  Department of Anesthesiology, Perioperative and Pain Medicine,
Center for Experimental Therapeutics and Reperfusion Injury, Brigham and Women’s
Hospital, Harvard Medical School, Boston, MA, USA
MARINA FRANCO MAGGI TAVARES  Institute of Chemistry, University of Sao Paulo (USP), Sao
Paulo, SP, Brazil
OLAF VAN TELLINGEN  Laboratory of Clinical Chemistry and Hematology, The Netherlands
Cancer Institute, Amsterdam, The Netherlands
AURELIEN THOMAS  Unit of Toxicology, CURML, CHUV Lausanne University Hospital,
HUG Geneva University Hospitals, Lausanne, Switzerland; Faculty of Biology and
Medicine, University of Lausanne, Lausanne, Switzerland
FEDERICO TORTA  Singapore Lipidomics Incubator (SLING), Department of Biochemistry,
YLL School of Medicine, National University of Singapore, Singapore, Singapore
xiv Contributors

BALJIT K. UBHI  SCIEX, Redwood City, CA, USA

HEIN W. VERSPAGET  Department of Biobanking, Leiden University Medical Center,
Leiden, The Netherlands; Parelsnoer Institute, Utrecht, The Netherlands
MARY E. WALKER  Lipid Mediator Unit, Biochemical Pharmacology, William Harvey
Research Institute, Barts and the London School of Medicine, Queen Mary University of
London, London, UK
MARKUS R. WENK  Singapore Lipidomics Incubator (SLING), Department of Biochemistry,
YLL School of Medicine, National University of Singapore, Singapore, Singapore
CRAIG E. WHEELOCK  Division of Physiological Chemistry 2, Department of Medical
Biochemistry and Biophysics, Karolinska Institutet, Stockholm, Sweden; Gunma University
Initiative for Advanced Research (GIAR), Gunma University, Gunma, Japan
LENNART J. VAN WINDEN  Laboratory of Clinical Chemistry and Hematology, The
Netherlands Cancer Institute, Amsterdam, The Netherlands
 Part I

Clinical Metabolomics and Lipidomics

 Chapter 1

Metabolomics as a Tool to Understand Pathophysiological

Processes
Julijana Ivanisevic and Aurelien Thomas

Abstract
Multiple diseases have a strong metabolic component, and metabolomics as a powerful phenotyping
technology, in combination with orthogonal biological and clinical approaches, will undoubtedly play a
determinant role in accelerating the understanding of mechanisms that underlie these complex diseases
determined by a set of genetic, lifestyle, and environmental exposure factors. Here, we provide several
examples of valuable findings from metabolomics-led studies in diabetes and obesity metabolism, neurode-
generative disorders, and cancer metabolism and offer a longer term vision toward personalized approach to
medicine, from population-based studies to pharmacometabolomics.

Key words Metabolomics, Obesity metabolism, Diabetes metabolism, Neurodegenerative diseases,

Cancer metabolism, Personalized medicine, Pharmacometabolomics, Population studies

1 Metabolome as a Complement to Other ‘Omes

Metabolomics is making a major impact in a wide variety of scien-

tific areas such as functional genomics, fundamental biochemistry,
toxicology, environmental sciences, food safety, drug dosage and
discovery, new-born screening, disease diagnostics, and biomedical
and clinical research in general [1, 2].
To get a complete picture about the potential of metabolomics
and the advantages of studying the metabolome in this post-
genomic era of biology, we should first take a look at the central
dogma of molecular biology (Fig. 1). Using high-throughput
‘omics technologies, we can measure the biological variability at
four main biochemical or functional levels of a system’s organiza-
tion, each one providing complementary information about the
phenotype. The genome or complete information encoded in
genes, represented by both the coding and noncoding sequences
of DNA, provides us with data on how an organism should be built
and maintained or what may happen. Compared to other ‘omes, the
genome is relatively static although subject to regulatory processes

Martin Giera (ed.), Clinical Metabolomics: Methods and Protocols, Methods in Molecular Biology, vol. 1730,
https://doi.org/10.1007/978-1-4939-7592-1_1, © Springer Science+Business Media, LLC 2018

3
4 Julijana Ivanisevic and Aurelien Thomas

Fig. 1 Metabolome—downstream of genome and exposome providing the closest measure of phenotype at
the molecular level. Figure was significantly modified from Ivanisevic & Siuzdak, 2015 and the permission for
re-use has been obtained from the Journal of NeuroimmunePharmacology [3]

such as epigenetic modifications that influence the gene expression

and transcription by switching it “on” or “off.” The transcriptome
or the full qualitative set of messenger RNA molecules transcribed
from the “expressed” part of the genome and proteome or the
complete set of proteins translated from messenger RNA are more
dynamic, thus providing the information on what is happening and
what makes it happen, respectively. Both are subject to posttran-
scriptional and posttranslational modifications, respectively. As the
final piece of the ‘omic puzzle, the metabolome is highly dynamic
providing the snapshots of information about what has happened
[4]. More explicitly in the biological context, while the information
in genes is inherited and genotypes describe the potential of the
system or what may happen, metabolites tell us what has really
happened and describe the current functional (physiological or
developmental) status of the system. Metabolites are the end pro-
ducts of genome, transcriptome, and proteome activity, thus repre-
senting the direct signatures of cellular activity and the reason why
the metabolome is addressed as the closest link to phenotype,
providing the direct and sensitive measure of the dynamic changes
at the molecular level [5, 6]. In addition to mediating biochemical
processes as substrates and products of energy conversion reactions
and signaling molecules (or secondary messengers), metabolites
also have far-reaching biochemical actions in the regulation of
epigenetic modifications and protein activity [7, 8].
A further important advantage of studying the metabolome lies
in the fact that in addition of being the downstream output of the
genome, the metabolome also directly reflects the upstream input
from the environment, allowing us to explore the interaction
between the genome and the environment [1]. Metabolite levels
are significantly influenced by synergistic effects of many different
 Metabolomics in Biomedical and Clinical Research 5

environmental risk factors, including the internal microbiome, oxi-

dative stress and inflammation, and the external climate, pollutants,
pathogens, food, drugs, and social interactions, we are exposed
to. These exposures we are subjected to throughout our life also
define our phenotype [9]. As the end product of the genome under
the direct influence of the environmental stressors, the metabolome
allows us to measure the integrated effect of both, providing the
assessment of the exposome (Fig. 1). Several metabolomics-led
studies have recently revealed the highly relevant influence that
the environment, and more particularly our microbiome (gastroin-
testinal, nasal, oral, urogenital, skin population, etc.), has on disease
development, in cardiovascular disease, obesity, cancer, etc.,
[10, 11]. It is also assumed that the microbiome’s metabolic activ-
ity actively influences our mind [12]!
Overall, the discovery of a dominant metabolic component in
many diseases (rather than genetic as widely accepted), such as in
cancer, atherosclerosis, diabetes, and neurodegenerative disorders,
has stimulated the rethinking toward a systems biology approach
[1]. By integration of metabolomics into systems biology, we will
go beyond the biomarker discovery and toward mechanistic under-
standing of pathophysiological processes [2, 13]. The advance-
ments in technology, computing power, and back-end
bioinformatics solutions [14–17] made it possible not only to
rapidly measure thousands of metabolites from only minimal
amounts of biological sample (biofluids, cells, and tissues) but
also to go from metabolites to pathways (Fig. 2) and to link these
altered identified pathways back to the phenotype in combination
with orthogonal molecular approaches (interfering with gene
expression, enzyme activity, cell signaling, etc.) and technologies
(epigenomics, transcriptomics, proteomics) [2].
As described in several case studies in the following section,
disease mechanisms can be better understood, and significantly
more variance explained, when viewed from the metabolite and
not only gene perspective [1]. In these lines, metabolomics is also
bringing new insights into medical approaches, from a reactive,
one-size-fits-all model of care toward a predictive, preventive, and
patient-centric model of care. As a powerful phenotyping technol-
ogy, metabolomics will play a key role in further development of
this personalized approach to medicine [19]. The advantage of
metabolomics is high throughput and scalability, thus offering the
possibility to acquire specific metabolic profiles for each individual
in the emerging population-based metabolomics studies [20]. The
great potential of population-based studies is the evaluation of the
effects of environmental exposures (the “exposome”), including
pharmaceutical and nutritional, on disease onset, diagnosis, and
progression.
6 Julijana Ivanisevic and Aurelien Thomas

Fig. 2 Discovery-based clinical metabolomic workflow, combining untargeted and targeted metabolomic
approach. From comprehensive untargeted profiling, through data processing, statistical analyses, metabolite
identification, and pathway analyses, to targeted tandem MS quantification. Figure was significantly modified
from Ivanisevic et al. Scientific Reports 2015 and the permission for re-use is specified under Creative
Commons License [18]

2 Metabolomic Technology and Technical Approaches

The human metabolome is the final level of cellular regulatory

processes, constituted of more than 30,000 low-molecular-weight
molecules (<1500 Da) with a great chemical diversity and a wide
concentration range [21]. Due to its complexity and size, compre-
hensive coverage of the metabolome is challenging but necessary
when studying phenotypes of complex biological systems
[22]. Integration of metabolomic strategies in the bioanalytical
research pipeline has created great opportunities to bring new
 Metabolomics in Biomedical and Clinical Research 7

biological and clinical insights, out of which some will be developed

in this chapter [23, 24].
Large-scale coverage of the metabolome requires the combina-
tion of multiple orthogonal and complementary analytical plat-
forms, non-exhaustively including liquid chromatography-mass
spectrometry (LC-MS), nuclear magnetic resonance (NMR), gas
chromatography-MS (GC-MS), direct injection MS (DIMS), and
capillary electrophoresis-MS (CE-MS) [25, 26]. Several metabo-
lomic workflows involving various combinations of these platforms
have been successfully demonstrated [27, 28]. For instance, GC-
and LC-MS analyses based on full-scan acquisition mode are highly
complementary, as each gives access to metabolites with different
physicochemical properties and usually a limited number of
detected species by both platforms [29]. GC provides efficient
separation of relatively polar metabolites with a molecular weight
range below 350 Da, primarily including amino and organic acids,
fatty acids, and carbohydrates, whereas LC-MS non-exhaustively
allows the detection of higher-molecular-weight compounds [26].
Despite this analytical diversity, emergence of ultrahigh pres-
sure LC (UHPLC) hyphenated to electrospray ionization source
high-resolution MS (ESI-HRMS) has deeply extended our capacity
of deciphering the metabolome [30]. The recent advancements and
innovations of HRMS in terms of sensitivity, resolution, and
throughput have driven the development of untargeted metabolo-
mics by overcoming technical limitations associated with the com-
plexity of this field [24]. The latest generation of instrumentation,
notably the Fourier transform MS, routinely reaches mass resolu-
tion above 200,000 (m/Δm, FWHM) at 1 ppm of mass accuracy
[31]. Associated with continual improvement of scan rate frequen-
cies and dynamic range, these platforms provide unprecedented
insights into the ultimate response of biological systems to genetic
or environmental changes [1, 32].
In combination to the high spectral capacities of HRMS, the
use of orthogonal UHPLC separation modes that involve different
physicochemical interactions is suitable since the chromatographic
peak capacity (i.e., the number of peaks that can be theoretically
separated with a chromatographic resolution of 1) increases drasti-
cally for complex biological extracts [22, 33]. In this way, HILIC
provides a good alternative to RP separation for polar and ionic
compounds and is attractive because both techniques use the same
type of eluent: a mixture of organic solvent (typically acetonitrile)
and water with a volatile buffer [34, 35]. Combination of both
modes enables the observation of metabolites associated to lipid
and central carbon metabolism from a single biological extract with
the highest number of unique detected features by RP and HILIC
in ESI positive and negative modes, respectively (Fig. 2) [36].
In order to better understand pathophysiological processes, a
main advantage of applying untargeted metabolomics is the
8 Julijana Ivanisevic and Aurelien Thomas

amount of data that can be generated in a single run. This advan-

tage also represents one of the main challenges in the field consid-
ering our capability to maximize the computational mining of this
big data. Gigabytes of data per run can be easily generated raising
the question of the amount of information which can indeed be
currently explored. To overcome this issue, different aspects have to
be considered and optimized ranging from data pre- and post-
processing, statistical analysis to metabolite assignment, database
curation, and biological interpretation. From a bioinformatic point
of view, computational resources have greatly been improved in the
last few years helping the scientific community with this fastidious
task. Notably, different Web-based software, such as XCMS Online
[37] and MetaboAnalyst [38], have reached a state of maturity for
accelerating and automating the computational workflow,
providing user-friendly tools for both novice and expert bioinfor-
maticians [2] .
Unlike other ‘omics approaches, annotation, identification, and
validation of metabolite features is still considered as a bottleneck
due to the wide physicochemical diversity of the metabolome.
Basically, metabolites cannot be “sequenced” per se leading to
challenging in silico prediction of a specific chemical structure.
However, UHPLC-HRMS provides high mass and MS/MS spec-
tral accuracies, which represent a key step for generating putative
structure attribution to a given peak of interest when used with an
adapted heuristic filtering procedure [39]. This information can be
even complemented by the use of authentic standard when available
in order to obtain the reference MS/MS spectrum and chro-
matographic retention time specific to each dedicated instrumenta-
tion and method [2]. Despite the current progress observed in the
database curation like for HMDB [40] and METLIN [16], only a
very small subset (roughly below 10%) of all the metabolite features
detected in an untargeted approach can be undoubtedly identified.
However, this matching score is perpetually improved with the
experimental characterization of the metabolome growing number
of records in databases and mass spectral libraries.
Facing these metabolite assignment issues in untargeted experi-
ments, a paradigm shift has been observed toward the emergence of
large-scale targeted metabolomics. Because of its high sensitivity,
selectivity, robustness, and throughput, triple quadrupole instru-
ments (QqQ) working in the multiple reaction monitoring (MRM)
mode provide an excellent approach in metabolomic profiling
allowing simultaneous targeted analysis of hundreds of metabolites
involved in multiple major metabolic pathways (Fig. 2)
[24, 41]. Considerable efforts have been made to increase the
number of measured metabolites, making this approach more pop-
ular within the clinical community [42]. The possibility of moni-
toring a broad range of identified metabolites is speeding up the use
 Metabolomics in Biomedical and Clinical Research 9

and application of metabolomics in clinical studies, bringing new

insights into personalized approaches to medicine.
In addition to the conventional metabolomic strategies dis-
cussed in this section, imaging MS is one of the most recent and
innovative approaches for direct analysis of tissues or individual
cells. Indeed, imaging MS will allow the mapping of hundreds of
metabolites in tissue sections while maintaining a high correlation
between the histological features and the obtained molecular
images [43–45]. Enriched by a decade of development, the tech-
nology has reached a stage of maturity bringing promising oppor-
tunities as a tool to understand pathophysiological processes
[28]. Many clinical applications of the technology have been
demonstrated, notably in cancer research, cardiovascular diseases,
and neurosciences, and some of them will be discussed in more
details further on in this chapter [46–49]. In the MALDI Biotyper®
era with the successful implementation in clinical microbiology
laboratories, imaging MS may find concrete applications in clinical
pathology by fostering the discovery of molecular signatures asso-
ciated to pathophysiological conditions.

3 Novel Biological Insights into Disease Pathology Led by Metabolomics

Multiple diseases, including diabetes, obesity, cancer, cardiovascu-

lar disease, and neurodegenerative disorders, have a dominant met-
abolic component or even cause in terms of metabolic switch or
reprogramming that likely occurs prior to disease onset. Applying
metabolomics, as a powerful phenotyping technology, in combina-
tion with orthogonal biological and clinical approaches, could
significantly accelerate the understanding of mechanisms that
underlie these complex diseases determined by a set of genetic,
lifestyle, and environmental risk factors [50, 51]. Here, we provide
several examples of valuable findings from metabolomics-led stud-
ies in diabetes and obesity metabolism, neurodegenerative disor-
ders, and cancer metabolism.

3.1 Obesity Incidence of obesity has doubled since 1980, making it a major
Metabolism public health concern worldwide. In 2014, more than 1.9 billion
adults were overweight and 600 million were obese, numbers that
are expected to increase in the future [52]. Obesity is a complex
chronic disorder with a multifactorial etiology, involving increased
intake of high caloric food and a decrease of physical activity that
arise from a combination of genetic and environmental factors.
Recently, genome-wide association studies have identified common
single-nucleotide polymorphisms (SNPs) associated with obesity
[53, 54]. While the catalogue of susceptibility loci in obesity
keeps growing, only few specific genes may currently support the
biological processes of the identified genetic association. For the
10 Julijana Ivanisevic and Aurelien Thomas

vast majority of loci, the causal relationship between genes and

pathways remains unknown [54]. Metabolomics has emerged as a
promising approach to better characterize molecular mechanisms
involved in pathophysiology of obesity [55, 56]. Studies in both
human and animal models have revealed strong associations
between obesity and different metabolic species, including amino
acids and acylcarnitines [57, 58]. For instance, Newgard et al.
found that the dysregulation of branched-chain amino acid
(BCAA) metabolism in humans contributes to the development
of obesity-related insulin resistance, ultimately leading to type
2 diabetes mellitus (T2DM) [59]. They validated their finding by
showing that rats fed on a high-fat diet supplemented by BCAA
developed insulin resistance despite having reduced food intake
compared to the high-fat diet group [59].
In the last decade, the potential role of adipose tissue (AT) in the
onset of obesity and associated metabolic disorders has gained atten-
tion [60, 61]. Initially viewed as a simple storage tissue for accumu-
lation of xenobiotics, AT is becoming more recognized as an active
metabolic and endocrine organ, essential for regulation of appetite,
energy metabolism, and inflammatory response [60, 62, 63]. How-
ever, the response of different AT depots to diet-induced obesity is
variable and increasingly studied. While visceral AT is more prone to
inflammation compared to subcutaneous AT, less is known about the
molecular-signature-induced phenotype that differs between these
two AT depots. Metabolomics studies have demonstrated significant
difference in the molecular compositions between AT depots, open-
ing the route for further research to decipher the pathophysiological
phenotype of visceral AT [55, 64–66].
In addition, metabolomics has shown promising results in
clinical research to assess and predict the response to weight loss
intervention [57]. For instance, alterations in saturated fatty acid,
monounsaturated fatty acid, and BCAA levels were observed in
serum samples of overweight and obese elder subjects demonstrat-
ing a reduction of weight of 7% after 8 weeks of energy-restricted
diets [67]. In another study, a metabolomic signature of BCAAs
and biologically associated metabolites predicted improvement in
insulin resistance, thus helping with the identification of patients
who will most likely benefit from a moderate weight lost [68]. For
severely obese patients, the Roux-en-Y gastric bypass (RYGB) is
considered as the most common and efficient bariatric surgery to
reduce body weight but also to solve comorbid conditions, such as
remission of T2DM [69, 70]. However, one third of postoperative
patients will still remain “nonrespondent” without significant
weight loss following the surgery [71, 72]. Due to this, bariatric
surgery is another field of interest in which metabolomics could
bring new opportunities to identify predictive biomarkers of “non-
respondent” individuals and thus help physicians in the selection of
candidate patients prior to surgery [57, 73, 74].
 Metabolomics in Biomedical and Clinical Research 11

3.2 Diabetes Type 2 diabetes mellitus (T2DM) is a major public health concern
Metabolism worldwide. According to estimations roughly half of 400 million
individuals suffering of T2DM are undiagnosed [75, 76]. Since the
disease remains asymptomatic for years, its clinical assessment is
difficult. Due to this, patients are often identified when complica-
tions, such as chronic kidney diseases and cardiovascular diseases,
among others, are already present. In order to prevent or postpone
the onset of T2DM, there is a need to develop predictive biomar-
kers to identify the individuals at risk for developing the disease.
Currently approved clinical and biological markers were not suffi-
cient to efficiently address the complexity of T2DM pathogenesis
[77, 78]. The addition of different panels of T2DM-associated
genetic markers allowed for a limited improvement in the predictive
capacity of usual clinical scores [79–81].
Looking at the complexity of T2DM pathophysiology involv-
ing numerous organs such as the pancreas, liver, blood vessels,
kidneys, and eyes, this multifaceted disease requires the develop-
ment and application of systems medicine approach including
metabolomics to increase translational potential toward clinical
applications [57, 82]. Metabolomics studies have been applied to
population-based cohorts to identify individuals at risk to develop
diabetes [83]. These studies emphasized the role of amino acid and
lipid metabolism in the underlying mechanism of the disease
[57]. For instance, Wang-Sattler et al. quantified 140 metabolites
in 4297 serum samples in the population-based cooperative health
research of Augsburg (KORA) [84]. Their study revealed glycine
and lysophosphatidylcholine (18:2) as strong predictors of glucose
tolerance, even 7 years before disease onset. In another study, a
panel of 196 metabolites enabled the identification of a metabolic
signature associated with the early manifestations of T2DM, char-
acterized by the significant increase of glyoxylate [85]. In a case-
control study, Wang et al. performed a metabolomic profiling of
180 metabolites in 189 individuals who developed new-onset dia-
betes during a 12-year follow-up period and 189 propensity-
matched control subjects from the same baseline examination
[86]. They identified a metabolic signature marked by five BCAA
(leucine, isoleucine, and valine) and aromatic amino acids (phenyl-
alanine and tyrosine) significantly correlated with diabetes onset. By
integrating three of these metabolites (isoleucine, tyrosine, and
phenylalanine), individuals in the top quartile exhibited fivefold
higher risk to become diabetes converters, thus demonstrating
the predictive value of this signature.
These studies are in agreement with the potential key role of
BCAA metabolism in adipose tissue accompanied with the increase
of circulating BCAA observed in obese and prediabetic patients
[87, 88]. This is emphasized by the high correlation between
insulin sensitivity and expression of BCAA metabolism genes
observed in adipose tissue of diabetic patients [57, 89]. These
12 Julijana Ivanisevic and Aurelien Thomas

examples demonstrate the great potential of metabolomics in the

study of metabolic diseases not only for determining new clinical
biomarkers but also to decipher potential biological pathways asso-
ciated with the onset and development of such diseases. However,
most of the population-based studies discussed in this section have
been carried out by targeted metabolomics focusing largely on
amino acid and lipid metabolism. Considering the methodological
advances in the field of metabolomics, we foresee that the applica-
tion of untargeted strategies could potentially bring additional
clues about the complexity of T2DM by expanding the metabo-
lome coverage.

3.3 Aging and In this last decade, the great emphasis has been placed on the
Age-Related understanding of biological mechanisms of aging representing the
Neurodegenerative greatest risk factor for the onset of multiple age-related declines in
Diseases organ function and thus the whole spectra of diseases (including
cancer, cardiovascular disease, neurological disorders, diabetes,
etc.). There is an obvious added value in targeting the processes
of aging itself to improve healthy longevity by delaying the onset
and progression of many different diseases [90]. It is assumed that
diabetes and obesity as consequences of excessive energy intake
accelerate brain aging and increase the risk for neurodegenerative
diseases (NDDs) and stroke. In today’s aging society, the NDDs,
including Alzheimer disease (AD), Parkinson disease (PD), and
amyotrophic lateral sclerosis (ALS), present one of the major health
care concerns [91]. According to the National Institute of Aging in
the USA, one in four individuals over 60 years of age are affected by
AD as the most common cause of dementia. Although clinical
studies have already revealed a great deal of information about
NDDs, the changes at the molecular level that cause these diseases
are poorly understood, and it is still unknown why some diseases,
like AD, are tenfold more frequent than the others (PD or ALS)
and why cognitive decline occurs more rapidly in some individuals
than in others [91]. Currently approved therapeutic strategies that
target the abnormal metabolism of amyloid and tau protein in AD
(followed by extracellular deposition of amyloid oligomer plaques
and intracellular formation of neurofibrillary tangles, respectively)
have so far been unsuccessful treating the symptoms or conse-
quences of disease progression rather than the causes of disease
itself. In this surge to understand the cause of NDDs, few recent
studies suggest the importance of exploring energy metabolism and
the role it plays in brain function [91–94]. It is well known that the
human brain is metabolically expensive using up to 20% of the total
body energy to sustain its function [95]. This is supported by the
evolutionary findings of massive increase of genes involved in
energy production in human cortex compared to nonhuman pri-
mates. Elevated expression of genes involved in cellular energy
metabolism and neuronal function, as highest energy-demanding
 Metabolomics in Biomedical and Clinical Research 13

cells, has led to specific metabolic profiles, highly sensitive to

changes in energy homeostasis and oxidative stress [91]. Energy
homeostasis is maintained by small-molecule metabolites (among
other actors at the molecular level) that mediate biochemical reac-
tions in the interconnected central carbon and signaling pathways.
Neurodegenerative disorders are thought to be associated with and
perhaps even caused by alterations in central carbon energy metab-
olism and derived oxidative stress [93, 96, 97]. However, these
metabolic alterations, beyond the reduced glucose metabolism that
was demonstrated in functional neuroimaging studies, remain
largely unexplored. In this context, the utilization of metabolomics
to comprehensively investigate metabolic alterations associated
with the pathophysiological process of NDDs is achieving consid-
erable interest [94]. Multiple metabolomics studies have explored
the metabolic traits of AD, PD, and ALS in several transgenic
animal models and small human cohorts, using different metabo-
lomic platforms, from NMR and GC-MS to CE-MS, DIMS, and
LC-MS. Various sample types, from biofluids such as plasma,
serum, urine, saliva, and CSF to brain tissue and several peripheral
organs, have been analyzed at different stages of disease severity
[93, 98–102]. Although mainly putative, the output of these stud-
ies implies the cellular traits of accelerated aging related to the
intrinsic neuronal vulnerability to oxidative stress. Alterations at
the metabolite and pathway level highlight the perturbed energy
production in mitochondria accompanied by excessive reactive
oxygen species (ROS) formation leading to oxidative damage to
DNA, electron transport chain proteins, and plasma membrane
redox system (due to the loss of function and structural integrity
of amino acids, nucleic acids, and lipids as a consequence of oxida-
tion) [50, 51]. These changes are thought to trigger the produc-
tion and appearance of disease-specific factors, amyloid-β and p-tau
proteins in AD, α-synuclein in PD, huntingtin in Huntington’s
disease, and Cu/Zn-superoxide dismutase (SOD) in ALS, further
inducing a neurodegenerative cascade of perturbed nutrient trans-
port, neurotransmission, and cell-to-cell communication across the
brain, leading to synapse dysfunction and loss [51, 103,
104]. Beyond these findings related to free radical theory of
aging, many new additional facets of mitochondrial dysfunction
associated with aging and NDDs are constantly being revealed
[51, 92, 105, 106]. These include a marked deregulation of mito-
chondrial aspartate metabolism [96] as well as stage- and region-
dependent deregulation of purine metabolism in postmortem fron-
tal cortex tissue of confirmed AD-affected brain [102]. Metabolic
profiling of biofluids from AD-confirmed patients has revealed the
affected metabolism of specific amino acids with implications on
carnitine synthesis and fatty-acid oxidation in mitochondria and the
potential influence on the neurotransmitter synthesis [93]. Two
interesting studies, although based on small number of samples
14 Julijana Ivanisevic and Aurelien Thomas

(<20 subjects in each group), have yielded compelling AD predic-

tion models applying the chemometrics on the acquired CE-MS
and LC-MS profiles of CSF and postmortem brain tissue
[107, 108]. The latest comparative analysis across multiple cohorts
within the Alzheimer Disease Metabolomics Consortium has
shown the alterations in sphingomyelins and ether-containing
phosphatidylcholines in preclinical biomarker-defined AD stages,
whereas the later symptomatic stages were characterized by altered
levels of acylcarnitines and several amines [94]. Overall, there is a
need to validate the acquired data in the independent cohorts of
clinically well-characterized patients at different stages of disease
and correlate the data with clinical manifestations.
By comprehensive evaluation of changes embedded in meta-
bolic networks, metabolomics has the potential to advance the
understanding of the mechanistic association between the energy
intake and brain function in relation to the aging process, oxidative
stress, and neurodegeneration. This will be essential for unraveling
new pathways and small-molecule targets for diagnostic and thera-
peutic purposes to prevent the pathogenesis before irreparable cell
death. Beyond the identification of small-molecule markers to
diagnose the disease, the strength of the metabolomic approaches
will be in the large-scale cohort studies of clinically well-
characterized patients and healthy aged controls (without cognitive
impairment). These population studies will allow us to correlate the
revealed metabolic alterations with differentially expressed proteins
and genes and clinical meta-data (including the patient’s gender,
body mass index, blood-brain barrier permeability, structural and
functional neuroimaging results, assessment of cognitive decline,
etc.) to provide the reliable clues about the mechanism of disease
onset and progression for further investigation using in vivo mod-
eling for design of efficient therapeutic strategies to slow down or
prevent the disease.

3.4 Mechanistic Despite population-wide efforts in whole genome sequencing,

Discoveries in Cancer transcript profiling, and single-nucleotide polymorphism (SNP)
Metabolism Research characterization, there is a lack of targetable disease genes, and it
has become evident that many diseases can be better understood
when viewed from the metabolic in addition to the genetic perspec-
tive [1, 109]. Cancer is a great example of a fairly complex disease
whose understanding has evolved significantly since its metabolic
component has been under tenacious investigation. Today, the
metabolic reprogramming is considered as the core hallmark of
cancer [110]. The first report of deregulated metabolism in cancer
dates back to the early 1930s when high glucose consumption and
lactate production, even in the presence of oxygen, were discovered
by Otto Warburg [111]. After decades of intensive genetic explo-
ration, the recent advancements in metabolomic technology have
contributed to rethinking of cancer as a metabolic disorder via (re)
 Metabolomics in Biomedical and Clinical Research 15

discovery of many additional oncometabolites. Instead of challeng-

ing exploration of millions of possible mutations, recent
metabolomics-assisted studies imply that the majority of the
cancer-associated genes code for well-known metabolic enzymes
and thus impact only several pathways, such as aerobic glycolysis,
glutaminolysis, and one-carbon metabolism [112]. These oncopath-
ways serve in the production of a number of oncometabolites whose
upregulation in tumor tissue is essential for rapid tumor growth
and propagation. Most oncometabolites play a key role in tumor-
igenesis by altering signaling pathways and cell division processes.
For example, the accumulation of organic acids, succinate, fuma-
rate, and 2-hydroxyglutarate leads to activation of cellular hypoxia
pathways in the presence of normal oxygen concentrations (termed
as pseudo-hypoxia), as well as far-reaching DNA methylation and
histone modifications [113]. Latter epigenetic changes affect gene
expression on a large scale, thus promoting the cancer develop-
ment. Reliance or even “addiction” of different types of cancer cells
on several nutrient oncometabolites is investigated for selective
targeting of cancer proliferation [112]. These metabolic vulnerabil-
ities include the dependence of most of the cancer types on glucose,
lactate, serine, and polyamines and more specifically the reliance of
Myc-activated tumors on glutamine, breast cancer on glycine syn-
thesis, acute lymphoblastic leukemia on asparagine, and brain and
prostate cancer on choline. The discovery of these metabolic vul-
nerabilities in cancer was made possible primarily due to the
advancements in HRMS-based metabolomics, MS-based imaging,
and magnetic resonance imaging (MRI) that allow us to detect and
measure a broad range of small-molecule metabolites in tumor
tissues. As a follow-up to comprehensive untargeted and targeted
approaches that highlight the potentially affected pathways, the
stable isotope-assisted metabolomic profiling and dynamic flux
analyses are gaining significant interest in cancer research to provide
further insights into direct pathway activity by tracing the nutrient
utilization in cancer. The most recent study published in Nature
Neuroscience demonstrates the application of nitrogen-labeling-
assisted tracing experiment to decipher how brain tumor-initiating
cells (BTICs) in glioma activate de novo purine synthesis to main-
tain self-renewal, proliferation, and tumor-forming capacity
[114]. The cancer and immunology field is progressing at the
moment due to application of isotopic profiling. Many efforts are
invested toward understanding how to regulate the nutrient uptake
and utilization in T cells that have an essential role in immunother-
apy against cancer [115]. Overall, the combined usage of different
metabolomic and genomic approaches is paving the way for manip-
ulation of cellular metabolism and design of efficient anticancer
therapies by targeting the well-known metabolic enzymes.
16 Julijana Ivanisevic and Aurelien Thomas

4 Toward Personalized Approach to Medicine: From Population-Based

Studies to Pharmacometabolomics

The understanding of molecular mechanisms associated with the

interindividual variability in response to drug treatment, including
the acute or chronic toxicities, and the discovery of associated
biomarkers represent key elements in the development of precision
or personalized medicine. Combination of “omics” technologies,
to analyze gene polymorphisms, epigenetic modifications, gene and
protein expressions, and metabolite concentrations and interac-
tions, is required to identify key players at the molecular level with
the greatest detectable effect in a pathophysiological state. The
increased reproducibility and reduced cost of “omics” approaches,
together with development of computational tools, have allowed us
to go toward multi-scale, network-embedded integration of differ-
ent types of molecular data. The ultimate goal of multi-scale data
integration is to link the changes at the metabolite level within
affected biochemical pathways to the responsible enzymes and the
underlying genetic alterations (Fig. 3).
Application of multi-omic approach in longitudinal
population-based studies is one of the most promising strategies
to better characterize and understand complex pathophysiological
state at the system’s level. Although genocentric in their begin-
nings, the population-based initiatives have recognized the value of
comprehensive and high-throughput nature of metabolomics, as a
powerful phenotyping technology that allows the cost-effective
measurement of a massive panel of metabolites in diverse biological
matrices from large population cohorts [1, 117]. In addition, the
data generated by metabolomics is remarkably consistent with the
data provided by genomic analysis. Integration of genome-wide
associations studies (GWAS) and metabolomic profiling has thus
revealed unprecedented insights into how genetic polymorphisms
influence human metabolome, providing strong association
between metabolic traits and loci of biomedical and pharmaceutical
interest [116, 118]. For example, combining GWAS and metabo-
lomics, Shin et al. reported a comprehensive exploration of genetic
loci influencing human metabolism based on 7824 participants
from two European population cohorts (Fig. 3) [116]. Their
study provided information on hundreds of single genotype-
metabolite associations, greatly expanding the assignment of gene
functions and disease-associated gene-metabolite links. Among
37 genetic loci associated with blood metabolite levels, Suhre
et al. identified a strong association between the rs662138 SNP
and isobutyrylcarnitine [118]. Among genetically determined
metabotypes (GDMs), they reported the polymorphism in the
SLC22A1 loci associated with the pharmacokinetics of the antidia-
betic drug metformin that can be used as a typical example of how
 Metabolomics in Biomedical and Clinical Research 17

75%

HB) ylcysteine

ine
lutam
inooctanoic acid

e
Co

Creatin

Pyrog
ybutyrate
tinin

eth

ine
e

ybutyr11786--m

rnit
ate (A

Ca
,4-dihydrox

ine
e

yl
X-12510--2-am

lycin

os
taro
X-

tyr
N-acetylalanine

cety e
utyrate

Glu
N-a Glycin

L-
lg

o-
Asparagine

Methio2-hydrox

ulf
50%

X-04499--3
Aspartate
7-m Paraxa

-s
2-aminob

ine
3-m-dimeylxan min e

-O
Alanine

nine
1,7 eth eobrohylline

1-m ylxannthinee

te
1-m Th eopffein te

eth

3-
e

Creatin

ate
1,3

e
ine

oa
Glu mine
id

eth thy thin e

42
nin
rolin
ethyl thin
m

te
,7-t

tan
tam

11
ta

Cyste
ylxa lura e
ce

ta
reo
Th Ca ylura

X-
Gluta

oninine
l]a

bu
rim

lac
xop

Se cety e
uratee

e
py

rin lth
nth te

ido

din

yl)
p- -(cy hox ya

e)
Beta
Er

eth
o

5-o
pr

2- -hy

at
ac st ya ce 4-

en
3 et rox

go

His etam

isti
in
l)

m
m d

e
Thre
eta ein ce ta ac
2
-y

ph
Th
th

na
ylh
-1

ethine
N-a
Ery hari te
ion

xy
m -S tammin eta

Ph3-p enynyll ol slanin-hy sine

ym P Quindrin

cin
in

Sa l su erinate

c
id -y in o m
lid

e te

enhen lacact ulfa e dro

3-m tid
4-a
ein

ro
thri n
o ip
op l)a o p id
rro

sin ola
cc lfa e

Ph he res yla 3-(4Tyro

Sta

e e
yd
te n am )
e
he ce ph he o
py

ta pio lut LA
at in
te

tol

(h
Ly ipec

yla ylp et ate te

na xo

c
ny ta en n s ph

ce ro ylg (P
hy
o n
pio 2-

lg m s ulf en
ha

Sacuroophlfat te l

P
2- -(

lu in u a o
hy G ro p
lyc

lic nid en e
dr e l)p -[3 to
25% yp

P -c en
yla e
ox er te sin ny --N yltr e
HW 4 o

pPh
yh lfa ro he 056

te
ES ipp -hyd l 2- Ibu su yty yp 3
s
co nin
AS 4- ura ro pho pro ol ox ox X-0 gly ure e
XX eth te xy
hip sph fen h en eth ydr C- yn tain
y4- ( h K e
vin Ca lph salicBen pu ate P -m 3-
3 -( b
y lph ch y ylu zoarate
te e n 3 an
Bra ph e
hin ne
dy en ol lsu rate te to )
kin s lf yp n HT rnit Proli
in, Hipol su ulfa ate ) Tr phain (5te te an tylo
a ph ce
γ-gγ-glu es-a
d p ur atete
lf
ate
10% to
yp ton cta ion te
Tr ero lela rop sulfa try
pto N-a
γ luta m
ta rg(9 S do le xyl tate roxy
p
γ-g γ -glu my ylt ) In do do ce hyd
luta γ -gluta tam lme hreo In -in lea 0--
3 do 10
m -g m y th n
γ-g ylle lutam ylvalglutaioninine In -12 ine
luta ucin yli lin m e X rull
m e s o e ine Cit
γ-g ylph leu
luta en c e
AD DSGE my ylala ine
lty a ine lline xyprolin MA
)
erat
e
pS G n Ure AD oval
ADS GEG DFXA rosineine in itru ro
Argomoc -hyd A+ xyis
GEG DFX EG DM -hydro
Glycy Pheny DF AEG GGV H ns-4 e e (S α
lvalin lalany XAEGGGGV R Tra ithin arginin
e lphe G R Orn ethyl ci ne itine ca rnitine
VR Dim Isoleu rn yryl
nyla
lani ylca Isobut
Aspar Le ne aler
uc Isov
X-1418 X- Py tylphe ylleuc
9--leuc14304--le roglutamnylalaniine ine
X-14450 ylalan ucylala ylglyc ne Leuc e rnitine
--pheny
lalanylleu ine nine ine Valinlbutyroylcacarnitine
cine m et hy aleroyl
Pro-hyCyclo(Leu 2- yisov rate
droxy-Pro –Pro) Amino acid Hydroxroxyisovale valerate
X-14208
--phenyl β-hyd yl-2-oxoopentanoate
X-12244--N alanyl 3-meth yl-2-ox
-acetylca
rnosineserine 4-methoxyisobutyrate ate)
Carbohydrate 2-hydr ate (4-oxovaler
Levulin l-2-oxobutyrate
Uridine 3-methy
Pseudouridine
Cofactor or vitamin Erythronate
Urate Allantoin Mannose
Mannitol
Xanthine Energy Fructose
Hypoxanthine Erythrose
Inosine
7-methylguanine
Lipid Glycerate 1,5-anhydroglucitol (1,5-AG)
ine Glucose
N2,N2-dimethylguanos 1,6-anhydroglucose
Nucleotide Pyruvate Lactate
Guanosine denosine
N1-methyla Threitol
mide
-pyridone-
4-carboxa
te
Peptide Arabinose
1-methyl-3 Estrone
3-sulfarol X-1159
X-12095--N Lathoste ol 3--
Cortisate Xenobiotic Ascorb ThreonaO-methylas
l disulf l ate (vi te corbate
α- dio stero tamin
β, 20 Cholertisone Bilirub C)
-p re gnan-3 Co oca) Bilirub in (Z
5α e (7-H in (E ,Z) Biliverdin
te noat fate Biliru ,E)
oles ul Hem bin (E X-117
o- 4-ch ,20α-dis e ,Z or 93--ox
3-ox 3β
oxy- egnan- Pan Z,E) idiz
hydr to ed bi
7-α- --5-α-prte 2 γ-to thenat
e
lirub
in
45 lfa α-to coph
X-114 ol disu lfate n c ero
β-di disu yeli Pyri ophero l
om
3β, 17 diol 1 ing lerate dox l
os ten- β,17β- ulfate ) sph Va ) ate
dr -3
4-an rostan -diol d HEA-S
is oyl lein Su
7β lmit oo in) Ma ccin
nd (D Pa on almit rin) ) C
α-k late itra ylc
5α-a n-3β, 1 lfate lfate (1-m op stea lein P eto te arn
e e su s u l n itin
dro st
ero
n ne ero -mo no no :0) ) Ac hosp gluta e
4-an ndrost rostero lyc l (1 -moonoli (6 ety ha rate
isoa Epiand ulfate
y lg ro (1 te 1:0 ) lph te
leo lyce rol (1-m roate (1 (7:0:0) C
G ho os
ydro s 1-ooylg lyceerol p 2 ) ph
Deh ne it ylg c Caanoaoate (110:0:0) T ly la
tero alm aro lgly c n te ( 9 ) Taaur coute ate
ros 1-p -ste leoy de pta ura te ( 8:0 G o r T au
And Un He Laapra natete ( U ly ur c so
1
in o Hy rso co olithhen deo rode
1-l C rgo ryla 1) G c o o x ox
n e D ly od de he c de yc yc
la p
Pe Ca 11:1 lin ol T e c e o n ho o h ho
( ho sit e G au oxyoch oxy xyc ode latexycholate late
ly ro c o ch ho ox 3 o
te ) oc ino in ine co c ho la o la yc -s la
oa n7 sph ho olam m
1 ylg le ylg lyc er p os a o ch oli ne

de ho la te lat te ho ulf te
en :1
1-e idon -my lyc oylg lyc ero oph hosp phonolaino oli ne
leo 1-o oleo oylg lyc cero ophoeth sphphooch holi

o p n la ox late te e late ate

1-p icos oylg risto erop lyce erop pho osp ho ch minsito ne

ec (12 ph hos tha ano

a ly ylg h ro h s ho ch oli e l

yc
2- ole oylglgly cer ph pho os ph oc

n d ro p
-u ate ce ro oe th ho
1-d 1-a lmito trien cero lyc osph phoosp pho cho olin ne
lin n y ly os o h s h
1- eca ito oylg ph cer eroppho osp

10 eno lgly ce sph nyle

oco rac ylg oylg pho ero oe sp ho cho lin e

la
h lyc ly s p th h ch li e

te
y ly o e
1-a ahex idon ero ceropho hosp anooch olin ne

c
2- aury ylca lcarn itin
d lm r ro ly c o h

de
pta pa a e ylg lgly cer rop

L o

no oylg oph sm
1- rach aenooylglyphos phoetha hoc lamolinee

X- Car Cis ylc ine

tetr lc rn
Ole lmit ylc

do 13
itolenoylg lyce rophoeth hoc lam linee

Pa tearo

die ar er la
2- ylgly ito roy lgly ce

5-
palm ido ylg ce ph sp no ho in

ad arn itin itine

S

sa ste lyc ylp 43 itine 4-de nitin

o alm a oy ly

oylg lyc rop osp ano holinine

Pe roph osphhochholinine
lyc erop hosp hoc lam e
ole -p ste le ylg

ec itin e

ico 1- ylg ito

dec sphooinosolinee

1-
Ac noyl
Deetylc carn

He ona Oc
3- ano oylc no
S noa holi ol

-n
lin 1 1- lino ito

e
e e
o
Ole teara te (1 ne

X-1 omo ate oco )

1- ro alm

de y a ylc
n - ar

x n ta
Dihinolen D 8:2n6 e

ca ar itin
y rn
8 )
c it
2- palm

ate te (1 5:0

ea -p
Non Marglmitate8:1n9:0)

hy lca rnit arn

no nit e

st 1
oate 7:0 )

Lino 83--st

299 -Lin [α o sape

ade ara (1 )

dr rn ine itin
can te (1 6:0

X-1
Dihom ecen Myrisate (2 (19:0 )
o-lin oate (1 tate 2:4n6))

1-
1-

oc itin
15-m
9:1n (14:0)
c

0-- ole r γ ; nta

ar e
leate eara
31
Dode
Octade

ce e

nit
2-hydr

Eic cosaapen0:3n3 or n6(n3 D

:4n3)

N-butyl

Do os (2
Isovalera

e
doc nate (18 eno
(1

)
te

no
Glycerol

ethy
tradecaoate (17:1ate (20:4n6

in
Choline
Chiro-inositol

2-hydroxypalmitate
2-hydroxystearate
Myo-inositol
Scyllo-inositol
9)
o
s

0:1n9 0:2n6)
h

osa he rae or

e
Glycerol 3-phosphate (G3P)
Glycerophosphorylcholine (GPC)

yl
a

(1

c
canedio
1n7)or 11)

ca
ate (14:1naridonate (18

lpalm
-

pe xae noic n6)

oxyglut
2

canedioarate
Pa

rn
1-

Adren
nta

oleate

nta noa ac

it
n7)
Palm noate (2oleate (2

e
ine
te

Tetracarboxy- ecanedio
he

m
itate

en te id (n

X-1
3- Hexad
1-

ate e
Propionylcarn

id
BA)
on

Ste ate

oa (D
:3
at
1-o

24
deca 4-met
e (16:

n3 te
X-1244 10-heptadec Arachid

(isob

te HA 6-DP
dieno

5)
tyrate (BH

41
(EP ; 2
a

--1
h

nedi
a

onad

ar w
rac

itoleat
s

A; 2:6n )

2-h
en

)]
1-a

oatehyl-5-pro

20
ith 2-

yd
Eicose
10-n

itine

:5n 3)

rox
3-hydroxybu

PA
Myristole

3)
2--5,8-te

met

ye
A
; 22
ate

ico
hylp

sa
:5 n3)

tetr
pyl-2

alm

ae
itate
-fura

no
ate
)
npro

(12

Explained variance
-H
Butyrylcarn

pano

ET

Estimated heritability
E)
ate (C

Common environmental components

MPF)
itine

Fig. 3 Map of associations between the single-nucleotide polymorphism (SNP) and metabolite levels in blood.
The contribution of genetic vs. environmental factors to metabolite level variance was estimated on the basis
of GWAS study results. The proportion of heritability explained by all SNPs associated with a given metabolite
at the genome-wide level is shown in red. Figure was reproduced from Shin et al. Nature genetics 2014 with
the permission for re-use from Nature publishing group [116]

the association between genetic variants and specific product meta-

bolites allows for the early determination of adverse pharmacologi-
cal effects [118].
In the era of pharmacogenomics and personalized medicine,
the term “pharmacometabolomics” has emerged to define the use
of metabolomics in measurement of drug or disease biomarker
signatures to predict and assess therapeutic response
[119–122]. The measured changes in endogenous and exogenous
metabolite levels reflect the end response of biological systems as a
combined effect of genetic determinants and environmental and
18 Julijana Ivanisevic and Aurelien Thomas

internal microbiome exposures. In the field of pharmacology and

toxicology, metabolomics is increasingly applied to determine sig-
natures of drug exposition, susceptibility, or toxicity (e.g., early
signature of hepatotoxicity) and to investigate, in relation to the
observed molecular patterns, the cellular mechanisms leading to
the adverse effects of xenobiotics and to the development of asso-
ciated pathologies [123]. Via monitoring of drug metabolites and
metabolism (through drug levels and intake), metabolomic assays
can considerably contribute to the assessment of toxicity and opti-
mization of drug doses.
A traditional view of the medicine is to apply a standardized
treatment following the “one size fits all” dogma to the entire
population of patients suffering from the same disease. However,
responder patients to a major drug treatment can differ from 25% in
the case of cancer chemotherapy to 80% in the case of Cox-2
inhibitors [124]. Such an interindividual variability in the response
rate depends on multiple factors involving the genetic and environ-
mental components, including dietary regimen, toxin exposures, or
potential drug-drug interactions [125]. Even slight modifications
of the pharmacokinetic parameters induced by this set of factors can
have important effect on pharmacodynamic properties of a drug.
Cytochrome P450 (CYP) enzymes are the main drug-metabolizing
systems in humans and the major source of variability in drug
pharmacokinetics and response [125]. Their expression is highly
influenced by a combination of factors, including genetic polymor-
phism, disease state, sex, age, etc. In addition, the pharmacokinetic
variability in drug response is also increased by the variability of
drug efflux and influx transporters [126]. Accordingly, the activity
modifications of these classes of proteins can induce numerous
pharmacological and toxicological consequences. For instance,
the risk assessment of drug-drug interactions in elderly patients or
patients under polymedication is a challenging task in clinical prac-
tice [127]. The in vivo assessment of the activities of enzymes and
transporters to adapt the therapeutic strategy, known as phenotyp-
ing, is already used in clinical setting. Phenotyping consists of the
controlled administration of a cocktail containing multiple probes,
each of them targeting one specific enzyme or transporter
[128, 129]. Unlike genotyping, phenotyping presents the advan-
tage of measuring the combined effect of genetic, environmental,
and endogenous factors on the activity of proteins involved in the
drug pharmacokinetic variability [130, 131]. To overcome the
potential issues with a drug cocktail administration in patients at
risk, the possibility of developing metabolomics-based endogenous
biomarkers is a promising alternative to current phenotyping assays.
For example, change in blood levels of an unidentified metabolite
was recently shown to be significantly associated with polymor-
phism in genomic region corresponding to the CYP3A5 gene
locus [132]. Another recent untargeted metabolomics study
 Metabolomics in Biomedical and Clinical Research 19

allowed the identification of a urinary metabolite associated with

CYP2D6 activity [133]. This association was validated in an adult
cohort with the observation of a strong decrease of the metabolite
abundance following the administration of fluoxetine as an inhibi-
tor of CYP2D6.
In a systems biology approach, the integration of genomics,
epigenomics, transcriptomics, proteomics, and metabolomics to
investigate the response of an organism to xenobiotics represents
a unique opportunity to accelerate the understanding of alterations
in biological pathways associated with pathophysiological traits and
the discovery of more classified biomarkers toward personalized
approach to the medicine [134]. The integration of metabolomics
to systems biology approach is rapidly evolving to provide a more
complete picture of the dynamics of molecular systems [135]. Due
to the fact that many diseases have a strong metabolic basis (that is
heavily influenced by external and internal environmental expo-
sures), in this post-genomic era, metabolomics will undoubtedly
play a determinant role, offering powerful phenotyping capacity for
more cost-effective models in drug discovery, efficient monitoring
of therapeutic response, and customization of drug dosing.

References

1. Wishart DS (2016) Emerging applications of Margaretha L, Robitaille AM,

metabolomics in drug discovery and precision Margineantu D, Fiehn O, Hockenbery D,
medicine. Nat Rev Drug Discov 15 Blau CA, Raftery D, Margolin AA, Hawkins
(7):473–484. https://doi.org/10.1038/ RD, Moon RT, Ware CB, Ruohola-Baker H
nrd.2016.32 (2015) The metabolome regulates the epige-
2. Johnson CH, Ivanisevic J, Siuzdak G (2016) netic landscape during naive-to-primed
Metabolomics: beyond biomarkers and human embryonic stem cell transition. Nat
towards mechanisms. Nat Rev Mol Cell Biol Cell Biol 17(12):1523–1535. https://doi.
17(7):451–459. https://doi.org/10.1038/ org/10.1038/ncb3264
nrm.2016.25 8. Sabari BR, Zhang D, Allis CD, Zhao Y (2017)
3. Ivanisevic J, Siuzdak G (2015) The role of Metabolic regulation of gene expression
metabolomics in brain metabolism research. through histone acylations. Nat Rev Mol
J Neuroimmune Pharmacol 10:391–395 Cell Biol 18(2):90–101. https://doi.org/10.
4. Schmidt CW (2004) Metabolomics: what’s 1038/nrm.2016.140
happening downstream of DNA. Environ 9. Siroux V, Agier L, Slama R (2016) The expo-
Health Perspect 112(7):A410–A415 some concept: a challenge and a potential
5. Zamboni N, Saghatelian A, Patti GJ (2015) driver for environmental health research. Eur
Defining the metabolome: size, flux, and reg- Respir Rev 25(140):124–129. https://doi.
ulation. Mol Cell 58(4):699–706. https:// org/10.1183/16000617.0034-2016
doi.org/10.1016/j.molcel.2015.04.021 10. Pedersen HK, Gudmundsdottir V, Nielsen
6. Patti GJ, Yanes O, Siuzdak G (2012) Innova- HB, Hyotylainen T, Nielsen T, Jensen BAH,
tion: metabolomics: the apogee of the omics Forslund K, Hildebrand F, Prifti E, Falony G,
trilogy. Nat Rev Mol Cell Biol 13 Le Chatelier E, Levenez F, Doré J, Mattila I,
(4):263–269. https://doi.org/10.1038/ Plichta DR, Pöhö P, Hellgren LI,
nrm3314 Arumugam M, Sunagawa S, Vieira-Silva S,
Jørgensen T, Holm JB, Trošt K, MetaHIT
7. Sperber H, Mathieu J, Wang Y, Ferreccio A, Consortium, Kristiansen K, Brix S, Raes J,
Hesson J, Xu Z, Fischer KA, Devi A, Wang J, Hansen T, Bork P, Brunak S,
Detraux D, Gu H, Battle SL, Showalter M, Oresic M, Ehrlich SD, Pedersen O (2016)
Valensisi C, Bielas JH, Ericson NG, Human gut microbes impact host serum
20 Julijana Ivanisevic and Aurelien Thomas

metabolome and insulin sensitivity. Nature Metabolomics 12(9):149. https://doi.org/

535(7612):376–381 10.1007/s11306-016-1094-6
11. Bucci M (2016) Gut microbiome: branching 20. Su LJ, Fiehn O, Maruvada P, Moore SC,
into metabolic disease. Nat Chem Biol 12 O’Keefe SJ, Wishart DS, Zanetti KA (2014)
(9):657–657. https://doi.org/10.1038/ The use of metabolomics in population-based
nchembio.2164 research. Adv Nutr 5(6):785–788. https://
12. Mayer EA, Knight R, Mazmanian SK, Cryan doi.org/10.3945/an.114.006494
JF, Tillisch K (2014) Gut microbes and the 21. Psychogios N, Hau DD, Peng J, Guo AC,
brain: paradigm shift in neuroscience. J Neu- Mandal R, Bouatra S, Sinelnikov I,
rosci 34(46):15490–15496. https://doi.org/ Krishnamurthy R, Eisner R, Gautam B,
10.1523/jneurosci.3299-14.2014 Young N, Xia J, Knox C, Dong E, Huang P,
13. Mamas M, Dunn WB, Neyses L, Goodacre R Hollander Z, Pedersen TL, Smith SR,
(2011) The role of metabolites and metabo- Bamforth F, Greiner R, McManus B, New-
lomics in clinically applicable biomarkers of man JW, Goodfriend T, Wishart DS (2011)
disease. Arch Toxicol 85(1):5–17. https:// The human serum metabolome. PLoS One 6
doi.org/10.1007/s00204-010-0609-6 (2):e16957. https://doi.org/10.1371/jour
14. Weber RJM, Lawson TN, Salek RM, Ebbels nal.pone.0016957
TMD, Glen RC, Goodacre R, Griffin JL, 22. Yanes O, Tautenhahn R, Patti GJ, Siuzdak G
Haug K, Koulman A, Moreno P, Ralser M, (2011) Expanding coverage of the metabo-
Steinbeck C, Dunn WB, Viant MR (2017) lome for global metabolite profiling. Anal
Computational tools and workflows in meta- Chem 83(6):2152–2161. https://doi.org/
bolomics: an international survey highlights 10.1021/ac102981k
the opportunity for harmonisation through 23. Zhang A, Sun H, Yan G, Wang P, Han Y,
galaxy. Metabolomics 13(2):12. https://doi. Wang X (2014) Metabolomics in diagnosis
org/10.1007/s11306-016-1147-x and biomarker discovery of colorectal cancer.
15. Benton HP, Ivanisevic J, Mahieu NG, Kurczy Cancer Lett 345(1):17–20. https://doi.org/
ME, Johnson CH, Franco L, Rinehart D, 10.1016/j.canlet.2013.11.011
Valentine E, Gowda H, Ubhi BK, 24. Thomas A, Lenglet S, Chaurand P, Deglon J,
Tautenhahn R, Gieschen A, Fields MW, Patti Mangin P, Mach F, Steffens S, Wolfender JL,
GJ, Siuzdak G (2015) Autonomous metabo- Staub C (2011) Mass spectrometry for the
lomics for rapid metabolite identification in evaluation of cardiovascular diseases based
global profiling. Anal Chem 87(2):884–891. on proteomics and lipidomics. Thromb Hae-
https://doi.org/10.1021/ac5025649 most 106(1):20–33. https://doi.org/10.
16. Tautenhahn R, Cho K, Uritboonthai W, 1160/TH10-12-0812
Zhu Z, Patti GJ, Siuzdak G (2012) An accel- 25. Dunn WB, Bailey NJ, Johnson HE (2005)
erated workflow for untargeted metabolomics Measuring the metabolome: current analytical
using the METLIN database. Nat Biotechnol technologies. Analyst 130(5):606–625.
30(9):826–828. https://doi.org/10.1038/ https://doi.org/10.1039/b418288j
nbt.2348 26. Dunn WB, Broadhurst D, Begley P, Zelena E,
17. Johnson CH, Ivanisevic J, Benton HP, Siuz- Francis-McIntyre S, Anderson N, Brown M,
dak G (2015) Bioinformatics: the next fron- Knowles JD, Halsall A, Haselden JN, Nicholls
tier of metabolomics. Anal Chem 87 AW, Wilson ID, Kell DB, Goodacre R,
(1):147–156. https://doi.org/10.1021/ Human Serum Metabolome Consortium
ac5040693 (2011) Procedures for large-scale metabolic
18. Ivanisevic J, Elias D, Deguchi H, Averell PM, profiling of serum and plasma using gas chro-
Kurczy M, Johnson CH, Tautenhahn R, matography and liquid chromatography cou-
Zhu Z, Watrous J, Jain M (2015) Arteriove- pled to mass spectrometry. Nat Protoc 6
nous blood metabolomics: a readout of intra- (7):1060–1083. https://doi.org/10.1038/
tissue metabostasis. Sci Rep 5:12757 nprot.2011.335
19. Beger RD, Dunn W, Schmidt MA, Gross SS, 27. Schweiger R, Baier MC, Persicke M, Muller C
Kirwan JA, Cascante M, Brennan L, Wishart (2014) High specificity in plant leaf metabolic
DS, Oresic M, Hankemeier T, Broadhurst DI, responses to arbuscular mycorrhiza. Nat
Lane AN, Suhre K, Kastenm€ uller G, Sumner Commun 5:3886. https://doi.org/10.
SJ, Thiele I, Fiehn O, Kaddurah-Daouk R, for 1038/ncomms4886
“Precision M, Pharmacometabolomics Task 28. Tugizimana F, Steenkamp PA, Piater LA,
Group”-Metabolomics Society I (2016) Dubery IA (2014) Multi-platform metabolo-
Metabolomics enables precision medicine: “a mic analyses of ergosterol-induced dynamic
white paper, community perspective”. changes in Nicotiana tabacum cells. PLoS
 Metabolomics in Biomedical and Clinical Research 21

One 9(1):e87846. https://doi.org/10. platform to process untargeted metabolomic

1371/journal.pone.0087846 data. Anal Chem 84(11):5035–5039.
29. Bouatra S, Aziat F, Mandal R, Guo AC, Wil- https://doi.org/10.1021/ac300698c
son MR, Knox C, Bjorndahl TC, 38. Xia J, Wishart DS (2011) Web-based infer-
Krishnamurthy R, Saleem F, Liu P, Dame ence of biological patterns, functions and
ZT, Poelzer J, Huynh J, Yallou FS, pathways from metabolomic data using Meta-
Psychogios N, Dong E, Bogumil R, boAnalyst. Nat Protoc 6(6):743–760.
Roehring C, Wishart DS (2013) The human https://doi.org/10.1038/nprot.2011.319
urine metabolome. PLoS One 8(9):e73076. 39. Eugster PJ, Glauser G, Wolfender JL (2013)
https://doi.org/10.1371/journal.pone. Strategies in biomarker discovery. Peak anno-
0073076 tation by MS and targeted LC-MS micro-frac-
30. Want EJ, Masson P, Michopoulos F, Wilson tionation for de novo structure identification
ID, Theodoridis G, Plumb RS, Shockcor J, by micro-NMR. Methods Mol Biol
Loftus N, Holmes E, Nicholson JK (2013) 1055:267–289. https://doi.org/10.1007/
Global metabolic profiling of animal and 978-1-62703-577-4_19
human tissues via UPLC-MS. Nat Protoc 8 40. Wishart DS, Jewison T, Guo AC, Wilson M,
(1):17–32. https://doi.org/10.1038/nprot. Knox C, Liu Y, Djoumbou Y, Mandal R,
2012.135 Aziat F, Dong E, Bouatra S, Sinelnikov I,
31. Junot C, Fenaille F, Colsch B, Becher F Arndt D, Xia J, Liu P, Yallou F, Bjorndahl T,
(2013) High resolution mass spectrometry Perez-Pineiro R, Eisner R, Allen F, Neveu V,
based techniques at the crossroads of meta- Greiner R, Scalbert A (2013) HMDB 3.0—
bolic pathways. Mass Spectrom Rev 33 The Human Metabolome Database in 2013.
(6):471–500. https://doi.org/10.1002/ Nucleic Acids Res 41(Database issue):
mas.21401 D801–D807. https://doi.org/10.1093/
32. Fiehn O (2002) Metabolomics – the link nar/gks1065
between genotypes and phenotypes. Plant 41. Yuan M, Breitkopf SB, Yang X, Asara JM
Mol Biol 48(1–2):155–171 (2012) A positive/negative ion-switching,
33. Wolfender JL, Marti G, Thomas A, Bertrand S targeted mass spectrometry-based metabolo-
(2014) Current approaches and challenges for mics platform for bodily fluids, cells, and fresh
the metabolite profiling of complex natural and fixed tissue. Nat Protoc 7(5):872–881.
extracts. J Chromatogr A 1382:136–164. https://doi.org/10.1038/nprot.2012.024
https://doi.org/10.1016/j.chroma.2014. 42. Cai Y, Weng K, Guo Y, Peng J, Zhu Z-J
10.091 (2015) An integrated targeted metabolomic
34. Thomas A, Deglon J, Steimer T, Mangin P, platform for high-throughput metabolite
Daali Y, Staub C (2010) On-line desorption profiling and automated data processing.
of dried blood spots coupled to hydrophilic Metabolomics 11(6):1575–1586. https://
interaction/reversed-phase LC/MS/MS sys- doi.org/10.1007/s11306-015-0809-4
tem for the simultaneous analysis of drugs and 43. Norris JL, Caprioli RM (2013) Analysis of
their polar metabolites. J Sep Sci 33 tissue specimens by matrix-assisted laser
(6–7):873–879. https://doi.org/10.1002/ desorption/ionization imaging mass spec-
jssc.200900593 trometry in biological and clinical research.
35. McCalley DV (2007) Is hydrophilic interac- Chem Rev 113(4):2309–2342. https://doi.
tion chromatography with silica columns a org/10.1021/cr3004295
viable alternative to reversed-phase liquid 44. Sun N, Ly A, Meding S, Witting M, Hauck
chromatography for the analysis of ionisable SM, Ueffing M, Schmitt-Kopplin P,
compounds? J Chromatogr A 1171 Aichler M, Walch A (2014) High-resolution
(1–2):46–55. https://doi.org/10.1016/j. metabolite imaging of light and dark treated
chroma.2007.09.047 retina using MALDI-FTICR mass spectrome-
36. Ivanisevic J, Zhu ZJ, Plate L, Tautenhahn R, try. Proteomics 14(7-8):913–923. https://
Chen S, O’Brien PJ, Johnson CH, Marletta doi.org/10.1002/pmic.201300407
MA, Patti GJ, Siuzdak G (2013) Toward 45. Thomas A, Charbonneau JL, Fournaise E,
‘omic scale metabolite profiling: a dual Chaurand P (2012) Sublimation of new
separation-mass spectrometry approach for matrix candidates for high spatial resolution
coverage of lipid and central carbon metabo- imaging mass spectrometry of lipids:
lism. Anal Chem 85(14):6876–6884. Enhanced information in both positive and
https://doi.org/10.1021/ac401140h negative polarities after
37. Tautenhahn R, Patti GJ, Rinehart D, Siuzdak 1,5-diaminonapthalene deposition. Anal
G (2012) XCMS Online: a web-based
22 Julijana Ivanisevic and Aurelien Thomas

Chem 84(4):2048–2054. https://doi.org/ Winkler TW, Wood AR, Workalemahu T,

10.1021/ac2033547 Faul JD, Smith JA, Zhao JH, Zhao W,
46. Ly A, Buck A, Balluff B, Sun N, Gorzolka K, Chen J, Fehrmann R, Hedman ÅK,
Feuchtinger A, Janssen KP, Kuppen PJ, van de Karjalainen J, Schmidt EM, Absher D,
Velde CJ, Weirich G, Erlmeier F, Langer R, Amin N, Anderson D, Beekman M, Bolton
Aubele M, Zitzelsberger H, McDonnell L, JL, Bragg-Gresham JL, Buyske S,
Aichler M, Walch A (2016) High-mass-reso- Demirkan A, Deng G, Ehret GB, Feenstra B,
lution MALDI mass spectrometry imaging of Feitosa MF, Fischer K, Goel A, Gong J, Jack-
metabolites from formalin-fixed paraffin- son AU, Kanoni S, Kleber ME,
embedded tissue. Nat Protoc 11 Kristiansson K, Lim U, Lotay V,
(8):1428–1443. https://doi.org/10.1038/ Mangino M, Leach IM, Medina-Gomez C,
nprot.2016.081 Medland SE, Nalls MA, Palmer CD,
47. Patterson NH, Alabdulkarim B, Lazaris A, Pasko D, Pechlivanis S, Peters MJ,
Thomas A, Marcinkiewicz MM, Gao ZH, Prokopenko I, Shungin D, Stančáková A,
Vermeulen PB, Chaurand P, Metrakos P Strawbridge RJ, Sung YJ, Tanaka T,
(2016) Assessment of pathological response Teumer A, Trompet S, van der Laan SW, van
to therapy using lipid mass spectrometry Setten J, Van Vliet-Ostaptchouk JV, Wang Z,
imaging. Sci Rep 6:36814. https://doi.org/ Yengo L, Zhang W, Isaacs A, Albrecht E,
10.1038/srep36814 Ärnlöv J, Arscott GM, Attwood AP,
Bandinelli S, Barrett A, Bas IN, Bellis C, Ben-
48. Aichler M, Walch A (2015) MALDI Imaging nett AJ, Berne C, Blagieva R, Bl€ uher M,
mass spectrometry: current frontiers and per- Böhringer S, Bonnycastle LL, Böttcher Y,
spectives in pathology research and practice. Boyd HA, Bruinenberg M, Caspersen IH,
Lab Invest 95(4):422–431. https://doi.org/ Chen YI, Clarke R, Daw EW, de Craen AJM,
10.1038/labinvest.2014.156 Delgado G, Dimitriou M, Doney ASF,
49. Hamilton LK, Dufresne M, Joppe SE, Eklund N, Estrada K, Eury E, Folkersen L,
Petryszyn S, Aumont A, Calon F, Barnabe- Fraser RM, Garcia ME, Geller F, Giedraitis V,
Heider F, Furtos A, Parent M, Chaurand P, Gigante B, Go AS, Golay A, Goodall AH,
Fernandes KJ (2015) Aberrant lipid metabo- Gordon SD, Gorski M, Grabe HJ,
lism in the forebrain niche suppresses adult Grallert H, Grammer TB, Gr€aßler J,
neural stem cell proliferation in an animal Grönberg H, Groves CJ, Gusto G,
model of Alzheimer’s disease. Cell Stem Cell Haessler J, Hall P, Haller T, Hallmans G,
17(4):397–411. https://doi.org/10.1016/j. Hartman CA, Hassinen M, Hayward C,
stem.2015.08.001 Heard-Costa NL, Helmer Q,
50. Tonnies E, Trushina E (2017) Oxidative Hengstenberg C, Holmen O, Hottenga JJ,
stress, synaptic dysfunction, and Alzheimer’s James AL, Jeff JM, Johansson Å, Jolley J,
disease. J Alzheimer’s Dis 57(4):1105–1121. Juliusdottir T, Kinnunen L, Koenig W,
https://doi.org/10.3233/jad-161088 Koskenvuo M, Kratzer W, Laitinen J,
51. Stranahan AM, Mattson MP (2012) Recruit- Lamina C, Leander K, Lee NR, Lichtner P,
ing adaptive cellular stress responses for suc- Lind L, Lindström J, Lo KS, Lobbens S,
cessful brain ageing. Nat Rev Neurosci 13 Lorbeer R, Lu Y, Mach F, Magnusson PKE,
(3):209–216. https://doi.org/10.1038/ Mahajan A, McArdle WL, McLachlan S,
nrn3151 Menni C, Merger S, Mihailov E, Milani L,
52. http://www.who.int/mediacentre/ Moayyeri A, Monda KL, Morken MA,
factsheets/fs311/en/ Mulas A, M€ uller G, M€ uller-Nurasyid M,
Musk AW, Nagaraja R, Nöthen MM, Nolte
53. Jiao H, Arner P, Hoffstedt J, Brodin D, IM, Pilz S, Rayner NW, Renstrom F, Rettig R,
Dubern B, Czernichow S, van’t Hooft F, Ried JS, Ripke S, Robertson NR, Rose LM,
Axelsson T, Pedersen O, Hansen T, Sorensen Sanna S, Scharnagl H, Scholtens S, Schuma-
TI, Hebebrand J, Kere J, Dahlman-Wright K, cher FR, Scott WR, Seufferlein T, Shi J, Smith
Hamsten A, Clement K, Dahlman I (2011) AV, Smolonska J, Stanton AV,
Genome wide association study identifies Steinthorsdottir V, Stirrups K, Stringham
KCNMA1 contributing to human obesity. HM, Sundström J, Swertz MA, Swift AJ,
BMC Med Genomics 4:51. https://doi.org/ Syv€anen AC, Tan ST, Tayo BO, Thorand B,
10.1186/1755-8794-4-51 Thorleifsson G, Tyrer JP, Uh HW,
54. Locke AE, Kahali B, Berndt SI, Justice AE, Vandenput L, Verhulst FC, Vermeulen SH,
Pers TH, Day FR, Powell C, Vedantam S, Verweij N, Vonk JM, Waite LL, Warren HR,
Buchkovich ML, Yang J, Croteau-Chonka Waterworth D, Weedon MN, Wilkens LR,
DC, Esko T, Fall T, Ferreira T, Gustafsson S, Willenborg C, Wilsgaard T, Wojczynski MK,
Kutalik Z, Luan J, M€agi R, Randall JC,
 Metabolomics in Biomedical and Clinical Research 23

Wong A, Wright AF, Zhang Q, LifeLines Bochud M, Boehm BO, Boerwinkle E, Born-
Cohort Study, Brennan EP, Choi M, stein SR, Bottinger EP, Bouchard C,
Dastani Z, Drong AW, Eriksson P, Franco- Cauchi S, Chambers JC, Chanock SJ, Cooper
Cereceda A, Gådin JR, Gharavi AG, Goddard RS, de Bakker PIW, Dedoussis G, Ferrucci L,
ME, Handsaker RE, Huang J, Karpe F, Franks PW, Froguel P, Groop LC, Haiman
Kathiresan S, Keildson S, Kiryluk K, CA, Hamsten A, Hui J, Hunter DJ,
Kubo M, Lee JY, Liang L, Lifton RP, Ma B, Hveem K, Kaplan RC, Kivimaki M, Kuh D,
McCarroll SA, McKnight AJ, Min JL, Moffatt Laakso M, Liu Y, Martin NG, M€arz W,
MF, Montgomery GW, Murabito JM, Melbye M, Metspalu A, Moebus S, Munroe
Nicholson G, Nyholt DR, Okada Y, JRB P, PB, Njølstad I, Oostra BA, Palmer CNA, Ped-
Dorajoo R, Reinmaa E, Salem RM, ersen NL, Perola M, Pérusse L, Peters U,
Sandholm N, Scott RA, Stolk L, Power C, Quertermous T, Rauramaa R,
Takahashi A, Tanaka T, van’t Hooft FM, Rivadeneira F, Saaristo TE, Saleheen D,
AAE V, Westra HJ, Zheng W, Zondervan Sattar N, Schadt EE, Schlessinger D, Slag-
KT, ADIPOGen Consortium, AGEN-BMI boom PE, Snieder H, Spector TD,
Working Group, CARDIOGRAMplusC4D Thorsteinsdottir U, Stumvoll M,
Consortium, CKDGen Consortium, GLGC, Tuomilehto J, Uitterlinden AG, Uusitupa M,
ICBP, MAGIC Investigators, MuTHER van der Harst P, Walker M, Wallaschofski H,
Consortium, MIGen Consortium, PAGE Wareham NJ, Watkins H, Weir DR, Wich-
Consortium, ReproGen Consortium, mann HE, Wilson JF, Zanen P, Borecki IB,
GENIE Consortium, International Endo- Deloukas P, Fox CS, Heid IM, O’Connell JR,
gene Consortium, Heath AC, Arveiler D, Strachan DP, Stefansson K, van Duijn CM,
SJL B, Beilby J, Bergman RN, Blangero J, Abecasis GR, Franke L, Frayling TM,
Bovet P, Campbell H, Caulfield MJ, McCarthy MI, Visscher PM, Scherag A,
Cesana G, Chakravarti A, Chasman DI, Willer CJ, Boehnke M, Mohlke KL, Lindgren
Chines PS, Collins FS, Crawford DC, Cupples CM, Beckmann JS, Barroso I, North KE,
LA, Cusi D, Danesh J, de Faire U, den Ruijter Ingelsson E, Hirschhorn JN, Loos RJF, Spe-
HM, Dominiczak AF, Erbel R, Erdmann J, liotes EK (2015) Genetic studies of body mass
Eriksson JG, Farrall M, Felix SB, Ferrannini E, index yield new insights for obesity biology.
Ferrières J, Ford I, Forouhi NG, Forrester T, Nature 518(7538):197–206. https://doi.
Franco OH, Gansevoort RT, Gejman PV, org/10.1038/nature14177
Gieger C, Gottesman O, Gudnason V, 55. Moreno-Navarrete JM, Jove M, Ortega F,
Gyllensten U, Hall AS, Harris TB, Hattersley Xifra G, Ricart W, Obis E, Pamplona R,
AT, Hicks AA, Hindorff LA, Hingorani AD, Portero-Otin M, Fernandez-Real JM (2016)
Hofman A, Homuth G, Hovingh GK, Metabolomics uncovers the role of adipose
Humphries SE, Hunt SC, Hyppönen E, tissue PDXK in adipogenesis and systemic
Illig T, Jacobs KB, Jarvelin MR, Jöckel KH, insulin sensitivity. Diabetologia 59
Johansen B, Jousilahti P, Jukema JW, Jula (4):822–832. https://doi.org/10.1007/
AM, Kaprio J, Kastelein JJP, Keinanen- s00125-016-3863-1
Kiukaanniemi SM, Kiemeney LA, Knekt P, 56. Abu Bakar MH, Sarmidi MR, Cheng KK, Ali
Kooner JS, Kooperberg C, Kovacs P, Kraja Khan A, Suan CL, Zaman Huri H, Yaakob H
AT, Kumari M, Kuusisto J, Lakka TA, (2015) Metabolomics – the complementary
Langenberg C, Marchand LL, Lehtim€aki T, field in systems biology: a review on obesity
Lyssenko V, M€annistö S, Marette A, Matise and type 2 diabetes. Mol Biosyst 11
TC, McKenzie CA, McKnight B, Moll FL, (7):1742–1774. https://doi.org/10.1039/
Morris AD, Morris AP, Murray JC, Nelis M, c5mb00158g
Ohlsson C, Oldehinkel AJ, Ong KK, PAF M,
Pasterkamp G, Peden JF, Peters A, Postma 57. Roberts LD, Koulman A, Griffin JL (2014)
DS, Pramstaller PP, Price JF, Qi L, Raitakari Towards metabolic biomarkers of insulin
OT, Rankinen T, Rao DC, Rice TK, Ridker resistance and type 2 diabetes: progress from
PM, Rioux JD, Ritchie MD, Rudan I, the metabolome. Lancet Diabetes Endocrinol
Salomaa V, Samani NJ, Saramies J, Sarzynski 2(1):65–75. https://doi.org/10.1016/
MA, Schunkert H, Schwarz PEH, Sever P, S2213-8587(13)70143-8
Shuldiner AR, Sinisalo J, Stolk RP, 58. Morris C, O’Grada C, Ryan M, Roche HM,
Strauch K, Tönjes A, Trégouët DA, Gibney MJ, Gibney ER, Brennan L (2012)
Tremblay A, Tremoli E, Virtamo J, Vohl The relationship between BMI and metabo-
MC, Völker U, Waeber G, Willemsen G, Wit- lomic profiles: a focus on amino acids. Proc
teman JC, Zillikens MC, Adair LS, Nutr Soc 71(4):634–638. https://doi.org/
Amouyel P, Asselbergs FW, Assimes TL, 10.1017/S0029665112000699
24 Julijana Ivanisevic and Aurelien Thomas

59. Newgard CB, An J, Bain JR, Muehlbauer MJ, study. Am J Clin Nutr 102(2):433–443.
Stevens RD, Lien LF, Haqq AM, Shah SH, https://doi.org/10.3945/ajcn.114.103804
Arlotto M, Slentz CA, Rochon J, Gallup D, 67. Perez-Cornago A, Brennan L, Ibero-
Ilkayeva O, Wenner BR, Yancy WS Jr, Baraibar I, Hermsdorff HH, O’Gorman A,
Eisenson H, Musante G, Surwit RS, Milli- Zulet MA, Martinez JA (2014) Metabolomics
ngton DS, Butler MD, Svetkey LP (2009) A identifies changes in fatty acid and amino acid
branched-chain amino acid-related metabolic profiles in serum of overweight older adults
signature that differentiates obese and lean following a weight loss intervention. J Physiol
humans and contributes to insulin resistance. Biochem 70(2):593–602. https://doi.org/
Cell Metab 9(4):311–326. https://doi.org/ 10.1007/s13105-013-0311-2
10.1016/j.cmet.2009.02.002 68. Shah SH, Crosslin DR, Haynes CS, Nelson S,
60. Makki K, Froguel P, Wolowczuk I (2013) Turer CB, Stevens RD, Muehlbauer MJ,
Adipose tissue in obesity-related inflamma- Wenner BR, Bain JR, Laferrere B,
tion and insulin resistance: cells, cytokines, Gorroochurn P, Teixeira J, Brantley PJ, Ste-
and chemokines. ISRN Inflamm vens VJ, Hollis JF, Appel LJ, Lien LF,
2013:139239. https://doi.org/10.1155/ Batch B, Newgard CB, Svetkey LP (2012)
2013/139239 Branched-chain amino acid levels are asso-
61. Sun K, Kusminski CM, Scherer PE (2011) ciated with improvement in insulin resistance
Adipose tissue remodeling and obesity. J Clin with weight loss. Diabetologia 55
Invest 121(6):2094–2101. https://doi.org/ (2):321–330. https://doi.org/10.1007/
10.1172/JCI45887 s00125-011-2356-5
62. La Merrill M, Emond C, Kim MJ, Antignac 69. Sjostrom L (2013) Review of the key results
JP, Le Bizec B, Clement K, Birnbaum LS, from the Swedish Obese Subjects (SOS) trial –
Barouki R (2013) Toxicological function of a prospective controlled intervention study of
adipose tissue: focus on persistent organic bariatric surgery. J Intern Med 273
pollutants. Environ Health Perspect 121 (3):219–234. https://doi.org/10.1111/
(2):162–169. https://doi.org/10.1289/ joim.12012
ehp.1205485 70. Buchwald H, Avidor Y, Braunwald E, Jensen
63. Frayn KN, Karpe F, Fielding BA, Macdonald MD, Pories W, Fahrbach K, Schoelles K
IA, Coppack SW (2003) Integrative physiol- (2004) Bariatric surgery: a systematic review
ogy of human adipose tissue. Int J Obes Relat and meta-analysis. JAMA 292
Metab Disord 27(8):875–888. https://doi. (14):1724–1737. https://doi.org/10.1001/
org/10.1038/sj.ijo.0802326 jama.292.14.1724
64. Hanzu FA, Vinaixa M, Papageorgiou A, 71. Liu SY, Wong SK, Lam CC, Yung MY, Kong
Parrizas M, Correig X, Delgado S, AP, Ng EK (2015) Long-term results on
Carmona F, Samino S, Vidal J, Gomis R weight loss and diabetes remission after
(2014) Obesity rather than regional fat laparoscopic sleeve gastrectomy for a mor-
depots marks the metabolomic pattern of adi- bidly obese Chinese population. Obes Surg
pose tissue: an untargeted metabolomic 25(10):1901–1908. https://doi.org/10.
approach. Obesity (Silver Spring) 22 1007/s11695-015-1628-4
(3):698–704. https://doi.org/10.1002/oby. 72. Parikh M, Pomp A, Gagner M (2007)
20541 Laparoscopic conversion of failed gastric
65. Cao H, Gerhold K, Mayers JR, Wiest MM, bypass to duodenal switch: technical consid-
Watkins SM, Hotamisligil GS (2008) Identifi- erations and preliminary outcomes. Surg
cation of a lipokine, a lipid hormone linking Obes Relat Dis 3(6):611–618. doi:S1550-
adipose tissue to systemic metabolism. Cell 7289(07)00569-2 [pii]. https://doi.org/10.
134(6):933–944. https://doi.org/10.1016/ 1016/j.soard.2007.07.010
j.cell.2008.07.048 73. Mutch DM, Fuhrmann JC, Rein D, Wiemer
66. Liesenfeld DB, Grapov D, Fahrmann JF, JC, Bouillot JL, Poitou C, Clement K (2009)
Salou M, Scherer D, Toth R, Habermann N, Metabolite profiling identifies candidate mar-
Bohm J, Schrotz-King P, Gigic B, kers reflecting the clinical adaptations asso-
Schneider M, Ulrich A, Herpel E, ciated with Roux-en-Y gastric bypass surgery.
Schirmacher P, Fiehn O, Lampe JW, Ulrich PLoS One 4(11):e7905. https://doi.org/10.
CM (2015) Metabolomics and transcrip- 1371/journal.pone.0007905
tomics identify pathway differences between 74. Laferrere B, Reilly D, Arias S, Swerdlow N,
visceral and subcutaneous adipose tissue in Gorroochurn P, Bawa B, Bose M, Teixeira J,
colorectal cancer patients: the ColoCare Stevens RD, Wenner BR, Bain JR, Muehl-
 Metabolomics in Biomedical and Clinical Research 25

bauer MJ, Haqq A, Lien L, Shah SH, Svetkey 2 diabetes. N Engl J Med 359
LP, Newgard CB (2011) Differential meta- (21):2208–2219. https://doi.org/10.1056/
bolic impact of gastric bypass surgery versus NEJMoa0804742
dietary intervention in obese diabetic subjects 82. Kussmann M, Morine MJ, Hager J,
despite identical weight loss. Sci Transl Med 3 Sonderegger B, Kaput J (2013) Perspective:
(80):80re82. https://doi.org/10.1126/ a systems approach to diabetes research. Front
scitranslmed.3002043 Genet 4:205. https://doi.org/10.3389/
75. Whiting DR, Guariguata L, Weil C, Shaw J fgene.2013.00205
(2011) IDF diabetes atlas: global estimates of 83. Klein MS, Shearer J (2016) Metabolomics
the prevalence of diabetes for 2011 and 2030. and type 2 diabetes: translating basic research
Diabetes Res Clin Pract 94(3):311–321. into clinical application. J Diabetes Res
https://doi.org/10.1016/j.diabres.2011.10. 2016:3898502. https://doi.org/10.1155/
029 2016/3898502
76. Huber CA, Schwenkglenks M, Rapold R, 84. Wang-Sattler R, Yu Z, Herder C, Messias AC,
Reich O (2014) Epidemiology and costs of Floegel A, He Y, Heim K, Campillos M,
diabetes mellitus in Switzerland: an analysis Holzapfel C, Thorand B, Grallert H, Xu T,
of health care claims data, 2006 and 2011. Bader E, Huth C, Mittelstrass K, Doring A,
BMC Endocr Disord 14:44. https://doi. Meisinger C, Gieger C, Prehn C, Roemisch-
org/10.1186/1472-6823-14-44 Margl W, Carstensen M, Xie L, Yamanaka-
77. Schmid R, Vollenweider P, Waeber G, Okumura H, Xing G, Ceglarek U, Thiery J,
Marques-Vidal P (2011) Estimating the risk Giani G, Lickert H, Lin X, Li Y, Boeing H,
of developing type 2 diabetes: a comparison of Joost HG, de Angelis MH, Rathmann W,
several risk scores: the Cohorte Lausannoise Suhre K, Prokisch H, Peters A, Meitinger T,
study. Diabetes Care 34(8):1863–1868. Roden M, Wichmann HE, Pischon T,
https://doi.org/10.2337/dc11-0206 Adamski J, Illig T (2012) Novel biomarkers
78. Marques-Vidal P, Schmid R, Bochud M, for pre-diabetes identified by metabolomics.
Bastardot F, von Kanel R, Paccaud F, Mol Syst Biol 8:615. https://doi.org/10.
Glaus J, Preisig M, Waeber G, Vollenweider 1038/msb.2012.43
P (2012) Adipocytokines, hepatic and inflam- 85. Padberg I, Peter E, Gonzalez-Maldonado S,
matory biomarkers and incidence of type Witt H, Mueller M, Weis T, Bethan B,
2 diabetes. the CoLaus study. PLoS One 7 Liebenberg V, Wiemer J, Katus HA, Rein D,
(12):e51768. https://doi.org/10.1371/jour Schatz P (2014) A new metabolomic signa-
nal.pone.0051768 ture in type-2 diabetes mellitus and its patho-
79. Vaxillaire M, Yengo L, Lobbens S, physiology. PLoS One 9(1):e85082. https://
Rocheleau G, Eury E, Lantieri O, Marre M, doi.org/10.1371/journal.pone.0085082
Balkau B, Bonnefond A, Froguel P (2014) 86. Wang TJ, Larson MG, Vasan RS, Cheng S,
Type 2 diabetes-related genetic risk scores Rhee EP, McCabe E, Lewis GD, Fox CS,
associated with variations in fasting plasma Jacques PF, Fernandez C, O’Donnell CJ,
glucose and development of impaired glucose Carr SA, Mootha VK, Florez JC, Souza A,
homeostasis in the prospective DESIR study. Melander O, Clish CB, Gerszten RE (2011)
Diabetologia 57(8):1601–1610. https://doi. Metabolite profiles and the risk of developing
org/10.1007/s00125-014-3277-x diabetes. Nat Med 17(4):448–453. https://
80. Lin X, Song K, Lim N, Yuan X, Johnson T, doi.org/10.1038/nm.2307
Abderrahmani A, Vollenweider P, 87. Newgard CB (2012) Interplay between lipids
Stirnadel H, Sundseth SS, Lai E, Burns DK, and branched-chain amino acids in develop-
Middleton LT, Roses AD, Matthews PM, ment of insulin resistance. Cell Metab 15
Waeber G, Cardon L, Waterworth DM, Moo- (5):606–614. https://doi.org/10.1016/j.
ser V (2009) Risk prediction of prevalent dia- cmet.2012.01.024
betes in a Swiss population using a weighted 88. Herman MA, She P, Peroni OD, Lynch CJ,
genetic score – the CoLaus Study. Diabetolo- Kahn BB (2010) Adipose tissue branched
gia 52(4):600–608. https://doi.org/10. chain amino acid (BCAA) metabolism modu-
1007/s00125-008-1254-y lates circulating BCAA levels. J Biol Chem
81. Meigs JB, Shrader P, Sullivan LM, McAteer 285(15):11348–11356. https://doi.org/10.
JB, Fox CS, Dupuis J, Manning AK, Florez 1074/jbc.M109.075184
JC, Wilson PW, D’Agostino RB Sr, Cupples 89. Sears DD, Hsiao G, Hsiao A, Yu JG, Court-
LA (2008) Genotype score in addition to ney CH, Ofrecio JM, Chapman J, Subrama-
common risk factors for prediction of type niam S (2009) Mechanisms of human insulin
26 Julijana Ivanisevic and Aurelien Thomas

resistance and thiazolidinedione-mediated 97. Kapogiannis D, Mattson MP (2011) Dis-

insulin sensitization. Proc Natl Acad Sci U S rupted energy metabolism and neuronal cir-
A 106(44):18745–18750. https://doi.org/ cuit dysfunction in cognitive impairment and
10.1073/pnas.0903032106 Alzheimer’s disease. Lancet Neurol 10
90. Kaeberlein M, Rabinovitch PS, Martin GM (2):187–198. https://doi.org/10.1016/
(2015) Healthy aging: the ultimate preventa- s1474-4422(10)70277-5
tive medicine. Science 350 98. Trushina E, Nemutlu E, Zhang S,
(6265):1191–1193. https://doi.org/10. Christensen T, Camp J, Mesa J, Siddiqui A,
1126/science.aad3267 Tamura Y, Sesaki H, Wengenack TM, Dzeja
91. Jove M, Portero-Otin M, Naudi A, Ferrer I, PP, Poduslo JF (2012) Defects in mitochon-
Pamplona R (2014) Metabolomics of human drial dynamics and metabolomic signatures of
brain aging and age-related neurodegenera- evolving energetic stress in mouse models of
tive diseases. J Neuropathol Exp Neurol 73 familial Alzheimer’s disease. PLoS One 7(2):
(7):640–657. https://doi.org/10.1097/ e32737. https://doi.org/10.1371/journal.
nen.0000000000000091 pone.0032737
92. Magistretti PJ, Allaman I (2015) A cellular 99. Han X, Rozen S, Boyle SH, Hellegers C,
perspective on brain energy metabolism and Cheng H, Burke JR, Welsh-Bohmer KA, Dor-
functional imaging. Neuron 86(4):883–901. aiswamy PM, Kaddurah-Daouk R (2011)
https://doi.org/10.1016/j.neuron.2015. Metabolomics in early Alzheimer’s disease:
03.035 identification of altered plasma sphingolipi-
93. Trushina E, Dutta T, Persson XM, Mielke dome using shotgun lipidomics. PLoS One 6
MM, Petersen RC (2013) Identification of (7):e21643. https://doi.org/10.1371/jour
altered metabolic pathways in plasma and nal.pone.0021643
CSF in mild cognitive impairment and Alzhei- 100. Mapstone M, Cheema AK, Fiandaca MS,
mer’s disease using metabolomics. PLoS One Zhong X, Mhyre TR, MacArthur LH, Hall
8(5):e63644. https://doi.org/10.1371/jour WJ, Fisher SG, Peterson DR, Haley JM,
nal.pone.0063644 Nazar MD, Rich SA, Berlau DJ, Peltz CB,
94. Toledo JB, Arnold M, Kastenm€ uller G, Tan MT, Kawas CH, Federoff HJ (2014)
Chang R, Baillie RA, Han X, Plasma phospholipids identify antecedent
Thambisetty M, Tenenbaum JD, Suhre K, memory impairment in older adults. Nat
Thompson JW, John-Williams LS, Med 20(4):415–418. https://doi.org/10.
MahmoudianDehkordi S, Rotroff DM, Jack 1038/nm.3466
JR, Motsinger-Reif A, Risacher SL, Blach C, 101. Kang J, Lu J, Zhang X (2015) Metabolomics-
Lucas JE, Massaro T, Louie G, Zhu H, based promising candidate biomarkers and
Dallmann G, Klavins K, Koal T, Kim S, pathways in Alzheimer’s disease. Pharmazie
Nho K, Shen L, Casanova R, Varma S, 70(5):277–282
Legido-Quigley C, Moseley MA, Zhu K, 102. Ansoleaga B, Jove M, Schluter A, Garcia-
Henrion MYR, van der Lee SJ, Harms AC, Esparcia P, Moreno J, Pujol A, Pamplona R,
Demirkan A, Hankemeier T, van Duijn CM, Portero-Otin M, Ferrer I (2015) Deregula-
Trojanowski JQ, Shaw LM, Saykin AJ, Weiner tion of purine metabolism in Alzheimer’s dis-
MW, Doraiswamy PM, Kaddurah-Daouk R ease. Neurobiol Aging 36(1):68–80. https://
(2017) Metabolic network failures in Alzhei- doi.org/10.1016/j.neurobiolaging.2014.
mer’s disease – a biochemical road map. Alz- 08.004
heimers Dement 13(9):965–984. https:// 103. Mattson MP (1998) Modification of ion
doi.org/10.1016/j.jalz.2017.01.020 homeostasis by lipid peroxidation: roles in
95. Mink JW, Blumenschine RJ, Adams DB neuronal degeneration and adaptive plasticity.
(1981) Ratio of central nervous system to Trends Neurosci 21(2):53–57
body metabolism in vertebrates: its constancy 104. Mattson MP, Gleichmann M, Cheng A
and functional basis. Am J Physiol 241(3): (2008) Mitochondria in neuroplasticity and
R203–R212 neurological disorders. Neuron 60
96. Paglia G, Stocchero M, Cacciatore S, Lai S, (5):748–766. https://doi.org/10.1016/j.
Angel P, Alam MT, Keller M, Ralser M, Astar- neuron.2008.10.010
ita G (2016) Unbiased metabolomic investi- 105. Payne BAI, Chinnery PF (2015) Mitochon-
gation of Alzheimer’s disease brain points to drial dysfunction in aging: Much progress but
dysregulation of mitochondrial aspartate many unresolved questions. Biochimica et
metabolism. J Proteome Res 15 Biophysica Acta 1847(11):1347–1353
(2):608–618. https://doi.org/10.1021/acs. 106. Mouchiroud L, Houtkooper RH, Moullan N,
jproteome.5b01020 Katsyuba E, Ryu D, Cantó C, Mottis A, Jo
 Metabolomics in Biomedical and Clinical Research 27

Y-S, Viswanathan M, Schoonjans K, 116. Shin SY, Fauman EB, Petersen AK,
Guarente L, Auwerx J (2013) The NAD(+)/ Krumsiek J, Santos R, Huang J, Arnold M,
sirtuin pathway modulates longevity through Erte I, Forgetta V, Yang TP, Walter K,
activation of mitochondrial UPR and FOXO Menni C, Chen L, Vasquez L, Valdes AM,
signaling. Cell 154(2):430–441. https://doi. Hyde CL, Wang V, Ziemek D, Roberts P,
org/10.1016/j.cell.2013.06.016 Xi L, Grundberg E, Multiple Tissue Human
107. Ibanez C, Simo C, Martin-Alvarez PJ, Expression Resource Consortium,
Kivipelto M, Winblad B, Cedazo-Minguez A, Waldenberger M, Richards JB, Mohney RP,
Cifuentes A (2012) Toward a predictive Milburn MV, John SL, Trimmer J, Theis FJ,
model of Alzheimer’s disease progression Overington JP, Suhre K, Brosnan MJ,
using capillary electrophoresis-mass spec- Gieger C, Kastenmuller G, Spector TD, Sor-
trometry metabolomics. Anal Chem 84 anzo N (2014) An atlas of genetic influences
(20):8532–8540. https://doi.org/10.1021/ on human blood metabolites. Nat Genet 46
ac301243k (6):543–550. https://doi.org/10.1038/ng.
108. Graham SF, Chevallier OP, Roberts D, 2982
Holscher C, Elliott CT, Green BD (2013) 117. Draisma HH, Pool R, Kobl M, Jansen R,
Investigation of the human brain metabo- Petersen AK, Vaarhorst AA, Yet I, Haller T,
lome to identify potential markers for early Demirkan A, Esko T, Zhu G, Bohringer S,
diagnosis and therapeutic targets of Alzhei- Beekman M, van Klinken JB, Romisch-
mer’s disease. Anal Chem 85(3):1803–1811. Margl W, Prehn C, Adamski J, de Craen AJ,
https://doi.org/10.1021/ac303163f van Leeuwen EM, Amin N, Dharuri H, Wes-
109. Wishart DS (2015) Is cancer a genetic disease tra HJ, Franke L, de Geus EJ, Hottenga JJ,
or a metabolic disease? EBioMedicine 2 Willemsen G, Henders AK, Montgomery
(6):478–479. https://doi.org/10.1016/j. GW, Nyholt DR, Whitfield JB, Penninx BW,
ebiom.2015.05.022 Spector TD, Metspalu A, Eline Slagboom P,
van Dijk KW, t Hoen PA, Strauch K, Martin
110. Hanahan D, Weinberg RA Hallmarks of can- NG, van Ommen GJ, Illig T, Bell JT,
cer: the next generation. Cell 144 Mangino M, Suhre K, McCarthy MI,
(5):646–674. https://doi.org/10.1016/j. Gieger C, Isaacs A, van Duijn CM, Boomsma
cell.2011.02.013 DI (2015) Genome-wide association study
111. Otto AM (2016) Warburg effect(s)—a bio- identifies novel genetic variants contributing
graphical sketch of Otto Warburg and his to variation in blood metabolite levels. Nat
impacts on tumor metabolism. Cancer Commun 6:7208. https://doi.org/10.
Metab 4:5. https://doi.org/10.1186/ 1038/ncomms8208
s40170-016-0145-9 118. Suhre K, Shin SY, Petersen AK, Mohney RP,
112. Yang M, Soga T, Pollard PJ (2013) Oncome- Meredith D, Wagele B, Altmaier E, CARDIo-
tabolites: linking altered metabolism with GRAM, Deloukas P, Erdmann J,
cancer. J Clin Invest 123(9):3652–3658. Grundberg E, Hammond CJ, de Angelis
https://doi.org/10.1172/JCI67228 MH, Kastenmuller G, Kottgen A,
113. Morin A, Letouze E, Gimenez-Roqueplo AP, Kronenberg F, Mangino M, Meisinger C,
Favier J (2014) Oncometabolites-driven Meitinger T, Mewes HW, Milburn MV,
tumorigenesis: From genetics to targeted Prehn C, Raffler J, Ried JS, Romisch-Margl W,
therapy. Int J Cancer 135(10):2237–2248. Samani NJ, Small KS, Wichmann HE, Zhai G,
https://doi.org/10.1002/ijc.29080 Illig T, Spector TD, Adamski J, Soranzo N,
114. Wang X, Yang K, Xie Q, Wu Q, Mack SC, Gieger C (2011) Human metabolic individu-
Shi Y, Kim LJ, Prager BC, Flavahan WA, ality in biomedical and pharmaceutical
Liu X, Singer M, Hubert CG, Miller TE, research. Nature 477(7362):54–60. https://
Zhou W, Huang Z, Fang X, Regev A, Suva doi.org/10.1038/nature10354
ML, Hwang TH, Locasale JW, Bao S, Rich JN 119. Kaddurah-Daouk R, Weinshilboum R, Phar-
(2017) Purine synthesis promotes mainte- macometabolomics Research Network
nance of brain tumor initiating cells in glioma. (2015) Metabolomic signatures for drug
Nat Neurosci 20(5):661–673. https://doi. response phenotypes: pharmacometabolo-
org/10.1038/nn.4537 mics enables precision medicine. Clin Phar-
115. Chang CH, Pearce EL (2016) Emerging con- macol Ther 98(1):71–75. https://doi.org/
cepts of T cell metabolism as a target of 10.1002/cpt.134
immunotherapy. Nat Immunol 17 120. Everett JR (2016) From metabonomics to
(4):364–368. https://doi.org/10.1038/ni. pharmacometabonomics: the role of meta-
3415 bolic profiling in personalized medicine.
28 Julijana Ivanisevic and Aurelien Thomas

Front Pharmacol 7:297. https://doi.org/10. 129. Dumond JB, Vourvahis M, Rezk NL, Patter-
3389/fphar.2016.00297 son KB, Tien HC, White N, Jennings SH,
121. Lewis JP, Yerges-Armstrong LM, Ellero- Choi SO, Li J, Wagner MJ, La-Beck NM,
Simatos S, Georgiades A, Kaddurah-Daouk R, Drulak M, Sabo JP, Castles MA, Macgregor
Hankemeier T (2013) Integration of pharma- TR, Kashuba AD (2010) A phenotype-
cometabolomic and pharmacogenomic genotype approach to predicting CYP450
approaches reveals novel insights into antipla- and P-glycoprotein drug interactions with
telet therapy. Clin Pharmacol Ther 94 the mixed inhibitor/inducer tipranavir/rito-
(5):570–573. https://doi.org/10.1038/ navir. Clin Pharmacol Ther 87(6):735–742.
clpt.2013.153 https://doi.org/10.1038/clpt.2009.253
122. Neavin D, Kaddurah-Daouk R, Weinshil- 130. Daali Y, Samer C, Deglon J, Thomas A,
boum R (2016) Pharmacometabolomics Chabert J, Rebsamen M, Staub C, Dayer P,
informs Pharmacogenomics. Metabolomics Desmeules J (2012) Oral flurbiprofen meta-
12(7). https://doi.org/10.1007/s11306- bolic ratio assessment using a single-point
016-1066-x dried blood spot. Clin Pharmacol Ther.
123. Nicholson JK, Connelly J, Lindon JC, https://doi.org/10.1038/clpt.2011.247
Holmes E (2002) Metabonomics: a platform 131. Bosilkovska M, Samer C, Deglon J,
for studying drug toxicity and gene function. Thomas A, Walder B, Desmeules J, Daali Y
Nat Rev Drug Discov 1(2):153–161. https:// (2016) Evaluation of mutual drug-drug inter-
doi.org/10.1038/nrd728 action within Geneva cocktail for cytochrome
124. Spear BB, Heath-Chiozzi M, Huff J (2001) P450 phenotyping using innovative dried
Clinical application of pharmacogenetics. blood sampling method. Basic Clin Pharma-
Trends Mol Med 7(5):201–204 col Toxicol 119(3):284–290. https://doi.
org/10.1111/bcpt.12586
125. Bosilkovska M, Samer CF, Deglon J,
Rebsamen M, Staub C, Dayer P, Walder B, 132. Krumsiek J, Suhre K, Evans AM, Mitchell
Desmeules JA, Daali Y (2014) Geneva cock- MW, Mohney RP, Milburn MV, Wagele B,
tail for cytochrome p450 and P-glycoprotein Romisch-Margl W, Illig T, Adamski J,
activity assessment using dried blood spots. Gieger C, Theis FJ, Kastenmuller G (2012)
Clin Pharmacol Ther 96(3):349–359. Mining the unknown: a systems approach to
https://doi.org/10.1038/clpt.2014.83 metabolite identification combining genetic
and metabolic information. PLoS Genet 8
126. Konig J, Muller F, Fromm MF (2013) Trans- (10):e1003005. https://doi.org/10.1371/
porters and drug-drug interactions: impor- journal.pgen.1003005
tant determinants of drug disposition and
effects. Pharmacol Rev 65(3):944–966. 133. Tay-Sontheimer J, Shireman LM, Beyer RP,
https://doi.org/10.1124/pr.113.007518 Senn T, Witten D, Pearce RE, Gaedigk A,
Gana Fomban CL, Lutz JD, Isoherranen N,
127. Kohler GI, Bode-Boger SM, Busse R, Thummel KE, Fiehn O, Leeder JS, Lin YS
Hoopmann M, Welte T, Boger RH (2000) (2014) Detection of an endogenous urinary
Drug-drug interactions in medical patients: biomarker associated with CYP2D6 activity
effects of in-hospital treatment and relation using global metabolomics. Pharmacoge-
to multiple drug use. Int J Clin Pharmacol nomics 15(16):1947–1962. https://doi.
Ther 38(11):504–513 org/10.2217/pgs.14.155
128. Chainuvati S, Nafziger AN, Leeder JS, 134. Waters MD, Fostel JM (2004) Toxicoge-
Gaedigk A, Kearns GL, Sellers E, Zhang Y, nomics and systems toxicology: aims and pro-
Kashuba AD, Rowland E, Bertino JS Jr spects. Nat Rev Genet 5(12):936–948.
(2003) Combined phenotypic assessment of https://doi.org/10.1038/nrg1493
cytochrome p450 1A2, 2C9, 2C19, 2D6, and
3A, N-acetyltransferase-2, and xanthine oxi- 135. Bersanelli M, Mosca E, Remondini D,
dase activities with the "Cooperstown 5+1 Giampieri E, Sala C, Castellani G, Milanesi L
cocktail". Clin Pharmacol Ther 74 (2016) Methods for the integration of multi-
(5):437–447. https://doi.org/10.1016/ omics data: mathematical aspects. BMC Bio-
S0009-9236(03)00229-7 informatics 17(Suppl 2):15. https://doi.org/
10.1186/s12859-015-0857-9
 Chapter 2

Metabolomics in Immunology Research

Bart Everts

Abstract
There is a growing appreciation that metabolic processes and individual metabolites can shape the function
of immune cells and thereby play important roles in the outcome of immune responses. In this respect, the
use of MS- and NMR spectroscopy-based platforms to characterize and quantify metabolites in biological
samples has recently yielded important novel insights into how our immune system functions and has
contributed to the identification of biomarkers for immune-mediated diseases. Here, these recent immu-
nological studies in which metabolomics has been used and made significant contributions to these fields
will be discussed. In particular the role of metabolomics to the rapidly advancing field of cellular immuno-
metabolism will be highlighted as well as the future prospects of such metabolomic tools in immunology.

Key words Mass spectrometry, NMR, Metabolite tracing, Immune cells, Immunometabolism,
Cellular metabolism

1 Introduction

Over the past three decades, it has become increasingly clear that
our immune system play a key role not only in the protection
against invading pathogens but also in many physiological pro-
cesses to maintain whole-body homeostasis as well as in the
aetiology of a large number of noninfectious diseases. Therefore,
there have been tremendous efforts to understand in great detail
how our immune system functions, to be able to identify targets
and design approaches to manipulate immune cells or immunolog-
ical pathways for therapeutic purposes. In addition, there is a major
interest in developing approaches to reliably track and predict
certain immune response parameters that could be used as diagnos-
tic and/or prognostic biomarkers for susceptibility, progression,
and outcome of immune-mediated diseases.
In recent years, the development and use of a range of analytical
platforms, including gas (GC) or liquid chromatography (LC)
coupled to mass spectrometry (MS) or nuclear magnetic resonance
(NMR) spectroscopy, to detect, characterize, and quantify intra- or

Martin Giera (ed.), Clinical Metabolomics: Methods and Protocols, Methods in Molecular Biology, vol. 1730,
https://doi.org/10.1007/978-1-4939-7592-1_2, © Springer Science+Business Media, LLC 2018

29
30 Bart Everts

extracellular metabolites and related metabolic pathways, have

made important contributions to our understanding of how our
immune system functions as well as to the identification of biomar-
kers of immunological diseases. I will here highlight some of the
recent immunological studies in which these metabolomic
approaches have provided new insights in these fields, with particu-
lar emphasis on the field of cellular immunometabolism. In addi-
tion, the pros and cons as well as the future prospects of such
metabolomic tools in immunology research will be discussed.

2 Metabolics as a Tool to Study Cellular Immunometabolism

It has long been appreciated, especially in the cancer field, that

changes in cellular activation and proliferation coincide with, and
are underpinned by, alterations in cellular metabolic state
[1, 2]. During the last couple of years, immunologists have come
to realize that also immune cell activation, proliferation, fate, and
function are closely linked to and dependent on activation of spe-
cific intracellular metabolic pathways. This has resulted in the
emergence of the currently burgeoning field of immunometabo-
lism [3]. Thus far, most studies have focused on addressing the role
of core metabolic pathways like glycolysis, lipid metabolism, mito-
chondrial fatty acid oxidation (FAO), and oxidative phosphoryla-
tion (OXPHOS) in the regulation of immune cell biology.
However, apart from the classical role of metabolism in bioenerget-
ics and biosynthesis, recent studies have also highlighted that meta-
bolites or enzymes within these pathways can additionally act as
signaling molecules or have metabolism-independent “moonlight-
ing” functions, respectively. This further illustrates the intricate and
multilayered regulation of immune cell biology by cellular meta-
bolic pathways. Importantly, several seminal studies with T cells and
macrophages have shown that targeting these metabolic pathways
can greatly influence their functional properties. This has led to the
emerging concept that manipulation of immune cell metabolism
could be a powerful approach to direct immune responses [4].
Several methods are available to assess cellular metabolism of
immune cells and range from fluorescently labeled mitochondrial
dyes or glucose and fatty acids that can be analyzed in individual
cells by flow cytometry to tracing of uptake and of enzymatic
conversion of radioactively labeled metabolites by scintillation
counting and to real-time analysis of oxygen consumption and
extracellular acidification as measures of mitochondrial OXPHOS
and glycolysis, respectively, using the Seahorse XF flux analyzer
[5, 3]. However, a major limitation of all of these techniques is
that they are targeted approaches that enable one to assess only the
activity of predefined metabolic pathways, whereas in many cases
unbiased metabolic analyses are needed to identify novel metabolic
 Immunometabolomics 31

signatures and study their role in immune cell biology. This is where
untargeted metabolomics has proven to be a valuable and essential
tool in the field of cellular immunometabolism. In the next sec-
tions, the recent studies will be discussed in which metabolomic
platforms have provided novel insights in the metabolic character-
istics and regulation of different types of immune cells. These
studies illustrate that metabolomics has become an important and
integral part of immune cell metabolism research.

2.1 Metabolomics The role of metabolism in immune cell function has been most
as a Tool to Study well-characterized in macrophages and T cells. Already more than a
Macrophage decade ago, it was observed that pro-inflammatory classically acti-
Metabolism vated (“M1”) macrophages and anti-inflammatory alternatively
activated (“M2”) macrophages have very distinct metabolic proper-
ties on which they depend for their polarization [6]. M1 macro-
phages are characterized by enhanced aerobic glycolysis and
reduced mitochondrial OXPHOS. The use of GC/MS-based
untargeted metabolic profiling has been instrumental in providing
new insights into how M1 macrophage metabolism shapes their
biology. MS-based metabolomics revealed a dramatic buildup of
both succinate, a Krebs cycle intermediate, and itaconate, a product
of citrate, in M1-polarized macrophages [7–9]. This led to the
characterization of succinate as being crucial for IL-1β expression
by macrophages [7], a process that was later identified to be antag-
onized by itaconate [9]. Importantly, inhibition of succinate or
itaconate synthesis reduced or enhanced the pro-inflammatory
nature of the targeted macrophages, respectively, and thereby out-
come of models of macrophage-driven diseases in vivo [7–9]. In
addition, targeted MS-based platforms have been used to specifi-
cally study the lipid profile of macrophages in response to
M1-polarizing stimuli. This revealed immediate responses in fatty
acid metabolism represented by increases in eicosanoid synthesis
and delayed responses characterized by sphingolipid and sterol
biosynthesis [10, 11].
In contrast to M1 macrophages, M2 macrophages are meta-
bolically characterized by enhanced mitochondrial OXPHOS
fueled by FAO. Using GC/MS-based global metabolic profiling,
it was shown that M2 macrophages have increased intracellular
levels of various monoacylglycerides, which was suggestive of
enhanced lipolysis in these cells [12]. This observation resulted in
the identification of a cell-intrinsic lysosomal lipolysis pathway that
is essential for alternative activation of macrophages presumably by
means of generating free fatty acids (FAs) for FAO [12]. Another
example of how MS-based metabolomics has contributed to our
understanding of how M2 polarization and cellular metabolism are
linked comes from a study in which an integrated network analysis
was performed of both transcriptomic and metabolomic data
[13]. This integrated approach resulted in the identification of
32 Bart Everts

M2-specific metabolic nodes/pathways that only based on tran-

scriptomic or metabolomic data analysis would have been missed.
Specifically, M2 polarization was found to activate glutamine catab-
olism and UDP-GlcNAc-associated modules. Correspondingly,
glutamine deprivation or inhibition of N-glycosylation decreased
M2 polarization [13].

2.2 Metabolomics The metabolic requirements for T-cell activation, proliferation,

as a Tool to Study effector cytokine production, and memory formation have been
T-Cell Metabolism extensively studied in recent years. In particular the interplay
between metabolism and these processes has been well-
characterized in CD8+ T cells [14]. One of the early studies linking
T-cell activation and proliferation to metabolic reprogramming
used MS-based metabolomics to characterize the metabolic
changes that occur in CD8+ T cells upon activation [15]. Preceding
cellular division, sharp increases in intracellular levels of many
metabolites were observed, such as amino acids, lipids, and nucleo-
tides, indicative of increased biosynthesis required for cell growth
and proliferation. This was dependent on transcription factor
c-Myc. Interestingly, using MS-based untargeted metabolomics,
upon the first cell division of CD8+ T cells, c-Myc, mTORC1
(a central regulator of anabolic metabolism), and several amino
acids were found to be unequally split between daughter cells.
This asymmetric division in metabolic regulators and metabolites
resulted in stronger glutaminolysis and glycolysis in the cells con-
taining higher levels of c-Myc and mTORC1 and subsequently in
superior proliferative capacity and effector function compared to
c-Myclow daughter cells [16]. In further support for an important
role for the need of c-Myc in effector T-cell function, another study
demonstrated, through the use of LC/MS-based targeted quanti-
fication of UDP-GlcNAc, that c-Myc regulates intracellular protein
O-GlcNAcylation at key stages of T-cell development and differen-
tiation [17]. As a consequence, loss of O-GlcNAc transferase
blocked T-cell progenitor renewal and T-cell clonal expansion.
The formation of memory cells requires distinct sets of meta-
bolic processes and is known to be favored by mitochondrial-
centered catabolic metabolism in which FAO plays an important
role [14]. Also here, metabolomics has provided important new
insights into the metabolic processes underpinning CD8+ T-cell
memory formation. For instance, targeted lipidomics has revealed
that analogous to M2 macrophages, memory CD8+ T cells use cell-
intrinsic lipolysis to support the metabolic programming necessary
for development [18]. Moreover, autophagy-deficient CD8+ T cells
(atg7
/
) failed to efficiently develop into memory T cells. This
was found to be the result of defects in FAO in atg7
/
cells as
determined by LC/MS metabolomics [19].
The metabolic programs required for CD4+ T-cell effector
function and memory formation are less well defined. However,
 Immunometabolomics 33

untargeted metabolome analysis revealed that also in early activated

CD4+ T cells, intracellular pools of metabolites, in particular lipids
and FAs, are elevated, presumably acting as a pool to generate cell
membranes that allows for cellular proliferation [20]. Consistent
with a role for mTORC1 in CD8+ T-cell activation, this increase in
lipids was found to be dependent on mTORC1 activity [20]. CD4+
T cells can differentiate into different T-helper (Th) cell subsets
including T follicular helper (Tfh) cells, important for aiding B cells
in producing antibodies, and regulatory T cells (Tregs) that are
known for their ability to suppress immune responses. Pathway
analyses of altered metabolites based on LC/MS-based untargeted
metabolomics in Tfh cells showed that amino acid metabolism,
amino sugar metabolism, and glycolytic intermediates trended
higher compared to Th1 or Th17 cells, while Tfh cells had lower
amount of TCA metabolites and intermediates in cysteine and
glycerophospholipid metabolism, suggesting that different Th cell
subsets have unique metabolic profiles [21]. Early studies have
suggested that Tregs are mostly oxidative and rely on FAO for
their differentiation and function [22]. More recently, it was
found that Tregs are in fact metabolically flexible and engage
glycolysis and anabolic metabolism in order to proliferate but
dampen these metabolic pathways to acquire an immunosuppres-
sive phenotype [23]. Mechanistically, MS-based metabolome anal-
ysis revealed that this is a direct result of the ability of Foxp3, the
master regulator of Treg differentiation, to suppress expression of
genes involved in glycolysis and anabolic metabolism [23]. Finally,
by using LC/MS, a recent study found that in activated CD4+ T
cells, arginine is selectively depleted [24]. Restoring arginine levels
in these cells promoted their ability to form memory cells, which
was associated with enhanced mitochondrial metabolism. This
might suggest that analogous to CD8+ T cells, memory formation
by CD4+ T cells is promoted by increased oxidative metabolism.

2.3 Metabolomics Thus far there have only been a handful of studies that have started
as a Tool to Study to explore the metabolic properties through the use of metabolo-
Metabolism of Other mics of other immune cells and include innate lymphoid cells
Immune Cells (ILCs), dendritic cells (DCs), and monocytes. For instance, type
2 ILCs (ILC2s), which contribute to the initiation and mainte-
nance of type 2 immune responses, were found to express high
levels of arginase 1 (Arg1), an enzyme that converts arginine into
ornithine [25]. Arg1 expression was shown to be essential for ILC2
function, as inhibition of Arg1 blocked ILC2 proliferation and
effector function. LC/MS-based metabolomics revealed that
Arg1 enzymatic activity was important for polyamine biosynthesis
and promoting glycolysis [25], thereby linking metabolic repro-
gramming to cellular function. Furthermore, metabolomic plat-
forms have been used to study metabolism of DCs, the cells that
are crucial for priming of T-cell responses. It was found that
34 Bart Everts

activation of pro-inflammatory DCs by toll-like receptor (TLR)

ligands resulted in rapid metabolic changes in these cells, character-
ized by increased glycolysis and TCA cycle activity as determined by
untargeted MS-based metabolic profiling [26]. The authors went
on to show that this glycolytic reprogramming was essential for
TLR-induced activation by means of supporting de novo FA syn-
thesis required for ER and Golgi membrane expansion and thereby
efficient expression of activation markers and cytokines. More
recently, it was demonstrated through the use of MS-based meta-
bolomics that also in monocytes TLR ligands induce glycolysis
[27]. Interestingly, in the same analysis, it was observed that LPS
and Pam3Cys, ligands for TLR4 and TLR2, oppositely affected
OXPHOS, suggesting that different TLR ligands can promote
distinct metabolic programs, possibly tailored to the type of inflam-
matory response these different TLR ligands induce.

2.4 Isotope Labeling Global metabolomics can quantify relative or absolute differences
as a Tool to Study in abundances of metabolites between different conditions, which
Immunometabolism may indicate altered activity of the associated enzymatic, nonenzy-
matic, or transport reactions. However, since an increase in metab-
olite concentration can be indicative of both increased activity of
metabolite-generating enzymes as well as decreased activity of
metabolite-consuming enzymes, concentration changes do not
allow one to draw conclusions on metabolic rates and direction of
fluxes. This requires dedicated tracing studies with isotope-labeled
metabolites. Radioactive isotopes such as 3H incorporated into
glucose or palmitic acid are still being used to trace uptake and
incorporation into lipids or oxidation of these metabolites in the
mitochondria of immune cells. A downside of this technique is not
only the fact that one has to work with radioactive material but also
that these radiolabels do not allow for tracing of metabolite-derived
atoms at high resolution into individual metabolites. This has
spurred the utilization of nonradioactive heavy isotope (most fre-
quently 13C or 15N)-labeled nutrients to examine intracellular
fluxes. These heavy isotopes can be followed and quantified into
individual intracellular metabolites using MS- or NMR-based tech-
niques [28]. Tracing/flux studies are a crucial tool for interpreting
differences in static metabolite profiles between different cells or
conditions that are generated by untargeted metabolomic
approaches. As such this technique has been applied in several
immunometabolic studies, such as those that have defined meta-
bolic fluxes in macrophages (tracing 13C-glucose [29] and 13C- and
15
N-glutamine [13]), T cells (tracing 13C-acetate [30], serine [31],
glutamine, or glucose [18, 32]), ILC2s (tracing 13C-arginine
[25]), and DCs (tracing 13C-glucose [26]).
 Immunometabolomics 35

3 Metabolomics as a Tool to Study Effects of Extracellular Metabolites

on Immune Cells

Apart from an important role for intracellular metabolites and

associated metabolic pathways in regulating the biology of immune
cells, it is also well appreciated that extracellular metabolites present
in the microenvironment of immune cells can greatly influence their
functional properties. In this respect extracellular metabolites can
affect immune cell function by acting as direct nutrients for these
cells or as signaling molecules. Some of these insights have been
gained from studies using metabolomics.

3.1 Metabolomics The nutrients available to immune cells in their microenvironment

as a Tool to Study are likely to dictate their functional properties. This can be particu-
Effects of Extracellular larly relevant in situations where local nutrient availability has
Nutrients on Immune become limiting. For instance, a recent study elegantly showed
Cell Function that in a tumor microenvironment, tumor-infiltrating CD8+ T
cells are impaired in their function, due to insufficient availability
of glucose that has been depleted by highly glycolytic tumor cells
[33]. Moreover, lactic acid buildup as a result of high glycolytic
rates and subsequent acidosis in tumor microenvironments has also
been described to impair the function of pro-inflammatory immune
cells [34] as well as to foster Treg activity [35]. These studies
focused only on effects of single metabolites. It is to be expected
that unbiased extracellular metabolomics will help identifying addi-
tional metabolites that are depleted or accumulating in tumor
microenvironments that may further explain the poor effector
function of many immune cells present in tumors. Also in other
conditions, such as in highly inflamed lymph nodes or poorly
vascularized sites, it is likely that local nutrient levels can signifi-
cantly drop to the extent that it could have a negative functional
impact on immune cells. Conversely, exposure of immune cells to
very nutrient- or lipid-rich environments such as those found in
adipose tissues in obese individuals or atherosclerotic plaques can
also have a profound impact on immune cell function
[36–38]. However, an in-depth understanding of this metabolic
interaction between immune cells and their environment is still
missing, and extracellular metabolome analyses will become an
important tool in the coming years to fully uncover the complex
interplay between nutrient/metabolite availability and immune cell
function.

3.2 Metabolomics The prototypical soluble mediators that mediate communication

as a Tool to Study between cells are a family of proteins called cytokines. However, in
Effects of Extracellular addition to these proteins, various metabolites can also serve as
Signaling Metabolites signaling molecules to immune cells. In this respect, the most
on Immune Cell well-known family of metabolites that by interacting with G-
Function protein-coupled receptors (GPCRs) act as signaling molecules in
36 Bart Everts

paracrine or autocrine manner are the eicosanoids which are lipids

made by the enzymatic or nonenzymatic oxidation of arachidonic
acid or other C20-polyunsaturated fatty acids. Classic examples
include the prostaglandins, thromboxanes, leukotrienes, and resol-
vins that depending on the species involved promote pro- or anti-
inflammatory responses by immune cells. Major producers of these
lipid mediators are macrophages. In several studies MS-based meta-
bolomics has proven to be instrumental in generating a compre-
hensive picture of the eicosanoid profile produced by macrophages
under different inflammatory conditions both in vitro and in vivo
(reviewed by [39]).
Apart from host cell-derived metabolites, it is becoming
increasingly clear that our microbiota, especially a variety of intesti-
nal bacteria, produce metabolites that regulate various physiologi-
cal processes including the immune system of the host. In this
respect, one of the most well-studied groups of metabolites that
acts as signaling molecules on immune cells and has been shown to
play a crucial role in the proper development of a well-balanced
immune system is the short-chain fatty acids (SCFAs). SCFAs are
the products of digested fibers by certain microbial species in the
gut and have been shown to primarily promote anti-inflammatory
responses in immune cells, such as Treg responses, and as such have
been implicated in protection against various inflammatory diseases
[40, 41]. In two seminal studies, NMR-based [42] or MS-based
[43] comparative metabolomics were used to identify butyrate as
the main SCFA that could promote Tregs, illustrating how the use
of metabolomics has contributed to this field. Moreover, several
human studies have used NMR- and MS-based metabolome ana-
lyses to link butyrate concentrations in feces and other body fluids,
with changes in microbial composition, host immune responses,
and outcome of various diseases such as atopic dermatitis [44],
cancer [45], ulcerative colitis [46], and gout [47]. Similar metabo-
lomic approaches have been used to identify and track concentra-
tions of other immune modulatory microbiota-derived metabolites
and are recently reviewed in great detail elsewhere [48, 41].
Most signaling metabolites are recognized by immune cells via
metabolite-specific GPCRs. Interestingly, many putative GPCRs
are currently still orphan receptors, indicating that there are likely
several more still unidentified host- and/or commensal microbiota-
derived metabolites that could serve as signaling mediators and
thereby potentially direct immune cell function [40]. For these
kinds of studies, unbiased comparative “exo”-metabolomics will
be an important tool to identify some of these new signaling
metabolites [48].
 Immunometabolomics 37

4 Metabolomics as a Tool for Identification of Biomarkers of Immune Responses

As most diseases are immune-driven or have at least an immuno-

logical component, there is currently great interest in identifying
reliable biomarkers that allow one to track the immune status of
diseased individuals. Such biomarkers can then be used as diagnos-
tic and/or prognostic markers for susceptibility, progression, and
outcome of immune-mediated diseases, which will provide essential
pieces of information for effective management of these diseases. In
this respect, metabolomic profiling using NMR spectroscopy of
biofluids is proven to be a highly promising approach for biomarker
identification as it is becoming clear that metabolite composition or
individual metabolites within biofluids can serve as reliable biomar-
kers for certain immune parameters in both infectious [49, 50] and
immune-mediated noninfectious diseases [51]. Moreover, a practi-
cal advantage is that it can be performed reliably on readily obtain-
able biofluids like urine, blood, and saliva, without the need for
acquisition of samples from diseased tissues that often requires
invasive techniques. Although the use of metabolomics as a tool
for biomarker identification of immune responses and immune-
mediated diseases is a recent development, it is likely to make
important contributions to this rapidly advancing field.

5 Future Perspectives

The recent appreciation that metabolism and immune responses are

tightly linked has spurred the use of MS- and NMR-based plat-
forms to quantitatively measure metabolic profiles in immunologi-
cal studies and is rapidly becoming part of the standard toolbox for
immunologists to study the immune system. Until now, most
studies in which metabolomics has been used to assess cellular
immunometabolism have employed MS-based platforms as these
currently have a higher sensitivity compared to NMR-based tech-
niques, making them better suited for studying immune cell metab-
olism of primary immune cells that are often only available in low
numbers. Conversely, NMR spectroscopy has been the primary
platform for analysis of biological samples used for biomarker iden-
tification and profiling, mostly because of the simplicity of sample
preparation that does not rely on prior separation of the analytes.
Moreover, the lower sensitivity of NMR spectrometry is less of an
issue for biofluids as these samples can often be obtained in larger
quantities. However, now with the development of NMR spectro-
meters with stronger magnets that result in higher resolution and
sensitivity of this platform, it is to be expected that this technique
will also be more implemented in cellular immunometabolomic
38 Bart Everts

studies. An important consideration is that due to the complexity of

the metabolome and the diverse properties of metabolites, no
single analytical platform can be applied to detect all metabolites
in a biological sample. Hence, integrated complementary platforms
will need to be used to reliably cover the full range of metabolites
present in biological samples [52].
With improved sensitivity of individual metabolomic platforms
as well as increasing integration of data from different metabolic
platforms, there is a strong need for well-defined computational
modeling pipelines to identify and correlate individual metabolites
or metabolic pathways to immunological parameters of interest.
Moreover, within the field of immunology as well as immunome-
tabolism, there is a growing trend to use systems biology, which
applies various unbiased “omics” approaches (such as proteomics,
transcriptomics, metabolomics), to address immunological and
immunometabolic questions. This makes the development of bio-
informatic tools that enable one to interpret as well as integrate
these kinds of complex data sets even more important. Recently, for
instance, a web-based tool for integrated analysis of steady-state
metabolomic and transcriptional data has been developed that
focuses on identification of the most changing metabolic networks
between two conditions of interest [53]. This has already proven to
be a valuable bioinformatic tool for studying macrophage metabo-
lism [13]. Bioinformatic pipelines such as this will be key to reveal
the full depth of metabolic information that is captured by these
different omics approaches. As such it is to be expected that in the
coming years, NMR- and MS-based metabolomic platforms in
conjunction with novel bioinformatic pipelines will further estab-
lish themselves as critical tools for unraveling the role of metabo-
lism in immune cell biology and identifying metabolic biomarkers
for immune responses and immune-related diseases.

Acknowledgments

This work was in part funded by an LUMC and Marie Curie

fellowship (#631585). I declare to have no competing interests.

References

1. Vander Heiden MG (2011) Targeting cancer 3. O’Neill LA, Kishton RJ, Rathmell J (2016) A
metabolism: a therapeutic window opens. Nat guide to immunometabolism for immunolo-
Rev Drug Discov 10(9):671–684. https://doi. gists. Nat Rev Immunol 16(9):553–565.
org/10.1038/nrd3504 https://doi.org/10.1038/nri.2016.70
2. Vander Heiden MG, Cantley LC, Thompson 4. O’Sullivan D, Pearce EL (2015) Targeting T
CB (2009) Understanding the Warburg effect: cell metabolism for therapy. Trends Immunol
the metabolic requirements of cell prolifera- 36(2):71–80. https://doi.org/10.1016/j.it.
tion. Science 324(5930):1029–1033. https:// 2014.12.004
doi.org/10.1126/science.1160809
 Immunometabolomics 39

5. Pelgrom LR, van der Ham AJ, Everts B (2016) to characterize lipid alterations in inflammatory
Analysis of TLR-induced metabolic changes in macrophages. J Proteome Res 16(4):
dendritic cells using the seahorse XF(e)96 1460–1469. https://doi.org/10.1021/acs.
extracellular flux analyzer. Methods Mol Biol jproteome.6b00848
1390:273–285. https://doi.org/10.1007/ 12. Huang SC, Everts B, Ivanova Y, O’Sullivan D,
978-1-4939-3335-8_17 Nascimento M, Smith AM, Beatty W, Love-
6. Van den Bossche J, O’Neill LA, Menon D Gregory L, Lam WY, O’Neill CM, Yan C,
(2017) Macrophage immunometabolism: Du H, Abumrad NA, Urban JF Jr, Artyomov
where are we (Going)? Trends Immunol. MN, Pearce EL, Pearce EJ (2014) Cell-
https://doi.org/10.1016/j.it.2017.03.001 intrinsic lysosomal lipolysis is essential for alter-
7. Tannahill GM, Curtis AM, Adamik J, Palsson- native activation of macrophages. Nat Immu-
McDermott EM, McGettrick AF, Goel G, nol 15(9):846–855. https://doi.org/10.
Frezza C, Bernard NJ, Kelly B, Foley NH, 1038/ni.2956
Zheng L, Gardet A, Tong Z, Jany SS, Corr 13. Jha AK, Huang SC, Sergushichev A,
SC, Haneklaus M, Caffrey BE, Pierce K, Lampropoulou V, Ivanova Y, Loginicheva E,
Walmsley S, Beasley FC, Cummins E, Chmielewski K, Stewart KM, Ashall J,
Nizet V, Whyte M, Taylor CT, Lin H, Masters Everts B, Pearce EJ, Driggers EM, Artyomov
SL, Gottlieb E, Kelly VP, Clish C, Auron PE, MN (2015) Network integration of parallel
Xavier RJ, O’Neill LA (2013) Succinate is an metabolic and transcriptional data reveals met-
inflammatory signal that induces IL-1beta abolic modules that regulate macrophage
through HIF-1alpha. Nature 496 polarization. Immunity 42(3):419–430.
(7444):238–242. https://doi.org/10.1038/ https://doi.org/10.1016/j.immuni.2015.02.
nature11986 005
8. Mills EL, Kelly B, Logan A, Costa AS, 14. van der Windt GJ, Pearce EL (2012) Metabolic
Varma M, Bryant CE, Tourlomousis P, Dabritz switching and fuel choice during T-cell differ-
JH, Gottlieb E, Latorre I, Corr SC, entiation and memory development. Immunol
McManus G, Ryan D, Jacobs HT, Szibor M, Rev 249(1):27–42. https://doi.org/10.
Xavier RJ, Braun T, Frezza C, Murphy MP, 1111/j.1600-065X.2012.01150.x
O’Neill LA (2016) Succinate dehydrogenase 15. Wang R, Dillon CP, Shi LZ, Milasta S,
supports metabolic repurposing of mitochon- Carter R, Finkelstein D, McCormick LL,
dria to drive inflammatory macrophages. Cell Fitzgerald P, Chi H, Munger J, Green DR
167(2):457–470.e413. https://doi.org/10. (2011) The transcription factor Myc controls
1016/j.cell.2016.08.064 metabolic reprogramming upon T lymphocyte
9. Lampropoulou V, Sergushichev A, activation. Immunity 35(6):871–882. https://
Bambouskova M, Nair S, Vincent EE, doi.org/10.1016/j.immuni.2011.09.021
Loginicheva E, Cervantes-Barragan L, Ma X, 16. Verbist KC, Guy CS, Milasta S, Liedmann S,
Huang SC, Griss T, Weinheimer CJ, Khader S, Kaminski MM, Wang R, Green DR (2016)
Randolph GJ, Pearce EJ, Jones RG, Diwan A, Metabolic maintenance of cell asymmetry fol-
Diamond MS, Artyomov MN (2016) Itaco- lowing division in activated T lymphocytes.
nate links inhibition of succinate dehydroge- Nature 532(7599):389–393. https://doi.
nase with macrophage metabolic remodeling org/10.1038/nature17442
and regulation of inflammation. Cell Metab 17. Swamy M, Pathak S, Grzes KM, Damerow S,
24(1):158–166. https://doi.org/10.1016/j. Sinclair LV, van Aalten DM, Cantrell DA
cmet.2016.06.004 (2016) Glucose and glutamine fuel protein
10. Dennis EA, Deems RA, Harkewicz R, O-GlcNAcylation to control T cell self-renewal
Quehenberger O, Brown HA, Milne SB, and malignancy. Nat Immunol 17
Myers DS, Glass CK, Hardiman G, (6):712–720. https://doi.org/10.1038/ni.
Reichart D, Merrill AH Jr, Sullards MC, 3439
Wang E, Murphy RC, Raetz CR, Garrett TA, 18. O’Sullivan D, van der Windt GJ, Huang SC,
Guan Z, Ryan AC, Russell DW, McDonald JG, Curtis JD, Chang CH, Buck MD, Qiu J, Smith
Thompson BM, Shaw WA, Sud M, Zhao Y, AM, Lam WY, DiPlato LM, Hsu FF, Birnbaum
Gupta S, Maurya MR, Fahy E, Subramaniam MJ, Pearce EJ, Pearce EL (2014) Memory
S (2010) A mouse macrophage lipidome. J Biol CD8(þ) T cells use cell-intrinsic lipolysis to
Chem 285(51):39976–39985. https://doi. support the metabolic programming necessary
org/10.1074/jbc.M110.182915 for development. Immunity 41(1):75–88.
11. Lee JW, Mok HJ, Lee DY, Park SC, Kim GS, https://doi.org/10.1016/j.immuni.2014.06.
Lee SE, Lee YS, Kim KP, Kim HD (2017) 005
UPLC-QqQ/MS-based lipidomics approach
40 Bart Everts

19. Xu X, Araki K, Li S, Han JH, Ye L, Tan WG, Nat Immunol 15(4):323–332. https://doi.
Konieczny BT, Bruinsma MW, Martinez J, org/10.1038/ni.2833
Pearce EL, Green DR, Jones DP, Virgin HW, 27. Lachmandas E, Boutens L, Ratter JM,
Ahmed R (2014) Autophagy is essential for Hijmans A, Hooiveld GJ, Joosten LA, Roden-
effector CD8(þ) T cell survival and memory burg RJ, Fransen JA, Houtkooper RH, van
formation. Nat Immunol 15(12):1152–1161. Crevel R, Netea MG, Stienstra R (2016)
https://doi.org/10.1038/ni.3025 Microbial stimulation of different Toll-like
20. Angela M, Endo Y, Asou HK, Yamamoto T, receptor signalling pathways induces diverse
Tumes DJ, Tokuyama H, Yokote K, Nakayama metabolic programmes in human monocytes.
T (2016) Fatty acid metabolic reprogramming Nat Microbiol 2:16246. https://doi.org/10.
via mTOR-mediated inductions of PPAR- 1038/nmicrobiol.2016.246
gamma directs early activation of T cells. Nat 28. Buescher JM, Antoniewicz MR, Boros LG,
Commun 7:13683. https://doi.org/10. Burgess SC, Brunengraber H, Clish CB,
1038/ncomms13683 DeBerardinis RJ, Feron O, Frezza C,
21. Zeng H, Cohen S, Guy C, Shrestha S, Neale G, Ghesquiere B, Gottlieb E, Hiller K, Jones RG,
Brown SA, Cloer C, Kishton RJ, Gao X, Kamphorst JJ, Kibbey RG, Kimmelman AC,
Youngblood B, Do M, Li MO, Locasale JW, Locasale JW, Lunt SY, Maddocks OD,
Rathmell JC, Chi H (2016) mTORC1 and Malloy C, Metallo CM, Meuillet EJ,
mTORC2 kinase signaling and glucose metab- Munger J, Noh K, Rabinowitz JD, Ralser M,
olism drive follicular helper T cell differentia- Sauer U, Stephanopoulos G, St-Pierre J, Ten-
tion. Immunity 45(3):540–554. https://doi. nant DA, Wittmann C, Vander Heiden MG,
org/10.1016/j.immuni.2016.08.017 Vazquez A, Vousden K, Young JD,
22. Gerriets VA, Rathmell JC (2012) Metabolic Zamboni N, Fendt SM (2015) A roadmap for
pathways in T cell fate and function. Trends interpreting (13)C metabolite labeling patterns
Immunol 33(4):168–173. https://doi.org/ from cells. Curr Opin Biotechnol 34:189–201.
10.1016/j.it.2012.01.010 https://doi.org/10.1016/j.copbio.2015.02.
23. Gerriets VA, Kishton RJ, Johnson MO, 003
Cohen S, Siska PJ, Nichols AG, Warmoes 29. Rodriguez-Prados JC, Traves PG, Cuenca J,
MO, de Cubas AA, MacIver NJ, Locasale JW, Rico D, Aragones J, Martin-Sanz P,
Turka LA, Wells AD, Rathmell JC (2016) Cascante M, Bosca L (2010) Substrate fate in
Foxp3 and Toll-like receptor signaling balance activated macrophages: a comparison between
Treg cell anabolic metabolism for suppression. innate, classic, and alternative activation. J
Nat Immunol 17(12):1459–1466. https:// Immunol 185(1):605–614. https://doi.org/
doi.org/10.1038/ni.3577 10.4049/jimmunol.0901698
24. Geiger R, Rieckmann JC, Wolf T, Basso C, 30. Balmer ML, Ma EH, Bantug GR, Grahlert J,
Feng Y, Fuhrer T, Kogadeeva M, Picotti P, Pfister S, Glatter T, Jauch A, Dimeloe S,
Meissner F, Mann M, Zamboni N, Sallusto F, Slack E, Dehio P, Krzyzaniak MA, King CG,
Lanzavecchia A (2016) L-arginine modulates T Burgener AV, Fischer M, Develioglu L, Belle R,
cell metabolism and enhances survival and anti- Recher M, Bonilla WV, Macpherson AJ,
tumor activity. Cell 167(3):829–842.e813. Hapfelmeier S, Jones RG, Hess C (2016)
https://doi.org/10.1016/j.cell.2016.09.031 Memory CD8(þ) T cells require increased con-
25. Monticelli LA, Buck MD, Flamar AL, Saenz centrations of acetate induced by stress for
SA, Tait Wojno ED, Yudanin NA, Osborne optimal function. Immunity 44
LC, Hepworth MR, Tran SV, Rodewald HR, (6):1312–1324. https://doi.org/10.1016/j.
Shah H, Cross JR, Diamond JM, Cantu E, immuni.2016.03.016
Christie JD, Pearce EL, Artis D (2016) 31. Ma EH, Bantug G, Griss T, Condotta S, John-
Arginase 1 is an innate lymphoid-cell-intrinsic son RM, Samborska B, Mainolfi N, Suri V,
metabolic checkpoint controlling type 2 inflam- Guak H, Balmer ML, Verway MJ, Raissi TC,
mation. Nat Immunol 17(6):656–665. Tsui H, Boukhaled G, Henriques da Costa S,
https://doi.org/10.1038/ni.3421 Frezza C, Krawczyk CM, Friedman A,
26. Everts B, Amiel E, Huang SC, Smith AM, Manfredi M, Richer MJ, Hess C, Jones RG
Chang CH, Lam WY, Redmann V, Freitas (2017) Serine is an essential metabolite for
TC, Blagih J, van der Windt GJ, Artyomov effector T cell expansion. Cell Metab 25
MN, Jones RG, Pearce EL, Pearce EJ (2014) (2):345–357. https://doi.org/10.1016/j.
TLR-driven early glycolytic reprogramming via cmet.2016.12.011
the kinases TBK1-IKKvarepsilon supports the 32. Blagih J, Coulombe F, Vincent EE, Dupuy F,
anabolic demands of dendritic cell activation. Galicia-Vazquez G, Yurchenko E, Raissi TC,
 Immunometabolomics 41

van der Windt GJ, Viollet B, Pearce EL, signaling of metabolites. Cell Metab 25
Pelletier J, Piccirillo CA, Krawczyk CM, (4):777–796. https://doi.org/10.1016/j.
Divangahi M, Jones RG (2015) The energy cmet.2017.03.008
sensor AMPK regulates T cell metabolic adap- 41. Lin L, Zhang J (2017) Role of intestinal micro-
tation and effector responses in vivo. Immunity biota and metabolites on gut homeostasis and
42(1):41–54. https://doi.org/10.1016/j. human diseases. BMC Immunol 18(1):2.
immuni.2014.12.030 https://doi.org/10.1186/s12865-016-0187-3
33. Chang CH, Qiu J, O’Sullivan D, Buck MD, 42. Furusawa Y, Obata Y, Fukuda S, Endo TA,
Noguchi T, Curtis JD, Chen Q, Gindin M, Nakato G, Takahashi D, Nakanishi Y,
Gubin MM, van der Windt GJ, Tonc E, Schrei- Uetake C, Kato K, Kato T, Takahashi M, Fukuda
ber RD, Pearce EJ, Pearce EL (2015) Meta- NN, Murakami S, Miyauchi E, Hino S,
bolic competition in the tumor Atarashi K, Onawa S, Fujimura Y, Lockett T,
microenvironment is a driver of cancer progres- Clarke JM, Topping DL, Tomita M, Hori S,
sion. Cell 162(6):1229–1241. https://doi. Ohara O, Morita T, Koseki H, Kikuchi J,
org/10.1016/j.cell.2015.08.016 Honda K, Hase K, Ohno H (2013) Commensal
34. Romero-Garcia S, Moreno-Altamirano MM, microbe-derived butyrate induces the differenti-
Prado-Garcia H, Sanchez-Garcia FJ (2016) ation of colonic regulatory T cells. Nature 504
Lactate contribution to the tumor microenvi- (7480):446–450. https://doi.org/10.1038/
ronment: mechanisms, effects on immune cells nature12721
and therapeutic relevance. Front Immunol 43. Arpaia N, Campbell C, Fan X, Dikiy S, van der
7:52. https://doi.org/10.3389/fimmu.2016. Veeken J, deRoos P, Liu H, Cross JR, Pfeffer K,
00052 Coffer PJ, Rudensky AY (2013) Metabolites
35. Angelin A, Gil-de-Gomez L, Dahiya S, Jiao J, produced by commensal bacteria promote
Guo L, Levine MH, Wang Z, Quinn WJ III, peripheral regulatory T-cell generation. Nature
Kopinski PK, Wang L, Akimova T, Liu Y, 504(7480):451–455. https://doi.org/10.
Bhatti TR, Han R, Laskin BL, Baur JA, Blair 1038/nature12726
IA, Wallace DC, Hancock WW, Beier UH 44. Song H, Yoo Y, Hwang J, Na YC, Kim HS
(2017) Foxp3 reprograms T cell metabolism (2016) Faecalibacterium prausnitzii subspecies-
to function in low-glucose, high-lactate envir- level dysbiosis in the human gut microbiome
onments. Cell Metab 25:1282. https://doi. underlying atopic dermatitis. J Allergy Clin
org/10.1016/j.cmet.2016.12.018 Immunol 137(3):852–860. https://doi.org/
36. Klein-Wieringa IR, Andersen SN, Kwekke- 10.1016/j.jaci.2015.08.021
boom JC, Giera M, de Lange-Brokaar BJ, van 45. Amiot A, Dona AC, Wijeyesekera A,
Osch GJ, Zuurmond AM, Stojanovic-Susulic V, Tournigand C, Baumgaertner I, Lebaleur Y,
Nelissen RG, Pijl H, Huizinga TW, Sobhani I, Holmes E (2015) (1)H NMR spec-
Kloppenburg M, Toes RE, Ioan-Facsinay A troscopy of fecal extracts enables detection of
(2013) Adipocytes modulate the phenotype of advanced colorectal neoplasia. J Proteome Res
human macrophages through secreted lipids. J 14(9):3871–3881. https://doi.org/10.1021/
Immunol 191(3):1356–1363. https://doi. acs.jproteome.5b00277
org/10.4049/jimmunol.1203074 46. Machiels K, Joossens M, Sabino J, De Preter V,
37. Ioan-Facsinay A, Kwekkeboom JC, Westhoff S, Arijs I, Eeckhaut V, Ballet V, Claes K, Van
Giera M, Rombouts Y, van Harmelen V, Hui- Immerseel F, Verbeke K, Ferrante M,
zinga TW, Deelder A, Kloppenburg M, Toes Verhaegen J, Rutgeerts P, Vermeire S (2014)
RE (2013) Adipocyte-derived lipids modulate A decrease of the butyrate-producing species
CD4þ T-cell function. Eur J Immunol 43 Roseburia hominis and Faecalibacterium praus-
(6):1578–1587. https://doi.org/10.1002/ nitzii defines dysbiosis in patients with ulcerative
eji.201243096 colitis. Gut 63(8):1275–1283. https://doi.org/
38. Gistera A, Hansson GK (2017) The immunol- 10.1136/gutjnl-2013-304833
ogy of atherosclerosis. Nat Rev Nephrol 47. Guo Z, Zhang J, Wang Z, Ang KY, Huang S,
13:368. https://doi.org/10.1038/nrneph. Hou Q, Su X, Qiao J, Zheng Y, Wang L, Koh E,
2017.51 Danliang H, Xu J, Lee YK, Zhang H (2016)
39. Tam VC (2013) Lipidomic profiling of bioac- Intestinal microbiota distinguish gout patients
tive lipids by mass spectrometry during micro- from healthy humans. Sci Rep 6:20602.
bial infections. Semin Immunol 25 https://doi.org/10.1038/srep20602
(3):240–248. https://doi.org/10.1016/j. 48. Donia MS, Fischbach MA (2015) HUMAN
smim.2013.08.006 MICROBIOTA. Small molecules from the
40. Husted AS, Trauelsen M, Rudenko O, Hjorth human microbiota. Science 349
SA, Schwartz TW (2017) GPCR-mediated
42 Bart Everts

(6246):1254766. https://doi.org/10.1126/ Puig L, Gonzalez-Alvaro I, Pinto-Tasende JA,

science.1254766 Blanco R, Rodriguez MA, Beltran A,
49. Saric J (2010) Interactions between immunity Correig X, Marsal S (2016) Urine metabolome
and metabolism - contributions from the met- profiling of immune-mediated inflammatory
abolic profiling of parasite-rodent models. Par- diseases. BMC Med 14(1):133. https://doi.
asitology 137(9):1451–1466. https://doi. org/10.1186/s12916-016-0681-8
org/10.1017/S0031182010000697 52. Zhang A, Sun H, Wang P, Han Y, Wang X
50. Munshi SU, Rewari BB, Bhavesh NS, Jameel S (2012) Modern analytical techniques in meta-
(2013) Nuclear magnetic resonance based bolomics analysis. Analyst 137(2):293–300.
profiling of biofluids reveals metabolic dysre- https://doi.org/10.1039/c1an15605e
gulation in HIV-infected persons and those on 53. Sergushichev AA, Loboda AA, Jha AK, Vincent
anti-retroviral therapy. PLoS One 8(5): EE, Driggers EM, Jones RG, Pearce EJ, Artyo-
e64298. https://doi.org/10.1371/journal. mov MN (2016) GAM: a web-service for
pone.0064298 integrated transcriptional and metabolic net-
51. Alonso A, Julia A, Vinaixa M, Domenech E, work analysis. Nucleic Acids Res 44(W1):
Fernandez-Nebro A, Canete JD, Ferrandiz C, W194–W200. https://doi.org/10.1093/
Tornero J, Gisbert JP, Nos P, Casbas AG, nar/gkw266
 Part II

LC-MS-Based Metabolomics
 Chapter 3

LC-MS-Based Metabolomics of Biofluids Using All-Ion

Fragmentation (AIF) Acquisition
Romanas Chaleckis, Shama Naz, Isabel Meister, and Craig E. Wheelock

Abstract
The field of liquid chromatography-mass spectrometry (LC-MS)-based nontargeted metabolomics has advanced
significantly and can provide information on thousandsofcompounds in biological samples. However, compound
identification remains a major challenge, which is crucial in interpreting the biological function of metabolites.
Herein, we present a LC-MS method using the all-ion fragmentation (AIF) approach in combination with a data
processing method using an in-house spectral library. For the purposes of increasing accuracy in metabolite
annotation, up to four criteria are used: (1) accurate mass, (2) retention time, (3) MS/MS fragments, and
(4) product/precursor ion ratios. The relative standard deviation between ion ratios of a metabolite in a biofluid
vs. its analytical standard is used as an additional metric for confirming metabolite identity. Furthermore, we
include a scheme to distinguish co-eluting isobaric compounds. Our method enables database-dependent
targeted as well as nontargeted metabolomics analysis from the same data acquisition, while simultaneously
improving the accuracy in metabolite identification to increase the quality of the resulting biological information.

Key words Metabolomics, Liquid chromatography-mass spectrometry (LC-MS), All-ion fragmenta-

tion (AIF), Metabolite annotation

Abbreviations

ACN Acetonitrile
AIF All-ion fragmentation
AM Accurate mass
CID Collision-induced dissociation
EIC Extracted ion chromatogram
HILIC Hydrophilic interaction liquid chromatography
LC-MS Liquid chromatography-mass spectrometry
MeOH Methanol
RP Reverse phase
RT Retention time

Romanas Chaleckis and Shama Naz contributed equally to this work.

Martin Giera (ed.), Clinical Metabolomics: Methods and Protocols, Methods in Molecular Biology, vol. 1730,
https://doi.org/10.1007/978-1-4939-7592-1_3, © Springer Science+Business Media, LLC 2018

45
46 Romanas Chaleckis et al.

1 Introduction

Mass spectrometry (MS)-based metabolomics has been used exten-

sively to investigate biological processes and perform biomarker
discovery work [1, 2]. In addition, it is considered an important
part of precision medicine initiatives [3]. However, metabolite
annotation is still a major challenge and significant bottleneck in
translating metabolomics data into biochemical context [4, 5]. Fol-
lowing the identification criteria proposed by the Metabolomics
Standard Initiative (MSI), the highest level of metabolite identifi-
cation is based upon matching two or more orthogonal properties
[e.g., accurate mass (AM), retention time (RT)/index, isotopic
pattern, MS/MS spectrum] of an authentic reference standard
analyzed under the same condition as the metabolite of interest [6].
Liquid chromatography-mass spectrometry (LC-MS) is one of
the most commonly applied techniques in the field of metabolo-
mics, offering high sensitivity and wide metabolite coverage [7]. In
LC-MS-based nontargeted metabolomics approaches, metabolites
are generally reported based on AM and RT matching; however,
these criteria are often insufficient to conclusively identify a metab-
olite due to co-elution and/or the presence of metabolites of
similar m/z and molecular formula. Accordingly, these data are
frequently complemented with MS/MS spectra to increase the
certainty in the identification of metabolites of interest. While
useful, this approach relies heavily upon the use of spectral data-
bases to confirm metabolite structure. The number and quality of
spectral databases has steadily increased, improving the confidence
in metabolite identification; however, additional resources and
efforts are needed for comprehensive structural confirmation.
Herein, we describe a LC-MS metabolomics method using an
all-ion fragmentation (AIF) approach. An AIF experiment uses the
low-energy collision-induced dissociation (CID) to acquire precur-
sor ion mass spectra, whereas the high-energy CID is used to obtain
product ion information by tandem mass spectrometry [8]. In the
subsequent data analysis, extracted ion chromatograms (EIC) from
any precursor or associated product ions of interest are extracted
from the low- or high-energy scan data. One EIC is chosen for
relative quantification (the quantifier ion) of the metabolite, and
further product ions from the same compound are used as qualifier
ions. This approach has the advantage of simultaneously acquiring
extensive fragmentation data on the sample, which can then be
searched against spectral databases for MS/MS-based confirma-
tion. For the purposes of the reported method, an in-house spectral
database was constructed using commercially available analytical
standards [9].
To further increase the metabolite identification specificity,
the product/precursor ion ratio was calculated, and the % relative
 Exploratory LC-MS Based Metabolomics of Biofluids 47

error was calculated between each analytical standard and the

corresponding compound in the biological matrix. The use of
specific quantifier and qualifier ions is also used to distinguish
co-eluting isobaric compounds. The described data acquisition
strategy enables a simultaneous combination of database-
dependent targeted and nontargeted metabolomics in combination
with improved accuracy in metabolite identification, thus increas-
ing the quality of the biological information acquired in a metabo-
lomics experiment.

2 Materials

2.1 Samples 1. Plasma and/or serum samples kept at
80  C until the day of
analysis (see Note 1).
2. Urine samples kept at
80  C until the day of analysis.

2.2 Chemicals The analytical standards for the construction and expansion of the
and Standards compound spectral database, as well as the internal standards, can
be purchased from Sigma-Aldrich (St. Louis, USA), Cayman
Chemical Company (Michigan, USA), Toronto Research Chemi-
cals (Ontario, Canada), Zhejiang Ontores Biotechnologies Co.,
Ltd. (Zhejiang, China), and Avanti Polar Lipids, Inc. (Alabama,
USA) depending upon availability. The standards are prepared at
1 mM concentrations in the appropriate solvent for dissolution,
stored at
20  C, and diluted appropriately on the day of analysis.

2.3 Solutions All solutions are prepared at room temperature (25  C) using
and Solvents LC-MS grade solvents and analytical grade reagents.
1. ESI-low concentration tuning mix (Agilent Technologies,
Santa Clara, USA) for calibrating the TOF-MS: For 100 mL
of solution, mix 10 mL of ESI-low concentration tuning mix,
85.5 mL acetonitrile (ACN), 4.5 mL water, and 3 μL HP-321
(Agilent Technologies, Santa Clara, USA).
2. An internal lock mass mixture (Agilent Technologies, Santa
Clara, USA) for continuous mass calibration during data acqui-
sition: For 500 mL of solution, mix 475 mL ACN, 25 mL
water, 0.2 mL purine [5 mM in (9:1, v/v) ACN/water], and
0.5 mL of HP-0921 [2.5 mM in (9:1, v/v) ACN/water].
3. Solvents for hydrophilic interaction liquid chromatography
(HILIC).
Solvent A (water with 0.1% formic acid): prepare by adding
1 mL formic acid to 1 L of water.
Solvent B (ACN with 0.1% formic acid): prepare by adding
1 mL formic acid to 1 L of ACN.
4. Solvents for the reversed phase (RP) chromatography.
48 Romanas Chaleckis et al.

Solvent A (water with 0.1% formic acid): prepare by adding

1 mL formic acid to 1 L of water.
Solvent B [(90:10, v/v) 2-propanol:ACN]: prepare by add-
ing 1 mL formic acid to 1 L of (90:10, v/v) 2-propanol:ACN.
5. A methanol (MeOH) crash solution for plasma/serum metab-
olite extraction or urine nonpolar metabolite extraction con-
taining internal standards (Table 1) (see Note 2).
6. An ACN crash solution for urine polar metabolite extraction
containing internal standards (Table 1) (see Note 2).

2.4 LC-MS System 1. Ultrahigh performance liquid chromatography (UHPLC)

1290 Infinity II system (including in-line filter 0.3 μm, Agilent
Technologies, Santa Clara, USA) with 1260 Infinity II isocratic
pump (including 1:100 splitter) coupled to a 6550 iFunnel
quadrupole-time of flight (Q-TOF) mass spectrometer with a
dual AJS electrospray ionization source (Agilent Technologies,
Santa Clara, USA).
2. For separation of polar metabolites, HILIC SeQuant® ZIC®-
HILIC column (100 mm  2.1 mm, 100 Å, 3.5 μm particle
size, Merck, Darmstadt, Germany) coupled to a guard column
(2.1 mm  2 mm, 3.5 μm particle size, Merck, Darmstadt,
Germany).
3. For separation of nonpolar metabolites, reversed phase
(RP) Zorbax Eclipse Plus C18, RRHD column
(100 mm  2.1 mm, 1.8 μm particle size, Agilent Technologies,
Santa Clara, USA) coupled to a guard column (5 mm  2 mm,
1.8 μm Agilent Technologies, Santa Clara, USA).

2.5 Software 1. Agilent MassHunter Acquisition for Q-TOF (version B.06.01,

Agilent Technologies, Santa Clara, USA) for acquiring
the data.
2. Agilent PCDL (version B.07.00, Agilent Technologies, Santa
Clara, USA) for managing the compound database (AM, RT,
and MS/MS spectra).
3. Agilent MassHunter TOF-Quant software (version B.07.00,
Agilent Technologies, Santa Clara, USA) for construction of
the data processing method and processing the acquired data
for targeted metabolite screening.

2.6 Other 1. Pipettes and tips (2–20, 20–200, 100–1000, 500–5000 μL).
Equipments 2. Specific gravity refractometer (Atago, Tokyo, Japan).
3. Eppendorf 1.5 mL tubes.
4. Vortex mixer.
5. Pyrex measuring cylinders (100 mL, 1000 mL).
Table 1
Composition of the crash solution (total volume of 500 mL in MeOH or ACN)

Volume for Concentration

Stock crash in
concentration Solvent for the solution (μL) crash solution Monoisotopic
Compound (μg/mL or ppm) stock solution in 500 mL (μg/mL or ppm) Formula mass
13
L-phenylalanine- C9,15N 200 Water 1250 0.5 C6H14[15N]4O2 175.1062
15 15
L-arginine- N4 500 Water 2600 2.6 C6H9[ N]3O2 178.0998
Uracil-15N2 500 Warm water/ 500 0.5 C5H11[15N]O2 114.0213
ammonia water
15
L-valine- N3 1200 Water 8333 20 C3[13C]H3D2O3 118.0760
13 15 13
L-tyrosine- C9 , N 133 Water 3759 1 [ C]2H7NO3S 191.1011
13
Taurine- C2 1000 Water 1250 2.5 C11D5H7N2O2 127.0214
15
L-asparagine- N2 1000 Water 1250 2.5 C9H14D3NO4 134.0476
Acetyl-d3-carnitine 1240 MeOH 202 0.5 C6H5D7O6 206.1346
13 15 13 15
Allantoin C, N 1300 Water 385 1 C3[ C]H6N3[ N] 160.0444
O3
Arachidonic acid-d8 100 Ethanol 5000 1 C20H24D8O2 312.2904
Linoleic acid-d11 100 Ethanol 5000 1 C18H21D11O2 291.3093
Docosahexaenoic acid-d5 100 Ethanol 5000 1 C22H27D5O2 333.2716
Eicosapentaenoic acid-d5 100 Ethanol 5000 1 C20H25D5O2 307.2560
S1P(d18:1/17:0) 456 MeOH 1000 0.186 C17H36NO5P 365.2331
SM(d18:1/17:0) 1000 MeOH 5000 10 C40H81N2O6P 716.5832
Exploratory LC-MS Based Metabolomics of Biofluids

Cer(d18:1/17:0) 500 MeOH 1000 1 C35H69NO3 551.5277

Palmitoyl-L-carnitine (N-methyl-d3) 100 MeOH 250 0.05 C23D3H42NO4 402.3537
49

Chenodeoxycholic acid-2,2,4,4,-d4 1000 MeOH 200 0.4 C24D4H36O4 396.3178

50 Romanas Chaleckis et al.

6. Centrifuge.
7. Evaporator-concentrator.
8. HPLC vials (fixed insert 300 μL amber, Agilent Technologies,
Santa Clara, USA) and screw caps (Agilent Technologies, Santa
Clara, USA).

3 Methods

3.1 Processing 1. Thaw the samples on ice (~1 h, depending on volume), fol-
of Plasma/Serum lowed by vortexing for 30 s.
Samples 2. Place 150 μL of plasma or serum sample in a 1.5 mL Eppendorf
tube (see Note 3).
3. Add 450 μL of ice-cold MeOH crash solution (sample/solvent
1:3, v/v) (see Note 4).
4. Vortex 10 s.
5. Centrifuge the sample to pellet the precipitate (13,000  g,
10 min, 4  C).
6. Transfer 70 μL of the supernatant containing extracted meta-
bolites to a new 1.5 mL Eppendorf tube (see Note 5).
7. Evaporate the samples using an evaporator-concentrator. Set
temperature at 30  C, and dry for 45 min (drying time will vary
depending on unit/volume) (see Note 6).
8. Reconstitute the evaporated samples on the day of the analysis
in 50 μL (8:2, v/v) ACN/water for HILIC and 50 μL MeOH
for RP metabolomics.
9. Centrifuge the reconstituted samples (13,000  g, 2 min,
4  C).
10. Transfer 35 μL to HPLC vial for analysis (see Note 7).

3.2 Processing 1. Thaw the samples on ice (~1 h, depending on volume).

of Urine Samples 2. Vortex 30 s.
3. Centrifuge the sample (13,000  g, 10 min, 4  C), and use the
supernatant in the following steps.
4. Measure the specific gravity by pipetting 100 μL of urine on the
specific gravity refractometer (see Note 8).
5. Normalize all the urine samples to the sample with the lowest
measured specific gravity using LC-MS grade water [10] (see
Note 9).
6. Place 20 μL of normalized urine sample in a 1.5 mL
Eppendorf tube.
7. Add 180 μL of crash solution containing internal standards
(sample/solvent 1:9, v/v).
 Exploratory LC-MS Based Metabolomics of Biofluids 51

8. Vortex 5 s.
9. Centrifuge the sample (13,000  g, 10 min, 4  C).
10. Transfer 40 μL of supernatant to HPLC vial for analysis (see
Note 10).

3.3 LC Parameters To obtain consistently high-quality data, regularly inspect and

maintain the system (see Note 11).
1. Maintain the samples in the autosampler module at 10  C (see
Note 12).
2. Injection volume 2 μL (injection loop 20 μL) (see Note 13).
3. The mobile phase gradient for both HILIC and RP chromatog-
raphy is presented in Tables 2 and 3, respectively (see Note 14).

Table 2
Gradient settings for HILIC chromatography

Time (min) Solvent A (%) Solvent B (%)

0 5 95
1.5 5 95
12 60 40
14 60 40
14.2 75 25
17 75 25
18 5 95
25 5 95

Table 3
Gradient settings for RP chromatography

Time (min) Solvent A (%) Solvent B (%)

0 95 5
3 95 5
5 70 30
18.5 2 98
20 2 98
20.5 95 5
25 95 5
52 Romanas Chaleckis et al.

4. The mobile phase gradient flow rate is 0.3 mL/min for HILIC
(see Note 15) and 0.4 mL/min for RP (see Note 16).
5. The column oven temperature is maintained at 25  C for
HILIC chromatography and 50  C for RP chromatography.

3.4 MS Parameters 1. Tune the MS system before each project using the ESI-low
concentration tuning mix solution.
2. The internal lock mass mixture is constantly infused at a flow
rate of 1 mL/min using an isocratic pump (split ESI spray/
return to the stock bottle 1:100; see Note 17) together with
the LC eluent for constant mass correction. Positive ionization
mode: purine ([MþH]+ m/z 121.0509), HP-0921 ([MþH]+
m/z 922.0098). Negative ionization mode: purine ([M
H]

m/z 119.0363), HP-0921 ([MþCOOH]
m/z 966.0007).
Use a detection window of 100 ppm and minimum height of
1000 counts.
3. Set the sheath and drying gas (nitrogen purity &gt; 99.999%)
flows to 8 L/min and 15 L/min, respectively. Set the tempera-
ture of the drying and sheath gas at 250  C, with the nebulizer
pressure at 35 psig. Set the voltages for positive and negative
ionization modes at þ3000 V and
3000 V, respectively.
4. Set the fragmentor voltages to 380 V at 0 eV, 185 V at 10 eV,
and 410 V at 30 eV.
5. Acquire the data in centroid mode with a mass range of
40–1200 m/z for HILIC and 40–1200 m/z for RP.
6. Perform MS acquisition in AIF mode, where full-scan high-
resolution data are acquired at three alternating collision ener-
gies (0 eV – full scan, 10 eV, and 30 eV) (see Note 18).
7. Set the data acquisition rate at six scans/s.

3.5 Data Processing A data processing method for database-dependent metabolite

and Metabolite screening was constructed in the Agilent TOF-Quant software
Identification (version B.07.00, Agilent Technologies, see Note 19) using precur-
sor and product ion information. The manually constructed data
processing method is useful in improving metabolite identification
as well as deconvoluting isobaric compounds. A list of 413 com-
pounds with AM, RT (HILIC and RP chromatography), fragmen-
tation, and ions ratios is provided in Naz et al. [9]. The level of
confidence in the identification is reflected by the ranking: Rank
1, AM and RT; Rank 2, AM, RT, and MS/MS; and Rank 3, AM,
RT, MS/MS, and ion ratio. Of the 413 compounds, 229 metabo-
lites (Rank 1, 40; Rank 2, 86; Rank 3, 99) were detected in plasma.
The steps for adding a compound to the data processing method
are explained using the example of N-acetylcarnosine (Fig. 1).
1. Inject and measure appropriately diluted analytical standards
using HILIC and/or RP chromatography in positive and/or
negative ionization modes.
 Exploratory LC-MS Based Metabolomics of Biofluids 53

Fig. 1 Metabolite identification with MS/MS fragments and ion ratio confirmation using all-ion fragmentation
(AIF) data. (a) N-acetylcarnosine spectra obtained from an analytical standard (HILIC column, positive
ionization mode) at 0 eV (upper panel), 10 eV (middle panel), 30 eV (lower panel). (b) N-acetylcarnosine
spectra obtained in plasma (HILIC column, positive ionization mode) at the elution time of 8.37 min at 0 eV
(upper panel), 10 eV (middle panel), 30 eV (lower panel). (c) Extracted ion chromatograms of N-acetylcarno-
sine precursor and MS/MS product ions overlaid in standard (left panel) and plasma (right panel). (d) Ion ratios
based on the peak areas in standard and plasma samples and the respective relative errors
54 Romanas Chaleckis et al.

2. Characterize the analytical standard to obtain the precursor ion

m/z and the retention time (RT) from the 0 eV scan. In the
example of N-acetylcarnosine, the [MþH]+ precursor ion is
detected at m/z 269.1244 and RT 8.37 min (Fig. 1a upper
panel).
3. When possible, select more than two product ions (from 10 to
30 eV scans; see Note 20) for each compound (as unique as
possible), and calculate their ratios with the precursor ion. In
the example of N-acetylcarnosine, three product ions can be
observed: m/z 110.0713 at 30 eV, m/z 156.0768 at 10 eV, and
m/z 251.1139 at 10 eV (Fig. 1a middle and lower panels). For
isobaric compounds, a compound-specific highly abundant
product ion is selected as the precursor ion. An example for
deconvoluting isobaric compounds is shown in Fig. 2.
4. For each metabolite, merge the quantifier ion (precursor ion),
qualifier ion(s) (product ions), their relative ion ratios, and RTs
into the data processing method. Use a mass error of 20 ppm,
Gaussian smoothing width of nine points, and a RT window
of 10%.
5. The compound identification is then ranked based on the
matching of the identification criteria as described above. In
the example of N-acetylcarnosine (Fig. 1c, d), it is a Rank
3 compound in plasma sample (see Note 21).
6. After data processing, the quantifier ion peak area of each
identified metabolite is exported in a .csv file format to be
further used for statistical analysis.

3.6 Nontargeted With the presented AIF approach, data can be processed for tar-
Metabolomics Data geted metabolite screening as described above, while at the same
Processing time also enabling nontargeted metabolomics analysis with the
data collected at 0 eV (full scan) (Fig. 3). The workflow below
can be used as a starting point from nontargeted to targeted
metabolomics.
1. Process the data: Detect masses, build (perform smoothing if
needed) and deconvolute chromatograms, remove isotopic
peaks, align the peaks to obtain a feature list for the experiment
(see Note 22).
2. Analyze the data to identify relevant peaks/features.
3. Perform database search to obtain candidate compounds, and
if MS/MS spectra are available, search for fragments in the
10 eV and 30 eV scans (see Note 23).
4. Acquire an analytical standard, and add it to the database-
dependent data processing method as described in
Subheading 3.5.
 Exploratory LC-MS Based Metabolomics of Biofluids 55

Fig. 2 Improving selectivity for co-eluting isobaric compounds using all-ion fragmentation (AIF) data. (a)
Standards of 1-, 3-, and 7-methylxanthines (HILIC method, positive ionization mode) have very close retention
times for the precursor [MþH]+ ion at 0 eV (first column). However, each of the methylxanthines has a distinct
product ion at 30 eV, m/z 110.0349, m/z 149.0458, and m/z 150.0298 for 1-, 3-, and 7-methylxanthines,
respectively (second to fourth columns). (b) In plasma (HILIC column, positive ionization mode), the precursor
ion for methylxanthines [MþH]+ m/z 167.0564 is detected at 2.7 min, 0 eV (first column). In the 30 eV scans,
product ion for 1-methylxanthine (m/z 110.0349, second column) and 7-methylxanthine (a low level of m/z
150.0298, fourth column), but no product ion for 3-methylxanthine, is observed, consistent with the literature
[11]. Integration of the specific fragment ions enables the relative quantification of each co-eluting isobaric
metabolite

4 Notes

1. While beyond the scope of the current protocol, sample collec-

tion and processing procedures are vital for data quality.
2. Compounds can be excluded/included as well as concentra-
tions adjusted as required by the metabolomics experiment.
56 Romanas Chaleckis et al.

AIF-based metabolomics
0eV
3 sequential Full scan @ 0eV
experiments Full scan @ 10eV 10eV
Full scan @ 30eV 30eV

Targeted metabolite screening

Non-targeted metabolite (database-dependent metabolite
profiling screening)

Acquire analytical standard

Process data
to a feature list (m/z and RT)
using XCMS, MZmine, Profinder Characterize analytical standard
-m/z and RT
-MS/MS and ion ratios

Data analysis
-Uni- and multivariate statistics Data processing method
-Database search -Precursor ion as quantifier
-Product ions as qualifiers

Tentative metabolite identification Metabolite identification

based on MS/MS in 10eV and 30eV Rank 1: m/z and RT
scans Rank 2: m/z, RT and MS/MS
Rank 3: m/z, RT, MS/MS and
ion ratios

Fig. 3 Proposed workflow that can be used as a starting point from nontargeted to targeted metabolomics

3. This volume can be adjusted depending upon how many ali-

quots are needed (e.g., 100 μL are sufficient for four aliquots).
4. Crash solution containing standards should be stored at

80  C until needed.
5. A maximum of six 70 μL aliquots can be obtained from 150 μL
plasma or serum sample.
6. If required, the sample can be stored after this step at
80  C.
7. Sometimes after reconstitution, a precipitate may form; avoid
transferring it to the HPLC vial.
8. Wipe off the prism with lint-free tissue between measurements.
9. Create a dilution scheme by using the urine sample with high-
est specific gravity and subsequent 2-, 5-, 10-, and 50-fold
dilutions.
 Exploratory LC-MS Based Metabolomics of Biofluids 57

10. Sometimes after adding crash solution, a precipitate (very little)

may form; avoid transferring it to HPLC vial. A maximum of
four 40 μL aliquots can be obtained from 20 μL urine sample.
11. Before running a new project, we recommend to clean the
autosampler needle, change the in-line filter, and prepare
fresh solvents.
12. For the strong needle wash, the current method uses ACN/-
water ratios of 9:1(v/v) and 1:9 (v/v) for HILIC and RP
chromatographic methods, respectively.
13. The draw speed for the syringe is 100 μL/min, the eject speed
is 400 μL/min, and the wait time after the draw is 1–2 s.
14. For the seal wash of the pumps, use (9:1, v/v) water/MeOH.
15. Expected back pressure at initial conditions 65–70 bars.
16. Expected back pressure at initial conditions 365 bars.
17. Expected back pressure ~23 bars.
18. To increase sensitivity, use two instead of three alternating
scans (e.g., 0 eV and 10 eV).
19. Alternatively, other metabolomics packages such as MS-DIAL
[12] can be adopted to the workflow.
20. Alternative energies can be used if desired.
21. We consider compounds confirmed by ion ratio if the % relative
error of ion ratio (peak area) for at least one product/product
pair is <25%.
22. The parameters for data processing (mass detection, deconvo-
lution, smoothing, etc.) are project and matrix dependent.
23. A list with a short description of online metabolomics data-
bases is available under http://metabolomicssociety.org/res
ources/metabolomics-databases.

Acknowledgments

We acknowledge the support of the Swedish Heart Lung Founda-

tion (HLF 20140469), the Swedish Research Council (2016-
02798), the Swedish Foundation for Strategic Research, the Kar-
olinska Institutet and AstraZeneca Joint Research Program in
Translational Science (ChAMP; Centre for Allergy Research High-
lights Asthma Markers of Phenotype), the Novo Nordisk Founda-
tion (TrIC NNF15CC0018486 and MSAM NNF15CC0018346),
and Gunma University Initiative for Advanced Research (GIAR).
This work was supported in part by The Environment Research and
Technology Development Fund (ERTDF) (Grant No 5-1752).
CEW was supported by the Swedish Heart Lung Foundation
(HLF 20150640).
58 Romanas Chaleckis et al.

References
1. Dunn WB, Broadhurst DI, Atherton HJ et al coupled to mass spectrometry. Nat Protoc
(2011) Systems level studies of mammalian 6:1060–1083. https://doi.org/10.1038/
metabolomes: the roles of mass spectrometry nprot.2011.335
and nuclear magnetic resonance spectroscopy. 8. Plumb RS, Johnson KA, Rainville P et al
Chem Soc Rev 40:387–426. https://doi.org/ (2006) UPLC/MS(E); a new approach for
10.1039/b906712b generating molecular fragment information
2. Kell DB, Oliver SG (2016) The metabolome for biomarker structure elucidation. Rapid
18 years on: a concept comes of age. Metabo- Commun Mass Spectrom 20:1989–1994.
lomics 12:148. https://doi.org/10.1007/ https://doi.org/10.1002/rcm.2550
s11306-016-1108-4 9. Naz S, Gallart-Ayala H, Reinke SN et al (2017)
3. Beger RD, Dunn W, Schmidt MA et al (2016) Development of an LC-HRMS metabolomics
Metabolomics enables precision medicine: “a method with high specificity for metabolite
white paper, community perspective”. Metabo- identification using all ion fragmentation
lomics 12:149. https://doi.org/10.1007/ (AIF) acquisition. Anal Chem 89
s11306-016-1094-6 (15):7933–7942. https://doi.org/10.1021/
4. Viant MR, Kurland IJ, Jones MR, Dunn WB acs.analchem.7b00925.
(2017) How close are we to complete annota- 10. Edmands WMB, Ferrari P, Scalbert A (2014)
tion of metabolomes? Curr Opin Chem Biol Normalization to specific gravity prior to anal-
36:64–69. https://doi.org/10.1016/j.cbpa. ysis improves information recovery from high
2017.01.001 resolution mass spectrometry metabolomic
5. Dias DA, Jones OAH, Beale DJ et al (2016) profiles of human urine. Anal Chem
Current and future perspectives on the struc- 86:10925–10931. https://doi.org/10.1021/
tural identification of small molecules in ac503190m
biological systems. Meta 6:46. https://doi. 11. Martı́nez-López S, Sarriá B, Baeza G et al
org/10.3390/metabo6040046 (2014) Pharmacokinetics of caffeine and its
6. Sumner LW, Amberg A, Barrett D et al (2007) metabolites in plasma and urine after consum-
Proposed minimum reporting standards for ing a soluble green/roasted coffee blend by
chemical analysis chemical analysis working healthy subjects. Food Res Int 64:125–133.
group (CAWG) metabolomics standards initia- https://doi.org/10.1016/j.foodres.2014.05.
tive (MSI). Metabolomics 3:211–221. https:// 043
doi.org/10.1007/s11306-007-0082-2 12. Tsugawa H, Cajka T, Kind T et al (2015)
7. Dunn WB, Broadhurst D, Begley P et al MS-DIAL: data-independent MS/MS decon-
(2011) Procedures for large-scale metabolic volution for comprehensive metabolome anal-
profiling of serum and plasma using gas chro- ysis. Nat Methods 12:523–526. https://doi.
matography and liquid chromatography org/10.1038/nmeth.3393
 Chapter 4

Lipid Mediator Metabolomics Via LC-MS/MS Profiling

and Analysis
Jesmond Dalli, Romain A. Colas, Mary E. Walker, and Charles N. Serhan

Abstract
Solid-phase extraction coupled with liquid chromatography tandem mass spectrometry provides a robust
and sensitive approach for the identification and quantitation of specialized pro-resolving mediators
(lipoxins, resolvins, protectins, and maresins), their pathway markers and the classic eicosanoids. Here,
we provide a detailed description of the methodologies employed for the extraction of these mediators from
biological systems, setup of the instrumentation, sample processing, and then the procedures followed for
their identification and quantitation.

Key words Lipid mediator metabololipidomics, Flux analysis, Eicosanoids, Resolvin, Protectin,
Maresin, Profiling, Liquid chromatography-tandem mass spectrometry

1 Introduction

The role of lipid mediators in regulating distinct aspects of the

body’s functions in humans and experimental systems is well
appreciated [1–4]. In this chapter, we shall detail the methodol-
ogies pioneered in the Serhan laboratory to obtain a snapshot of
the dynamic pathways for the four major bioactive metabolomes
that include the arachidonic acid, eicosapentaenoic acid, and doc-
osahexaenoic acid metabolomes [5–12]. These methodologies
provide insights into mechanisms activated during inflammation
as well as new leads into underlying causes of disease by measur-
ing the flux down each of the major bioactive metabolomes
[5–11]. The methodologies that will be discussed herein employ
C18-based extraction (Fig. 1) to enrich for lipid mediators in
biological systems. These are then coupled with reversed-phase
liquid chromatography electrospray tandem mass spectrometry
that allows for the separation, identification, and quantitation of
these molecules [6, 10] (Figs. 2 and 3). Given that the structures
of these mediators are conserved throughout evolution, these
methodologies are applicable to biological material from

Martin Giera (ed.), Clinical Metabolomics: Methods and Protocols, Methods in Molecular Biology, vol. 1730,
https://doi.org/10.1007/978-1-4939-7592-1_4, © Springer Science+Business Media, LLC 2018

59
60 Jesmond Dalli et al.

Protein precipitation

Addition of Deuterated
Internal Standards in
MeOH

45 min at -20°C to allow

protein precipitation

Solid Phase Extraction Auto-extractor (C18)

1 2 3 4 5 6
H 2O Sample H 2O Hexane MF
Conditioning

Conditioning

Washing
Loading

Washing

Eluting
Waste

Automated Suspend in
Evaporation phase

MF evaporation
using N2

Fig. 1 Sample preparation. Tissues from various origins (human or animal tissues or fluids, mice, planaria, cell
culture, etc.) should be homogenized in ice-cold MeOH containing deuterated internal standards. Samples
should always be kept on ice to prevent mediator isomerization. For biological fluids 4 volumes of methanol
should be added to the samples. Homogenized tissues and biological fluids should then be kept at
20  C for
45 min to allow for protein precipitation. Samples should then be centrifuged and supernatant collected and
acidified to pH 3.5 with HCl prior to solid-phase extraction. The eluting fraction containing the bioactive lipid
mediators is dried under a gentle flux of nitrogen (N2) and resuspended in phase prior injection. MeOH
methanol, H2O water, MF methyl formate, N2 nitrogen

experimental systems such as tunicates [13], mice [5, 7], and

baboons [9] as well as humans [6, 8, 11, 14, 15]. Thus, they
facilitate the direct translation of findings made in experimental
systems to humans and vice versa.
 Lipid Mediator Profiling 61

Auto-injector LC (RP-C18)
100

MeOH (%)
0 Time (min) 22
Separation according to polarity

MS-MS (Qtrap 5500)

Q1 Q2 Q3 D
E
E M/Z- m2/z- T
S M/Z-
m1/z-
M/Z- m1/z- I
E
III m3/z- C
m2/z- m3/z- I
T
Selection of Further Selection of O
parent ion fragmentation the pair ions R

MRM scan EPI scan

Fig. 2 Schematic of instrument setup for mediator identification. Samples in phase are injected using a HPLC-
MS-MS system; no more than 40 μL should be injected. LC liquid chromatography, RP reverse phase, MeOH
methanol, ESI electron spay ionization, m/z mass to charge ratio, MRM multiple reaction monitoring, EPI
enhanced product ion

2 Materials

All solvents should be LC-MS or analytical grade. All stocks should

be handled using zero dead volume Hamilton syringes. Ensure that
the Hamilton syringes are appropriately cleaned using isopropanol
and methanol; this should be done by aliquoting at least 20 syringe
volumes with each solvent. Please note that these compounds are
sensitive to light, oxygen, and heat. They should be stored at
62 Jesmond Dalli et al.

Database IDENTIFICATION
Authentic and synthetic
Matching RT and ≥ 6 ions in MS-MS spectrum against database for each LM
standards of each LM
and pathway markers Synthetic LXB4 Synthetic LXB4 LXB4in sample
(>50) 115 351
115 351
235 235
-H OH OH
COO- -H
COO-
+H +H
221 HO OH 221 HO OH
-H
LXB4 in sample +H +H -H 333
251 279 279
251
289
221 315 307
177 207 233 271 333 115 221 279
115 189 217 289 235 271 315
251279 307 217

Standard curves QUANTIFICATION MULTIVARIATE ANALYSIS

Authentic and synthetic Using individual PCA, PLS-DA and OPLS-DA
standard curves standard curves to
quantify the identified
LM and pathway
markers

AUC pg

Score plot Loading plot

Fig. 3 Lipid mediator identification and quantification. Identification of bioactive LM is performed by matching
retention time (RT) and at least six diagnostic ions from the MS-MS spectrum with those of synthetic or
authentic standards for each mediator. Quantitation is achieved using linear regression curves that are
constructed using synthetic or authentic standards for each mediator of interest. Interrelationship(s) for
identified mediators within each condition and between different conditions is further investigated using
multivariate analysis. LM lipid mediator, RT retention time, AUC area under the curve, PCA principal
component analysis, PLS-DA partial least square-discriminant analysis, OPLS-DA orthogonal partial least
square-discriminant analysis

20  C or
80  C, and exposure to light should be minimized.

Each time the stock is used, purge the vial with a gentle stream of
nitrogen very briefly before closing it to prevent oxidative degrada-
tion. DO NOT use these compounds with DMSO because the
SPM is sensitive to isomerization and oxidation in this solvent.
Please see Notes 1–9 in the notes section for important precautions
to be taken in handling lipid mediators and materials.

2.1 Sample 1. Methanol (MeOH) is used for protein precipitation and sample
Preparation and Solid- extraction as well as solid-phase extraction (SPE) cartridge
Phase Extraction equilibration.
2. For washing of SPE cartridges, use n-hexane.
3. Methyl formate is used as SPE elution solvent.
4. Isolute 500 mg/3 mL C18 SPE columns are used (Biotage).
5. Deuterium-labelled internal standards used for quantification
can be found in Table 1.
 Lipid Mediator Profiling 63

Table 1
List of deuterated internal standards, their complete stereochemistries, and source

Abbreviation Trivial name Full stereochemistry Source

5S-HETE-d8 Deuterium-labelled 5S-hydroxy-eicosa-6E,8Z,11Z, Cayman Chemical
5S-HETE 14Z-tetraenoic-5,6,8,9,11,12,
14,15-d8 acid
LTB4-d4 Deuterium-labelled 5S,12R-dihydroxy-eicosa-6Z,8E, Cayman Chemical
leukotriene B4 10E,14Z-tetraenoic-6,7,14,15-d4 acid
LXA4-d5 Deuterium-labelled 5S,6R,15S-trihydroxy-eicosa-7E,9E,11Z, Cayman Chemical
lipoxin A4 13E-tetraenoic-19,19,20,20,20-d5 acid
RvD2-d5 Deuterium-labelled 7S,16R,17S-trihydroxy-docosa- Cayman Chemical
resolvin D2 4Z,8E,10Z,12E,14E,19Z-hexaenoic-
21,21,22,22,22-d5 acid
PGE2-d4 Deuterium-labelled 9-oxo-11α,15S-dihydroxy-prosta- Cayman Chemical
prostaglandin E2 5Z,13E-dien-1-oic-3,3,4,4-d4 acid

2.2 Liquid 1. Water containing 0.01% acetic acid is used as solvent A.

Chromatography 2. MeOH containing 0.01% acetic acid is used as solvent B.
Tandem Mass
3. LC-MS standards (see Tables 2, 3, 4, and 5).
Spectrometry
4. A Poroshell 120 EC-18 4.6 mm  100 mm  2.7 μm reversed-
phase column is employed (Agilent).

3 Methods

3.1 Preparing 1. Blank the UV spectrophotometer using the solvent of interest.

Standard Mixes 2. Aliquot a known volume of the stock using a Hamilton syringe.
for Internal Standards
3. Measure the UV absorbance ensuring that a characteristic chro-
and Standard Curves
mophore is observed for the molecule of interest (i.e., diene,
3.1.1 Determining triene, or tetraene) [6].
the Concentration 4. Calculate absorption using a three-point dropline.
of Synthetic Material
5. Calculate concentration using the Beer-Lambert law.
Concentration ¼ absorbance/extinction coefficient*dilution
(see Table 6 for λmaxMEOH and extinction coefficients for each
conjugated double-bond system).

3.2 Tuning of the LC- 1. Using synthetic standards, first tune the probe position to maxi-
MS-MS for Lipid mize signal.
Mediator Profiling 2. Next, adjust electrode position.
3. Afterward, in the instrument parameter window, adjust curtain
gas, collisionally activated dissociation (CAD) gas, electrode
64 Jesmond Dalli et al.

Table 2
List of arachidonic acid metabolome, their complete stereochemistries, and source

Abbreviation Trivial name Full stereochemistry Source

AA Arachidonic acid eicosa-5Z,8Z,11Z,14Z- Cayman Chemical
tetraenoic acid
PGD2 Prostaglandin D2 9α,15S-dihydroxy-11-oxo-prosta-5Z, Cayman Chemical
13E-dien-1-oic acid
PGE2 Prostaglandin E2 9-oxo-11α,15S-dihydroxy-prosta-5Z, Cayman Chemical
13E-dien-1-oic acid
PGF2a Prostaglandin F2a 9α,11α,15S-trihydroxy-prosta-5Z, Cayman Chemical
13E-dien-1-oic acid
TxB2 Thromboxane B2 9α,11,15S-trihydroxy-thromba-5Z, Cayman Chemical
13E-dien-1-oic acid
LTB4 Leukotriene B4 5S,12R-dihydroxy-eicosa-6Z,8E,10E, Cayman Chemical
14Z-tetraenoic acid
LXA4 Lipoxin A4 5S,6R,15S-trihydroxy-eicosa-7E,9E, Cayman Chemical
11Z,13E-tetraenoic acid
LXB4 Lipoxin B4 5S,14R,15S-trihydroxy-eicosa-6E,8Z, Cayman Chemical
10E,12E-tetraenoic acid
15-HETE 15S-hydroxy-eicosa-5Z,8Z,11Z, Cayman Chemical
13E-tetraenoic acid
12-HETE 12-hydroxy-eicosa-5Z,8Z,10E, Cayman Chemical
14Z-tetraenoic acid
5-HETE 5S-hydroxy-eicosa-6E,8Z,11Z, Cayman Chemical
14Z-tetraenoic acid
Δ6-trans-LTB4 Δ6-trans- 5S,12R-dihydroxy-eicosa-6E,8E, Cayman Chemical
leukotriene B4 10E,14Z-tetraenoic acid
5S,12S-diHETE 5S,12S-dihydroxy-eicosa-6E,8E, Biogenic Synthesis
10E,14Z-tetraenoic acid

voltage, source temperature, and the source gases one at a time

to identify optimal source settings.
4. Subsequently tune the compound parameters (declustering
potential, entrance potential, collision energy, and collision cell
exit potential) for each compound and for each transition.
5. It is suggested that for each compound (both pathway markers
and mediators) at least two transitions are used.

3.3 Automated Lipid 1. Ensure that samples have not been freeze-thawed; fresh sam-
Mediator Extraction ples are preferred.
3.3.1 Sample 2. For frozen samples allow to defrost on ice.
Preparation (Fig. 1) 3. For tissues dissociate gently in ice-cold MeOH ensuring that
the samples remain cold throughout the process.
 Lipid Mediator Profiling 65

Table 3
List of eicosapentaenoic acid metabolome, their complete stereochemistries, and source

Abbreviation Trivial name Full stereochemistry

EPA eicosa-5Z,8Z,11Z,14Z, Cayman Chemical
17Z-pentaenoic acid
RvE1 Resolvin E1 5S,12R,18R-trihydroxy-eicosa- Cayman Chemical
6Z,8E,10E,14Z,16E-pentaenoic acid
RvE2 Resolvin E2 5S,18R-dihydroxy-eicosa- Biogenic Synthesis
6Z,8E,11E,14E,16Z-pentaenoic acid
RvE3 Resolvin E3 17R,18R/S-dihydroxy-eicosa- Custom Synthesis (Dr Makoto
5Z,8Z,11Z,13E,15E-pentaenoic acid Arita, Riken Institute Japan)
18-HEPE 18-hydroxy-eicosa- Cayman Chemical
5Z,8Z,11Z,14Z,16E-pentaenoic acid
15-HEPE 15-hydroxy-eicosa- Cayman Chemical
5Z,8Z,11Z,13E,17Z-pentaenoic acid
12-HEPE 12-hydroxy-eicosa- Cayman Chemical
5Z,8Z,10E,14Z,17Z-pentaenoic acid
5-HEPE 5-hydroxy-eicosa-6E,8Z,11Z,14Z, Cayman Chemical
17Z-pentaenoic acid

Table 4
List of docosahexaenoic acid metabolome, their complete stereochemistries, and source

Abbreviation Trivial name Full stereochemistry

DHA docosa-4Z,7Z,10Z,13Z,16Z, Cayman Chemical
19Z-hexaenoic acid
RvD1 Resolvin D1 7S,8R,17S-trihydroxy-docosa- Cayman Chemical
4Z,9E,11E,13Z,15E,
19Z-hexaenoic acid
RvD2 Resolvin D2 7S,16R,17S-trihydroxy-docosa- Cayman Chemical
4Z,8E,10Z,12E,14E,
19Z-hexaenoic acid
RvD3 Resolvin D3 4S,7R,17S-trihydroxy-docosa- Cayman Chemical
5Z,7E,9E,13Z,15E,
19Z-hexaenoic acid
RvD4 Resolvin D4 4S,5R,17S-trihydroxy-docosa- Custom Synthesis
6E,8E,10Z,13Z,15E, (Dr Charles Serhan
19Z-hexaenoic acid
RvD5 Resolvin D5 7S,17S-dihydroxy-docosa- Cayman Chemical
4Z,8E,10Z,13Z,
15E,19Z-hexaenoic acid
(continued)
66 Jesmond Dalli et al.

Table 4
(continued)

Abbreviation Trivial name Full stereochemistry

RvD6 Resolvin D6 4S,17S-dihydroxy-docosa- Biogenic Synthesis
5E,7Z,10Z,13Z,
15E,19Z-hexaenoic acid
MaR1 Maresin 1 7R,14S-dihydroxy-docosa- Cayman Chemical
4Z,8E,10E,12Z,
16Z,19Z-hexaenoic acid
4S,14S- 4S,14S-dihydroxy-docosa- Biogenic Synthesis
diHDHA 5Z,7E,10E,12Z,
16E,19E-hexaenoic acid
7S,14S- 7S,14S-dihydroxy-docosa- Biogenic Synthesis
diHDHA 4Z,8E,10E,12Z,
16E,19E-hexaenoic acid
PD1 Protectin D1 10R,17S-dihydroxy-docosa- Custom Synthesis
4Z,7Z,11E,13E, (Dr Charles Serhan)
15Z,19Z-hexaenoic acid
10S,17S- Protectin Dx 10S,17S-dihydroxy-docosa- Cayman Chemical
diHDHA 4Z,7Z,11E,13Z,
15E,19Z-hexaenoic acid
22-OH-PD1 22-OH-protectin 10R,17S,20-trihydroxy-docosa- Custom Synthesis
D1 4Z,7Z,11E,13E,15Z, (Dr Trond
19Z-hexaenoic acid V. Hansen, University of
Oslo)
17-HDHA 17-hydroxy-docosa- Cayman Chemical
4Z,7Z,10Z,13Z,15E,
19Z-hexaenoic acid
14-HDHA 14S-hydroxy-docosa- Cayman Chemical
4Z,7Z,10Z,12E,16Z,
19Z-hexaenoic acid
13-HDHA 13-hydroxy-docosa- Cayman Chemical
4Z,7Z,10Z,14E,16Z,
19Z-hexaenoic acid
7-HDHA 7-hydroxy-docosa- Cayman Chemical
4Z,8E,10Z,13Z,16Z,
19Z-hexaenoic acid
4-HDHA 4-hydroxy-docosa- Cayman Chemical
5E,7Z,10Z,13Z,16Z,
19Z-hexaenoic acid
 Lipid Mediator Profiling 67

Table 5
List of aspirin-triggered mediators, their complete stereochemistries, and source

Abbreviation Trivial name Full stereochemistry

AT-LXA4 Aspirin-triggered 5S,6R,15R-trihydroxy-eicosa- Cayman Chemical
lipoxin A4 7E,9E,11Z,13E-tetraenoic acid
AT-LXB4 Aspirin-triggered 5S,14R,15R-trihydroxy-eicosa- Cayman Chemical
lipoxin B4 6E,8Z,10E,12E-tetraenoic acid
AT-RvD1 Aspirin-triggered 7S,8R,17R-trihydroxy-docosa- Cayman Chemical
resolvin D1 4Z,9E,11E,13Z,15E,19Z-hexaenoic acid
AT-RvD3 Aspirin-triggered 4S,7R,17R-trihydroxy-docosa- Custom Synthesis
resolvin D3 5Z,7E,9E,13Z,15E,19Z-hexaenoic acid (Dr Charles Serhan)
AT-PD1 Aspirin-triggered 10R,17R-dihydroxy-docosa- Custom Synthesis
protectin D1 4Z,7Z,11E,13E,15Z,19Z-hexaenoic (Dr Charles Serhan)
acid

Table 6
Extinction coefficient and λmaxMEOH for distinct conjugated double-bond systems

Double-bond system Extinction coefficient λmaxMEOH

Conjugated diene (e.g., monohydroxy acids) 25,000 235
Two conjugated dienes (e.g., RvD5) 25,000 240
Diene-triene conjugated system (e.g., RvE1) 40,000 271
Conjugated triene (e.g., PD1) 40,000 269
Conjugated tetraene (e.g., RvD1) 50,000 301

4. For liquid samples (e.g., cell culture preparations and plasma)

add four equal volumes of ice-cold methanol containing
deuterium-labelled internal standards to each sample.
5. Place at
20  C for 45 min to allow for protein precipitation.
6. Centrifuge samples at 2000  g for 10 min at 4  C.
7. Collect supernatant.
8. Transfer tubes to TurboVap Evaporator.
9. Ensure that the water bath is set to 37  C.
10. Turn nitrogen feed on and set the flow rate to no more than
15 psi.
11. Ensure that the lid is closed.
12. Methanol volume should be evaporated to less than 1 mL
using a steady nitrogen stream.
68 Jesmond Dalli et al.

13. Centrifuge samples at 2000  g for 10 min at 4  C.

14. Samples are now ready for extraction. At this stage samples
should not be stored.

3.3.2 Lipid Mediator 1. Add new C18 columns on the rack.

Extraction 2. Add the collection plates containing elution tubes.
3. Make sure run-through plate is at position D.
4. Click “run single method.”
5. Select the extraction method.
6. Click “prepare run.”
7. Go to solvent feeder (5), and prime each individually (make
sure, the containers are filled, especially the MF).
8. Go to extraction media (3) and select how many columns you
will use.
9. Go to solvent tips (1) and sample tips (2), and check tips are
filled as displayed on screen. If necessary, add or delete missing
columns (or add tips).
10. Click “run method.”
11. Monitor that the extraction procedure is initiated correctly.
12. At the end of the extraction, transfer the eluted samples (i.e.,
the MF fraction) to glass tubes, rinse the collection tube or the
collecting well once with MeOH, and add to the sample.
13. Samples are ready for drying on TurboVap.

3.3.3 Solvent 1. Switch the TurboVap Evaporator on.

Evaporation 2. Ensure that the water bath is set to 37  C.
3. Turn nitrogen feed on and set the flow rate to no more than
15 psi.
4. Place 10 mL conical borosilicate tubes containing the methyl
formate fraction or methanol fraction obtained during solid-
phase extraction in the TurboVap.
5. Ensure that the lid is closed.
6. When the solvent is more than 95% evaporated, rinse the walls
of the tube using methyl formate.
7. When the solvent is more than 95% evaporated, rinse the walls
of the tube using methanol.
8. When the solvent is completely evaporated, add 40 μL of
methanol/water (1:1).
9. Centrifuge the tube at 2000  g for 2 min.
10. Carefully transfer the supernatant to an injection vial insert.
11. Place the insert in a clearly labelled 1.5 mL Eppendorf tube.
 Lipid Mediator Profiling 69

12. Centrifuge the insert at 10,000  g for no more than 2 min.

13. Transfer the supernatant to an injection vial insert, and place
the insert in a clearly labelled sample vial.
14. Samples are now ready for LC-MS-MS profiling.

3.4 Chromatography 1. Using a Poroshell 120 EC-18 (4.6 mm  100 mm  2.7 μm)
and water containing 0.01% acetic acid as solvent A, and meth-
anol containing 0.01% acetic acid as solvent B, the following
gradient should be used to chromatographically separate the
mediators from their isomers and pathway markers.
2. Column temperature should be set at 50  C.
3. Equilibrate the column with mobile phase at 80:20 (A:B).
4. This should be ramped to 50:50 (A:B) over 12 s.
5. The gradient should be maintained for 2 min.
6. Then to 80:20 over the subsequent 9 min.
7. This should be maintained for the next 3.5 min.
8. Then ramped to 98:2 (A:B).
9. Finally maintain this for 5.4 min to wash the column.
10. The follow rate should be maintained at 0.5 mL/min through-
out the experiment.

3.5 Data Analysis 1. In analyst click on build quantitation method.

Using Analyst 2. Select a file that contains a standard mix.
Quantitate Tool
3. Assign which transitions correspond to the internal standards.
3.5.1 Setting Up 4. Assign the internal standards that will be used for identification
a Quantitation Method and quantitation of each molecule.
5. In the integrate tab, select a retention time window of not more
than 5 s.
6. Save the method.

3.5.2 Analyzing the Data 1. Click the quantitation wizard, and follow the instructions to
load your samples and the appropriate analysis method
obtained as detailed above.
2. Start by integrating the deuterium-labelled internal standards
making a note of any drifts from the expected retention times.
3. If retention time drifts are observed, correct the expected
retention times for each of the analyte according to the reten-
tion time drift observed in the respective internal standards
being used for identification (e.g., if d5-LXA4 is observed to
elute 2 s after the anticipated retention time, then LXA4
expected retention time will also shift by 2 s).
70 Jesmond Dalli et al.

4. After determining the expected retention times for each of the

samples, integrate each of the mediators. Here all peaks that are
clearly visible even if not more than three times higher than
baseline should be integrated.
5. After integrating all the peaks in explore mode, search for
MS-MS spectra for molecules where a peak was obtained in at
least one transition in the same region of the chromatogram
where the peak was recorded.
6. Compare the tandem mass spectrum obtained in the sample
with that obtained from the synthetic/authentic compound
matching at least six ions in the MS-MS spectrum, with one
of the ions being derived from a backbone break. These can be
compared and matched to the Spectra Book available for all of
these mediators and pathways on the Serhan lab website.
7. For those molecules where the MS-MS spectrum is a positive
match, proceed to quantitation using a standard curve con-
structed using either synthetic or authentic standard mixes with
concentrations determined as detailed above.
8. For molecules where standards are not available, molecules that
carry similar physical properties may be used.
9. If an MS-MS spectrum is not obtained for a molecule or group
of molecules and additional sample is available, this could be
rerun looking specifically for the molecule(s) where a spectrum
was not obtained.
10. If no additional sample is available then the identification cri-
teria are not fulfilled.

4 Notes

1. Concentrated HCl and acetic acid should be handled in a fume-

hood. Gloves and eye protection need to be worn at all times
when handling these acids. Also always add the acid to water to
avoid injury.
2. All standard stocks should be stored under nitrogen, shielded
from light, and at
20  C for short-term storage and
80  C for
long-term storage.
3. Each time the stock is used, purge the vial with a gentle stream of
nitrogen very briefly before closing it to prevent oxidative
degradation.
4. All solvents are flammable; they should be handled with care in a
fumehood and kept away from open flames.
5. All compound stocks should be handled using zero dead volume
Hamilton syringes.
 Lipid Mediator Profiling 71

6. Ensure that the Hamilton syringes are appropriately cleaned

using isopropanol and methanol; this should be done by aliquot-
ing at least 20 syringe volumes with each solvent.
7. Only use nitrogen or an inert gas to evaporate solvents; do not
use air since this will lead to mediator oxidation.
8. Upon solvent evaporation and suspension of samples in water/
methanol, samples need to be profiled within 24 h to avoid
isomerization of the mediators. At all times samples should be
kept at 4  C and not frozen to maximize mediator integrity.
9. DO NOT DMSO at any stage since SPM is sensitive to isomeri-
zation and oxidation in this solvent.

Acknowledgments

This work was supported by funding from the European Research

Council (ERC) under the European Union’s Horizon 2020
research and innovation program (Grant number: 677542), a Sir
Henry Dale Fellowship jointly funded by the Wellcome Trust and
the Royal Society (Grant number: 107613/Z/15/Z), and the
Barts Charity (Grant number: MGU0343). CN Serhan is sup-
ported by the National Institutes of Health, USA Grant Number
P01GM095467.

References
1. Basil MC, Levy BD (2016) Specialized 6. Colas RA, Shinohara M, Dalli J, Chiang N,
pro-resolving mediators: endogenous regula- Serhan CN (2014) Identification and signature
tors of infection and inflammation. Nat Rev profiles for pro-resolving and inflammatory
Immunol 16(1):51–67. https://doi.org/10. lipid mediators in human tissue. Am J Physiol
1038/nri.2015.4 Cell Physiol 307(1):C39–C54. https://doi.
2. Haworth O, Buckley CD (2015) Pathways org/10.1152/ajpcell.00024.2014
involved in the resolution of inflammatory 7. Dalli J, Colas RA, Arnardottir H, Serhan CN
joint disease. Semin Immunol 27 (2017) Vagal regulation of group 3 innate lym-
(3):194–199. https://doi.org/10.1016/j. phoid cells and the Immunoresolvent PCTR1
smim.2015.04.002 controls infection resolution. Immunity 46
3. Miyata J, Arita M (2015) Role of omega-3 fatty (1):92–105. https://doi.org/10.1016/j.
acids and their metabolites in asthma and aller- immuni.2016.12.009
gic diseases. Allergol Int 64(1):27–34. https:// 8. Dalli J, Colas RA, Quintana C, Barragan-
doi.org/10.1016/j.alit.2014.08.003 Bradford D, Hurwitz S, Levy BD, Choi AM,
4. Serhan CN, Chiang N, Dalli J (2015) The Serhan CN, Baron RM (2017) Human sepsis
resolution code of acute inflammation: novel eicosanoid and proresolving lipid mediator
pro-resolving lipid mediators in resolution. temporal profiles: correlations with survival
Semin Immunol 27(3):200–215. https://doi. and clinical outcomes. Crit Care Med 45
org/10.1016/j.smim.2015.03.004 (1):58–68. https://doi.org/10.1097/CCM.
5. Chiang N, Fredman G, Backhed F, Oh SF, 0000000000002014
Vickery T, Schmidt BA, Serhan CN (2012) 9. Dalli J, Kraft BD, Colas RA, Shinohara M, Fre-
Infection regulates pro-resolving mediators denburgh LE, Hess DR, Chiang N, Welty-
that lower antibiotic requirements. Nature Wolf K, Choi AM, Piantadosi CA, Serhan CN
484(7395):524–528. https://doi.org/10. (2015) The regulation of proresolving lipid
1038/nature11042 mediator profiles in baboon pneumonia by
72 Jesmond Dalli et al.

inhaled carbon monoxide. Am J Respir Cell 13. Knight J, Taylor GW, Wright P, Clare AS, Row-
Mol Biol 53(3):314–325. https://doi.org/ ley AF (1999) Eicosanoid biosynthesis in an
10.1165/rcmb.2014-0299OC advanced deuterostomate invertebrate, the sea
10. Dalli J, Serhan CN (2012) Specific lipid medi- squirt (Ciona intestinalis). Biochim Biophys
ator signatures of human phagocytes: micro- Acta 1436(3):467–478
particles stimulate macrophage efferocytosis 14. Rathod KS, Kapil V, Velmurugan S, Khambata
and pro-resolving mediators. Blood 120(15): RS, Siddique U, Khan S, Van Eijl S, Gee LC,
e60–e72. https://doi.org/10.1182/blood- Bansal J, Pitrola K, Shaw C, D’Acquisto F,
2012-04-423525 Colas RA, Marelli-Berg F, Dalli J, Ahluwalia A
11. Weiss GA, Troxler H, Klinke G, Rogler D, (2017) Accelerated resolution of inflammation
Braegger C, Hersberger M (2013) High levels underlies sex differences in inflammatory
of anti-inflammatory and pro-resolving lipid responses in humans. J Clin Invest 127
mediators lipoxins and resolvins and declining (1):169–182. https://doi.org/10.1172/
docosahexaenoic acid levels in human milk dur- JCI89429
ing the first month of lactation. Lipids Health 15. Titos E, Rius B, Lopez-Vicario C, Alcaraz-
Dis 12:89. https://doi.org/10.1186/1476- Quiles J, Garcia-Alonso V, Lopategi A, Dalli J,
511X-12-89 Lozano JJ, Arroyo V, Delgado S, Serhan CN,
12. Serhan CN, Petasis NA (2011) Resolvins and Claria J (2016) Signaling and immunoresol-
protectins in inflammation resolution. Chem ving actions of resolvin D1 in inflamed human
Rev 111(10):5922–5943. https://doi.org/ visceral adipose tissue. J Immunol 197
10.1021/cr100396c (8):3360–3370. https://doi.org/10.4049/
jimmunol.1502522
 Chapter 5

UHPSFC/ESI-MS Analysis of Lipids

Miroslav Lı́sa and Michal Holčapek

Abstract
This new analytical approach for high-throughput and comprehensive lipidomic analysis of biological
samples using ultrahigh-performance supercritical fluid chromatography (UHPSFC) with electrospray
ionization–mass spectrometry (ESI-MS) is based on lipid class separation using 1.7 μm particle bridged
ethylene hybrid silica columns and a gradient of methanol–water–ammonium acetate mixture as a modifier.
The method enables a fast separation of 30 nonpolar and polar lipid classes within 6-min analysis time
covering six main lipid categories including fatty acyls, glycerolipids, glycerophospholipids, sphingolipids,
sterols, and prenols. Individual lipid species within lipid classes are identified based on positive- and
negative-ion full scan and tandem mass spectra measured with high mass accuracy and high resolving
power. The method is used for the quantitative analysis of lipid species in biological tissues using internal
standards for each lipid class. This high-throughput, comprehensive, and accurate UHPSFC/ESI-MS
method is suitable for the lipidomic analysis of large sample sets in clinical research.

Key words Supercritical fluid chromatography, Ultrahigh-performance supercritical fluid chromatog-

raphy, Mass spectrometry, Electrospray ionization, Lipids, Lipidomics, Glycerolipids, Glyceropho-
spholipids, Sphingolipids

1 Introduction

Supercritical fluid chromatography (SFC) is a chromatographic

technique, which uses a supercritical fluid heated above critical
temperature and pressurized above critical pressure as the mobile
phase [1]. The supercritical fluid combines some unique properties
of gases and liquids, such as low viscosity similarly to gases, high
density similarly to liquids, and diffusivity between gases and
liquids. These properties enable fast analyses with high resolving
power in SFC. Any fluid above the critical point can be used as the
mobile phase in SFC, but carbon dioxide is almost exclusively
employed due to its favorable physicochemical properties, such as
relatively low values of critical characteristics (7.4 MPa and 31
C),
which can be easily achieved using conventional chromatographic
instrumentation. Organic modifiers are added to carbon dioxide to
enhance its polarity, solvating power, and elution strength, such as

Martin Giera (ed.), Clinical Metabolomics: Methods and Protocols, Methods in Molecular Biology, vol. 1730,
https://doi.org/10.1007/978-1-4939-7592-1_5, © Springer Science+Business Media, LLC 2018

73
74 Miroslav Lı́sa and Michal Holčapek

methanol, ethanol, acetonitrile, 2-propanol, or their mixtures.

Conventional LC columns with 3–5 μm particle size have been
mainly used in SFC, but in recent years, sub-2 μm particles
providing ultrahigh-performance SFC (UHPSFC) separations
demonstrate their benefits, such as faster analyses and higher
efficiency.
The UHPSFC/ESI-MS coupling is a powerful tool for the
sensitive identification/quantitation in complex mixtures. The
organic solvent is usually added after the column to the mobile
phase as a makeup liquid to ensure ESI ionization.
SFC of lipids can be performed in two separation modes, i.e.,
the lipid class separation according to the class polarity or the
separation of individual lipid species based on the fatty acyl chain
composition [2]. The lipid class separation is used for the analysis of
total lipid extracts, where lipids are separated into lipid classes
according to their polarities, while species within the class differing
only in fatty acyl chain composition are not separated, and they
elute together in one chromatographic peak of particular lipid class
[3, 4]. Lipid classes are identified based on retention times and class
characteristic fragment ions in ESI mass spectra. Averaged mass
spectra of lipid class chromatographic peaks are used for the deter-
mination of lipid species level, i.e., the number of carbon atoms and
double bonds of attached fatty acyl/alkyls. For the quantitative
analysis, nonendogenous lipid species for each lipid class are
added before the extraction step as internal standards (IS) [5, 6]
to cover the extraction efficiency and possible fluctuation of MS
signal. The lipid class separation provides the coelution of lipid
species with the lipid class IS in one peak under identical mobile
phase and matrix composition, which is especially important for
accurate quantitative measurements.
This protocol presents a new analytical strategy for the high-
throughput and comprehensive lipidomic analysis of biological
samples applicable for large-scale clinical studies. High-throughput
UHPSFC/ESI-MS method using sub-2 μm particle bridged ethyl-
ene hybrid silica column is used for the lipid class separation of a
wide range of nonpolar and polar lipids in one analysis including the
identification and quantitation of individual lipid species using
ESI-MS.

2 Materials

Solutions and samples are prepared using acetonitrile, 2-propanol,

methanol, n-hexane, and chloroform stabilized with 0.5–1% etha-
nol (all HPLC/MS or HPLC grade). Ammonium acetate, deio-
nized water, methanol (all HPLC/MS grade), and carbon dioxide
4.5 grade (99.995%) are used for UHPSFC/ESI-MS analysis.
 UHPSFC/ESI-MS Analysis of Lipids 75

2.1 Lipid Standards Qualitative standards of lipid class representatives used for
UHPSFC/ESI-MS method development, i.e., CE, TG, FA,
1,3-DG, 1,2-DG, and 1-MG containing oleoyl acyls and choles-
terol, can be purchased from Sigma-Aldrich, and Cer, PG, PE,
LPG, PI, LPE, CL, LPI, PA, PC, pPC, ePC, PS, LPA, SM, LPC,
and LPS containing oleoyl acyls, sphingosine, sphinganine, GlcCer
d18:1/16:0, LacCer d18:1/16:0, S1P d17:1, desmosterol, and
DHEA can be purchased from Avanti Polar Lipids (Alabaster, AL,
USA). Internal standards (IS) used for quantitation, i.e., CE 19:0,
TG 19:0/19:0/19:0, FA 14:0, DG 19:0/0:0/19:0, and MG
19:0/0:0/0:0, are purchased from Nu-Chek Prep (Elysian, MN,
USA); D7-cholesterol, Cer d18:1/17:0, GlcCer d18:1/12:0, PG
14:0/14:0, LacCer d18:1/12:0, PE 14:0/14:0, LPG 14:0/0:0,
LPE 14:0/0:0, PC 22:1/22:1, PC 14:0/14:0, SM d18:1/17:0,
and LPC 17:0/0:0 are purchased from Avanti Polar Lipids.

2.2 Solutions 1. Stock solutions of individual qualitative and quantitative stan-

dards: prepare individual stock solutions at a concentration of
2 mg/mL using a (1:4, v/v) chloroform–2-propanol mixture.
2. The solution of all qualitative standards is prepared by mixing
of 20 μL of each standard diluted to the appropriate concentra-
tion with a (7:3, v/v) hexane–chloroform mixture for
UHPSFC/ESI-MS analysis.

2.3 Instruments UHPSFC experiments were performed on an Acquity UPC2

instrument (Waters, Milford, MA, USA). HPLC 515 pump
(Waters) was used as a makeup pump. The hybrid quadrupole–tra-
velling wave ion mobility–time-of-flight mass spectrometer Synapt
G2Si (Waters) was used for MS experiments.

3 Methods

3.1 Sample Total lipid extracts of tissues (typically 25 mg) or plasma (typically
Preparation 25 μL) are prepared according to the modified Folch procedure [7]
using a chloroform–methanol–water system, subsequently evapo-
rated by a gentle stream of nitrogen and redissolved in 1 mL of a
(1:1, v/v) chloroform–2-propanol mixture. 10 μL of the extract
stock solution is diluted into 1 mL of a (7:3, v/v) hexane–chloro-
form mixture for nontargeted UHPSFC/ESI-MS analysis. The
same procedure is used for the preparation of the total lipid extract
for UHPSFC/ESI-MS quantitation with the addition of IS into the
sample before the extraction. 2 μL of LPE and DG; 4 μL of TG,
Cer, PG, LPC, and MG; and 40 μL of FA, GlcCer, CE,
D7-cholesterol, sulfatide, PE, PI, PC, and SM stock solutions of
IS were added.
76 Miroslav Lı́sa and Michal Holčapek

3.2 UHPSFC 1. Prepare the modifier by mixing methanol and water in the ratio
Conditions of 99:1 (v/v) with the addition of 30 mM of ammonium
acetate.
2. Condition the Acquity BEH UPC2 column (100  3 mm,
1.7 μm, Waters) using a flow rate of 1.9 mL/min consisting
of 99% carbon dioxide and 1% of the modifier, column temper-
ature 60
C, and the active back pressure regulator (ABPR)
1800 psi for 20 min before the first injection (see Note 1).
3. Inject 1 μL of sample with the gradient of the modifier: 0 min,
1%; 5 min, 51%; 6, min 51%; 7 min, 1%; and 10 min, 1% (see
Note 2) (Fig. 1).
4. Wash the injector needle with a (2:2:1, v/v/v) hexane–2-pro-
panol–water mixture after each injection.

3.3 UHPSFC/ESI-MS 1. Prepare a mixture of (99:1, v/v) methanol–water as a makeup

Analysis liquid, flush a makeup pump, and set up the flow rate of
0.25 mL/min.
2. Calibrate the MS instrument using a sodium formate solution
prepared by mixing 10 μL 1 M NaOH, 20 μL formic acid, and
20 mL (1:1, v/v) water–acetonitrile.
3. Connect the UHPSFC instrument with the mass spectrometer
via the commercial interface kit (Waters) composed of two
T-pieces enabling the backpressure control and mixing of col-
umn effluent with a makeup liquid.
4. Collect MS data in the centroid high-resolution mode for
6 min using 0.15 s scan speed in the mass range m/z
50–1600 (see Note 3) with the following setting of tuning
parameters: capillary voltages 3.0 kV and 2.5 kV for positive-
ion and negative-ion modes (see Note 4), respectively, the
sampling cone 20 V, the source offset 90 V, the source temper-
ature 150
C, the drying temperature 500
C, the cone gas
flow 0.8 L/min, the drying gas flow 17 L/min, and the neb-
ulizer gas flow 4 bar. Use 10 μL/min of leucine enkephalin as
the lock mass for all experiments with the acquisition of lock
mass spectra each 30 s for 0.1 s.
5. Identify lipid classes and individual lipid species (see Note 5)
based on their retention times (Fig. 2) (see Note 6), observed
ions (see Note 7), and characteristic fragments (see Note 8) in
ESI mass spectra.

3.4 UHPSFC/ESI-MS 1. Perform UHPLC/ESI-MS analysis of prepared lipidomic

Quantitation of Lipids extracts.
2. Process data using MarkerLynx XS software (see Note 9), i.e.,
combine together individual ESI scans corresponding to lipid
 UHPSFC/ESI-MS Analysis of Lipids 77

Fig. 1 Positive-ion UHPSFC/ESI-MS chromatograms of the mixture of lipid class standards (a) and a total lipid
extract of porcine brain (b). UHPSFC conditions: Acquity BEH UPC2 column (100  3 mm, 1.7 μm, Waters),
the flow rate 1.9 mL/min, the column temperature 60
C, the ABPR pressure 1800 psi, and the gradient of
methanol–water mixture (99:1, v/v) containing 30 mM of ammonium acetate as the modifier: 0 min, 1%,
5 min, 51%, 6 min, 51%. Peak annotation: CE, cholesteryl esters; TG, triacylglycerols; FA, fatty acids; DG,
diacylglycerols; MG, monoacylglycerols; DHEA, dehydroepiandrosterone; Cer, ceramides; GlcCer, glucosyl-
ceramides; HexCer, hexosylceramides; PG, phosphatidylglycerols; LacCer, lactosylceramides; pPE, 1-alke-
nyl-2-acyl phosphatidylethanolamines (plasmalogens); ePE, 1-alkyl-2-acyl phosphatidylethanolamines
(ethers); PE, phosphatidylethanolamines; LPG, lysophosphatidylglycerols; PI, phosphatidylinositols; LPE,
lysophosphatidylethanolamines; CL, cardiolipins; LPI, lysophosphatidylinositols; PA, phosphatidic acids;
PC, phosphatidylcholines; pPC, 1-alkenyl-2-acyl phosphatidylcholines; ePC, 1-alkyl-2-acyl phosphatidyl-
cholines; PS, phosphatidylserines; LPA, lysophosphatidic acids; S1P, sphingosine-1-phosphate; SM, sphin-
gomyelins; LPC, lysophosphatidylcholines; LPS, lysophosphatidylserines. Reprinted with permission from
(Lı́sa M, Holčapek M (2015). Analytical Chemistry 87 (14):7187–7195). Copyright (2015) American Chemical
Society
Fig. 2 Effects of DB number and fatty acyl chain length on the retention behavior of lipids using UHPSFC/ESI-
MS. Reconstructed ion chromatograms from the analysis of porcine brain extract: (a) TG with 54 carbon atoms
and the different number of DB, (b) TG with 2 DB and the different number of carbon atoms, and (c) PC with
1 DB and the different number of carbon atoms. Reprinted with permission from (Lı́sa M, Holčapek M (2015).
Analytical Chemistry 87 (14):7187–7195). Copyright (2015) American Chemical Society
 UHPSFC/ESI-MS Analysis of Lipids 79

class peaks with the peak separation set to 50 mDa and an

intensity threshold of 3000 for markers extraction.
3. Detect lipid species within extracted markers according to
accurate m/z values, and perform isotopic correction of ion
intensities (see Note 10).
4. Calculate concentrations of lipids from corrected ion intensities
related to the intensity of lipid class IS.

4 Notes

1. A properly conditioned column is necessary before the first

injection to achieve appropriate and stable retention times.
Stable column temperature and system pressure have to be
achieved as well. The UHPSFC/ESI-MS method provides an
excellent intraday stability of retention times in the
range  0.01 min. A small continuous reduction of retention
times of mainly polar lipids can be observed during multiple
consecutive days. This phenomenon of small retention shift on
SFC columns is probably given by the formation of silyl ethers
on the particle surface with alcohols from the mobile phase
leading to the decrease of hydrophilicity [7], but it does not
affect the identification and quantitation of lipids. The process
is reversible and silanols can be hydroxylated again using water.
For this reason, the column is regenerated daily by five consec-
utive injections of 10 μL of water and acetonitrile into 100% of
CO2 at 0.5 mL/min flow rate or occasionally is flushed with a
(1:1, v/v) water–acetonitrile mixture, which minimizes this
effect.
2. In this UHPSFC method, lipids are mainly separated according
to their polarity into lipid classes. Retention times increase with
the increased lipid polarity. In total, up to 30 lipid classes from
six different lipid categories covering a wide range of nonpolar
and polar lipids are separated within 6 min analysis. Under
optimized UHPSFC conditions, most of the lipid classes are
baseline or at least partially separated. Similar to LC separa-
tions, the strong peak tailing is observed for UHPSFC of acidic
lipids, such as PA, PS, LPA, S1P, and LPS, which complicates
their precise quantitation.
3. The employed m/z range should be defined by the molecular
weights of lipid molecules analyzed.
4. Polarity of acquisition depends on the monitored lipid classes.
Most of the lipid classes provide appropriate signal in the
positive-ion mode. The combination of positive- and
negative-ion modes is used for the identification of lipids with
a higher confidence.
80 Miroslav Lı́sa and Michal Holčapek

5. Retention times from the UHPSFC analysis and ESI mass

spectra measured with high mass accuracy and high resolving
power in both positive- and negative-ion modes are used for
the unambiguous identification of individual lipid species. The
total lipid extract is separated into lipid classes using UHPSFC
enabling the direct identification of lipid class based on the
comparison of retention times with standards. And then, the
averaged mass spectrum of a lipid class chromatographic peak is
used for the determination at lipid species level, i.e., the num-
ber of carbon atoms and DB of attached fatty acyl/alkyls. The
partial separation of individual species within lipid classes
according to fatty acyl lengths and the number of DB provide
another supporting information for their identification. The
fatty acyl composition of individual species can be determined
using [RCOO] ions in negative-ion MS/MS spectra [8].
6. In addition to the class separation, lipid species within individ-
ual lipid classes are partially separated according to the fatty acyl
composition (Fig. 2). Retention times of all lipids increase with
increased DB number, as demonstrated on reconstructed ion
chromatograms of TG with 54 carbon atoms and different
number of DB (Fig. 2a). A different situation is observed for
species differing in fatty acyl lengths, where the retention
behavior of lipids differs according to the individual lipid clas-
ses. Retention times of TG increase with the fatty acyl length
(Fig. 2b), which is observed also for other nonpolar lipid
classes, such as CE, FA, DG, MG, fatty amides, and Cer. On
the other hand, retention times of polar lipids decrease for
longer fatty acyls, which can be demonstrated on the separation
of PC species (Fig. 2c).
7. The most abundant ions in positive-ion UHPSFC/ESI-MS
full-scan mass spectra are protonated molecules [MþH]+
(base peaks for fatty amides, sphinganine, LacCer, PE, LPE,
CL, PC, SM, and LPC), adducts with ammonium ion
[MþNH4]+ (TG and coenzyme Q10) and neutral losses of
water [MþH-H2O]+ (DG, cholesterol, MG, Cer, sphingosine,
and LacCer), attached fatty acyls [MþH-acyl]+ (CE) or phos-
phoglycerol [MþH- H2PO4CH2CHOHCH2OH]+ (PG and
LPG), and sulfo [MþH-SO3]+ (sulfatides) groups. UHPSFC/
ESI-MS mass spectra provide relatively low abundance of
sodium adduct ions, which reduces the risk of incorrect identi-
fication between [MþH]+ and [MþNa]+ ions, because the
difference Δm/z 22 also corresponds to an additional two
methylene units minus 3 DB. The data are also correlated
with negative-ion mass spectra and retention times of lipid
species within lipid classes to unambiguously confirm the iden-
tification of all reported lipids.
 UHPSFC/ESI-MS Analysis of Lipids 81

8. Positive-ion MS/MS spectra of identified classes provide well-

known characteristic fragment ions and neutral losses observed
in HPLC/MS, such as the phosphocholine fragment ion m/z
184 ([H2PO4CH2CH2N(CH3)3]+) for moieties containing
choline (PC, LPC, and SM), m/z369 ([MþH-H2O]+ or
[MþH-acyl]+) for cholesterol-containing lipids (cholesterol
and CE), or fragment ions corresponding to ceramide bases
(Cer, GlcCer, and LacCer). The neutral loss of phosphoetha-
nolamine Δm/z ¼ 141 (H2PO4CH2CH2NH2) is observed for
PE and neutral losses of fatty acyls for TG, DG, and MG. In the
negative-ion mode, base peaks of spectra are mostly deproto-
nated molecules [MH], except for DG, PC, SM, and LPC,
providing mainly adduct ions with acetate [MþCH3COO].
Relatively high abundances of [MþCH3COO] ions are also
observed for Cer, GlcCer, and LacCer (70–95%). CL species
exhibit [M2H]2 ions, which can be used for their identifica-
tion even in the lower mass range, such as m/z ¼ 50–1000
usually used in lipidomic analyses. Negative-ion MS/MS spec-
tra show mainly [RCOO] ions corresponding to the fatty
acyl/alkyl composition.
9. Ion intensities on ESI full-scan mass spectra of lipid class peaks
are used for the data evaluation of UHPSFC/ESI-MS
experiments.
10. A contribution of Mþ2 isotope intensity of species with one
more double bond (Δm/z ¼ 2) has to be subtracted to achieve
more accurate quantitative results. We use our Excel macro
script LipidQuant for detection of lipid species based on
accurate m/z values, isotopic correction, and calculation of
concentrations.

Acknowledgments

This work was supported by the grant project LL1302 sponsored

by the Ministry of Education, Youth and Sports of the Czech
Republic.

References
1. Nováková L, Perrenoud AG-G, Francois I, Anal Chem 61:192–206. https://doi.org/10.
West C, Lesellier E, Guillarme D (2014) Modern 1016/j.trac.2014.04.017
analytical supercritical fluid chromatography 3. Lesellier E, Gaudin K, Chaminade P, Tchapla A,
using columns packed with sub-2 micrometer Baillet A (2003) Isolation of ceramide fractions
particles: a tutorial. Anal Chim Acta 824:18–35. from skin sample by subcritical chromatography
https://doi.org/10.1016/j.aca.2014.03.034 with packed silica and evaporative light scattering
2. Čajka T, Fiehn O (2014) Comprehensive analy- detection. J Chromatogr A 1016(1):111–121.
sis of lipids in biological systems by liquid https://doi.org/10.1016/s0021-9673(03)
chromatography-mass spectrometry. Trends 01323-2
82 Miroslav Lı́sa and Michal Holčapek

4. Lı́sa M, Holčapek M (2015) High-throughput ionization tandem mass spectrometry. J Lipid

and comprehensive lipidomic analysis using Res 50(3):574–585. https://doi.org/10.
ultrahigh-performance supercritical fluid 1194/jlr.D800028-JLR200
chromatography-mass spectrometry. Anal 7. Fairchild JN, Brousmiche DW, Hill JF, Morris
Chem 87(14):7187–7195. https://doi.org/ MF, Boissel CA, Wyndham KD (2015) Chro-
10.1021/acs.analchem.5b01054 matographic evidence of Silyl ether formation
5. Schuhmann K, Almeida R, Baumert M, (SEF) in supercritical fluid chromatography.
Herzog R, Bornstein SR, Shevchenko A (2012) Anal Chem. https://doi.org/10.1021/
Shotgun lipidomics on a LTQ Orbitrap mass ac5035709
spectrometer by successive switching between 8. Lı́sa M, Cı́fková E, Holčapek M (2011) Lipido-
acquisition polarity modes. J Mass Spectrom 47 mic profiling of biological tissues using off-line
(1):96–104. https://doi.org/10.1002/jms. two-dimensional high-performance liquid chro-
2031 matography mass spectrometry. J Chromatogr A
6. Wiesner P, Leidl K, Boettcher A, Schmitz G, 1218(31):5146–5156. https://doi.org/10.
Liebisch G (2009) Lipid profiling of FPLC- 1016/j.chroma.2011.05.081
separated lipoprotein fractions by electrospray
 Chapter 6

LC-MS/MS Analysis of Lipid Oxidation Products in Blood

and Tissue Samples
Yiu Yiu Lee and Jetty Chung-Yung Lee

Abstract
Oxygenated lipid products of non-cyclooxygenase derivatives, namely, prostanoids such as, isoprostanes
and isofurans, are formed in vivo through lipid autoxidation. Insofar it has been marked as novel biomarkers
of oxidative stress in the biological systems. Elevations of these oxidized products are associated with several
diseases. This chapter describes the preparation and measurement of the products, including newly
identified F2-dihomo-isoprostanes and dihomo-isofurans, from plasma and tissue samples using the liquid
chromatography-tandem mass spectrometry approach.

Key words Lipid peroxidation products, Isoprostanes, Isofurans, Liquid chromatography-tandem

mass spectrometry

1 Introduction

Oxidative stress is highly related to various metabolic diseases

associated to cardiovascular, neurodegenerative, pulmonary disor-
ders, and cancer. The measurement of eicosanoid-based oxidized
lipid products is claimed to be the most reliable indicator of oxida-
tive damage in diseases [1–4]. These oxidized lipid products are
generated nonenzymatically and typically initiated by the action of
free radicals and reactive oxygen species on polyunsaturated fatty
acids (PUFA) and in certain diseases catalyzed by elements such as
iron. Among the oxidized PUFA products, F2-isoprostanes
(F2-IsoPs) from arachidonic acid are the most commonly measured
due to their specificity and sensitivity [5]. These compounds are
non-cyclooxygenase derivatives of prostanoids that are formed in
situ, esterified to phospholipid, and released as free form into
circulation via phospholipase A2-catalyzed hydrolysis [6, 7].
Depending on the parent PUFA, different types of IsoPs have
been discovered and quantified in numerous pathological condi-
tions [8]. These include F2-IsoPs and isofurans (IsoFs) from ara-
chidonic acid, F2-dihomo-IsoPs and dihomo-IsoFs from adrenic

Martin Giera (ed.), Clinical Metabolomics: Methods and Protocols, Methods in Molecular Biology, vol. 1730,
https://doi.org/10.1007/978-1-4939-7592-1_6, © Springer Science+Business Media, LLC 2018

83
84 Yiu Yiu Lee and Jetty Chung-Yung Lee

acid, and F4-neuroprostanes (NeuroPs) and neurofurans (Neu-

roFs) from docosahexaenoic acid [8]. Among them, 15-F2t-IsoP
is the most represented and measured IsoP [9] and has been
corroborated by the European Food Safety Authority (EFSA) as
biomarkers for oxidative damage in cardiovascular health
[10]. Besides being oxidative stress biomarkers, some autoxidized
products have been reported to be bioactive, but the research
progression in their bioactivities has been slow, partly due to the
lack of chemical standards available in the commercial market.
The quantitation of oxidized PUFA products originates from
the measurement of 15-F2t-IsoP using gas chromatography-mass
spectrometry (GC-MS) in negative chemical ionization (NCI). The
merit of GC-NCI-MS over many other analytical methods is owed
to its superior sensitivity and is recognized as the “gold standard” in
targeted lipid profiling (see preceding chapter by Barden and Mori).
However, isomeric compounds or different compounds with the
exact mass and elution time cannot be effectively differentiated in
GC-NCI-MS. With recent advancements in MS, the use of
LC-MS/MS in multiple reaction monitoring (MRM) mode has
received great attention for the ease of sample preparation, rela-
tively lesser run time, and high selectivity which enables the separa-
tion of isomeric compounds.
In this chapter, the preparation procedure and analysis of oxi-
dized PUFA products in plasma and tissue samples for LC-MS/MS
measurement are described. The method is applicable for the mea-
surement of all IsoPs and IsoFs; however, we will focus on the
measurement of F2-dihomo-IsoPs and dihomo-IsoFs. These are
newly identified oxidized lipid products and potentially a robust
biomarker for neurological disorders such as Rett syndrome and
epilepsy [10–12].

2 Materials

Prepare all reagents using ultrapure water. All solvents should be at

least high-performance liquid chromatography (HPLC) grade. All
solvents and buffers prepared should be stored at room tempera-
ture, unless otherwise stated. It is not encouraged to use sodium
azide as buffer preservative for the potential of ion suppression in
the LC-MS/MS measurement.

2.1 Lipid Extraction 1. Ethylenediaminetetraacetic acid (EDTA)-primed syringe:

expose the syringe inner wall with 5 M EDTA (pH 8.0) and
subsequently empty the syringe before blood withdrawal.
2. Phosphate-buffered saline (PBS): dissolve 8 g sodium chloride
(NaCl), 0.2 g KCl, 1.44 g Na2PO4, and 0.24 g KH2PO4 in
800 mL of ultrapure water. Then adjust to pH 7.4 with HCl
and top up to 1 L.
 LC-MS/MS Analysis of Lipid Autoxidation Products in Biological Samples 85

3. 20 ml EPA glass vials with PTFE septa screw cap (VWR).

4. 1 M methanolic potassium hydroxide (KOH): dissolve 5.6 g of
KOH pellets per 100 mL of absolute methanol.
5. 5% (w/v) butylated hydroxytoluene (BHT): dissolve 5 g of
BHT per 100 mL of absolute ethanol.
6. 1% (w/v) indomethacin: dissolve 1 g of indomethacin per
100 mL of ethanol.
7. Folch solution: mix chloroform and methanol in a ratio of 2:1
and spike with 0.005% (w/v) BHT (i.e., 1 mL 5% BHT in 1 L
Folch solution). Store at
20  C.
8. A cocktail of internal standards (ISTD): accurately dilute con-
centrated stock ISTD (e.g., 15-F2t-IsoP-d4, 4-F4t-NeuroP-d4,
10-F4t-NeuroP-d4) to 1 ng/μL in absolute methanol. Store at

80  C.
9. 0.9% NaCl: dissolve 0.9 g of NaCl per 100 mL ultrapure water.
10. 0.45 μm PTFE syringe filters (13 mm diameter) (Membrane
Solutions).

2.2 Solid-Phase 1. 5 M HCl: add 41.059 mL of 37% stock HCl to ultrapure water
Extraction (SPE) per 100 mL final volume.
2. 40 mM formic acid: add 0.159 mL of stock formic acid to
ultrapure water per 100 mL final volume.
3. 20 mM formic acid: 1:1 dilute from 40 mM formic acid with
ultrapure water.
4. Mixed-mode reversed-phase/strong anion-exchange (MAX)
cartridges (Oasis).
5. Visiprep™ SPE Vacuum Manifold (Supelco).
6. 2% ammonium hydroxide (NH4OH): add 3.5 mL of 56.6%
stock NH4OH per 100 mL final volume of ultrapure water.
7. Elution solvent: for every 100 mL, add 70 mL of n-hexane,
29.4 mL of ethanol, and 0.4 mL of acetic acid. Store at room
temperature in amber bottle.

2.3 LC-ESI-MS/MS 1. LC solvent A: add 1 mL of formic acid to ultrapure water per

Analysis 1 L final volume.
2. LC solvent B: add 1 mL of formic acid to absolute acetonitrile
per 1 L final volume.
3. Kinetex reverse phase C18 column (2.6 μ, 100 Å,
150  2.1 mm) (Phenomenex).
4. Ultrahigh-performance liquid chromatography (UHPLC) sys-
tem (Agilent) (1290 Infinity).
5. Hybrid triple quadrupole/linear ion trap (QqQLIT) mass
spectrometer (MS) (AB Sciex (QTRAP 3200)).
86 Yiu Yiu Lee and Jetty Chung-Yung Lee

3 Methods

3.1 Extraction 1. Use 22 G needle with a 10 mL EDTA-primed syringe tube for

of Lipid Autoxidized blood collection in rats via cardiac puncture, followed by
Products from Blood immediate euthanization (see Note 1).
and Tissue Samples 2. Transfer the collected blood into an EDTA-primed polypro-
3.1.1 Blood Samples
pylene Falcon™ tube and gently shake the tube with hand (see
Note 1). Separate the plasma (top layer) from whole blood by
centrifugation at 3000  g for 15 min, 4  C.
3. Following centrifugation, dispense 1 mL aliquots of blood
plasma into prechilled microcentrifuge tubes and spike with
20 μL of 5% (w/v) BHT and 1% (w/v) indomethacin (see
Note 2). Briefly vortex the samples and proceed to hydrolysis
step immediately or store at
80  C until analysis (see Note 3).
4. On the day of analysis, thaw 1 mL (one vial) of frozen plasma
on ice and transfer to a clean 20 mL glass vial containing 1 mL
of 1 M methanolic KOH and add 50 μL of IS cocktail. Mix the
plasma mixture by aspiration, and incubate in 50  C oven
for 1 h.

3.1.2 Tissue Samples 1. Briefly rinse the freshly harvested organs with ice-cold PBS
containing BHT (0.005%, w/v) and (0.005%, w/v) indometh-
acin (see Note 2). Cleaned tissue samples not analyzed imme-
diately should be snap frozen with liquid nitrogen and stored at

80  C (see Note 3).
2. On the day of tissue lipid extraction, excise a portion of tissues
from the frozen organ, and grind into powder using mortar
and pestle in the presence of liquid nitrogen. Weigh about
0.2–0.5 g of tissue powder into a clean 15 mL Falcon™ tube
before it becomes moist, and add 10 mL of ice-cold Folch
solution and 50 μL of ISTD cocktail (see Note 4).
3. Homogenize the tissue using a Polytron benchtop homoge-
nizer at 24,000 rpm for four bursts of 10 s, while placing the
Falcon tube on ice. The homogenizer probe should pass
through multiple washes of ultrapure water and methanol to
prevent cross contamination between samples.
4. Firmly screw cap the Falcon™ tubes and place them on ice in a
rocking motion for at least 30 min, and vortex every 10 min to
enhance extraction efficiency.
5. Introduce phase separation by adding 2 mL of 0.9% NaCl to
the homogenate, shake vigorously, and centrifuge at 3000  g
for 10 min, 4  C.
6. Transfer the lower phase using a glass pipette and filter through
a 0.45 μm PTFE filter membrane into a new 20 mL glass vial
(see Note 5).
 LC-MS/MS Analysis of Lipid Autoxidation Products in Biological Samples 87

7. Further add 6 mL of chloroform to the remaining homoge-

nates, vortex, and repeat the centrifugation and filtration steps
(steps 5 and 6).
8. Pool the lower phases together and dry under a gentle stream
of nitrogen, on a heat block set to 37  C (see Note 6).
9. To the dried lipid extracts, add 1 mL of 1 M methanolic KOH,
and sonicate until all extracts dissociate from the bottom of the
vial. Then, add 1 mL of PBS, mix thoroughly by aspiration,
flush the vial with nitrogen, and incubate in darkness at room
temperature overnight (see Note 7).

3.2 Solid-Phase 1. Ensure the sample is cooled to room temperature, and add
Extraction (SPE) 0.5 mL methanol, 0.2 mL 5 M HCl, 2.7 mL 40 mM formic
acid, and 4 mL 20 mM formic acid, in succession, vortex, and
transfer to a new 15 mL Falcon™ tube. Centrifuge at 3000  g
for 10 min to remove any protein precipitates (see Note 8).
2. Mount Oasis MAX 3 cc extraction cartridges onto a SPE
Vacuum Manifold and precondition with 2 mL of methanol
and 2 mL of 20 mM formic acid.
3. Load the acidified samples and allow passing the solution at a
flow of about one drop per second to maximize the interaction
of analytes with the cartridge sorbents.
4. Wash the unwanted and weakly bound compounds by adding
2 mL of 2% NH4OH, followed by 2 mL of 20 mM formic acid.
Dry the cartridges by increasing air flow for at least 5 min to
remove all water content (see Note 9).
5. Elute all lipid analytes by passing through 2 mL of n-hexane
and twice with 2 mL of elution solvent into a clean 20 mL glass
vial. Dry with nitrogen gas to approximately 1 mL of volume
before transferring to the autosampler vial.
6. Continue drying the SPE extracts in the autosampler vials until
complete dryness and reconstitute with 200 μL of acetonitrile
(see Note 10).
7. Allow at least 5 min of incubation on bench in the dark before
transferring to autosampler spring insert (see Note 11). Firmly
screw the cap and store at 4  C until ready for analysis within
the day.

3.3 LC-ESI-MS/MS 1. Prepare solvent A (0.1% formic acid in ultrapure water) and B
(0.1% formic acid in acetonitrile), filtered and degassed and
store at room temperature.
2. Clean the LC C18 RP column by flushing 98% solvent B at
0.2 mL/min until the bin pump pressure is stabilized. Then,
equilibrate the column for at least 30 min with 20% solvent A
immediately before the start of the analysis.
88 Yiu Yiu Lee and Jetty Chung-Yung Lee

3. The elution gradient program begins with 20% solvent B for

2 min and ramp to 98% solvent B over an 8 min period and hold
for 5 min. Then reduce solvent B to 20% for another 5 min to
equilibrate for the next run. Set the flow rate to 0.3 mL/min
and the injection volume to 10 μL. Fit the C18 column in the
column oven and maintain at 30  C.
4. QTRAP 3200 hybrid triple quadrupole-linear ion trap mass
spectrometry is equipped with a Turbo V ESI source operated
in negative (
) mode with the following settings: curtain gas
(CUR) ¼ 10 psi, ion spray voltage (IS) ¼
4500 V,
CAD ¼ HIGH, temperature (TEM) ¼ 500  C, ion source
gas 1 (GS1, nebulizer) ¼ 30 psi (nitrogen), GS2
(heater) ¼ 30 psi (nitrogen), interface heater (ihe) ¼ on,
CEP ¼
10 V, and CXP ¼
10 V.
5. To achieve maximum selectivity, QTRAP 3200 MS is operated
in scheduled multiple reaction monitoring mode (MRM) with a
retention time window of 1 min. The transitions of each analyte
are chosen by infusing their pure standards into MS and gener-
ate MS/MS spectra from which the values are selected and
optimized. Optimal declustering potential (DP) and collision
energy (CE) of each transition are listed in Table 1. For illus-
tration, only F2-dihomo-IsoPs and dihomo-IsoFs that were
first characterized by our laboratory are shown (Fig. 1).
6. The amount of endogenous autoxidation products is then
calculated from normalized response using relative response
factors (RFFs) with respect to internal standards. The RFFs
are usually predetermined using peak areas and masses of deut-
erated standards versus analyte standards. This is often deter-
mined by the construction of a calibration line with at least five
varying analyte concentrations (typically 0.01–1 ng/μL), each
having a fixed concentration of internal standard (see Note 12).
The calibration line slope plotting Areaanalyte/AreaIS against
Massanalyte/MassIS represents the RRF.
7. The final concentration of an autoxidized product from the
sample is then calculated as (Areaanalyte  MassIS)/
(AreaIS  RRF) (see Note 13).

4 Notes

1. Blood drawn into the syringe must be accompanied with

appropriate needle size to minimize rupturing of the red
blood cells that can cause hemolysis. Plasma contaminated
with hemolyzed red blood cells will lead to IsoPs artifacts
[13]. Similarly, the needle must be removed before the transfer
 LC-MS/MS Analysis of Lipid Autoxidation Products in Biological Samples 89

Table 1
The availability of some isomeric autoxidized products currently in the market

Transition
Autoxidized products (m/z) DP (eV) CE (eV) ISTD Supplier
Arachidonic acid
5-F2t-IsoP 353 ! 115
50
25 5-F2t-IsoP-d4 Cayman
8-F2t-IsoP 353 ! 127
50
25 15-F2t-IsoP-d4 Cayman
12-F2t-IsoP 353 ! 151
50
25 15-F2t-IsoP-d4 NA
15-F2t-IsoP 353 ! 193
50
25 15-F2t-IsoP-d4 Cayman
Isofurans 369 ! 115
45
20 15-F2t-IsoP-d4 Na
Adrenic acid
7-F2t-Dihomo-IsoP 381 ! 143
35
34 15-F2t-IsoP-d4 Durand
17-F2t-Dihomo-IsoP 381 ! 237
65
22 15-F2t-IsoP-d4 Durand
8
7(RS)-ST-Δ -11-Dihomo-IsoF 397 ! 201
60
32 15-F2t-IsoP-d4 Durand
17(RS)-10-epi-SC-Δ15- 397 ! 155
70
30 15-F2t-IsoP-d4 Durand
11-Dihomo-IsoF
Eicosapentaenoic acid
5-F3t-IsoP 351 ! 115
50
25 15-F2t-IsoP-d4 Durand
8-F3t-IsoP 351 ! 127
55
35 15-F2t-IsoP-d4 Durand
11-F3t-IsoP 351 ! 167
50
25 15-F2t-IsoP-d4 NA
12-F3t-IsoP 351 ! 151
50
25 15-F2t-IsoP-d4 NA
15-F3t-IsoP 351 ! 193
50
25 15-F2t-IsoP-d4 Cayman
18-F3t-IsoP 351 ! 233
50
25 15-F2t-IsoP-d4 NA
Docosahexaenoic acid
4-F4t-NeuroP 377 ! 101
60
30 4-F4t-NeuroP-d4 Durand
7-F4t-NeuroP 377 ! 141
60
30 4-F4t-NeuroP-d4 NA
10-F4t-NeuroP 377 ! 153
55
40 10-F4t-NeuroP-d4 Durand
11-F4t-NeuroP 377 ! 178
60
30 4-F4t-NeuroP-d4 NA
13-F4t-NeuroP 377 ! 221
60
30 4-F4t-NeuroP-d4 NA
14-F4t-NeuroP 377 ! 138
60
30 4-F4t-NeuroP-d4 Durand
17-F4t-NeuroP 377 ! 98
60
30 4-F4t-NeuroP-d4 NA
20-F4t-NeuroP 377 ! 58
60
30 4-F4t-NeuroP-d4 NA
5
4(RS)-ST-Δ -8-NeuroF 393 ! 325
55
32 4-F4t-NeuroP-d4 Durand
Neurofurans 393 ! 193
35
20 4-F4t-NeuroP-d4 NA
(continued)
90 Yiu Yiu Lee and Jetty Chung-Yung Lee

Table 1
(continued)

Transition
Autoxidized products (m/z) DP (eV) CE (eV) ISTD Supplier
ISTD
5-F2t-IsoP-d4 357 ! 115
50
25 – Cayman
15-F2t-IsoP-d4 357 ! 197
50
25 – Cayman
4-F4t-IsoP-d4 381 ! 101
60
30 – Durand
10-F4t-IsoP-d4 381 ! 157
55
40 – Durand
DP declustering potential, CE collision energy, ISTD internal standard, NA not available, Cayman Chemicals, USA,
IBMM Dr. Thierry Durand, Institut des Biomolécules Max Mousseron, Montpellier, France

100 6.6
143
HO OH
COO
7(RS)-7-F2t-Dihomo-IsoP m/z 381 → 143
HO
0

100 6.6
HO 237
COO
17(RS)-17-F2t-Dihomo-IsoP m/z 381 → 237
Relative Abundance,%

HO OH

100 6.6 OH 201

COO
7(RS)-ST-Δ8-11-Dihomo-IsoF O OH m/z 397 → 201
HO

100 7.0 OH
155
COO
10-epi-17(RS)-SC-Δ15-11-Dihomo-IsoF O m/z 397 → 155
HO
OH
0

0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 11.0 12.0 13.0 14.0 15.0 16.0 17.0 18.0 19.0 20.0
Time,min

Fig. 1 Chromatograms of F2-dihomo-IsoPs and F2-dihomo-IsoFs in a typical LC-MS/MS analytical run. Dotted
red line indicates the proposed site of fragmentation

of the blood into the 15 mL Falcon™ tube as failure to remove

the needle may induce severe hemolysis.
2. It is known that improper handling of blood and tissue samples
can lead to artificial increase of IsoPs and IsoFs; therefore, it is
important to treat the samples appropriately using anti-
cyclooxygenase (indomethacin) and antioxidants (BHT)
before storing away.
LC-MS/MS Analysis of Lipid Autoxidation Products in Biological Samples 91

3. For long-term storage, samples are recommended to be kept at

80  C as higher temperature, like
20  C, has shown to
induce significant lipid autoxidation.
4. Wear personal protective equipment when handling liquid
nitrogen. To avoid the tissue powder from building up mois-
ture, carefully add in more liquid nitrogen and grind until dry.
5. Sample matrices filtered through PTFE membranes eliminate
the inclusion of sample tissues and allow easier sample cleanups
at later stages of sample preparation.
6. Adding heat to the glass vial while drying enhances the rate of
solvent volatility. A higher temperature, e.g., 40  C, may be
applied to the glass vial when large volume of solvent is present.
It is essential to make sure that the vial is completely dry before
further processing.
7. Hydrolysis of lipid compounds is achievable by placing the vials
at 50–60  C for a period of 30 min to 2 h. It is found that the
best hydrolyzing method for tissues is to incubate the sample
overnight at room temperature to allow longer time for the
release of the products embedded in the tissue matrix.
8. Acidification of sample is essential to unionize the analytes to
maximize the reversed-phase retention of the MAX SPE car-
tridge. Also, note that it is necessary to perform centrifugation
prior to loading onto the SPE cartridge to prevent clots.
9. It is advisable to remove all remaining small amount of water
after washes as failure to do so may lead to the elution of
contaminants in the subsequent steps.
10. Complete dryness of the lipid extracts is critical, or the final
analyte concentrations may be affected.
11. In some instances, precipitates may form during mobile phase
resuspension. Passing through a small diameter PTFE mem-
brane (low holdup volume) can help to resolve this problem
and prevent clotting of HPLC needle and column.
12. It is advisable to use the deuterated form of the exact analyte as
the respective internal standard, i.e., 4-F4-NeuroP-d4 as the IS
for 4-F4-NeuroP detection and quantitation. However, the
availability of deuterated IS is lacking in the commercial market
for many compounds; therefore, it is common to use the most
available one, 15-F2t-IsoP-d4, as the IS.
13. The calculated concentration represents only single injection
and is subjected to technical repeats (n ¼ 3). Also, the calcu-
lated concentration requires normalization to the total volume
of lipid extract and the weight of tissue samples used when
expressing the amount of analyte per gram of tissue.
92 Yiu Yiu Lee and Jetty Chung-Yung Lee

References
1. Mallat Z, Nakamura T, Ohan J et al (1999) The 8. Galano JM, Lee JCY, Gladine C et al (2015)
relationship of hydroxyeicosatetraenoic acids Non-enzymatic cyclic oxygenated metabolites
and F2-isoprostanes to plaque instability in of adrenic, docosahexaenoic, eicosapentaenoic
human carotid atherosclerosis. J Clin Invest and a-linolenic acids; bioactivities and potential
103:421–427 use as biomarkers. Biochim Biophys Acta
2. Practico D, Lee VMY, Trojanowski JQ et al 1851:446–455
(1998) Increased F2-isoprostanes in Alzhei- 9. Kadiiska MB, Gladen BC, Baird DD et al
mer’s disease: evidence for enhanced lipid per- (2005) Biomarkers of oxidative stress study
oxidation in vivo. FASEB J 12:1777–1783 II. Are oxidation products of lipids, proteins,
3. Lee CYJ, Seet RCS, Huang SH et al (2009) and DNA markers of CCl4 poisoning? Free
Different patterns of oxidized lipid products in Radic Biol Med 38:698–710
plasma and urine of dengue fever, stroke, and 10. De Felice C, Signorini C, Durand T et al
Parkinson’s disease patients: cautions in the use (2011) F2-dihomo-isoprostanes as potential
of biomarkers of oxidative stress. Antioxid early biomarkers of lipid oxidative damage in
Redox Signal 11:407–420 Rett syndrome. J Lipid Res 52:2287–2297
4. Basu S, Harris H, Wolk A et al (2016) Inflam- 11. Medina S, De Miguel-Elı́zaga I, Oger C et al
matory F2-isoprostane, prostaglandin- F2α, (2015) Dihomo-isoprostanes—nonenzymatic
pentraxin 3 levels and breast cancer risk: The metabolites of AdA – are higher in epileptic
Swedish Mamography Cohort. Prostaglandins patients compared to healthy individuals by a
Leukot Essent Fatty Acids 113:28–32 new ultrahigh pressure liquid chromatogra-
5. Milne GL, Gao B, Terry ES et al (2013) Mea- phy–triple quadrupole–tandem mass spectrom-
surement of F2-isoprostanes and isofurans etry method. Free Radic Biol Med 79:154–163
using gas chromatography-mass spectrometry. 12. De La Torre A, Lee YY, Oger C et al (2014)
Free Radic Biol Med 59:36–44 Synthesis, discovery, and quantitation of
6. Morrow JD, Awad JA, Boss HJ et al (1992) dihomo-isofurans: biomarkers for in vivo adre-
Non-cyclooxygenase-derived prostanoids (F2- nic acid peroxidation. Angew Chem Int Ed
isoprostanes) are formed in situ on phospholi- 53:6249–6252
pids. Proc Natl Acad Sci U S A 13. Gruber J, Tang SY, Jenner AM et al (2009)
89:10721–10725 Allantoin in human plasma, serum, and nasal-
7. Awad JA, Morrow JD, Takahashi K et al (1993) lining fluids as a biomarker of oxidative stress:
Identification of non-cyclooxygenase-derived avoiding artifacts and extasblishing real in vivo
prostanoid (F2-Isoprostane) metabolite in concentrations. Antioxid Redox Signal
human urine and plasma. J Biol Chem 11:1767–1776
268:4161–4169
 Chapter 7

Serum Testosterone by Liquid Chromatography Tandem

Mass Spectrometry for Routine Clinical Diagnostics
Lennart J. van Winden, Olaf van Tellingen, and Huub H. van Rossum

Abstract
In clinical diagnostics, samples containing low testosterone cannot be analyzed by random access immu-
noassays normally available at clinical laboratories. For these samples, sensitive and specific LC-MS/MS-
based testosterone methods are required. An LC-MS/MS-based testosterone assay is described that was
developed and validated for routine clinical application.

Key words Testosterone, LC-MS/MS, Steroids, Hormone

1 Introduction

Testosterone is the most important male sex hormone (androgen)

in terms of potency and amounts secreted [1]. In clinical diagnos-
tics, circulating testosterone concentrations are used for the diag-
nosis and follow-up of various diseases such as hypogonadism,
polycystic ovary syndrome, and precocious or delayed puberty
[2]. Furthermore, androgen deprivation therapy aimed at suppres-
sing testosterone is the most successful first-line treatment for
advanced prostate cancer [3]. Monitoring proper androgen sup-
pression and identification of resistance to androgen deprivation
therapy requires measurement of circulating testosterone
concentrations.
In clinical laboratories, analysis of testosterone is performed
mainly by automated, random access immunoassays. The advan-
tages of these routine assays are their high throughput and low
turnaround time properties. Furthermore, they do not require
specialized machinery (HPLC and mass spectrometer systems)
and expertise [3]. However, for low concentrations of testosterone
as observed in pediatric, female and castrated male samples, these
immunoassays lack sensitivity and specificity [3]. For such samples,
application of liquid chromatography-tandem mass spectrometry

Martin Giera (ed.), Clinical Metabolomics: Methods and Protocols, Methods in Molecular Biology, vol. 1730,
https://doi.org/10.1007/978-1-4939-7592-1_7, © Springer Science+Business Media, LLC 2018

93
94 Lennart J. van Winden et al.

(LC-MS/MS)-based testosterone assays is required for accurate

testosterone analysis. In addition, recognizing the limitations of
the testosterone immunoassays, leading journals now require tes-
tosterone analysis by LC-MS when testosterone is used as a major
end point in studies [4].
Our hospital serves a relatively large population of prostate
cancer patients. For routine application of testosterone analysis by
LC-MS/MS in clinical diagnostics, we developed and validated a
liquid chromatography-tandem mass spectrometry-based testoster-
one assay. Besides the analytical characteristics, the practical aspects
of minimization of the hands-on time and minimization of poten-
tial errors were taken into account when the assay was designed.

2 Materials

2.1 Stock Solutions 1. Preparation of testosterone stock solution: dissolve 10 mg tes-

and Calibration Curve tosterone (Sigma-Aldrich, St. Louis, MO, USA) in 3.473 mL
DMSO to a concentration of approximately 10 mmol/L.
2. Preparation of internal standard (IS) working solution: dissolve
10 mg testosterone-d5 (CDN isotopes, Pointe-Claire, QC,
Canada) in 3.473 mL DMSO to a concentration of approxi-
mately 10 mmol/L.
3. For a concentration of 10 μmol/L, 10 μL stock solution is
mixed with 9990 μL methanol. Subsequently, final testoster-
one concentrations of 1, 3.3, 10, 33, 100, 333, and
1000 nmol/L are established by serial dilution in methanol
(see Notes 1 and 2).
4. The calibration curve is standardized against the NIST stan-
dard reference material (SRM) 971 (see Note 3).

2.2 Other Reagents All solutions prepared in this method are of analytical grade or
and Materials better, and ultrapure water (purified deionized water, with a sensi-
tivity of 15 MΩ/cm at 25  C) is used.
1. In total, three serum-based control levels are used. The quality
control (QC) samples are made from leftover patient materials,
and pools are aimed at levels representing (1) (chemically)
castrated men, (2) female normal range, and (3) adult male
concentrations.
2. Storage: patient serum samples are stored for up to 1 week at
4  C before analysis. QC samples are stored long term at
20  C (see Note 4).
3. Preparation of system suitability test (SST) working solution:
for a concentration of 10 μmol/L, 10 μL stock solution is
mixed with 9990 μL methanol. Subsequently, a final testoster-
one concentration of 5 nmol/L is established by dilution.
 Serum Testosterone for Routine Diagnostics 95

4. Extraction solvent: methyl tert-butyl ether (MTBE) is used to

extract steroids from human serum.
5. Injection solution: methanol and water are mixed to a 7:3 ratio
(see Note 5).

2.3 Liquid 1. HPLC system: liquid chromatography is performed on a Dio-

Chromatography- nex Ultimate 3000 HPLC system consisting of a vacuum
Tandem Mass degasser, pump, and an autosampler or equivalent. The auto-
Spectrometry System sampler needle can pierce the cap of plastic microvials (see Note
6) and has an injection volume range of 1–100 μL.
2. Mobile phase: the mobile phase is composed of water (contain-
ing 0.1% formic acid) (hereafter referred to as mobile phase A)
and methanol (hereafter referred to as mobile phase B).
3. Separation column: reversed-phase C-18 column (5 μm parti-
cle size, 2.1  100 m internal diameter, Agilent Zorbax Extend
C-18) (see Note 7).
4. Ion source: Turbo V electrospray ionization (ESI).
5. Mass spectrometer: API4000 (AB Sciex, Concord, ON,
Canada) triple quadrupole mass spectrometer or equivalent or
higher-end MS should be used.

3 Methods

3.1 Preparation 1. Add 250 μL human serum to a 2 mL Safe-Lock microvial (see

of Samples Notes 8 and 9).
and Controls 2. Add 25 μL calibration standard (see Note 10) in 1.5 mL tubes.
3. Add 100 μL SST working solution in 1.5 mL tubes.
4. Add 10 μL IS working solution to calibrators, blanks, and
serum samples. The SST, calibrator, blank, and double blank
samples are excluded from steps 4–8 (see Note 11).
5. Add 1 mL of extraction solvent, MTBE, to the serum (see
Notes 12 and 13).
6. Mix for 15 min on an orbital shaker (see Note 14 and 15).
7. Centrifuge at 3000  g for 5 min at room temperature (RT).
8. Snap freeze the human serum using dry ice and ethanol (see
Notes 13 and 16).
9. Decant the extraction solvent into a 1.5 mL soft cap micro-
tube (see Note 6).
10. Dry all samples in a SpeedVac concentrator in combination
with a vapor trap (see Note 17).
11. Reconstitute the sample extract in 100 μL injection solution.
96 Lennart J. van Winden et al.

12. Mix (400 rpm) for 10 min at RT (see Note 18).

13. Centrifuge the samples at 18,213  g for 5 min at RT (see
Note 19).

3.2 Liquid 1. Purge the system with 50% mobile phase A and 50% mobile
Chromatography phase B at 1 mL/min (see Note 20).
2. Equilibrate the column at least 10 min prior to run with 20%
mobile phase A and 80% mobile phase B (see Note 21).
3. Apply a run time of 5 min with an isocratic mobile phase flow.
4. Set the injection volume to 50 μL.
5. Program an autosampler wash cycle between each sample to
minimize cross-contamination.

3.3 Tandem Mass 1. Set ESI settings according to the values listed in Table 1.
Spectrometry Settings 2. Measure in multiple reaction monitoring (MRM) mode. Select
the quantifier and qualifier transitions for testosterone and
testosterone-d5 (Table 2).
3. Tune for optimal MRM settings. The obtained optimal MRM
settings for the collision energy (CE), declustering potential
(DP), entrance potential (EP), and cell exit potential (CXP) for
both the mass transitions of our system are listed in Table 2 (see
Note 22).
4. Set the dwell time to 100 ms.

Table 1
General settings of the electrospray ionization (ESI) source

General settings
Nebulizer gas (psi) 50
Curtain gas (psi) 20
Ion spray (V) 5500
ESI temp. ( C) 500

Table 2
Multiple reaction monitoring (MRM) settings

MRM transition (m/z)

Q1 Q3 CE (V) DP (V) EP (V) CXP (V) Dwell time (ms)

Quantifier Testosterone 289.5 109.1 35 90 10 7 100
Testosterone-d5 294.5 113.1 35 90 10 7 100
Qualifier Testosterone 289.5 97.1 35 110 10 7 100
Testosterone-d5 294.5 100.1 35 110 10 7 100
 Serum Testosterone for Routine Diagnostics 97

3.4 LC-MS Run 1. The analytical run starts with three system suitability tests
and Batch Design (SST) containing 2 nmol/L testosterone standard and IS (see
Note 23).
2. Next, eight calibrators including a blank are measured in
duplicate.
3. Double blank samples in duplicate are introduced after the
calibrators (see Note 24).
4. Two sets of QC samples (three levels, singular samples) are
used: one set placed before and one set placed after the patient
samples (see Notes 25 and 26).
5. Patient samples are scheduled in between the sets of QC sam-
ples and run in duplicate (see Note 27).

3.5 Data Analysis 1. Starting data analysis, chromatography of the quantifier and
qualifier transitions are checked for their similarity. Aberrant
peak shapes between quantifier and qualifier transitions, such as
shouldering or twin peaks, are an indicator of interference (see
Note 28).
2. Next, in quantification mode, integration of peaks is automati-
cally performed and manually reviewed and corrected when
inaccurate automated integration was observed. Integration
of the analyte should match integration of the corresponding
IS (see Note 29).
3. For calibration, the testosterone/IS ratio is calculated using the
quantifier transition.
4. A calibration curve is generated using a linear regression, and a
1/x2 weighting is applied (see Notes 3 and 30).
5. Patient and QC results are individually calculated using the
calibration curve.
6. QCs are individually reviewed and a 2SD control rule is applied
for every control.
7. For patient samples, the duplicate results obtained are aver-
aged, and the average concentrations are reported as testoster-
one concentration.
8. The difference between the duplicate sample results relative to
the mean concentration was calculated and used to control for
sample handling errors (see Note 31).

4 Notes

1. The calibration curve range is 0.1–100 nmol/L. By using a

concentrated calibrator stock, the standard volumes added to
the calibrator samples can be decreased accordingly. This
method prolongs the usage of one set of standard stock.
98 Lennart J. van Winden et al.

2. Generally, a calibration curve is made in the same matrix as the

QC and patient samples. However, a human serum pool with
undetectable levels of testosterone is difficult to obtain in the
required volumes. For this reason, we verified whether the
methanol calibration matrix was suitable as an alternative to
the serum matrix and chose to use methanol as calibration
matrix.
3. The assay was standardized against the serum-based NIST
(National Institute of Standards and Technology) reference
material SRM 971. The two reference material samples were
analyzed in four independent runs, and both SRM 971 stan-
dards were diluted in various concentrations to confirm true-
ness throughout the measuring range.
4. Serum testosterone stability was tested as part of the method
validation. Testosterone stability in collected serum was at least
1 week when stored at room temperature, at least 2 weeks when
stored at 4  C, and at least 2 months when stored at 20  C.
Furthermore, when serum was not separated from the separa-
tor gel containing collection tube after centrifugation, testos-
terone was stable for at least 1 week when stored at 4  C
[5, 6]. Stability was defined as a  6% bias from the fresh
testosterone sample.
5. Injection solution can be stored at RT. However, to avoid
contamination we prepared the injection solution on a weekly
basis.
6. Specifically, we used Brand soft cap micro-tubes (Wertheim,
Germany) in combination with the cap piercing of the auto-
sampler. Therefore, no certified glass sample vials are needed.
7. Our assay was validated with an Agilent Zorbax Extend C-18
column, since this column is also used for other applications
run on the LC-MS/MS system.
8. A 250 μL sample size was chosen as this offers a sufficiently low
lower limit of quantification (LLOQ) of 0.17 nmol/L with
acceptable sample consumption. At this level, a total CV of
5.88% and a mean signal-to-noise ratio (S/N) of 14 are
obtained. The application of higher-end mass spectrometers
[2, 7] or incorporation of derivatization methods [8, 9] allows
higher assay sensitivity, although derivatization processes are
often more time-consuming and therefore not preferred for
routine clinical diagnostics.
9. For this step, we used 2 mL Safe-Lock Eppendorf tubes (Ham-
burg, Germany) that provide additional protection against
spilling during the mixing step. We find that normal tubes
often lack cap security, resulting in sample loss.
 Serum Testosterone for Routine Diagnostics 99

10. Note that by reconstitution of sample extract of 250 μL serum

in 100 μL injection solution, the samples are concentrated 2.5
times. This is accounted for by adding a volume of 25 μL
standard working solution into the calibrator samples.
11. SST, calibrator, blank, and double blank samples do not con-
tain serum. Therefore, no extraction is needed for these
samples.
12. MTBE is highly volatile: this results in inaccurate pipetting
when using air displacement pipettes. We find that using a
repeater pipette significantly increases the accuracy of distribut-
ing equal volumes.
13. Precautionary measures are accounted for by working in a
fume hood.
14. As a reference, we performed sample mixing with an IKA KS
130 basic orbital shaker (IKA, Staufen im Breisgau, Germany).
15. We find that by placing the sample vials sideways in the shaker,
the mixing is performed with improved efficiency. By increas-
ing the surface area, the immiscible liquids distribute more
equally inside the tube.
16. For snap freezing a Styrofoam container is used. First, a per-
meable micro-tube rack is placed in the container. Subse-
quently, dry ice is placed around the rack. Ethanol is then
poured into the container. Importantly, the ethanol should
cover the aqueous phase in the micro-tubes. Freezing the
aqueous phase usually takes 20–40 s, depending on the volume
of ethanol and the amount of dry ice. A good indicator is the
formation of a projection on the surface of the frozen serum.
17. As a reference, we used a Savant SC210A SpeedVac concentra-
tor (Thermo Fisher Scientific, Waltham, MA, USA) in combi-
nation with refrigerated vapor trap RVT4104 (Thermo Fisher
Scientific, Waltham, MA, USA) to dry the organic phase after
extraction.
18. The samples should not be mixed sideways as this step focuses
on reconstitution of the dried sample, which is mainly posi-
tioned at the bottom of the tube.
19. By briefly centrifuging the samples, droplets of dissolved sam-
ple located in the upper part of the tube are repositioned at the
bottom of the tube. Additionally, in the case of increased
turbidity in the samples, which is often present in human
serum, this step acts as purification of the sample. After inten-
sive centrifugation, the turbidity extract is pelleted at the bot-
tom of the tube. As a result, the turbidity extract remains in the
tube during injection of the sample.
100 Lennart J. van Winden et al.

3.9e5
3.8e5 Testosterone 289.5->109.1 m/z
3.6e5 Testosterone 289.5->97.1 m/z
3.4e5 Testosterone-d5 294.5->113.1 m/z
3.2e5 Testosterone-d5 294.5->100.1 m/z
3.0e5
2.09
2.8e5

2.6e5
Intensity (cps)

2.4e5

2.2e5

2.0e5

1.8e5

1.6e5

1.4e5

1.2e5

1.0e5

8.0e4

6.0e4

4.0e4

2.0e4

0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 4.0 4.2 4.4 4.6 4.8 5.0
Time (min)

Fig. 1 Testosterone chromatogram. Chromatogram of a calibrator sample containing 25.4 nmol/L testoster-
one. The sample is spiked with 2 nmol/L internal standard (IS). Blue represents the quantifier testosterone
transition, red represents the qualifier testosterone transition, green represents the quantifier testosterone IS
transition, and gray represents the qualifier testosterone IS transition

20. Generally, the system is purged for at least 1 min. This results in
clearance of any air bubbles present in the tubing leading to the
pumps.
21. Testosterone and the corresponding IS have a tested retention
time around 2.1 min using a 80:20 mobile phase B:A ratio in
our system setup (Fig. 1).
22. Details on the general settings and MRM settings of testoster-
one and testosterone-d5 used in our method are given in
Tables 1 and 2, respectively. Note that the parameters of the
IS corresponding to the analyte are identical. Firstly, variations
in parameter values between analyte and IS could result in
signal differences troubling quantification. Secondly, replace-
ment of hydrogens by deuterium yields a molecule that has
virtually identical chemical-physical properties. Therefore, tun-
ing both molecules should result in similar values for all
parameters.
23. An SST is performed in triplicate, prior to the analytical run, to
confirm proper system functioning before starting an
analytical run.
 Serum Testosterone for Routine Diagnostics 101

24. Normally, double blank samples are used to check the matrix
and LC-MS/MS system for contaminants and interferences. In
this case, however, the calibration curve is generated in metha-
nol, and double blank samples are introduced to identify a
possible contamination in the IS working solution or the injec-
tion solution.
25. The QC samples serve as a measure of the quality for each run.
When a significant aberration of QC concentration is detected
(exceeding a 2SD rule), the run and obtained results require
additional evaluation before the obtained results can be
released.
26. Assay precision was tested by measuring three QC levels in
quadruplicate for ten individual runs. Total coefficient of vari-
ation (CV) was 4.7, 4.1, and 3.3% for testosterone concentra-
tions of 0.4, 3.1, and 7.6 nmol/L, respectively.
27. We find that increasing the number of samples reduces the
practicality of the sample pretreatment. Therefore, we recom-
mend to divide large patient serum sample sets (i.e., > 40 patient
serum samples) over multiple runs.
28. No interference was detected from DHT, androstenedione,
DHEA, 17-OHP, cortisol, hemolysate, bilirubin, and Intrali-
pid. Only epitestosterone tested positive for interference.
However, to date, no epitestosterone interference (i.e., a peak
at a retention time of 2.4 min) has been observed in any patient
sample analyzed.
29. Generally, the Analyst® software package (Version 1.6.2) inte-
grates peaks automatically based on predetermined settings. In
general, these settings should provide correct integration in
more than 90% of samples, but at low concentrations (smaller
peaks), fluctuation may result in poor automatic recognition of
peak start and ending. Correct where needed, and ascertain
that the peaks of the analyte and the corresponding IS are
integrated similarly.
30. The applied 1/x2 weighting was selected based on overall
trueness and linearity results obtained during assay validation.
We determined that a 1/x2 weighting outperformed a 1/x
weighting assay performance, especially for lower testosterone
concentrations that we considered most relevant. A similar
observation has recently been reported [2].
31. Patient serum samples are measured in duplicate to detect
laboratory handling errors. In the case of inconsistent results
(difference of duplicate results > 15%), the sample should be
considered for reanalysis.
102 Lennart J. van Winden et al.

References
1. Marshall SK, Bangert WJ (2008) Clinical chem- chromatography-tandem mass spectrometry:
istry, 6th edn. Elsevier Science, Edinburgh interference from blood collection tubes. Clin
2. Wang HW, Gay Y, Botelho GD, Caudill JC, Biochem 45(18):1706–1709
Vesper SP (2014) Total testosterone quantitative 7. Methlie P et al (2013) Multisteroid LC-MS/MS
measurement in serum by LC-MS/MS. Clin assay for glucocorticoids and androgens, and its
Chim Acta 8(5):583–592 application in Addison’s disease. Endocr Con-
3. van Rossum HH, Bergman AM, Lentjes E nect 2(12):125–136
(2015) Analytical challenges and potential appli- 8. Bui HN et al (2010) Serum testosterone levels
cations of sex steroid hormone analysis in breast measured by isotope dilution-liquid chromatog-
and prostate cancer patients. Chromatographia raphy-tandem mass spectrometry in postmeno-
78(5–6):359–365 pausal women versus those in women who
4. Handelsman DJ, Wartofsky L (2013) Require- underwent bilateral oophorectomy. Ann Clin
ment for mass spectrometry sex steroid assays in Biochem 47:248–252
the journal of clinical endocrinology and metabo- 9. Kushnir MM et al (2010) Liquid
lism. J Clin Endocrinol Metab 98(10):3971–3973 chromatography-tandem mass spectrometry
5. Hepburn S et al (2016) Sex steroid hormone assay for androstenedione, dehydroepiandros-
stability in serum tubes with and without separa- terone, and testosterone with pediatric and
tor gels. Clin Chem Lab Med 54(9):1451–1459 adult reference intervals. Clin Chem 56
6. Shi RZ, van Rossum HH, Bowen RAR (2012) (7):1138–1147
Serum testosterone quantitation by liquid
 Chapter 8

LC-MS/MS Analysis of Bile Acids

Sabrina Krautbauer and Gerhard Liebisch

Abstract
Besides their role as lipid solubilizers, bile acids (BAs) are increasingly appreciated as bioactive molecules.
They bind to G-protein-coupled receptors and nuclear hormone receptors. So they control their own
metabolism and act on lipid and energy metabolism. Here we describe a simple, accurate, and fast liquid
chromatography-tandem mass spectrometry (LC-MS/MS) method for the quantification of BAs in human
plasma/serum.

Key words Tandem mass spectrometry, Electrospray ionization, Bile acids, Quantification, Internal
standards, Method, Liquid chromatography, Isomer separation

1 Introduction

Bile acids (BAs) are primarily synthesized from cholesterol in the

liver. Cholesterol is catabolized by oxidation, shortening of the side
chain to form BAs. The free BAs are finally conjugated by glycine
and taurine, respectively [1, 2]. Amphiphatic BAs are essential to
solubilize dietary lipids and vitamins to promote their absorption.
In humans most abundant BAs comprise the primary BAs cholic
acid (CA) and chenodeoxycholic acid (CDCA) and the secondary
BAs deoxycholic acid (DCA), lithocholic acid (LCA), and urso-
deoxycholic acid (UDCA) generated by deconjugation and dehy-
droxylation by intestinal bacteria. BAs are effectively reabsorbed
and transported back to the liver to again enter enterohepatic
circulation [3].
Besides their role as lipid solubilizers, BAs are increasingly
studied due to their signaling functions. They can bind to G-
protein-coupled receptors and nuclear hormone receptors to con-
trol their own metabolism and to influence lipid and energy metab-
olism [4]. To study BA function in detail, it is necessary to use
methods for their quantification covering the structural diversity of
this group. For quantification of BAs including also their

Martin Giera (ed.), Clinical Metabolomics: Methods and Protocols, Methods in Molecular Biology, vol. 1730,
https://doi.org/10.1007/978-1-4939-7592-1_8, © Springer Science+Business Media, LLC 2018

103
104 Sabrina Krautbauer and Gerhard Liebisch

conjugates, liquid chromatography-tandem mass spectrometry

(LC-MS/MS) is considered as the method of choice.
In this chapter we describe a simple, accurate, and fast LC-MS/
MS method for the analysis of the main BAs in human plasma/
serum. This approach uses protein precipitation as simple sample
preparation. For accurate quantification stable isotope-labeled stan-
dards are included for the major species. The method provides
baseline separation of isomeric BAs within a run time of 6 min
due to application of core-shell material at basic pH of the mobile
phase [5, 6].

2 Materials

Only highest purity solvents and standards should be used. Milli-Q

water purified to a specific resistance of 18.2 MΩ is used.

2.1 Standards Dissolve the following BA standards (Table 1) at a concentration of

20 mg/mL (deuterated standards at 1 mg/mL) in methanol in
10 mL screw cap glass vessels. Store standards at
80  C.

2.2 Reagents 1. 1 M hydrochloric acid: Fill 100 mL water into a screw cap
bottle and add 6.35 mL of concentrated hydrochloric acid
(37%) and store in 100 mL reagent bottle (see Note 1).
2. Mobile phase A: Dissolve 771 mg ammonium acetate in
900 mL water and 100 mL methanol (measure volume in
appropriate graduated cylinder), add 1000 μL ammonium
hydroxide (25%), and store in 1 L reagent bottle.
3. Mobile phase B: Dissolve 771 mg ammonium acetate in
1000 mL methanol (measure volume in appropriate graduated
cylinder), add 1000 μL ammonium hydroxide (25%), and store
in 1 L reagent bottle.
4. Combined solution of internal standards: Pipette the following
volume of 1 mg/mL standard solutions in a 10 mL volumetric
flask, and add methanol to a volume of 10 mL. Store standards
below
18  C when used within 6 months (for long-term
storage
80  C). 15 μL each of D4-LCA, D4-GUDCA, and
D5-TLCA, 20 μL each of D4-GDCA and D5-TDCA, 50 μL
each of D4-CA and D4-DCA, 75 μL each of D4-CDCA and
D4-UDCA, and 200 μL each of D4-GCA, D4-GCDCA,
D4-GLCA, D5-TCA, D5-TCDCA, and D5-TUDCA.
5. Matrix calibrator: Human serum double charcoal-stripped,
delipidized Golden West Biologicals (Temecula, CA, USA) is
used as level 0, and pooled serum of healthy donors is used as
level 1. Levels 2–5 are spiked up to the following concentra-
tions (level 5 concentration; see Notes 2 and 3): 30 μM TCA,
 Bile Acid Analysis 105

Table 1
Analytical standards

Internal
Analyte Abbreviation standard Possible supplier
Cholic acid CA Sigma-Aldrich (Taufkirchen,
Chenodeoxycholic acid CDCA Germany)
Deoxycholic acid DCA
Lithocholic acid LCA
Ursodeoxycholic acid UDCA
Hyodeoxycholic acid HDCA
Glycocholic acid GCA
Glycochenodeoxycholic acid GCDCA
Glycodeoxycholic acid GDCA
Taurocholic acid TCA
Taurochenodeoxycholic acid TCDCA
Taurodeoxycholic acid TDCA
Taurolithocholic acid TLCA
Tauroursodeoxycholic acid TUDCA
Taurohyodeoxycholic acid THDCA
D4-CDCA x
D4-LCA x
D4-CA x
D4-DCA x
Glycolithocholic acid GLCA Steraloids (Newport, USA)
Glycoursodeoxycholic acid GUDCA
Glycohyodeoxycholic acid GHDCA
D4-UDCA x
D4-GCDCA x
D4-GCA x
D4-GLCA x CDN Isotopes (Pointe-Claire,
D4-GUDCA x QC, Canada)
D4-GDCA x
D5-TCA x Toronto Research Chemicals
D5-TUDCA x (Toronto, Canada)
D5-TCDCA x
D5-TDCA x
D5-TLCA x

30 μM TUDCA, 20 μM THDCA, 40 μM TCDCA, 16 μM

TDCA, 4 μM TLCA, 70 μM GCA, 100 μM GUDCA, 70 μM
GHDCA, 40 μM GCDCA, 16 μM GDCA, 4 μM GLCA,
16 μM CA, 22 μM UDCA, 8 μM HDCA, 8 μM CDCA,
6 μM DCA, 2 μM LCA. Store in 5 mL aliquots at
80  C.
6. Working aliquots of 0.5 mL are kept at
18  C for up to
6 months.
7. Serum quality controls (QCs): As low control, pooled serum of
healthy donors; as high control, pooled serum of healthy
106 Sabrina Krautbauer and Gerhard Liebisch

donors spiked with 1/10 of the highest calibrator level. Store

in 5 mL aliquots at
80  C. Working aliquots of 0.5 mL are
kept at
18  C for up to 6 months.

2.3 Liquid 1. MS/MS: Hybrid triple quadrupole/linear ion trap mass spec-
Chromatography- trometer, 4000 QTRAP® (Concord, Ontario, Canada)
Tandem Mass equipped with Turbo V ion spray source (see Note 4).
Spectrometry 2. LC: Autosampler, HTC PAL autosampler (CTC Analytics,
(LC-MS/MS) Zwingen, Switzerland) with additional 6-port valve, Agilent
1200 binary pump, 1200 isocratic pump, degasser (Wald-
bronn, Germany).
3. LC column: Core-shell column NUCLEOSHELL RP18,
50  2 mm, 2.7 μm (Macherey-Nagel, D€ uren, Germany)
equipped with a 0.5 μm prefilter (Upchurch Scientific, Oak
Harbor, WA, USA).
4. Glass autosampler vials (1.5 mL) with 200 μL glass insert.

3 Methods

3.1 Sample 1. Place 100 μL sample (patient material, QCs, calibrators, inter-
Preparation nal standard blank ¼ water) in 1.5 mL Eppendorf tube.
2. Add 20 μL of internal standard mix.
3. Mix briefly.
4. Add 30 μL 1 M HCl.
5. Add 1 mL acetonitrile.
6. Mix thoroughly about 1 min.
7. Centrifuge 15 min at 14,000  g.
8. Transfer 1 mL supernatant into 1.5 mL Eppendorf tube.
9. Remove solvent in a vacuum concentrator.
10. Reconstitute in 100 μL (3/7; v/v) methanol/water.
11. Mix thoroughly about 1 min.
12. Place for 10 min into ultrasonic bath.
13. Centrifuge 15 min at 14,000  g.
14. Transfer clear supernatant into autosampler vial (1.5 mL) with
200 μL glass insert (see Note 5).

3.2 Liquid Samples are analyzed by LC-MS/MS together with calibrators,

Chromatography- QCs including double blank ((3/7; v/v) methanol/water), and
Tandem Mass internal standard blank (see Subheading 3.1) in each batch.
Spectrometry 1. LC settings: Purge gradient pump with mobile phases A and
(LC-MS/MS) B. Wash solvent 1: 50/50 (v/v) methanol/2-propanol. Wash
solvent 2: 50/50 (v/v) methanol/water.
 Bile Acid Analysis 107

2. Install core-shell column NUCLEOSHELL RP18,

50  2 mm, 2.7 μm (see Note 6).
3. Set column oven temperature: 50  C.
4. The following gradient is used: 80% A for 0.05 min, a stepwise
linear decrease to 53% A at 0.1 min, to 49% A at 2.0 min and to
28% A at 4.5 min. The flow rate was set to 0.5 mL/min. A
column wash at 100% B for 0.5 min and re-equilibration at
100% A for 0.6 min were performed at a flow rate of 0.8 mL/
min.
5. Isocratic pump delivers methanol at a flow rate of 250 μL/min.
6. Divert valve directs column flow only from 65 to 290 s into the
mass spectrometer.
7. Inject volume: 5 μL.
8. Mass spectrometer settings (for 4000 QTRAP): Ensure the
following parameters are set: ion source, Turbo V ion spray
source; ionization mode, negative; ion spray voltage,
4500 V;
ion source heater temperature, 450  C; source gas 1, 40 psi;
source gas 2, 35 psi; curtain gas, 20 psi; collision gas, 6 psi;
entrance potential,
10 V; dwell time, 30 msec; operate both
Q1 and Q3 at unit resolution; analysis mode, MRM (for mass
transitions, see Table 2).

Table 2
Mass transitions and retentions times of the individual BAs including their internal standards

BA species RT [min] Q1 [m/z] Q3 [m/z] DP [V] CE [V] CXP [V] IS

CA 2.63 407.3 407.3
135
30
11 D4-CA
UDCA 1.96 391.3 391.3
135
30
11 D4-UDCA
HDCA 2.19 391.3 391.3
135
30
11 –
CDCA 3.64 391.3 391.3
135
30
11 D4-CDCA
DCA 3.83 391.3 391.3
135
30
11 D4-DCA
LCA 4.55 375.3 375.3
135
30
11 D4-LCA
GCA 2.57 464.3 74
105
72
1 D4-GCA
GUDCA 1.87 448.3 74
115
70
1 D4-GUDCA
GHDCA 2.11 448.3 74
115
70
1 –
GCDCA 3.57 448.3 74
115
70
1 D4-GCDCA
GDCA 3.78 448.3 74
115
70
1 D4-GDCA
GLCA 4.45 432.3 74
105
64
1 D4-GLCA
TCA 2.54 514.3 80
185
116
1 D5-TCA
(continued)
108 Sabrina Krautbauer and Gerhard Liebisch

Table 2
(continued)

BA species RT [min] Q1 [m/z] Q3 [m/z] DP [V] CE [V] CXP [V] IS

TUDCA 1.86 498.3 80
185
116
1 D5-TUDCA
THDCA 2.10 498.3 80
185
116
1 –
TCDCA 3.52 498.3 80
185
116
1 D5-TCDCA
TDCA 3.73 498.3 80
185
116
1 D5-TDCA
TLCA 4.41 482.3 80
165
108
1 D5-TLCA
D4-CA 2.61 411.3 411.3
135
30
11
D4-CDCA 3.62 395.3 395.3
135
30
11
D4-DCA 3.81 395.3 395.3
135
30
11
D4-LCA 4.53 379.3 379.3
135
30
11
D4-UDCA 1.94 395.3 395.3
135
30
11
D4-GCA 2.55 468.3 74
105
72
1
D4-GCDCA 3.55 452.3 74
115
70
1
D4-GDCA 3.76 452.3 74
115
70
1
D4-GUDCA 1.85 452.3 74
115
70
1
D4-GLCA 4.43 436.3 74
105
64
1
D5-TCA 2.52 519.3 80
185
116
1
D5-TUDCA 1.84 503.3 80
185
116
1
D5-TCDCA 3.50 503.3 80
185
116
1
D5-TDCA 3.71 503.3 80
185
116
1
D5-TLCA 4.39 487.3 80
165
108
1

9. Data analysis: Integrate peaks by instrument software (here

Analyst Software 1.4.2.; Applied Biosystems, Darmstadt, Ger-
many) (see Fig. 1). Quantification is based on ratio of area
counts of analyte to internal standard (see Table 2). Calibration
lines were generated by linear regression with 1/x weighting
(see Notes 7 and 8).
10. Quality check: Check for baseline separation of isomeric BA
species, e.g., CDCA and DCA including their conjugates (see
Fig. 1). Check purity of the deuterated BAs: When interference
on the analyte transitions is present, quantification needs to
consider that. For calibration lines the coefficient of determi-
nation R2 should be above 0.99, and individual calibrator
 Bile Acid Analysis 109

Fig. 1 Chromatogram of bile acid species. The chromatograms show unconjugated (a) glyco- (b) and tauro-
conjugates (c). Displayed are chromatograms from a matrix calibrator after normalization to the highest peak

levels should not deviate more than 15% (more than 20% at
LOQ) from the target value. Concentrations of QCs BA levels
have to be in a defined range (recommended maximal deviation
<15% from the target value; <20% at the LOQ).
110 Sabrina Krautbauer and Gerhard Liebisch

4 Notes

1. Always add the acid slowly to the water not the other way
round.
2. High calibrator concentrations for UDCA and its conjugates
are needed for cholestatic patients treated with UDCA. Lower
calibrator levels of all BAs may be used when only control
subjects are analyzed.
3. Concentrations of endogenous BAs are calculated from the
spiked calibrators. BA concentrations of level 0 are
below LOD.
4. The method needs a sensitive triple-quadrupole mass spec-
trometer equipped with electrospray ion source.
5. Check for air bubbles in the glass insert and remove these if
present.
6. Clean the column before storage with acetonitrile/water (3/1,
v/v) because basic solvent will decrease column lifetime.
7. For HDCA and its conjugates, stable isotope-labeled standards
are not available. Due to low levels in our patients, we do not
report these values.
8. For BA species without matching stable isotope-labeled inter-
nal standard, analytical performance of internal standards
should be tested with matrix QCs as described in [6].

Acknowledgments

This work was supported by the “Stiftung f€

ur Pathobiochemie und
Molekulare Diagnostik” and by the European Union’s FP7
Programme MyNewGut (Grant Agreement Number 613979).

References
1. Hofmann AF, Hagey LR (2008) Bile acids: Nat Rev Endocrinol 10(8):488–498. https://
chemistry, pathochemistry, biology, pathobiol- doi.org/10.1038/nrendo.2014.60
ogy, and therapeutics. Cell Mol Life Sci 65 5. Scherer M, Gnewuch C, Schmitz G, Liebisch G
(16):2461–2483 (2009) Rapid quantification of bile acids and
2. Thomas C, Pellicciari R, Pruzanski M, Auwerx J, their conjugates in serum by liquid
Schoonjans K (2008) Targeting bile-acid signal- chromatography-tandem mass spectrometry. J
ling for metabolic diseases. Nat Rev Drug Dis- Chromatogr B Analyt Technol Biomed Life Sci
cov 7(8):678–693. https://doi.org/10.1038/ 877(30):3920–3925. https://doi.org/10.
nrd2619 1016/j.jchromb.2009.09.038
3. Dawson PA, Karpen SJ (2015) Intestinal trans- 6. Krautbauer S, Buechler C, Liebisch G (2016)
port and metabolism of bile acids. J Lipid Res 56 Relevance in the use of appropriate internal stan-
(6):1085–1099. https://doi.org/10.1194/jlr. dards for accurate quantification using LC-MS/
R054114 MS: tauro-conjugated bile acids as an example.
4. Kuipers F, Bloks VW, Groen AK (2014) Beyond Anal Chem 88(22):10957–10961. https://doi.
intestinal soap—bile acids in metabolic control. org/10.1021/acs.analchem.6b02596
 Chapter 9

LC-MS/MS Analysis of Triglycerides in Blood-Derived

Samples
Madlen Reinicke, Susen Becker, and Uta Ceglarek

Abstract
The increasing interest in the analysis of triglyceride (TG) species and the individual fatty acid
(FA) composition requires expeditious and reliable quantification strategies. The utilization of flow injec-
tion analysis (FIA) coupled to quadrupole tandem mass spectrometry (MS/MS) for the simultaneous
quantitation of TG and identification of FA composition facilitates the multiplexed verification of various
biomarkers from small sample quantities. Enzymatic methods based on saponification and glycerol analysis
are not suited for the determination of the FA distribution in TGs. This protocol proposes a procedure for
the establishment of a relative quantitation method for middle- to high-abundance plasma TGs and the
corresponding FA composition. Essential topics as FIA-MS/MS method development as well as sample
preparation and validation strategies are described in detail.

Key words Triglyceride species, Fatty acid distribution, Neutral loss experiments, Relative quantifica-
tion, Biomarker validation

1 Introduction

Reliable quantitation of triglyceride (TG) species and the

corresponding fatty acid (FA) distribution is a basic requirement
for biomarker verification studies in human body fluids. Routine
diagnostic quantification methods of TGs are based on hydrolysis
and spectrophotometric determination of total glycerol concentra-
tion. Thus, a differentiation of the TG molecular species and infor-
mation about the FA distribution are not available. Due to the
association of FAs with the development of cardiovascular diseases
and diabetes, detailed information about the kind of FAs bound
within the TG is required [1, 2]. The identification of TG molecu-
lar species in human body fluids is challenging because of the great
variability of the FAs relating to the degree of saturation, the chain
length, and the various combination possibilities of FAs at the
glycerol backbone. The close resemblance of the TG species
requires laborious chromatographic separation [3, 4]. We describe

Martin Giera (ed.), Clinical Metabolomics: Methods and Protocols, Methods in Molecular Biology, vol. 1730,
https://doi.org/10.1007/978-1-4939-7592-1_9, © Springer Science+Business Media, LLC 2018

111
112 Madlen Reinicke et al.

a flow injection analysis (FIA) coupled to quadrupole tandem mass

spectrometry (MS/MS) method without chromatography benefit-
ting from the advantages of fast measurement and mass separation
analysis performed in neutral loss (NL) scan mode using a triple
quadrupole mass spectrometer. In the first dimension, a defined
range of mass-to-charge ratios (m/z) is scanned. The scan range of
the second dimension is defined by a m/z offset corresponding to
the FA of interest. Signals for TG species are detected for a limited
number of NL-specific ions occurring in both scan dimensions.
Combining and monitoring of such FA-specific mass transitions
provide the basis for the high specificity in complex sample matrices
[5–9].Thereby, quantification is dependent on the implementation
of internal standards to compensate for sample losses during sample
preparation and matrix effects in the ionization process. For this
purpose, mostly deuterated analogs of the TG species are used.
Important for the deuterated analogs is a structurally stable posi-
tion of the deuterium atoms to ensure stable labeled fragments
during electrospray ionization.
The presented protocol for TG quantification by FIA-MS/MS
can be easily implemented for middle- to high-abundance plasma
TGs requiring only small sample quantities [10].

2 Materials

2.1 Reagents 1. Solvents: only use high purity solvents, ideally of MS grade.
and Consumables For example, ULC-MS grade methanol, ULC-MS grade
2-propanol, and Rotisolv® HPLC grade toluene can be used.
2. Protein precipitation: use 200 μL glass micro-inserts with
spring, 1.5 mL polypropylene microtube for centrifugation,
and 10 mM ammonium acetate in (50/50, v/v) toluene/
methanol.
3. LC-MS/MS analysis: use 200 μL glass micro-inserts in combi-
nation with 2.0 mL short thread autosampler vials with
slotted cap.

2.2 HPLC and Mobile 1. The HPLC instrument consists of an autosampler and a binary
Phases micro-pump system.
2. Isocratic conditions with a flow gradient of (10–500 μL/min)
are used.
3. The mobile phase consists of (90/10, v/v) 2-propanol/
methanol.

2.3 Mass 1. The employed mass spectrometer is triple quadrupole (e.g.,

Spectrometer API 4000, SCIEX).
2. Ionization source: electrospray ionization in the positive mode.
3. An analysis software Analyst (SCIEX) can be used.
 Triglyceride Analysis 113

Table 1
Standards and internal standards with abbreviations for total carbon/number of double bonds
and the FA composition

Total carbon/number
Standard of double bonds FA composition
Standards
Trimyristin glyceride TG 42:0 TG (14:0/14:0/14:0)
Tripalmitin glyceride TG 48:0 TG (16:0/16:0/16:0)
1,3-palmitin-2-stearin glyceride TG 50:0 TG (16:0/18:0/16:0)
Tristearin glyceride TG 54:0 TG (18:0/18:0/18:0)
Triolein glyceride TG 54:3 TG (18:1/18:1/18:1)
Triarachidonin glyceride TG 60:12 TG (20:4/20:4/20:4)
(1,1,2,3,3)-deuterium-labeled (d5) internal standards
1,3-myristin-2-palmitolein glyceride d5-TG 44:1 d5-TG (14:0/16:1/14:0)
1,3-pentadecanoin-2-olein glyceride d5-TG 48:1 d5-TG (15:0/18:1/15:0)
1,3-palmitin-2-stearin glyceride d5-TG 50:0 d5-TG (16:0/18:0/16:0)
1,3-nonadecanoin-2-laurin glyceride d5-TG 50:0 d5-TG (19:0/12:0/19:0)
1,3-heptadecanoin-2-heptadecenoin glyceride d5-TG 51:1 d5-TG (17:0/17:1/17:0)
1,3-eicosadienoin-2-linolenin glyceride d5-TG 58:7 d5-TG (20:2/18:3/20:2)
1,3-linolein-2-linolein glyceride d5-TG 58:10 d5-TG (20:4/18:2/20:4)
1,3-arachidin-2-eicosenoin glyceride d5-TG 60:1 d5-TG (20:0/20:1/20:0)
1,3-eicosapentaenoin-2-docosahexaenoin glyceride d5-TG 62:16 d5-TG (20:5/22:6/20:5)

2.4 Standards, Standards and internal standards are given in Table 1. For method
Internal Standards, development, human serum or EDTA plasma is required.
and Human Specimen

3 Methods

The following protocol summarizes basic steps in the method

development of a quantitative assay. Starting from the identification
of characteristic mass transitions, all mandatory requirements for
a relative quantification of middle- to high-abundance plasma
TGs and the FA composition by MS/MS are described (see Notes
1 and 2).
114 Madlen Reinicke et al.

Table 2
Possible NL transitions with the corresponding FA

Neutral loss [Da] Corresponding FA

245.24 Myristic acid C14:0
259.25 Pentadecylic acid C15:0
273.27 Palmitic acid C16:0
271.25 Palmitoleic acid C16:1
301.30 Stearic acid C18:0
299.28 Oleic acid C18:1
297.27 Linoleic acid C18:2
295.25 Linolenic acid C18:3
321.27 Arachidonic acid C20:4

3.1 Selection In the presence of ammonium ions (10 mM ammonium acetate),

of Mass Transitions TGs form the molecular ions [MþNH4]+ in positive mode ESI.
and HPLC Optimization The ammonium adducts were detected by neutral loss scans, char-
acterized by the loss of the FA and ammonia (NH3) as neutral
molecules during the fragmentation process, leaving a positively
charged molecule with the mass difference of the FA of interest.
The respective NL transitions can be calculated by its monoisotopic
mass in addition to the loss of NH3, e.g., myristic acid C14:0,
which has a monoisotopic mass of 228.21 Da and 17.03 Da for
the loss of NH3, yielding in a respective NL of 245.24 Da (see
Notes 2 and 3). An example of possible NL transitions with the
corresponding FA is given in Table 2.
After MS tuning and selection of the neutral loss transitions,
the HPLC part of the method has to be optimized. Below a
detailed procedure applying a triple quadrupole instrument is
described:
1. Calculate potential NL fragments of the fatty acid residues in
the TG species of interest.
2. Dissolve TG standards in toluene, to yield stock solutions of
1–5 mg/mL, and apply ultrasound if standards are less soluble
(see Note 4).
3. Prepare working standards in (50/50, v/v) toluene/methanol.
4. Prepare tuning solutions of 250 ng/mL or lower in (50/50,
v/v) toluene/methanol containing 10 mM ammonium acetate.
5. Activate the tune mode and directly infuse the tuning solution
at 10 μL/min using the following parameters: curtain gas,
10 psi; collision gas, high; nebulizer gas, 20 psi; heater gas,
0 psi; ion spray voltage, 5500 V; source temperature, 20  C;
and polarity, positive.
 Triglyceride Analysis 115

6. Perform a Q1 scan to identify the precursor ion, e.g., m/z

350–1000.
7. Tune the declustering potential in a Q1 multiple ion scan, e.g.,
0–200 V.
8. Identify fragment ions in a product ion scan performed with
ramped collision energy in multichannel analysis (MCA) mode.
Compare detected ions with the expected fragment ions iden-
tified before by the calculated neutral loss.
9. Tune the collision energy and the collision cell entrance and
exit potential in a multiple reaction monitoring (MRM) scan by
applying the ramped mode for each potential.
10. Optimize curtain gas, nebulizer gas, heater gas, ion spray volt-
age, source temperature, and collision gas by infusion of the
TG standard solution in a constant flow of eluent operated
under FIA conditions via a tee. Choose a flow rate which will
be applied in the final method. If necessary, repeat this step
during FIA optimization.
11. Create an analyzing method containing the optimized neutral
loss and MS parameters. An optimized flow gradient and stan-
dard gas and temperature settings are given below.
The injection volume was 40 μL, and the mobile phase was
2-propanol/methanol (90/10, v/v) with a flow gradient of
10 μL/min (0–3.5 min) followed by 500 μL/min (3.5–5 min).
The mass range was set from m/z 700 to m/z 1000 with a scan
rate of 1.5 s. Experimental conditions for electrospray ioniza-
tion of TGs in positive-ion mode with collisional activation by
nitrogen gas were as follows: curtain gas, 20 psi; collision gas,
high; declustering potential, 70 V; collision cell entrance
potential, 10 V; collision energy, 35 V; collision cell exit poten-
tial, 7 V; nebulizer gas, 30 psi; heater gas, 50 psi; ion spray
voltage, 5500 V; source temperature. 300  C.
12. Optimize HPLC parameters and settings including analytical
eluents, modifier, gradient, flow rate, column temperature, and
sample volume. Refer to the basic literature on HPLC method
development for guidance [11].

3.2 Internal One of the simplest approaches for the relative quantification of TG
Standardization species by means of MS/MS is the FIA combined with internal
and Quality Controls standardization. Search the literature for reference values or pub-
lished concentration levels of the target analyte in human plasma or
serum, e.g., [10, 12, 13].
1. Internal standard mix
Prepare an internal standard mix in (50/50, v/v) toluene/
methanol containing 10 mM ammonium acetate with concen-
trations similar to expected concentrations in human
116 Madlen Reinicke et al.

specimens. Note that signal intensity and peak area of the

internal standard should be comparable to the ones of the
monitored analyte.
2. Quality controls
For quality assessment and guarantee of batch comparabil-
ity, quality controls should be carried along in the analytical
process. Preferably, two to three concentration levels should be
investigated (low, middle, high), and blank matrix should be
preferred. TG standards can be used for the preparation of
quality controls. Human specimens, which should be stored
as aliquots to avoid freeze-thaw cycles, can be used as quality
control. Ring trial materials or the like with known target
values can also be used. Avoid multiple freeze-thaw cycles and
store internal standard mix and quality controls aliquoted,
ready for use.

3.3 Sample 1. Protein precipitation carried out in a glass micro-insert with

Pretreatment spring which is placed in a 1.5 mL polypropylene microtube for
of Human Specimens later centrifugation.
and Control Material 2. Place 245 μL of precipitation solvent containing 150 ng/mL
internal standard d5-TG mixture and 10 mM ammonium
acetate in the glass micro-insert with spring.
3. Add 5 μL of plasma/serum or standard or control material and
close the tube tightly.
4. After mixing for 30 s, spin down the precipitate at 13,000  g
for 10 min.
5. Transfer the clear supernatant in a new glass micro-insert
placed in an autosampler vial.

3.4 Sample Analysis 1. Run the developed MS/MS method (Subheading 3.1).
2. For quality assessment of the analysis, several aliquots of a
quality control should be carried along in the sample prepara-
tion procedure and be measured accordingly (see Note 5).

3.5 Data Analysis TG species were relatively quantified using the intensity ratio of the
and Relative summed NL intensities of each ion peak with that of the internal
Quantification standard, e.g., TG d5–50:0 (16:0/18:0/16:0), after correction for
two 13C isotope effects and the different fragmentation behavior of
the cleaved neutral acyl residues in the positive-ion mode as the
following [5, 10]:
1. Select a representative time window in the total ion current
spectrum, e.g., 2 min.
2. Select “show spectrum” to get a NL spectrum of every neutral
loss experiment included in the method.
3. To get the overall intensity of the TG species, the data of every
single NL scan needs to be summed up. Use, e.g., the Analyst
 Triglyceride Analysis 117

(SCIEX) options “overlay” and “sum overlays” to sum all NL

experiments. Use, e.g., “List data,” and select the peak list to
save the m/z and the corresponding peak intensity, e.g., as txt.
file (see Note 6).
4. The m/z for each ion peak corresponds to a TG species. The
peak list can be reduced to the m/z for the TG species of
interest including the internal standards and the peak intensity
of the corresponding [M-2]-peak for isotopic correction. An
example peak list with TG species, ion peak, and corresponding
[M-2]-peak is given in Table 3.

Table 3
Possible peak list with the m/z of the ion peak and corresponding
[M-2]-peak for 19 TG species and 4 internal standards

Triglyceride m/z [M+NH4]+ m/z [M-2]+

TG 46:1 794.7 792.7
TG 46:0 796.7 794.7
TG 48:2 820.7 818.7
TG 48:1 822.7 820.7
TG 48:0 824.7 822.7
TG 50:3 846.8 844.8
TG 50:2 848.8 846.8
TG 50:1 850.8 848.8
TG 50:0 852.8 850.8
TG 52:5 870.8 868.8
TG 52:4 872.8 870.8
TG 52:3 874.8 872.8
TG 52:2 876.8 874.8
TG 52:1 878.8 876.8
TG 54:6 896.8 894.8
TG 54:5 898.8 896.8
TG 54:4 900.8 898.8
TG 54:3 902.8 900.8
TG 54:2 904.8 902.8
d5-TG 44:1 771.7 769.7
d5-TG 48:1 827.7 825.7
d5-TG 50:0 857.8 855.8
d5-TG 58:10 949.8 947.8
118 Madlen Reinicke et al.

5. For further calculations, the summed NL intensities of each ion

peak are corrected through multiplication by the following
correction factors [5] (see Note 7): Z1 is the first 13C isotope
correction factor, where the difference between the carbon
number of a given TG species and the internal standard is
considered with the total acyl carbon number n in the TG
species of interest and the total acyl carbon number s of the
internal standard, which is, e.g., 50 for d5-TG 16:0/18:0/16:0:

1 þ 0:011n þ 0:011 2nðn1Þ

2

Z1 ¼
1 þ 0:011s þ 0:011 2s ðs1Þ
2

Z2 is the second 13C isotope correction factor and results from

the effect caused by the overlapping of the [Mþ2] isotope peak
with the molecular ion peak of a species, which has a 2-Da
higher mass. The general correction factor for this type of 13C
isotope effect is as follows:
I M 2 m ðm  1 Þ
Z2 ¼ 1   0:0112 
IM 2
where m is the total acyl carbon number in the TG species with
lower molecular mass and IM-2 and IM are peak intensities of
ions at molecular weight (M-2) and M, respectively. The signal
intensity of the cleaved neutral acyl residue is influenced by the
chain length and the degree of unsaturation. It has been shown
that the longer the acyl chain lengths and the lower the unsa-
turation index, the lower the intensity of TG species in
positive-ion ESI-MS. This effect is considered in the third
correction factor (Z3):
Z 3 ¼ 4:4979 þ 0:3441p  0:1269q  4:845  103 p  q
þ 9:9  104 q 2

where p is the total double bond number in the TG species and q is

the total acyl carbon number in the three acyl chains of the TG.
6. The m/z for each ion peak corresponds to a TG species. The
relative amount of the TG species can now be calculated using
the intensity ratio of the corrected summed NL intensities of
each ion peak and the corrected intensity of the internal stan-
dard, e.g., d5-TG 50:0 (16:0/18:0/16:0) with known concen-
tration (see Note 8).
7. The relative FA composition for a specific TG species charac-
terized by a corresponding m/z can be determined via the ratio
of the relative ion peak intensity of each NL for a certain m/z
and the corresponding summed NL intensity.
 Triglyceride Analysis 119

3.6 Validation Assess the validity of the complete method including sample prepa-
of the Analytical ration with the following experiments:
Process
1. Determine the limit of detection (LOD) and lower limit of
quantification (LLOQ) in a serial dilution or more appropri-
ately by spiking blank matrix according to concentration levels.
By the use of a serial dilution, the method’s linearity and the
linear dynamic concentration range can be assessed simulta-
neously. A signal-to-noise ratio of 3 is commonly accepted for
LOD estimation. LLOQ is defined as the analyte concentration
which can be measured with a coefficient of variation <20% (see
Note 9).
2. Assess within-day and between-day precision at two to three
concentration levels to test the reliability of the entire method.
Within-day precision: Prepare ten replicates per sample and
measure them in one run. Between-day precision: Prepare
and measure one replicate per sample in a single run on 10 con-
secutive working days.
3. Determine the recovery rate by spiking specified amounts of
TG standard in blank matrix or human material. Analyze these
samples as well as the starting material. Compare the difference
in calculated concentration of spiked sample and concentration
of starting material with the amount of analyte added.

4 Notes

1. The described method is a screening method with relative

quantification of the TG species containing the FAs implemen-
ted in the method, independent of the frequency and position
of the FA residue at the glyceride backbone. An absolute quan-
tification for a limited selection of TGs with a distinct FA
composition is possible if using an external calibration and
setting up a different MS/MS experiment, e.g., multiple reac-
tion monitoring.
2. An extension of the number of neutral loss experiments is
possible according to the monoisotopic mass of the FA of
interest, limited by the total scan time and the peak width.
3. All MS experiments can be performed in a modified way on
other MS platforms, e.g., linear ion trap mass spectrometers or
quadrupole time-of-flight mass spectrometers. However, spe-
cific experiments like enhanced product ion scans using the ion
trap function for ion accumulation can only be performed by
the use of hybrid triple quadrupole/linear ion trap mass
spectrometers.
4. Toluene and toluene/methanol had been used for standard
dilution and sample preparation due to the better solubility of
120 Madlen Reinicke et al.

TGs and higher ionization efficiency during the ESI process

than chloroform. The high vapor pressure of chloroform sup-
ports the evaporation process and negatively affects, e.g., the
sample preparation, solvent handling, and eluent storage.
Using tetrahydrofuran and n-hexane, we observed good TG
solubility but worse ionization on account of its lacking protic
properties.
5. If the quantification is performed via an external standard
calibration set, run the calibration standards including a blank
at the very beginning of each batch. The measurement of the
external standard calibration set should be repeated at the end
of each batch for quality assessment of the analysis.
6. For later calculations of the relative amount of each FA in a
specific TG species, save every single NL experiment with the
m/z and the peak intensity as txt.file, use, e.g., “List data,” and
select the peak list.
7. The selection of m/z for each ion peak including the
[M-2]-peak and calculation of the correction factors can be
done either manually using a common calculation program,
e.g., Microsoft Office Excel, or program a software, e.g.,
RStudio.
8. Several internal standards for different mass ranges could
be used.
9. Alternatively, a signal-to-noise ratio of 10 is commonly
accepted for LLOQ estimation.

Acknowledgment

This publication is supported by LIFE—Leipzig Research Center

for Civilization Diseases—Leipzig University. LIFE is funded by
means of the European Union, by the European Regional Devel-
opment Fund (ERDF), and by the Free State of Saxony within the
framework of the excellence initiative.

References

1. Erkkil€a A, de Mello VDF, Risérus U, Laakso- of fatty acids in triacylglycerols by hybrid mass
nen DE (2008) Dietary fatty acids and cardio- spectrometry. J Lipid Res 54(1):290–305.
vascular disease: an epidemiological approach. https://doi.org/10.1194/jlr.D031484
Prog Lipid Res 47:172–187. https://doi.org/ 4. Baiocchi C, Medana C, Dal Bello F (2015)
10.1016/j.plipres.2008.01.004 Analysis of regioisomers of polyunsaturated
2. Rhee EP et al (2011) Lipid profiling identifies a triacylglycerols in marine matrices by HPLC/
triacylglycerol signature of insulin resistance HRMS. Food Chem 166:551–560. https://
and improves diabetes prediction in humans. J doi.org/10.1016/j.foodchem.2014.06.067
Clin Invest 121(4):1402–1411. https://doi. 5. Han X, Gross RW (2001) Quantitative analysis
org/10.1172/JCI44442 and molecular species fingerprinting of triacyl-
3. Nagy K, Sandoz L, Destaillats F, Schafer O glyceride molecular species directly from lipid
(2013) Mapping the regioisomeric distribution extracts of biological samples by electrospray
 Triglyceride Analysis 121

ionization tandem mass spectrometry. Anal spectrometry for structural elucidation of tria-
Biochem 295(1):88–100. https://doi.org/ cylglycerols: assignment of fatty acyl groups on
10.1006/abio.2001.5178 the glycerol backbone and location of double
6. Krank J, Murphy RC, Barkley RM, bonds. J Am Soc Mass Spectrom 21
Duchoslav E, McAnoy A (2007) Qualitative (4):657–669. https://doi.org/10.1016/j.
analysis and quantitative assessment of changes jasms.2010.01.007
in neutral glycerol lipid molecular species 10. Sander M, Becker S, Thiery J, Ceglarek U
within cells. In: Brown HA (ed) Methods in (2015) Simultaneous identification and quan-
enzymology, vol 432. Academic Press, USA, pp tification of triacylglycerol species in human
1–20. https://doi.org/10.1016/S0076-6879 plasma by flow-injection electrospray ioniza-
(07)32001-6. tion tandem mass spectrometry. Chromatogra-
7. Fernandez C, Sandin M, Sampaio JL, phia 78(5):435–443. https://doi.org/10.
Almgren P, Narkiewicz K, Hoffmann M, 1007/s10337-014-2782-x
Hedner T, Wahlstrand B, Simons K, 11. Meyer VR (2010) Practical high-performance
Shevchenko A, James P, Melander O (2013) liquid chromatography. John Wiley & Sons,
Plasma lipid composition and risk of developing Ltd, Chichester, UK
cardiovascular disease. PLoS One 8(8):e71846 12. Leiker TJ, Barkley RM, Murphy RC (2011)
8. Schwudke D, Oegema J, Burton L, Entchev E, Analysis of diacylglycerol molecular species in
Hannich JT, Ejsing CS, Kurzchalia T, Shev- cellular lipid extracts by normal-phase LC-elec-
chenko A (2005) Lipid profiling by multiple trospray mass spectrometry. Int J Mass Spec-
precursor and neutral loss scanning driven by trom 305:103–108. https://doi.org/10.
the data-dependent acquisition. Anal Chem 78 1016/j.ijms.2010.09.008
(2):585–595. https://doi.org/10.1021/ 13. Quehenberger O et al (2010) Lipidomics
ac051605m. reveals a remarkable diversity of lipids in
9. Hsu F-F, Turk J (2010) Electrospray ioniza- human plasma. J Lipid Res 51:3299–3305.
tion multiple-stage linear ion-trap mass https://doi.org/10.1194/jlr.M009449
 Chapter 10

LC-MS/MS Analysis of the Epoxides and Diols Derived

from the Endocannabinoid Arachidonoyl Ethanolamide
Amy A. Rand, Patrick O. Helmer, Bora Inceoglu, Bruce D. Hammock,
and Christophe Morisseau

Abstract
Liquid chromatography-tandem mass spectrometry (LC-MS/MS) is a useful tool to characterize the
behavior of natural lipids within biological matrices. We report a LC-MS/MS method developed specifically
to analyze CYP products of the arachidonoyl ethanolamide (anandamide, AEA), the epoxyeicosatrienoic
acid ethanolamides (EET-EAs) and their hydrolyzed metabolites, and the dihydroxyeicosatrienoic acid
ethanolamides (DHET-EAs). This method was used to measure EET-EA biotransformation to DHET-EAs
by two human epoxide hydrolases: the soluble EH (sEH) and the microsomal EH (mEH). In general, sEH
and mEH substrate preference was similar, based on kcat/KM. The 14,15-EET-EA and 11,12-EET-EA were
the most efficiently hydrolyzed, followed by 8,9-EET-EA and 5,6-EET-EA. The method was also used to
detect endogenous levels of these lipids in mouse tissues, although levels were below the instrumental
detection limit (0.1–3.4 nM). Because both AEA and EETs are biologically active, the method described
herein will be invaluable in revealing the role(s) of EET-EAs in vivo.

Key words Liquid chromatography mass spectrometry, Anandamide, CYP450, Epoxide hydrolase,
EET-EAs, DHET-EAs, Limit of -detection, Enzyme kinetics, Bioactivity

1 Introduction

The endocannabinoid arachidonoyl ethanolamide (anandamide,

AEA) is a natural lipid-signaling molecule found in most tissues
that mediates neurological, immune, and cardiovascular functions
by primarily targeting cannabinoid receptors [1]. While it is primar-
ily metabolized by fatty acid amide hydrolase (FAAH) to arachido-
nic acid, it is also metabolized through several cytochrome P450s
(CYP450), including CYP3A4, CYP2B6, CYP2D6, CYP4F2,
CYP2J2, and CYP4X1 [2–4]. AEA is converted to monooxyge-
nated products in mouse liver and brain microsomes [5]. These
products include the epoxyeicosatrienoic acid ethanolamides
(EET-EAs), which were detected, upon incubation with AEA, in

Martin Giera (ed.), Clinical Metabolomics: Methods and Protocols, Methods in Molecular Biology, vol. 1730,
https://doi.org/10.1007/978-1-4939-7592-1_10, © Springer Science+Business Media, LLC 2018

123
124 Amy A. Rand et al.

human liver microsomes and brain mitochondrial fractions [6], as

well as in bovine and porcine heart microsomes [4].
While the biological function of EET-EAs needs to be further
explored, it is anticipated that they have important physiological
roles due to their structural similarity to the epoxyeicosatrienoic
acids (EETs) [7]. EETs are formed from CYP oxidation of arachi-
donic acid and have far-reaching physiological effects involved in
pain, inflammation, and kidney and cardiovascular diseases
[8, 9]. The 5,6-EET-EA is a potent agonist for the cannabinoid
receptor 2 [10], which may mediate brain and liver pathological
outcomes such as neurodegeneration and fibrosis, respectively, by
triggering anti-inflammatory responses [2]. While endogenous
levels of EET-EAs are uncertain and may be produced in an
activity-dependent manner, they are transformed by epoxide
hydrolases (EHs) to their corresponding dihydroxyeicosatrienoic
acid ethanolamides (DHET-EAs) [6, 10]. The biological activity of
DHET-EAs has not yet been reported but might be similar to the
less bioactive diols from EETs [11]. sEH inhibitors, which enhance
levels of EETs and other epoxy fatty acids, have similar activities to
cannabinoid receptor 2 agonists [12].
Analysis of EET-EAs and DHET-EAs is important to further
our understanding of their biological activity, metabolic fate, and
distribution within tissues. Here, we use two sample preparation
methods for analysis of EET-EAs and DHET-EAs by LC-MS/MS.
We report a sample dilution LC-MS/MS method to examine the
hydrolysis of EET-EAs to DHET-EAs by two human epoxide
hydrolases: the soluble EH (sEH) and the microsomal EH
(mEH). While the EET-EAs had similar turnover with both sEH
and mEH (Table 1), enzyme efficiency for each substrate (kcat/KM)
varied; the 14,15-EET-EA and 11,12-EET-EA were the most effi-
ciently hydrolyzed followed by 8,9-EET-EA and 5,6-EET-EA

Table 1
Kinetic constants of recombinant purified human sEH and mEH for the hydrolysis of all the
regioisomers of 14,15-EET-EAs

Enzyme Substrate KM (μM) r2 kcat (s
1) kcat/KM (s
1 μM
1)

Human sEH 14,15-EET-EA 16  1 0.99 3.32  0.02 0.20  0.01
11,12-EET-EA 19  6 0.92 6.0  0.8 0.31  0.07
8,9-EET-EA 28  3 0.99 1.9  0.1 0.068  0.009
5,6-EET-EA 22  4 0.96 0.22  0.02 0.010  0.002
Human mEH 14,15-EET-EA 3.8  0.4 0.99 0.61  0.02 0.16  0.02
11,12-EET-EA 1.2  0.3 0.99 0.17  0.01 0.15  0.04
8,9-EET-EA 6.2  0.6 0.99 0.10  0.01 0.016  0.002
5,6-EET-EA 10.2  1.4 0.99 0.11  0.01 0.010  0.002
Constants (KM and kcat) were calculated by nonlinear fitting of the Michaelis-Menten equation using the enzyme kinetic
module of SigmaPlot version 11. Results are value  error
 LC-MS/MS Analysis of EET-EAs and DHET-EAs 125

Fig. 1 Determination of the kinetic constants for 5,6-EET-EA; 8,9-EET-EA; 11,12-

EET-EA; and 14,15-EET-EA, with the human sEH ([HsEH]final ¼ 4 nM) in NaPO4
buffer (0.1 M pH 7.4 containing 0.1 mg/mL of BSA) at 37  C. Values were
obtained from sextuplicate analysis (average  standard deviation)

(Fig. 1 and Table 1). We also use solid-phase extraction (SPE),

paired with the reported LC-MS/MS method, to detect endoge-
nous levels of EET-EAs and DHET-EAs in mouse liver, kidney,
spleen, and brain tissue. Although levels of the EET-EAs and
DHET-EAs were below the instrumental limits of detection, the
method can be applied to further work using enzymatic inhibitors
and/or transgenic mice to raise levels of EET-EAs and DHET-EAs
and to understand EET-EA and DHET-EA endogenous distribu-
tion and biological function in endocannabinoid signaling.

2 Materials

All solvents used should be Optima grade or 18 MΩ ultrapure

water, unless otherwise indicated.

2.1 Standards 1. 5,6-; 8,9-; 11,12-; and 14,15-EET-EA and standard solutions
for Liquid (Cayman Chemical, Ann Arbor, MI, USA): Make up as 5 μM
Chromatography- solution in methanol and store at
20  C under nitrogen gas
Tandem Mass until use (see Note 1).
Spectrometry 2. AEA standard solution (Cayman Chemical, Ann Arbor, MI,
(LC-MS/MS) USA): Make up as 5 μM solution in methanol and store at

20  C under nitrogen gas until use.
126 Amy A. Rand et al.

3. 5,6-; 8,9-; 11,12-; and 14,15-DHET-EA standards were

prepared by enzymatic hydrolysis from their respective
EET-EAs. In 10  75 mm glass tube, to 100 μL of purified
recombinant mouse sEH [13] ([MsEH]final ¼ 0.6 μM,
0.04 mg/mL) in sodium phosphate buffer (0.1 M pH 7.4)
containing 0.1 mg/mL of BSA, 1 μL of solutions of individual
EET-EAs in DMSO was added ([S]final ¼ 0–2.5 μM). The
tubes were sealed and incubated at 37  C for 2 h. The reaction
was then stopped by adding 100 μL of chilled methanol con-
taining 0.5 μM of AEA as internal standard. The mixture was
then centrifuged (3000  g  5 min) to remove any precipi-
tated salt. The supernatant was transferred to vials, which were
kept at
20  C until LC-MS/MS analysis.

2.2 Tissue Collection 1. Liver, spleen, brain, and kidney tissues from saline-perfused
Swiss Webster mice were collected just after sacrifice of the
animals and immediately stored at
80  C until analysis. All
procedures and animal care adhered to the National Institutes
of Health Guide for the Care and Use of Laboratory Animals
(NIH Publications 8th Edition 2011) and were performed in
accordance with the protocols approved by the International
Animal Use and Care Committee (IACUC) of the University
of California, Davis.
2. The whole liver, spleen, brain, and kidney were used from mice
(n ¼ 3). Frozen tissues were cut into small pieces with a clean
razor blade. Tissues were weighted (100–260 mg) and allo-
cated to 2 mL polypropylene microtubes. Tissues were sus-
pended in 400 μL ice-cold methanol containing 0.1%
concentrated acetic acid.
3. Antioxidant solution was prepared by combining 0.6 mg/mL
butylated hydroxytoluene (BHT), 0.6 mg/mL triphenylpho-
sphine (TPP), and 0.6 mg/mL ethylenediaminetetraacetic acid
(EDTA) in DMSO to give a final concentration of 0.2 mg/mL
in methanol. A 10 μL aliquot was added to each tissue sample
to prevent autoxidation.

2.3 Solid-Phase 1. Oasis HLB 3 mL 60 mg cartridges (Waters Corporation, Mil-

Extraction (SPE) ford, MA, USA).
2. Water/methanol (8:2) with 0.1% acetic acid: Fill a 1 L
graduated cylinder with 800 mL pure water, 200 mL metha-
nol, and 1 mL concentrated acetic acid and store in a 1 L
reagent bottle.
3. 30% glycerol solution: Measure 6 mL glycerol and 14 mL
methanol in a 25 mL graduated cylinder and store in a 20 mL
glass screw-cap vial.
 LC-MS/MS Analysis of EET-EAs and DHET-EAs 127

4. Methanol: Store methanol in a 500 mL reagent bottle.

5. Ethyl acetate: Store ethyl acetate in a 500 mL reagent bottle.
6. SPE vacuum manifold processing station.

2.4 Liquid 1. LC-MS/MS: We used a Waters Acquity Ultra Performance

Chromatography- Liquid Chromatograph coupled to a Waters Xevo TQ-S Mass
Tandem Mass Spectrometer (Waters Corporation, Milford, MA, USA).
Spectrometry 2. Liquid chromatography column: Kinetex 1.7 μm XB-C18
(LC-MS/MS) (100 Å,150  2.1 mm) with a SecurityGuard ULTRA car-
tridge for C18 UPLC, sub-2 μm with 2.1 mm internal dia-
meters (Phenomenex, Torrance, CA, USA).
3. Water with 0.1% acetic acid and 1 mM ammonium acetate: Fill
a 1 L graduated cylinder with 999 mL LC-MS grade water,
1 mL acetic acid, and 77 mg ammonium acetate and store in a
1 L LC glass solvent bottle.
4. Methanol with 0.1% acetic acid and 1 mM ammonium acetate:
Fill a 1 L graduated cylinder with 999 mL LC-MS grade
methanol, 1 mL acetic acid, and 77 mg ammonium acetate
and store in a 1 L LC glass solvent bottle.
5. Glass screw top vials: 2 mL, 32  12 mm, and 9 mm Amber
Glass Screw Thread Vials are used.
6. Glass inserts for screw top vials: Use a 0.150 mL glass insert
with attached plastic spring purchased.
7. Caps for screw top vials: Use 9 mm thread cap with bonded
plastic PTFE/silicone septa with pre-slit.

3 Methods

3.1 Processing 1. Kinetic parameters for the 5,6-EET-EA; 8,9-EET-EA; 11,12-

of Enzyme Kinetic EET-EA; and 14,15-EET-EA were determined under steady-
Samples state conditions using partially purified recombinant human
sEH (95% pure) and human mEH (80% pure) [14, 15]. Just
before usage, the human sEH was diluted in chilled sodium
phosphate buffer (0.1 M pH 7.4) containing 0.1 mg/mL BSA
to a final concentration of 4 nM (0.25 μg/mL). Whereas, the
human mEH was diluted in chilled Tris/HCl buffer (0.1 M,
pH 9.0) containing 0.1 mg/mL of BSA to a final concentration
of 200 nM (10 μg/mL). Enzyme solutions were kept on ice
until usage.
2. A series of substrate (individual EET-EA) stock solutions were
prepared in DMSO. In general, 50–100 μL of eight substrate
concentrations ranging from 0 to 5 mM were prepared.
3. In 10  75 mm glass tubes, to 100 μL of enzyme solution or
buffer for control samples, 1 μL of solutions of individual
128 Amy A. Rand et al.

EET-EAs in DMSO was added ([S]final ¼ 0–50 μM). The

mixture was sealed and incubated at 37  C for 15–120 min
depending on substrate and enzyme. Incubation time was
optimized to ensure the enzymatic reaction is in steady-state
conditions, with less than 5% of the initial substrate metabo-
lized during the whole course of the hydrolysis. The reaction
was then stopped by adding 100 μL of chilled methanol con-
taining 0.5 μM of AEA as internal standard.
4. The mixture was then centrifuged (3000  g  5 min) to
remove any precipitated salt. The supernatant was transferred
to a vial, which was kept at
20  C until LC-MS/MS analysis.
5. The quantity of product (DHET-EA) formed was determined
by LC-MS/MS analysis (Subheading 3.3). The kinetic con-
stants (KM and kcat) were calculated by nonlinear fitting of the
Michaelis-Menten equation (v ¼ (kcat.[E].[S])/(KM + [S]))
using the enzyme kinetic module of SigmaPlot 11.0. Reactions
were done in hexaplicate (n ¼ 6).

3.2 Tissue Extraction 1. Between 100 and 260 mg of tissues are weighted in 2 mL
of EET-EAs and plastic vials. Then, 10 μL of the BHT, TPP, and EDTA antioxi-
DHET-EAs dant solution was added as well as 400 μL ice-cold methanol
containing 0.1% acetic acid. To correct for possible analyte loss
throughout the extraction, 5 μL of 1 μM AEA was added as an
internal standard (see Note 2) just prior to tissue
homogenization.
2. Tissue samples are homogenized by placing two steel beads
into each vial and shaken using a bead beater operating at
30 Hz for 10 min (see Note 3). The resulting homogenate
solution was centrifuged at 13,200  g  10 min.
3. The supernatant is collected into 2 mL polypropylene vials, and
the remaining tissue pellet is washed with 100 μL ice-cold
methanol containing 0.1% acetic acid by vortexing for 1 min
to break up and resuspend the pellet (see Note 4). Samples are
centrifuged again at 13, 200  g  10 min. This supernatant
was combined with the first to give a total volume approximat-
ing 0.5 mL. Samples are diluted with 1.5 mL ice-cold water
and stored on ice before loading onto the SPE cartridges (see
Note 5).
4. Oasis HLB cartridges (3 mL, 60 mg) are mounted on a vacuum
manifold processing station under vacuum (~5 bar). Cartridges
are preconditioned to remove impurities with 3 mL ethyl ace-
tate and 6 mL methanol [16].
5. Cartridges are equilibrated by loading 6 mL of water/metha-
nol (8:2) with 0.1% acetic acid under atmospheric pressure (see
Note 6).
 LC-MS/MS Analysis of EET-EAs and DHET-EAs 129

6. Samples are loaded onto the cartridge and left to filter at

atmospheric pressure through the cartridge. The approximate
flow rate is approximately 1 drop/3 s, monitored by watching
the drip rate at the bottom of the cartridge. If the rate is less
than this, light vacuum can be applied (see Note 7).
7. Polar substances from the tissue extracts are eluted through the
column using 6 mL of water/methanol (8:2) with 0.1%
acetic acid.
8. Apply approximately 10 bar vacuum to the manifold to dry the
cartridges (20 min).
9. While cartridges are drying, prepare 2 mL polypropylene
Eppendorf vials for the sample eluent, by adding 6 μL 30%
glycerol solution in methanol. Label each vial with their respec-
tive sample name and place in the vacuum manifold, just below
each sample cartridge.
10. EET-EAs, DHET-EAs, and AEA are eluted with 0.5 mL meth-
anol followed by 1.5 mL ethyl acetate into the 2 mL Eppendorf
vials at atmospheric pressure. Apply a 10 bar vacuum to elute
the remaining solvent from the cartridge.
11. Samples are evaporated under vacuum using a ScanVac Speed-
Vac (Neutec Group Inc., Farmingdale, NY) for approximately
2 h (see Note 8).
12. Samples are reconstituted in 50 μL methanol in a 2 mL poly-
propylene Eppendorf tube, filtered using Durapore PVDF
0.1 μm filter units (Millipore, Cork, Ireland), and centrifuged
at 13,200  g  10 min (see Note 9).
13. Filtrates are transferred to 0.15 mL glass inserts in a 2 mL
screw top vial and cap. Samples are stored at
20  C until
LC-MS/MS analysis.
14. Analyte recovery was validated by performing a spike-and-
recovery test, the values of which are given in Table 2 (see
Note 10).

3.3 LC-MS/MS 1. Samples are analyzed on a Waters Acquity Ultra Performance

Analysis Liquid Chromatograph coupled to a Waters Xevo TQ-S Mass
Spectrometer in positive electrospray ionization mode. Samples
are injected (5 μL) and separated using a Phenomenex Kinetex
column (150  2.1 mm; 1.7 μm) at 40  C using the following
mobile phase gradient, consisting of LC-MS grade water
(A) and optima LC-MS grade methanol (B) each containing
0.1% acid and 0.1 mM ammonium acetate: initial conditions of
70:30 A/B for 2.9 min (t ¼ 2.9 min), changing to 45:55 at
3 min (t ¼ 3 min) and decreasing to 35:65 over 5.5 min
(t ¼ 8.5 min), decreasing to 28:72 over 4 min (t ¼ 12.5 min),
decreasing to 18:82 over 2.5 min (t ¼ 15 min), decreasing to
130 Amy A. Rand et al.

Table 2
LC-MS/MS parameters, instrumental limits of detection and quantification, and analyte recovery
from mouse liver tissue

Cone Collision Limit of Limit of Analyte

Retention voltage energy detection quantification recovery
Analyte time (min) MRM transition (V) (V) (nM) (nM) (%)

14,15-DHET-EA 12.56 383.0 > 285.0 (165.0) 16 12 (32) 3.4 11 163  35

11,12-DHET-EA 13.22 383.0 > 285.0 (165.0) 16 12 (32) 1.7 5.7 117  15

8,9-DHET-EA 14.13 383.0 > 285.0 (165.0) 16 12 (32) 2.9 9.8 88  8

5,6-DHET-EA 15.13 383.0 > 285.0 (165.0) 16 12 (32) 5.4 18 167  18

14,15-EET-EA 15.77 346.3 > 62.2 (285.2) 56 16 (16) 0.11 0.38 93  11

11,12-EET-EA 16.17 346.3 > 62.2 (285.2) 56 16 (16) 1.6 5.4 82  11

8,9-EET-EA 16.40 346.3 > 62.2 (285.2) 56 16 (16) 0.42 1.4 77  14

5,6-EET-EA 16.74 346.3 > 62.2 (285.2) 56 16 (16) 0.79 2.6 11  3

AEA 17.79 348.1 > 62.2 15 25

Analytes measured were all the regioisomers of DHET-EAs and EET-EAs, extracted from tissues using AEA as an internal
standard. Mass spectra were collected in multiple reaction monitoring (MRM) mode, with two analyte transitions used
for quantification and qualification (shown in brackets). Limits of detection and quantification were obtained from linear
regression by calculating the concentration at which the signal is three and ten times greater than the noise, respectively.
Recoveries were calculated after spiking a known concentration (25 nM) of analytes into mouse liver tissue and extracted
using solid-phase extraction (SPE) prior to LC-MS/MS analysis. Values were obtained from triplicates and are given as
average  standard error

5:95 over 1.5 min (t ¼ 16.5 min), holding at 5:95 for 1.5 min
(t ¼ 18 min), reverting to initial conditions of 70:30 in 0.1 min
(t ¼ 18.1 min), and re-equilibrating for 2.9 min (t ¼ 21 min).
Mass spectral analysis was accomplished using a capillary volt-
age of 3 kV, a desolvation temperature of 200  C, a desolvation
gas (nitrogen 99.99%) flow of 800 L/h, a cone gas flow of
150 L/h, nebulizer pressure of 6 bar, and collision gas flow of
0.15 mL/min. LC-MS/MS parameters MRM transitions for
each analyte are shown in Table 2, with analyte dwell times of
25 ms. A representative spectrum of the separated EET-EAs
and DHET-EAs is shown in Fig. 2.
2. EET-EAs and DHET-EAs are detected by retention time and
mass transitions, values of which are given in Table 2.
3. Analyte concentration was calculated using a matrix-matched
and non-matrix-matched standard curve with known concen-
trations (2.5–60 nM) (see Notes 10 and 11). The area of all
analytes was achieved through peak integration and dividing its
value by the area of the AEA internal standard.
 LC-MS/MS Analysis of EET-EAs and DHET-EAs 131

Fig. 2 LC-MS/MS total ion chromatograph of a 60 nM standard mixture contain-

ing all the regioisomers of DHET-EAs and EET-EAs

4 Notes

1. Store lipid standards under nitrogen at
20  C to minimize

oxidation.
2. Deuterated standards are not commercially available for
EET-EAs and DHET-EAs; therefore, we used AEA (Cayman
Chemical, Ann Arbor, MI, USA) as an internal standard for
quantifying EET-EAs and DHET-EAs, due to its similar size
and structure. We used non-labeled AEA since endogenous
levels of AEA were not detected in any tissue. Alternatively,
the d4-AEA (Cayman Chemical, Ann Arbor, MI, USA) can be
used as an internal standard for extracting EET-EAs and
DHET-EAs, especially if there are levels of endogenous AEA
in the tissue samples.
3. Ensure vials are tightly capped; otherwise, the sample mixture
will be lost during homogenization.
132 Amy A. Rand et al.

4. Extracting the pellet twice with methanol increases the analyte

recovery from the tissue matrix.
5. Dilute samples with water before loading onto the SPE col-
umns. This will make the sample more polar and facilitate
nonpolar analyte adsorption onto the cartridge stationary
phase.
6. Upon equilibration, it is important that the solvent does not go
below the column frit at the top of the stationary phase. Doing
so will leave the top of the stationary phase dryer than the rest,
changing its physical composition and thus the sorption behav-
ior of the analytes, leading to variations in analyte recovery.
7. Generally, extracting the sample at a flow rate of 1 drop/3 s is
acceptable. However, samples may be eluted through the col-
umn at slower rates depending on the matrix and cartridge
composition. If this is the case, light vacuum can be applied
to maintain the desired flow rate.
8. Ensure all solvents are removed prior to reconstitution.
9. Filtering removes any particulates from the samples to maintain
LC column integrity.
10. Analyte recovery and matrix-matched calibration: To deter-
mine the extraction efficiency of the method described above,
we performed a spike-and-recovery test. The EET-EA and
DHET-EA stock solutions were diluted and combined to
form a solution having a final concentration of 500 nM. The
liver (approximately 600 mg) was cut and homogenized in
methanol. 100 μL of the homogenate was spiked with 2.5 μL
of the 500 nM EET-EA and DHET-EA solution in triplicate.
5 μL of the 500 nM AEA internal standardwas also added to
the spiked tissue samples. The rest of the unspiked liver
homogenate was divided and extracted to create a matrix-
matched calibration curve. All tissue samples (spiked and
unspiked) were extracted and analyzed using the SPE-LC-
MS/MS- method described above. Samples were reconstituted
in 50 μL methanol. Matrix-matched standards were made by
adding 0–6 μL of the 500 nM EET-EA and DHET-EA stock
solution to generate a concentration range from 0 to 60 nM, as
well as 5 μL of the 500 nM AEA. The final volume for each
standard was brought to 50 μL with methanol. A non-matrix-
matched calibration curve was produced in methanol (rather
than reconstituted matrix) using the same concentrations
(0–60 nM with 50 nM AEA).
11. Matrix-matched calibration curves are used to correct for
matrix effects within the samples. In our case, there was no
significant matrix suppression or enhancement when compared
to the non-matrix-matched calibration curve; therefore, either
curve can be used for quantification.
 LC-MS/MS Analysis of EET-EAs and DHET-EAs 133

Acknowledgments

This work received support in part from the National Institute of

Environmental Health Sciences R01 ES002710 and NIEHS Super-
fund Program P42 ES004699. A. Rand acknowledges support
from the OSCB training grant NIH/NIEHS T32 CA108459 and
from the 2016 AACR Judah Folkman Fellowship for Angiogenesis
Research, Grant Number 16-40-18-RAND.

References
1. Mechoulam R, Parker LA (2013) The endo- 9. Morisseau C, Hammock BD (2013) Impact of
cannabinoid system and the brain. Annu Rev soluble epoxide hydrolase and epoxyeicosa-
Psychol 64:21–47 noids on human health. Annu Rev Pharmacol
2. Snider NT, Walker VJ, Hollenberg PF (2010) Toxicol 53:37–58. https://doi.org/10.1146/
Oxidation of the endogenous cannabinoid ara- annurev-pharmtox-011112-140244
chidonoyl ethanolamide by the cytochrome 10. Snider NT, Nast JA, Tesmer LA, Hollenberg
P450 monooxygenases: physiological and PF (2009) A cytochrome P450-derived epox-
pharmacological implications. Pharmacol Rev ygenated metabolite of anandamide is a potent
62:136–154 cannabinoid receptor 2-selective agonist. Mol
3. Sridar C, Snider NT, Hollenberg PF (2011) Pharmacol 75:965–972
Anandamide oxidation by wild-type and poly- 11. Bellien J, Joannides R (2013) Epoxyeicosatrie-
morphically expressed CYP2B6 and CYP2D6. noic acid pathway in human health and dis-
Drug Metab Dispos 39:782–788 eases. J Cardiovasc Pharmacol 61:188–196
4. McDougle DR, Kambalyal A, Meling DD, Das 12. Wagner K, Inceoglu B, Hammock BD (2011)
A (2014) Endocannabinoids anandamide and Soluble epoxide hydrolase inhibition, epoxyge-
2-arachidonoylglycerol are substrates for nated fatty acids and nociception. Prostaglan-
human CYP2J2 epoxygenase. J Pharmacol dins Other Lipid Mediat 96:76–83
Exp Ther 351:616–627 13. Argiriadi MA, Morisseau C, Hammock BD,
5. Bornheim LM, Kim KY, Chen B, Correia MA Christianson DW (1999) Detoxification of
(1995) Microsomal cytochrome P450- environmental mutagens and carcinogens:
mediated liver and brain anandamide metabo- structure, mechanism, and evolution of liver
lism. Biochem Pharmacol 50:677–686 epoxide hydrolase. Proc Natl Acad Sci
6. Snider NT, Kornilov AM, Kent UM, Hollen- 96:10637–10642
berg PF (2007) Anandamide metabolism by 14. Morisseau C, Bernay M, Escaich A et al (2011)
human liver and kidney microsomal cyto- Development of fluorescent substrates for
chrome p450 enzymes to form hydroxyeicosa- microsomal epoxide hydrolase and application
tetraenoic and epoxyeicosatrienoic acid to inhibition studies. Anal Biochem
ethanolamides. J Pharmacol Exp Ther 414:154–162
321:590–597 15. Morisseau C, Wecksler AT, Deng C et al
7. Zelasko S, Arnold WR, Das A (2015) Endo- (2014) Effect of soluble epoxide hydrolase
cannabinoid metabolism by cytochrome P450 polymorphism on substrate and inhibitor selec-
monooxygenases. Prostaglandins Other Lipid tivity and dimer formation. J Lipid Res
Mediat 116:112–123 55:1131–1138
8. Imig JD, Hammock BD (2009) Soluble epox- 16. Yang J, Schmelzer K, Georgi K, Hammock BD
ide hydrolase as a therapeutic target for cardio- (2009) Quantitative profiling method for oxy-
vascular diseases. Nat Rev Drug Discov lipin metabolome by liquid chromatography
8:794–805 electrospray ionization tandem mass spectrom-
etry. Anal Chem 81:80–85
 Chapter 11

Sphingolipid Analysis in Clinical Research

Bo Burla, Sneha Muralidharan, Markus R. Wenk, and Federico Torta

Abstract
Sphingolipids are the most diverse class of lipids due to the numerous variations in their structural
components. This diversity is also reflected in their extremely different functions. Sphingolipids are not
only constituents of cell membranes but have also emerged as key signaling molecules involved in a variety
of cellular functions, such as cell growth and differentiation, proliferation, and apoptotic cell death.
Lipidomic analyses in clinical research have identified pathways and products of sphingolipid metabolism
that are altered in several human pathologies. In this article, we describe how to properly design a lipidomic
experiment in clinical research, how to handle plasma and serum samples for this purpose, and how to
measure sphingolipids using liquid chromatography-mass spectrometry.

Key words Sphingolipids, Mass spectrometry, Lipidomics, Sphingolipidomics, Ceramide, Sphingo-

myelin, Glucosylceramide, Sphingosine-1-phosphate, Clinical mass spectrometry, Quality control

1 Introduction

Sphingolipids (SL) represent a diverse class of molecules that

includes lipids with both structural (e.g., sphingomyelins) and
signaling functions (e.g., sphingosine-1-phosphate). They can be
found in cell membranes and lipoproteins in the blood [1].
The first step in the de novo synthesis of SL takes place in
the endoplasmic reticulum, involves the condensation of a
serine residue with a fatty acyl-CoA, and is catalyzed by serine
palmitoyltransferase (SPT). The product of this reaction is
3-ketodihydrosphingosine. A subsequent enzymatic reaction gen-
erates the long-chain base (LCB) dihydrosphingosine that then can
be N-acylated with fatty acids of various lengths to give dihydrocer-
amides and, after desaturation, ceramides, the precursors of com-
plex SL (such as sphingomyelins, glycosphingolipids, and
ceramide-1-phosphate). The final irreversible step for SL degrada-
tion, and recycling to glycerolipids, is catalyzed by the enzyme S1P
lyase that degrades sphingosine-1-phosphate (S1P) into a fatty
aldehyde (hexadecenal) and phosphoethanolamine [1].

Martin Giera (ed.), Clinical Metabolomics: Methods and Protocols, Methods in Molecular Biology, vol. 1730,
https://doi.org/10.1007/978-1-4939-7592-1_11, © Springer Science+Business Media, LLC 2018

135
136 Bo Burla et al.

Since numerous LCBs, with different carbon chain length,

position, and number of double bonds and hydroxylations, are
acylated with different fatty acids (between C14 and C26), SL
include extremely diverse molecules (>300 molecular species iden-
tified). This heterogeneity is even increased by the different mod-
ifications on the headgroup (glycosylation). SL are lower in
abundance (<20%) when compared to their glycerolipid counter-
parts, but they are very active and play key roles in the regulation of
cell and tissue functions.
Many sphingolipid and phospholipid functions have been
linked to pathological processes. In particular, genetic disorders
linked to the sphingolipid pathway are associated with cardiovascu-
lar and metabolic diseases [2]. Advances in mass spectrometry-
based lipidomics have greatly expanded our understanding of the
extent and complexity of lipid dysregulation in disease conditions.
This technology enables the measurement of several hundred indi-
vidual lipid species in thousands of samples for each study
[3, 4]. Lipidomics in population-based studies has helped to clarify
that concentrations of specific SL are altered in several metabolic
disorders and may serve as prognostic and diagnostic markers
[5]. Ceramides have been implicated in mediating or regulating
numerous central cellular processes such as proliferation, differen-
tiation and senescence, stress responses, apoptosis, and inflamma-
tion [6]. Harmful stimuli can activate these bioactive signaling
lipids leading to numerous pathophysiological states, including
inflammatory diseases such as obesity [7].
More specifically, four molecular ceramides have been found to
be closely linked to cardiovascular conditions [8] and can be used as
markers in stratifying a patient’s risk for fatal outcome of coronary
artery disease (CAD).
Alanine and glycine can also be used, instead of serine, as
substrates by SPT and condensed with fatty acyl-CoA to produce
deoxysphingolipids. Mutations in SPT can alter its preference for
the substrate, and, as a consequence, elevated levels of these lipids
can be present in plasma. This is the case of hereditary sensory and
autonomic neuropathy type 1, in which increased plasma levels of
deoxysphingolipids characterize this rare genetic disease [9]. Deox-
ysphingolipid levels are also elevated in type 2 diabetes mellitus.
Thus targeting deoxysphingolipid synthesis could complement the
currently available therapies for this pathology [10].
Altered glycosphingolipid metabolism is associated with com-
mon diseases of the kidney, in particular diabetic nephropathy,
polycystic kidney disease, and renal cell carcinoma. Patients with
these diseases show high levels of glucosylceramide, lactosylcera-
mide, and gangliosides in the kidneys. More importantly from a
therapeutic point of view, inhibition of glucosylceramide synthase
 Clinical Sphingolipidomics 137

with the use of small molecules reverses the abnormal renal phe-
notypes [11]. An important sphingolipid that has been involved in
many clinical research studies and functions as a regulator of pro-
cesses such as immune cell trafficking, angiogenesis, and vascular
permeability is S1P [12]. Our group has recently published a new
very sensitive method for its detection that expanded the number
of S1P molecular species usually detected in biological
samples [13].
In MS-based lipidomics, the two most common approaches
include analysis of lipids either by direct infusion (shotgun MS) or
after chromatographic separation (LC-MS/MS). Both approaches
have their limitations and advantages [14]. As the purpose of this
manuscript is to guide users in the analysis of specific molecules
(sphingolipids) that are present at low levels in plasma and serum,
we will describe here an approach based on LC-MS/MS, the most
commonly used technique for sensitive targeted studies. Several
specific methods for sphingolipid extraction have been published
in the last few years and have improved the detection of the low
abundant species [15, 16]. In this manuscript, we limit our
approach to a simple and fast single-phase extraction protocol
that is effective for many lipid classes. We think that when dealing
with high number of samples, as is often the case in clinical studies,
a fast extraction procedure might be the best choice [17].
Blood serum and plasma are the sample types most com-
monly used in clinical research studies, and for this reason we
will focus our method description on this kind of samples. Please
notice that plasma and serum are different matrices that have
different sphingolipid concentrations. One should be aware that
they are not comparable, and a choice driven by the scope of the
study should be made before starting the sample collection. As in
any clinical research study, the selection of the subjects should be
balanced, knowing that age, gender, and ethnicity have an impact
on the SL profiles [18, 19]. The level of specific lipids can also be
affected by diet, medications, circadian rhythm, and other factors
that should be documented as metadata to improve the final
analysis. All these considerations have been described and care-
fully considered in previous reports, as the use of lipidomics (and
metabolomics) in clinical research studies has been the focus of
many publications in the last few years [3, 20, 21].We describe
here a recommended approach for sphingolipidomics studies in
clinical research samples, including suggestions about sampling,
storage, analytical design, use of quality controls, and data analy-
sis. Two different analytical methods, one for extraction/analysis
of the most common sphingolipid species and one specific for
S1P (for which higher sensitivity is required), are described [13].
138 Bo Burla et al.

2 Materials

2.1 Blood Collection 1. Plasma: Blood collection container (e.g., BD Vacutainer and
Sarstedt S-Monovette®) with spray-dried K2EDTA (violet cap)
or buffered sodium citrate solution (0.109 M/3.2%, light
blue cap). Heparin is not recommended as anticoagulant.
2. Serum: Blood collection container (e.g., BD Vacutainer and
Sarstedt S-Monovette®) with or without spray-coated silica as
clot activator.
3. Biosafety cabinet.
4. Centrifuges for blood collection tubes.

2.2 Sample 1. Chemical fume hood.

Extraction 2. Eppendorf® Safe-Lock 2 mL polypropylene tubes.
and Derivatization
3. Ultrasonic bath.
4. Eppendorf ThermoMixer® or equivalent.
5. Microcentrifuge.

2.3 Liquid 1. UHPLC system: Agilent 1290 Infinity LC (Agilent Technolo-

Chromatography- gies, Santa Clara, USA).
Mass Spectrometry 2. Triple-quadrupole mass spectrometer: Agilent 6490 or
Systems 6495 QQQ.
3. Chemically inert autosampler vials (glass or polypropylene)
with airtight PTFE-sealed screw septum caps.
4. Agilent ZORBAX RRHD Eclipse Plus C18, 95 Å, 2.1

100 mm, 1.8 μm UPLC column.
5. Waters ACQUITY UPLC BEH C18 HILIC 130 Å, 2.1

100 mm, 1.7 μm UPLC column.
6. Schott Duran® glass bottles for preparation and storage of
mobile phases.

2.4 Reagents All compounds should be of analytical or better grade quality.

and Solutions Solvents (acetonitrile, methanol, 2-propanol, formic acid) must
be of LC-MS grade. Ultrapure water should be of LC-MS grade
or Milli-Q water (18 MΩ cm).
Internal standards (ISTDs)
1. S1P-13C2D2 solution: Dissolve vial content (1 mg) of D-ery-
thro-Sphingosine-1-phosphate-13C2,D2 (# S681502, Tor-
onto Research Chemicals, Toronto, Canada) in 10 mL
methanol by vortexing and sonication until solution is clear.
Store at 20 C.
 Clinical Sphingolipidomics 139

Table 1
Composition of the BUME + ISTD mix

Concentration in
Internal standards (IS) Monoisotopic mass Chemical formula BUME + ISTDs (nM)
Sph d17:1 285.2668 C17H35NO2 250
Sph d17:0 287.2824 C17H37NO2 250
S1P d17:1 365.2331 C17H36NO5P 250
S1P d17:0 367.2488 C17H38NO5P 250
LacCer d18:1/12:0 805.5551 C45H79NO3 250
SM d18:1/12:0 646.5050 C35H71N2O6P 250
GluCer d18:1/12:0 643.5023 C36H69NO8 250
Cer d18:1/12:0 481.4495 C30H59NO3 250
C1P d18:1/12:0 561.4158 C30H60NO6P 250
13 13
S1P d18:1 C2D2 383.2487 C16 C2H36D2NO5P 52.1

2. Cer/Sph mixture II: Ceramide/Sphingoid Internal Standard

Mixture II (#LM-6005, Avanti Polar Lipids, USA). Store at
20 C.
Sample processing and lipid extraction
3. BHT solution 100 mM: Dissolve 220.4 mg of 3,5-di-tert-4-
butylhydroxytoluene in 10 mL ethanol. Store in aliquots in
microtubes at 20 C.
4. BUME: Mix 1 volume 1-butanol (HPLC grade) with 1 volume
methanol and add 5 mM (315.3 mg/L) ammonium formate
(LC-MS grade). Store at room temperature in a glass bottle.
5. BUME + ISTDs: Add 100 μL Cer/Sph mixture II and 2 μL
S1P-13C2D2 solution to 9.9 mL BUME (see Table 1).
6. TMS-diazomethane solution: (Trimethylsilyl)diazomethane,
2 M solution in hexanes (e.g., #AC385330250, ACROS
Organics). Store at room temperature in the original bottle in
a safety cabinet or ventilated solvent cupboard.
7. Acetic acid, glacial (>99%).
8. Acetic acid 10% in methanol (for inactivation of excess TMS
solution).
Mobile phases for sphingolipid LC-MS analyses
9. Mobile phase A for sphingolipid analysis: Mix 600 mL metha-
nol with 400 mL ultrapure water and 2 mL formic acid
(98–100%) in a glass bottle, and then add 1.33 mL 7.5 M
140 Bo Burla et al.

ammonium acetate solution (e.g., #A2706, Sigma-Aldrich).

Mix well and sonicate for 10 min.
10. Mobile phase B for sphingolipid analysis: Mix 600 mL meth-
anol with 400 mL of 2-propanol and 2 mL formic acid in a
glass bottle, and then add 1.33 mL 7.5 M ammonium acetate
solution. Mix well and sonicate for 10 min.
Mobile phases for S1P LC-MS analyses
11. Ammonium formate buffer 25 mM pH 4.6: Dissolve
1575 mg ammonium formate (LC-MS grade) in approxi-
mately 800 mL ultrapure water. Adjust the pH to 4.6 by
adding formic acid 100% (approximately 100 μL). Adjust
volume to 1 L with ultrapure water and store in a glass bottle
at 4 C in dark.
12. Mobile phase A for S1P analysis: Mix 500 mL acetonitrile and
500 mL of ammonium formate buffer 25 mM pH 4.6 in a
glass bottle. Sonicate for 10 min.
13. Mobile phase B for S1P analysis: Mix 950 mL acetonitrile and
50 mL of ammonium formate buffer 25 mM pH 4.6 in a glass
bottle. Sonicate for 10 min.

3 Methods

3.1 Experimental The study design is a crucial part of the study that affects its
Study Design outcome and the interpretation of the results. To ensure reliable
results, the study must be carefully designed, taking into consider-
ation different aspects, such as composition of the cohorts, sample
collection, storage, and analysis order.
1. Age, gender, ethnicity, BMI, diet, medications, and other pos-
sible confounding factors should be recorded and considered
when designing the analytical workflow and analyzing the final
data to reliably establish associations between sphingolipid
levels and clinical parameters.
2. A consistent sampling procedure during the whole study is
essential since different collecting procedures for plasma or
serum have been shown to result in different metabolite isola-
tions (see below and notes). For this reason, a strict and well-
documented protocol must be followed, and eventual varia-
tions should be recorded.
3. Samples should be stored preferably at 80 C to minimize
metabolite degradation. Freeze-thaw cycles may affect the con-
centration of specific molecules so the number of these cycles
must be the same for all the samples and well documented.
 Clinical Sphingolipidomics 141

4. In case of large studies, it is recommended to divide them into

smaller batches, to facilitate experimental procedures and allow
for instrument cleaning between batches when necessary. Each
batch should represent the study population. This is achieved
by using a stratified randomization procedure, to exclude the
risk of bias or correlation between metabolite levels and analysis
order [20].

3.2 Blood Collection The choice between plasma and serum depends on the aims and
and Preparation practical aspects of the study (see Notes 1 and 2). Blood collection
of Plasma and Serum and subsequent processing of drawn blood must be performed
uniformly to minimize variability in measured analytes caused by
preanalytical effects (see Notes 3–5) [22]. Blood should be treated
as potentially infectious and handled with appropriate safety mea-
sures (see Note 1).
1. Collect fresh blood by venipuncture into collection tubes
without (for serum) or with (for plasma) anticoagulants (see
Notes 6–8).
2. For plasma preparation: Place tubes with anticoagulated whole
blood on ice immediately after collection (see Note 9).
3. For serum preparation: Allow collected blood to clot undis-
turbed at room temperature for 60 min in collection tubes
without clot activator and 30 min for tubes with silica clot
activator (see Note 10).
4. Centrifuge as soon as possible at 2000
g (4 C) for 15 min
(plasma) or 10 min (serum), respectively (see Notes 11 and 12).
5. Transfer supernatants into new polypropylene tubes without
disturbing the cellular fraction.
6. Add 1 μL 100 mM BHT solution in ethanol per 1 mL plasma
to prevent oxidation [17], proceed with sample preparation, or
store at 80 C until further processing.

3.3 Preparation Preparation of samples in large studies represents a nontrivial chal-

of Plasma/Serum lenge. Here below are some suggestions for sample preparation.
and BQC Samples
1. Thaw plasma/serum samples (study samples) at room temper-
ature, making sure samples are transferred to ice just before
they are completely thawed.
2. Vortex all study samples to remove concentration gradients
that may have resulted from the thawing process and to ensure
sample is completely thawed.
3. Centrifuge study samples at 12,000
g for 10 min (4 C) to
sediment precipitated material.
4. Stock BQC: Pool aliquots from the supernatant of each study
sample (or just a stratified representative subset of all the study
142 Bo Burla et al.

samples) to generate a stock BQC. Prepare a sufficient volume

of stock BQC, calculated based on the total estimated number
of BQC samples that will be analyzed.
5. Study and BQC samples: Transfer 20 μL supernatant from each
centrifuged study sample to a new 2 mL Eppendorf Safe-Lock
tubes. Transfer 20 μL supernatant from the stock BQC to a
new tube every five or ten samples. Two or more BQCs should
be added in the beginning and at the end of each batch.
6. Proceed directly with lipid extraction or store aliquots at
80 C until extraction.

3.4 Lipid Extraction Once aliquots with the required amount of all the samples have
been prepared, two general strategies can be considered: (1) extract
lipids from all the samples at the same time, divide them in batches
(according to rules mentioned in the previous paragraph), freeze all
of them, and analyze each batch separately, i.e., at different days, or
(2) divide the samples in several batches (according to rules men-
tioned in the previous paragraph), and extract lipids from them
every day prior to analysis [21].
The lipid extraction procedure described below is based on the
method described by Alshehry et al. [17].
1. Study and BQC samples: Add 200 μL BUME + ISTDs to all
aliquots (20 μL, see Notes 13–15).
2. Blanks: 200 μL BUME + ISTDs (for the blank + ISTD sam-
ples) and 200 μL BUME (for blank  ISTD samples) are added
into empty tubes of the same type used for the study samples.
Blanks can be added to the sample sequence (e.g., one
blank + ISTD after each BQC) to monitor contamination
from the extraction solvent, ISTD mixes, extraction procedure,
and carry-over.
3. Vortex all samples (including blank  ISTD and blank + ISTD)
for 30 s, and sonicate them for 30 min ensuring that the
temperature of the water bath does not exceed 20 C.
4. After sonication, samples are centrifuged for 10 min at
14,000
g (20 C) to precipitate proteins and salts, and then
the supernatant containing the sphingolipids is transferred into
new tubes or autosampler vials.
5. TQCSL samples: Pool aliquots of equal volumes from all
extracted BQCs to prepare the stock TQCSL. Estimate the
required volume based on injection sequence. Aliquot stock
TQCSL into two vials, one used as TQCSL for the LC-MS
analysis and the other for the initial conditioning of the
LC-MS system.
 Clinical Sphingolipidomics 143

6. TQCSL dilution series: A serial dilution series, e.g., 1:2, 1:4,

and 1:8 with BUME, is prepared from the stock TQCSL.
7. Extracts may be directly transferred to autosampler vials for
LC-MS analysis or stored at 80 C until needed (see Notes 16
and 17). The samples can also be dried with nitrogen or in
centrifugal concentrator and reconstituted with a desired vol-
ume of mobile phase A before analysis.

3.5 Derivatization This procedure is based on the method described in (Narayanas-

wamy et al. [13]). Consult safety information detailed in Notes 18
and 19 before performing this procedure.
1. 90 μL lipid extract from study samples, BQCs, and blanks is
transferred to 2 mL Eppendorf Safe-Lock polypropylene tubes
in the same sequence as performed for the lipid extraction.
2. 30 μL of TMS-diazomethane 2 M solution in hexanes is added
for derivatization of S1P (see Notes 18 and 19).
3. This mixture is incubated for 20 min at 750 rpm at 22 C using
a thermal shaker.
4. The derivatization reaction is stopped after 20 min by adding
1 μL of acetic acid 100% (see Note 19).
5. Samples are vortexed for 5 s and centrifuged for 10 min at
14,000
g (20  C) and the supernatant transferred to new
vials/tubes.
6. TQCS1P samples: Pool aliquots of equal volume from all deri-
vatized BQCs to prepare the stock TQCS1P. Estimate the
required volume based on the sample sequence. Aliquot stock
TQCS1P into two vials, one used as TQCS1P for the LC-MS
analysis and the other one for the initial conditioning of the
LC-MS system.
7. TQCS1P dilution series: A serial dilution series, e.g., 1:2, 1:4,
and 1:8 with BUME, is prepared from the stock TQCS1P.
8. Extracts may be directly transferred to vials for LC-MS analysis
or stored at 80 C until needed. The samples can also be dried
with nitrogen or in centrifugal concentrator and reconstituted
with a desired volume of the starting mobile phase before
analysis.

3.6 Liquid The protocol described below is based on a modified previously

Chromatography- described method by Wang et al. [23].
Mass Spectrometry
1. Samples are analyzed by reversed-phase liquid chromatography-
Analysis electrospray ionization (ESI) MS. Agilent 1290 UHPLC system
3.6.1 LC-MS Setup equipped with an Agilent ZORBAX RRHD Eclipse Plus C18,
for Sphingolipid Analysis 95 Å, 2.1
100 mm, 1.8 μm, set at 40  C, is used as a
chromatographic system to separate the sphingolipid species.
144 Bo Burla et al.

The flow rate is 0.4 mL/min. Mobile phases A and B are mixed
according to the following gradient: 0% B to 10% B from 0 to
3 min, 10 to 40% B from 3 to 5 min, 40 to 55% B from 5 to
5.3 min, 55 to 60% B from 5.3 to 8 min, 60 to 80% B from 8 to
8.5 min, 80% B from 8.5 to 10.5 min, 80 to 90% B from 10.5 to
16 min, 90% B from 10.5 to 19 min, and 90 to 100% B from
19 to 22 min, and re-equilibrate at 0% B from 22.1 to 25.0 min.
The total run time is 25.0 min.
2. AJS ESI source parameters: Sheath gas temperature and flow
are set to 200 C and 12 L/min, respectively; nebulizer pres-
sure, 25 psi; capillary voltage and nozzle voltage, 3500 V and
500 V, respectively; dry gas temperature and flow, 200 C and
15 L/min, respectively; and the delta EMV, 200 V. Positive
high-/low-pressure RF of the iFunnel is set to 210/110.
3. For this analysis, the triple-quadrupole mass spectrometer is
operated in positive ionization dynamic multiple reaction mon-
itoring (dMRM) mode. For ceramides (Cer), hexosylceramides
(HexCer), and dihexosylceramides (Hex2Cer), the long-chain
base product ions generated from intact precursors and from
precursors after in-source water loss are monitored. For sphin-
gomyelins (SM) only the product ion of m/z 184 is monitored.
MRM transitions, collision energies, and expected dMRM
retention time are listed in Table 2. Retention time windows
are set to 1.5; cycle time is set to 1000 ms, keeping a minimum
dwell time of 10 ms.
4. Injection volume for all injected samples is 2 μL (see Note 20).

3.6.2 LC-MS Setup The protocol described below is a modified version of a previously
for S1P Analysis described method by Narayanaswamy et al. [13].
1. S1P analysis is performed using hydrophilic interaction liquid
chromatography (HILIC)-ESI MS on Agilent 1290 UHPLC
connected to an Agilent 6495 mass spectrometer. A Waters
ACQUITY BEH HILIC 130 Å, 1.7 μm, 2.1
100 mm, col-
umn maintained at 60 C is used for the analysis. The flow rate
is 0.4 mL/min. Mobile phases A and B are mixed according to
the following gradient: 99.9% B to 40% B from 0 to 5 min,
40 to 10% B from 5 to 5.5 min,10% B from 5.5 to 6.5 min, and
re-equilibrate at 99.9% B from 6.6 to 9.6 min. The total run
time is 9.6 min.
2. AJS ESI source parameters: The sheath gas temperature and
flow are set at 400 C and 12 L/min, respectively; nebulizer
pressure, 25 psi; capillary voltage, 3500 V; nozzle voltage,
500 V; dry gas temperature and flow, 200 C and 12 L/min,
respectively; delta EMV, 200; positive high-/low-pressure RF,
200/110.
 Clinical Sphingolipidomics 145

Table 2
Parameters defined in the MS dMRM method for sphingolipid analysis

Ret time Collision

Compound name Precursor ion Product ion (min) energy
C1P d18:1/12:0 (ISTD) 562.4 264.1 9.8 25
C1P d18:1/12:0 (ISTD) 544.4 264.1 9.8 25
C1P d18:1/12:0 (ISTD) 562.4 464.2 9.8 10
C1P d18:1/16:0 618.5 264.2 11.9 25
C1P d18:1/16:0 600.5 264.2 11.9 25
C1P d18:1/18:0 646.5 264.2 12.9 25
C1P d18:1/18:0 628.5 264.2 12.9 25
C1P d18:1/26:0 758.6 264.2 16.9 25
C1P d18:1/26:0 740.6 264.2 16.9 25
Cer d18:1/12:0 (ISTD) 482.4 264.2 10.6 25
Cer d18:1/12:0 (ISTD)(-H2O) 464.4 264.2 10.6 25
Cer d18:0/14:0 512.5 266.2 13.4 25
Cer d18:0/14:0(-H2O) 494.5 266.2 13.4 25
Cer d18:1/14:0 510.5 264.2 11.8 25
Cer d18:1/14:0(-H2O) 492.5 264.2 11.8 25
Cer d18:1/14:1 508.5 264.2 10.6 25
Cer d18:1/14:1(-H2O) 490.5 264.2 10.6 25
Cer d18:2/14:0 508.5 262.2 10.6 25
Cer d18:2/14:0(-H2O) 490.5 262.2 10.6 25
Cer d18:0/15:0 526.5 266.2 14 25
Cer d18:0/15:0(-H2O) 508.5 266.2 14 25
Cer d18:0/16:0 540.5 266.2 14.6 25
Cer d18:0/16:0(-H2O) 522.5 266.2 14.6 25
Cer d18:1/16:0 538.5 264.2 12.5 25
Cer d18:1/16:0(-H2O) 520.5 264.2 12.5 25
Cer d18:1/16:1 536.5 264.2 11.5 25
Cer d18:1/16:1(-H2O) 518.5 264.2 11.5 25
Cer d18:2/16:0 536.5 262.2 11.8 25
Cer d18:2/16:0(-H2O) 518.5 262.2 11.8 25
Cer d18:0/17:0 554.5 266.2 15.2 25
(continued)
146 Bo Burla et al.

Table 2
(continued)

Ret time Collision

Compound name Precursor ion Product ion (min) energy
Cer d18:0/17:0(-H2O) 536.5 266.2 15.2 25
Cer d18:0/18:0 568.5 266.2 15.8 25
Cer d18:0/18:0(-H2O) 550.5 266.2 15.8 25
Cer d18:0/18:1 566.5 266.2 14.6 25
Cer d18:0/18:1(-H2O) 548.5 266.2 14.6 25
Cer d18:1/18:0 566.5 264.2 14 25
Cer d18:1/18:0(-H2O) 548.5 264.2 14 25
Cer d18:1/18:1 564.5 264.2 12.7 25
Cer d18:1/18:1(-H2O) 546.5 264.2 12.7 25
Cer d18:2/18:0 564.5 262.2 13 25
Cer d18:2/18:0(-H2O) 546.5 262.2 13 25
Cer d18:2/18:1 562.5 262.2 11.8 25
Cer d18:2/18:1(-H2O) 544.5 262.2 11.8 25
Cer d18:0/20:0 596.5 266.2 17 25
Cer d18:0/20:0(-H2O) 578.5 266.2 17 25
Cer d18:1/20:0 594.5 264.2 15.2 25
Cer d18:1/20:0(-H2O) 576.5 264.2 15.2 25
Cer d18:1/20:1 592.5 264.2 14 25
Cer d18:1/20:1(-H2O) 574.5 264.2 14 25
Cer d18:2/20:0 592.5 262.2 14.2 25
Cer d18:2/20:0(-H2O) 574.5 262.2 14.2 25
Cer d18:2/20:1 590.5 262.2 13 25
Cer d18:2/20:1(-H2O) 572.5 262.2 13 25
Cer d18:0/22:0 624.6 266.2 18.2 25
Cer d18:0/22:0(-H2O) 606.6 266.2 18.2 25
Cer d18:0/22:1 622.5 266.2 17 25
Cer d18:0/22:1(-H2O) 604.5 266.2 17 25
Cer d18:1/22:0 622.6 264.2 16.5 25
Cer d18:1/22:0(-H2O) 604.6 264.2 16.5 25
Cer d18:1/22:1 620.6 264.2 15.37 25
(continued)
 Clinical Sphingolipidomics 147

Table 2
(continued)

Ret time Collision

Compound name Precursor ion Product ion (min) energy
Cer d18:1/22:1(-H2O) 602.6 264.2 15.37 25
Cer d18:2/22:0 620.6 262.2 15.6 25
Cer d18:2/22:0(-H2O) 602.6 262.2 15.6 25
Cer d18:2/22:1 618.6 262.2 14.4 25
Cer d18:2/22:1(-H2O) 600.6 262.2 14.4 25
Cer d18:0/23:0 638.6 266.2 18.8 25
Cer d18:0/23:0(-H2O) 620.6 266.2 18.8 25
Cer d18:0/23:1 636.6 266.2 17.6 25
Cer d18:0/23:1(-H2O) 618.6 266.2 17.6 25
Cer d18:1/23:0 636.6 264.2 17.2 25
Cer d18:1/23:0(-H2O) 618.6 264.2 17.2 25
Cer d18:1/23:1 634.6 264.2 16 25
Cer d18:1/23:1(-H2O) 616.6 264.2 16 25
Cer d18:2/23:0 634.6 262.2 16.2 25
Cer d18:2/23:0(-H2O) 616.6 262.2 16.2 25
Cer d18:2/23:1 632.6 262.2 15 25
Cer d18:2/23:1(-H2O) 614.6 262.2 15 25
Cer d18:0/24:0 652.6 266.2 19.4 25
Cer d18:0/24:0(-H2O) 634.6 266.2 19.4 25
Cer d18:0/24:1 650.6 266.2 18.2 25
Cer d18:0/24:1(-H2O) 632.6 266.2 18.2 25
Cer d18:1/24:0 650.6 264.2 17.9 25
Cer d18:1/24:0(-H2O) 632.6 264.2 17.9 25
Cer d18:1/24:1 648.6 264.2 16.7 25
Cer d18:1/24:1(-H2O) 630.6 264.2 16.7 25
Cer d18:2/24:0 648.6 262.2 16.9 25
Cer d18:2/24:0(-H2O) 630.6 262.2 16.9 25
Cer d18:2/24:1 646.6 262.2 15.6 25
Cer d18:2/24:1(-H2O) 628.6 262.2 15.6 25
Cer d18:0/25:0 666.6 266.2 20 25
(continued)
148 Bo Burla et al.

Table 2
(continued)

Ret time Collision

Compound name Precursor ion Product ion (min) energy
Cer d18:0/25:0(-H2O) 648.6 266.2 20 25
Cer d18:0/25:1 664.6 266.2 18.8 25
Cer d18:0/25:1(-H2O) 646.6 266.2 18.8 25
Cer d18:1/25:0 (ISTD) 646.4 264.2 18.5 25
Cer d18:1/25:0 (ISTD)(-H2O) 664.4 264.2 18.5 25
Cer d18:0/26:0 680.6 266.2 19.4 25
Cer d18:0/26:0(-H2O) 662.6 266.2 19.4 25
Cer d18:0/26:1 678.6 266.2 18.2 25
Cer d18:0/26:1(-H2O) 660.6 266.2 18.2 25
Cer d18:1/26:0 678.6 264.2 19.6 25
Cer d18:1/26:0(-H2O) 660.6 264.2 19.6 25
Cer d18:1/26:1 676.6 264.2 18.9 25
Cer d18:1/26:1(-H2O) 658.6 264.2 18.9 25
Cer d18:2/26:0 676.6 262.2 18.1 25
Cer d18:2/26:0(-H2O) 658.6 262.2 18.1 25
Cer d18:2/26:1 674.6 262.2 16.9 25
Cer d18:2/26:1(-H2O) 656.6 262.2 16.9 25
Hex2Cer d18:1/12:0 (ISTD) 788.5 264.2 9.9 45
Hex2Cer d18:1/12:0 (ISTD)(-H2O) 806.5 264.2 9.9 45
Hex2Cer d18:1/16:0 862.7 264.2 11.2 45
Hex2Cer d18:1/16:0(-H2O) 844.7 264.2 11.2 45
Hex2Cer d18:1/18:0 890.7 264.2 12.4 45
Hex2Cer d18:1/18:0(-H2O) 872.7 264.2 12.4 45
Hex2Cer d18:1/18:1 888.7 264.2 11.4 45
Hex2Cer d18:1/18:1(-H2O) 870.7 264.2 11.4 45
Hex2Cer d18:1/20:0 918.7 264.2 13.6 45
Hex2Cer d18:1/20:0(-H2O) 900.7 264.2 13.6 45
Hex2Cer d18:1/22:0 946.7 264.2 14.6 45
Hex2Cer d18:1/22:0(-H2O) 928.7 264.2 14.6 45
Hex2Cer d18:1/22:1 944.7 264.2 13.5 45
(continued)
 Clinical Sphingolipidomics 149

Table 2
(continued)

Ret time Collision

Compound name Precursor ion Product ion (min) energy
Hex2Cer d18:1/22:1(-H2O) 926.7 264.2 13.5 45
Hex2Cer d18:1/23:0 960.7 264.2 15.2 45
Hex2Cer d18:1/23:0(-H2O) 942.7 264.2 15.2 45
Hex2Cer d18:1/24:0 974.7 264.2 15.7 45
Hex2Cer d18:1/24:0(-H2O) 956.7 264.2 15.7 45
Hex2Cer d18:1/24:1 972.7 264.2 14.6 45
Hex2Cer d18:1/24:1(-H2O) 954.7 264.2 14.6 45
HexCer 30:1 d18:1/12:0 (ISTD) 644.5 264.2 10 45
HexCer 30:1 d18:1/12:0 (ISTD)(-H2O) 626.5 264.2 10 45
HexCer 32:1 d18:1/14:0 672.6 264.2 10.5 45
HexCer 32:1 d18:1/14:0(-H2O) 654.6 264.2 10.5 45
HexCer d18:0/16:0 702.6 266.2 14 45
HexCer d18:0/16:0(-H2O) 684.6 266.2 14 45
HexCer d18:0/16:1 700.6 266.2 12.9 45
HexCer d18:0/16:1(-H2O) 682.6 266.2 12.9 45
HexCer d18:1/16:0 700.6 264.2 11.52 45
HexCer d18:1/16:0(-H2O) 682.6 264.2 11.52 45
HexCer d18:1/16:1 698.6 264.2 10.5 45
HexCer d18:1/16:1(-H2O) 680.6 264.2 10.5 45
HexCer d18:2/16:0 698.6 262.2 10.5 45
HexCer d18:2/16:0(-H2O) 680.6 262.2 10.5 45
HexCer d18:1/18:0 728.6 264.2 12.9 45
HexCer d18:1/18:0(-H2O) 710.6 264.2 12.9 45
HexCer d18:1/18:1 726.6 264.2 11.7 45
HexCer d18:1/18:1(-H2O) 708.6 264.2 11.7 45
HexCer d18:0/20:0 758.6 266.2 15.1 45
HexCer d18:0/20:0(-H2O) 740.6 266.2 15.1 45
HexCer d18:1/20:0 756.6 264.2 14.1 45
HexCer d18:1/20:0(-H2O) 738.6 264.2 14.1 45
HexCer d18:1/20:1 754.6 264.2 13 45
(continued)
150 Bo Burla et al.

Table 2
(continued)

Ret time Collision

Compound name Precursor ion Product ion (min) energy
HexCer d18:1/20:1(-H2O) 736.6 264.2 13 45
HexCer d18:2/20:0 754.6 262.2 13 45
HexCer d18:2/20:0(-H2O) 736.6 262.2 13 45
HexCer d18:0/22:0 786.6 266.2 17.5 45
HexCer d18:0/22:0 786.6 266.2 17.5 45
HexCer d18:0/22:0(-H2O) 768.6 266.2 17.5 45
HexCer d18:0/22:0(-H2O) 768.6 266.2 17.5 45
HexCer d18:0/22:1 784.6 266.2 16.3 45
HexCer d18:0/22:1(-H2O) 766.6 266.2 16.3 45
HexCer d18:1/22:0 784.6 264.2 15.2 45
HexCer d18:1/22:0(-H2O) 766.6 264.2 15.2 45
HexCer d18:1/22:1 782.6 264.2 14 45
HexCer d18:1/22:1(-H2O) 764.6 264.2 14 45
HexCer d18:2/22:0 782.6 262.2 14 45
HexCer d18:2/22:0(-H2O) 764.6 262.2 14 45
HexCer d18:0/23:0 800.7 266.2 18.1 45
HexCer d18:0/23:0(-H2O) 782.7 266.2 18.1 45
HexCer d18:1/23:0 798.6 264.2 15.9 45
HexCer d18:1/23:0(-H2O) 780.6 264.2 15.9 45
HexCer d18:1/23:1 796.6 264.2 14.7 45
HexCer d18:1/23:1(-H2O) 778.6 264.2 14.7 45
HexCer d18:0/24:0 814.7 266.2 18.6 45
HexCer d18:0/24:0 814.7 266.2 18.6 45
HexCer d18:0/24:0(-H2O) 796.7 266.2 18.6 45
HexCer d18:0/24:0(-H2O) 796.7 266.2 18.6 45
HexCer d18:0/24:1 812.7 266.2 17.5 45
HexCer d18:0/24:1(-H2O) 794.7 266.2 17.5 45
HexCer d18:1/24:0 812.7 264.2 16.5 45
HexCer d18:1/24:0(-H2O) 794.7 264.2 16.5 45
HexCer d18:1/24:1 810.7 264.2 15.3 45
(continued)
 Clinical Sphingolipidomics 151

Table 2
(continued)

Ret time Collision

Compound name Precursor ion Product ion (min) energy
HexCer d18:1/24:1(-H2O) 792.7 264.2 15.3 45
HexCer d18:1/25:0 826.7 264.2 17.1 45
HexCer d18:1/25:0(-H2O) 808.7 264.2 17.1 45
HexCer d18:1/25:1 824.7 264.2 16 45
HexCer d18:1/25:1(-H2O) 806.7 264.2 16 45
HexCer d18:0/26:0 842.7 266.2 18.7 45
HexCer d18:0/26:0(-H2O) 824.7 266.2 18.7 45
HexCer d18:0/26:1 842.7 266.2 17.7 45
HexCer d18:0/26:1(-H2O) 824.7 266.2 17.7 45
HexCer d18:1/26:0 840.7 264.2 17.7 45
HexCer d18:1/26:0(-H2O) 822.7 264.2 17.7 45
HexCer d18:1/26:1 838.7 264.2 16.5 45
HexCer d18:1/26:1(-H2O) 820.7 264.2 16.5 45
SM 32:0 677.53 184.1 11.5 25
SM 30:1 (ISTD) 647.42 184.1 10 25
SM 32:1 675.53 184.1 10.8 25
SM 32:2 673.52 184.1 10.1 25
SM 33:0 691.56 184.1 11.8 25
SM 33:1 689.56 184.1 11.2 25
SM 33:2 687.56 184.1 10.5 25
SM 34:0 705.58 184.1 12 25
SM 34:1 703.58 184.1 11.5 25
SM 34:2 701.56 184.1 10.9 25
SM 35:0 719.58 184.1 12.7 25
SM 35:1 717.59 184.1 12 25
SM 36:0 733.58 184.1 13.3 25
SM 36:1 731.61 184.1 12.6 25
SM 36:2 729.59 184.1 11.8 25
SM 37:0 747.61 184.1 14 25
SM 37:1 745.61 184.1 13.2 25
(continued)
152 Bo Burla et al.

Table 2
(continued)

Ret time Collision

Compound name Precursor ion Product ion (min) energy
SM 38:0 761.6 184.1 14.5 25
SM 38:1 759.64 184.1 13.8 25
SM 38:2 757.61 184.1 13 25
SM 39:1 773.6 184.1 14.4 25
SM 40:0 789.67 184.1 15.7 25
SM 40:1 787.67 184.1 15.1 25
SM 40:2 785.65 184.1 14.2 25
SM 40:3 783.65 184.1 13.3 25
SM 41:1 801.68 184.1 15.8 25
SM 41:2 799.67 184.1 14.8 25
SM 41:3 797.67 184.1 13.8 25
SM 42:0 817.7 184.1 16.9 25
SM 42:1 815.7 184.1 16.5 25
SM 42:2 813.68 184.1 15.5 25
SM 42:3 811.67 184.1 14.5 25
SM 43:1 829.7 184.1 17.1 25
SM 43:2 827.7 184.1 16.1 25
SM 44:1 843.7 184.1 17.9 25
SM 44:2 841.7 184.1 16.9 25
Sph d16:1 272.24 254.2 5.7 5
Sph d16:1 272.24 236.1 5.7 10
Sph d16:1 272.24 224.1 5.7 21
Sph d17:0 (ISTD) 288.32 270.3 6.4 20
Sph d17:0 (ISTD) 288.32 252.2 6.4 20
Sph d17:0 (ISTD) 288.32 240.2 6.4 20
Sph d17:1 (ISTD) 286.32 268.3 6.2 5
Sph d17:1 (ISTD) 286.32 250.2 6.2 17
Sph d17:1 (ISTD) 286.32 238.2 6.2 20
Sph d18:0 302.32 284.2 6.65 20
Sph d18:0 302.32 266.2 6.65 20
(continued)
 Clinical Sphingolipidomics 153

Table 2
(continued)

Ret time Collision

Compound name Precursor ion Product ion (min) energy
Sph d18:0 302.32 254.2 6.65 21
Sph d18:1 300.3 282.2 6.45 5
Sph d18:1 300.3 264.2 6.45 17
Sph d18:1 300.3 252.2 6.45 21
Sph d18:2 298.25 280.2 6 5
Sph d18:2 298.25 262.2 6 17
Sph d18:2 298.25 250.2 6 21

Table 3
Parameters defined in the MS MRM method for S1P analysis

Compound name Precursor ion Product ion

S1P d18:1 13C2D2 (ISTD) 440.33 113.0
13
S1P d18:1 C2D2 (ISTD) 440.33 60.08
S1P d18:0 438.34 113.0
S1P d18:0 438.34 60.08
S1P d18:1 436.31 113.0
S1P d18:1 436.31 60.08
S1P d18:2 434.29 113.0
S1P d18:2 434.29 60.08
S1P d17:0a 424.3 113.0
a
S1P d17:0 424.3 60.08
a
S1P d17:1 422.3 113.0
a
S1P d17:1 422.3 60.08
a
see Note 13

3. The triple-quadrupole (QQQ) mass spectrometer is operated

in positive ionization multiple reaction monitoring (MRM)
mode. Two product ions (from the tetramethylated S1P that
is formed after TMS-diazomethane derivatization) are moni-
tored at m/z 60.08 and m/z 113 (see Table 3). The fragmen-
tor, collision energy, and cell accelerator voltage are set to
380 V, 29 V, and 3 V, respectively.
4. Injection volume for all samples is 2 μL (see Note 20).
154 Bo Burla et al.

3.6.3 LC-MS Analyses 1. LC-MS system performances have to meet established quality
criteria (see Note 21).
2. Column is well conditioned before starting the analysis (see
Note 22).
3. The LC-MS system should first be equilibrated by injecting
three extracted blanks and then injecting TQC samples until
retention times and signals are stable. Ensure that all the peaks
are well captured within the dMRM windows for the sphingo-
lipid method (see Table 2), or adjust retention time windows if
necessary.
4. Injection sequence of samples, BQC, TQC, and extracted
blanks should depend on the experimental design. We recom-
mend the following injection sequence:
(a) TQC dilution series starting with the 1:8 diluted TQC.
(b) Samples, BQCs, and blanks in the same sequence used for
extraction.
(c) Insert TQCs in the sequence to monitor instrument per-
formance (e.g., after each blank + ISTD).

3.7 Data Analysis 1. MS raw data are imported into a MRM data analysis software
for small molecules, e.g., Agilent MassHunter Quantitative
3.7.1 Peak Integration
Analysis, and MRM chromatograms are extracted.
of MRM Chromatograms
2. Determine peak areas of quantifier transitions by peak integra-
tion. Transitions with fragments from intact precursors are
used as quantifiers and transitions with fragments from precur-
sors after water loss as qualifiers. For HexCer and Hex2Cer, it is
opposite; the transitions with intact precursors are used as
qualifiers, and those of precursors with water loss as qualifier.
For sphingomyelins (SM) no qualifier transitions were moni-
tored in this protocol (see Table 2). For S1P transitions, the
product ion at m/z 60.08 is used as a quantifier and the one at
m/z 113 as a qualifier. Ensure that the correct peak is picked,
that a co-eluting qualifier peak is present, and that integration
borders are determined correctly (see Notes 23–25).

3.7.2 Normalization 1. Normalization with internal standards is done by calculating

and Quantification the peak area ratio of the lipid of interest and the corresponding
internal standard (ISTD), in a linear interval covered by previ-
ously measured calibration curves [13, 16]. Lipids are normal-
ized with the corresponding class-specific compounds:
ceramides are normalized with Cer d18:1/12:0; HexCer with
GluCer d18:1/12:0; Hex2Cer with LacCer d18:1/12:0; SM
with SM d18:1/12:0; C1P with C1P d18:1/12:0; sphingo-
sines d18:1, d18:2, and d16:1 with Sph d17:1; sphingosine
d18:0 with Sph d17:0; and all S1Ps with S1P-13C2D2 d18:1
(see Notes 13 and 26).
 Clinical Sphingolipidomics 155

2. Isotope correction of the S1P quantifier transitions (m/z

60.03) is done by using the following formulas:

RAS1P d18:0 ¼ RArawS1P d18:0  RArawS1P d18:1
0:03161

RAS1P d18:1 ¼ RArawS1P d18:1  RArawS1P d18:2
0:03156
RA, relative abundance (corrected for isotope interference);
RAraw, raw relative abundance (see Note 27).
Calculation of absolute S1P d18:1 concentrations in the
study sample can be done using the following formula:
 
ν ISTD
c S1Pd18:1 ¼ RAS1Pd18:1

c ISTD
ν Sample
RAcorrS1P d18:1, relative abundance of S1P d18:1; νISTD, vol-
ume of S1P-13C2D2 ISTD solution (BUME + ISTD); νSample,
sample volume; cISTD, molar concentration of S1P d18:1 in the
sample; cISTD, molar concentration of S1P-13C2D2 in the ISTD
(BUME + ISTD) solution (see Note 28 and Table 1).
3. See Note 29 for calculation of concentrations of the other lipids
measured with the sphingolipid and S1P methods.

3.7.3 Quality Control The obtained raw dataset of lipid abundances should be filtered
based on quality control (QC) parameters to exclude potential
artifacts and noisy signals in downstream data analysis. The criteria
for QC filtering and parameters depend on the aims, requirements,
and practical aspects of the study.
1. Filtering of lipids exhibiting high variations in QC samples:
The coefficient of variation (CoV) in the BQC samples yields
information on the variation caused by experimental and ana-
lytical effects. We usually tend to exclude lipids with a %CoV of
>20–25%, depending on the study. %CoV is calculated using
the following formula:


StdDev RALipid X in all BQCs
%CoV Lipid X ¼

100
Mean RA Lipid X in all BQCs

RA, relative abundance.

2. Filtering of lipids that have high background signals. The
signal-to-blank peak area ratio of a lipid between sample and
corresponding blank yields information on interfering signals
from the extraction solvent, ISTD mix, extraction procedure,
and carry-over. We usually exclude lipids that have a signal-to-
blank ratio <10.
3. Filtering of lipids that do not linearly respond to dilution:
Significantly lower or higher response to dilution of extracts
may indicate saturation effects, interfering contaminants from
156 Bo Burla et al.

the lipid extraction or LC system, other artifacts from the

LC-MS system, or errors in data processing, i.e., peak picking
and integration. We usually exclude lipids that respond to
dilutions of TQC with a Pearson correlation coefficient < 0.8
(see also Note 20).
4. The relative abundances of lipid in BQCs can be used to
monitor, and to potentially correct, for drifts and batch effects
during analyses. We routinely plot relative abundances of each
lipid and the principal components 1–3 from a principal com-
ponent analysis (PCA) of all relative lipid abundances in all
BQCs against the analysis order to visualize and detect drifts
and batch effects.

4 Notes

1. Personal protection equipment (PPE) should be worn during

all experimental steps. Human blood, plasma, and serum sam-
ples should be treated as potentially infectious and handled in a
biosafety cabinet. All steps involving organic solvents should be
performed under a chemical hood, except when organic extrac-
tion solvent is added to the plasma samples. Concentrated
formic acid should be added into water/solvents and not in
reverse order.
2. During blood coagulation, platelets, leukocytes, and erythro-
cytes release extracellular vesicles and various compounds,
including lipids such as sphingosine-1-phosphate, and enzymes
that may change lipid profiles ex vivo. Plasma obtained from
anticoagulated blood can therefore be considered as the closest
to the blood plasma in vivo [22, 24, 25].
3. Blood should always be taken around the same time of the day
for all the samples, and fasting status (time between the last
meal and blood draw) should also be comparable between the
study subjects, to reduce diurnal variations and effect of food
intake on plasma lipids [26, 27].
4. Samples from matched reference groups should be taken for
each study and at times distributed throughout the study dura-
tion to avoid biases, if applicable. Comparisons with samples or
with lipidomic results obtained from other studies may be
difficult due to differences in the used methodologies that
may lead to incorrect conclusions.
5. Collection of capillary blood, e.g., by fingerprick, is not recom-
mended as blood may be contaminated with skin cells, tissue
fluids, and disinfectants. These procedures often result in
hemolytic samples, especially when the puncture point is
 Clinical Sphingolipidomics 157

squeezed (see Note 12), and clotting before mixing with the
anticoagulant [28].
6. Proper venipuncture protocols should be applied to prevent
hemolysis and clotting [29]. Discarding the first 1–3 mL of
blood with a dedicated discard tube may reduce the risk of
hemolysis and clots. Special care should be taken when blood
is collected via infusion catheters, in which case discarding the
first 1–3 mL blood reduces the risk of hemolysis and mixing of
infusion solution into the collected blood [30].
7. Heparin has been shown to introduce higher variability in the
measurements of certain sphingolipids [27].
8. When using blood collection containers with liquid anticoagu-
lant solution (i.e., citrate tubes), care must be taken not to
under- or overfill the tubes to avoid dilution errors and ensure
a reproducible final concentration of anticoagulant. The dilu-
tion by the anticoagulant solution must be considered in the
calculation of the lipid concentrations.
9. When collected whole blood is not kept on ice before centrifu-
gation, specific lipid levels may increase ex vivo, i.e., sphingo-
sine-1-phosphate [22].
10. Clotting time of blood from patients with disorders affecting
coagulation and/or who are under anticoagulant therapy may
considerably vary. In this case samples from all subjects should
be allowed to clot for the same time.
11. Low centrifugation forces or short centrifugation times may
lead to residual platelets in the plasma, which may affect levels
of lipids abundant in platelets, such as sphingosine-1-phos-
phate [31]. High centrifugation forces may cause hemolytic
serum preparations and alter the lipid levels as well.
12. Hemolytic samples may have altered profiles of specific lipids
that are highly abundant in erythrocytes, e.g., sphingosine-1-
phosphate, and must therefore be avoided or measured with
caution [22].
13. To analyze S1P d17:0 and S1P d17:1, the Cer/Sph mixture II
cannot be used, as it contains S1P d17:0 and S1P d17:1
standards, which will interfere with endogenous species. In
this case, a separate extraction with BUME containing only
S1P-13C2D2 as internal standard has to be prepared.
14. Before using internal standard solutions stored at 20 C or
80 C, allow solutions to equilibrate to room temperature,
sonicate for approximately 10 min, and mix well to ensure
compounds are fully dissolved.
15. Polypropylene tubes of some other brands may not be com-
patible with the butanol/methanol extraction solvent and/or
may cause interferences in LC-MS analyses. It is recommended
158 Bo Burla et al.

to test specific tubes by analyzing blanks extracted in these

tubes (see Subheading 3.4, step 2).
16. Note that these extracts can be used for other lipid analyses as
well, as, for example, in phospholipid and neutral lipid
analyses [17].
17. When extracts that were stored at 80 C will be analyzed, they
should be equilibrated at room temperature for 1 h, briefly
vortexed, sonicated for 10 min in a sonication bath, and vor-
texed again, before analysis.
18. (Trimethylsilyl)diazomethane (TMS-diazomethane) solution
is highly toxic by inhalation and skin contact, and all experi-
mental steps using this reagent must therefore be handled with
appropriate safety measures, i.e., working always under a fume
hood and wearing personal protection equipment. People
handling this compound should consult the safety data sheet
and should be accordingly trained. To minimize risk of larger
spills, only the estimated total volume of TMS-diazomethane
required for an experiment should be transferred from the
stock bottle to a glass or polypropylene tube using a glass
syringe. Aliquots are then taken from this tube. Excessive,
unused TMS-diazomethane should be neutralized by slowly
adding 10% acetic acid in methanol. This reaction is exother-
mic and produces nitrogen gas; therefore only little amounts of
10% acetic acid should be added at once.
19. After adding TMS-diazomethane, the solution in the tubes
turns yellow. After incubation, it may become less colored,
and after addition of acetic acid, the solution must become
colorless, indicating inactivation of TMS-diazomethane. The
amount of acetic acid added is enough to inactivate the
TMS-diazomethane in the tube. However, in case the sample
is still yellow, do not proceed with analyses, verify your proce-
dure, and add more acetic acid for full inactivation.
20. Injection volumes may change depending on the LC-MS sys-
tem used. We advise to check if signal intensities of lipid species
in the TQC dilution series follow the dilution pattern, espe-
cially when using a different LC-MS than the one described in
this article. When signals of lipids do not linearly respond to the
dilution, a signal saturation effect may be present, in which case
lowering the injection volume or diluting the samples may be
helpful (see Subheading 3.7, step 10).
21. LC-MS instrument performances may deteriorate in time,
especially during long sample sequences, due to accumulation
of contaminants in the system. We suggest using the TQC
samples to monitor the system condition and to perform
maintenance/cleaning procedures when necessary. At this
stage, it will be important to document when these procedures
 Clinical Sphingolipidomics 159

took place during the analysis, to be able to explain and correct

possible batch effects at the end of the analysis (see Subheading
3.7, step 11).
22. The column should be fully equilibrated prior to the start of
the sample analysis. The column can be equilibrated for 30 min
using the initial gradient conditions. A minimum of three
chromatographic runs using blanks prior to the start of TQC
injections is suggested.
23. Interferences from [M + 2] isotopes, in-source fragmentation
products, and unknown compounds may be present in the
chromatograms [16]. Particularly, in plasma and serum,
d18:0 sphingolipids are usually less abundant or not detectable
compared to d18:1 sphingolipids, whose M + 2 isotope can
cause interferences that may be misinterpreted as d18:0 sphin-
golipid peaks. For the sphingolipid method data, it is therefore
important to ensure that retention times of picked peaks are in
accordance with the chain length and desaturation of the lipid
species. In reversed-phase separation, lipid species with longer
chains elute later, whereas species with higher number of desa-
turations (double bonds in the acyl and sphingosine chains)
elute earlier [16]. In the S1P method all S1P species, including
the S1P-13C2D2 d18:1 internal standard, nearly co-elute in
overlapping elution profiles due to the HILIC separation
[13]. As a consequence, isotope correction has to be per-
formed for S1P d18:0 and d18:1 (see Note 27). Peaks that
do not at least partially co-elute with S1P-13C2D2 d18:1
indicate interferences and must not be considered (see
Note 23) [13].
24. Presence of corresponding co-eluting qualifier peaks should be
verified for all the integrated quantifier peaks. Compounds
without corresponding co-eluting qualifier may be interfer-
ences from unknown compounds and should not be
integrated.
25. In-source sugar loss may occur from HexCer and Hex2Cer
species and generate ceramide precursors. However, these
interferences can be identified by their shorter retention time
relative to corresponding “true” ceramide species present in
the sample.
26. Internal standards for each lipid class are used to normalize for
the differential extraction and ionization efficiencies of differ-
ent lipid classes.
27. For S1P d18:0 it is necessary to correct for the isotopic inter-
ference from M + 2 ions of S1P d18:1, which is usually higher
in abundance than S1P d18:0. For S1P d18:1 the isotopic
interference from S1P d18:2 M + 2 is usually less than 1% in
plasma and could therefore be omitted. No isotope correction
160 Bo Burla et al.

is needed for the lipids in the sphingolipid method, as those

lipids whose M + 2 isotopes could interfere are separated by
retention time.
28. Accurate quantification can be achieved for S1P d18:1 using
the isotope-labeled S1P-13C2D2 d18:1 internal standard com-
bined with HILIC chromatography. For all other lipid species
in the sphingolipids and S1P method matrix, chromatographic
and ionization effects may affect calculated concentrations as
only one internal standard is used per lipid class. The co-elution
of S1P species due to the HILIC separation in the S1P method
based may compensate for some of these effects and may
therefore allow a more accurate quantification of different
S1P species with only one internal standard (see Note 29).
29. Hypothesizing that the analyte and the selected internal stan-
dard have identical response factors, this general equation can
be used to calculate the concentration of an analyte in the
sample:
AreaAnalyte ν ISTD
c Analyte ¼

c ISTD
AreaISTD ν Sample
AreaAnalyte, peak area of analyte; AreaAnalyte, peak area of inter-
nal standard (ISTD); VISTD, volume of ISTD solution; VSample,
sample volume; cAnalyte, molar concentration of the analyte in
the sample; cISTD, molar concentration of the ISTD in the
ISTD solution (see Table 1).

References
1. Merrill AH (2011) Sphingolipid and glyco- 5. Sillence DJ, Platt FM (2003) Storage diseases:
sphingolipid metabolic pathways in the era of new insights into sphingolipid functions.
sphingolipidomics. Chem Rev 111 Trends Cell Biol 13(4):195–203. https://doi.
(10):6387–6422. https://doi.org/10.1021/ org/10.1016/S0962-8924(03)00033-3
cr2002917 6. Aburasayn H, Batran RA, Ussher JR (2016)
2. Iqbal J, Walsh MT, Hammad SM et al (2017) Targeting ceramide metabolism in obesity.
Sphingolipids and lipoproteins in health and Am J Physiol Endocrinol Metab 311(2):
metabolic disorders. Trends Endocrinol Metab. E423–E435. https://doi.org/10.1152/
https://doi.org/10.1016/j.tem.2017.03.005 ajpendo.00133.2016
3. Hyötyl€ainen T, Ahonen L, Poho P et al (2017) 7. Fucho R, Casals N, Serra D et al (2017) Cer-
Lipidomics in biomedical research-practical amides and mitochondrial fatty acid oxidation
considerations. Biochim Biophys Acta. in obesity. FASEB J 31(4):1263–1272.
https://doi.org/10.1016/j.bbalip.2017.04. https://doi.org/10.1096/fj.201601156R
002 8. Laaksonen R, Ekroos K, Sysi-Aho M et al
4. Zhao Y-Y, Wu S-P, Liu S et al (2014) Ultra- (2016) Plasma ceramides predict cardiovascu-
performance liquid chromatography-mass lar death in patients with stable coronary artery
spectrometry as a sensitive and powerful tech- disease and acute coronary syndromes beyond
nology in lipidomic applications. Chem Biol LDL-cholesterol. Eur Heart J. https://doi.
Interact 220:181–192. https://doi.org/10. org/10.1093/eurheartj/ehw148
1016/j.cbi.2014.06.029
 Clinical Sphingolipidomics 161

9. Penno A, Reilly MM, Houlden H et al (2010) predisposition shape a healthy plasma lipidome.
Hereditary sensory neuropathy type 1 is caused Sci Rep 6:27710. https://doi.org/10.1038/
by the accumulation of two neurotoxic sphin- srep27710
golipids. J Biol Chem 285(15):11178–11187. 20. Dunn WB, Wilson ID, Nicholls AW et al
https://doi.org/10.1074/jbc.M109.092973 (2012) The importance of experimental design
10. Zuellig RA, Hornemann T, Othman A et al and QC samples in large-scale and MS-driven
(2014) Deoxysphingolipids, novel biomarkers untargeted metabolomic studies of humans.
for type 2 diabetes, are cytotoxic for insulin- Bioanalysis 4(18):2249–2264. https://doi.
producing cells. Diabetes 63(4):1326–1339. org/10.4155/bio.12.204
https://doi.org/10.2337/db13-1042 21. Kohler I, Verhoeven A, Derks RJ et al (2016)
11. Shayman JA (2016) Targeting glycosphingoli- Analytical pitfalls and challenges in clinical
pid metabolism to treat kidney disease. Neph- metabolomics. Bioanalysis 8(14):1509–1532.
ron 134(1):37–42. https://doi.org/10.1159/ https://doi.org/10.4155/bio-2016-0090
000444926 22. Yin P, Peter A, Franken H et al (2013) Preana-
12. Kunkel GT, Maceyka M, Milstien S et al (2013) lytical aspects and sample quality assessment in
Targeting the sphingosine-1-phosphate axis in metabolomics studies of human blood. Clin
cancer, inflammation and beyond. Nat Rev Chem 59(5):833–845. https://doi.org/10.
Drug Discov 12(9):688–702. https://doi. 1373/clinchem.2012.199257
org/10.1038/nrd4099 23. Wang J-R, Zhang H, Yau LF et al (2014)
13. Narayanaswamy P, Shinde SA, Sulc R et al Improved sphingolipidomic approach based
(2014) Lipidomic ‘deep profiling’: an on ultra-high performance liquid chromatog-
enhanced workflow to reveal new molecular raphy and multiple mass spectrometries with
species of signaling lipids. Anal Chem 86 application to cellular neurotoxicity. Anal
(6):3043–3047. https://doi.org/10.1021/ Chem 86(12):5688–5696. https://doi.org/
ac4039652 10.1021/ac5009964
14. Blanksby SJ, Mitchell TW (2010) Advances in 24. Ono Y, Kurano M, Ohkawa R et al (2013)
mass spectrometry for lipidomics. Annu Rev Sphingosine 1-phosphate release from platelets
Anal Chem (Palo Alto, Calif) 3:433–465. during clot formation: close correlation
https://doi.org/10.1146/annurev.anchem. between platelet count and serum sphingosine
111808.073705 1-phosphate concentration. Lipids Health Dis
15. Sullards MC, Liu Y, Chen Y et al (2011) Anal- 12:20. https://doi.org/10.1186/1476-
ysis of mammalian sphingolipids by liquid 511X-12-20
chromatography tandem mass spectrometry 25. Yu Z, Kastenmuller G, He Y et al (2011) Dif-
(LC-MS/MS) and tissue imaging mass spec- ferences between human plasma and serum
trometry (TIMS). Biochim Biophys Acta metabolite profiles. PLoS One 6(7):e21230.
1811(11):838–853. https://doi.org/10. https://doi.org/10.1371/journal.pone.
1016/j.bbalip.2011.06.027 0021230
16. Wang J-R, Zhang H, Yau LF et al (2014) 26. Chua EC-P, Shui G, Lee IT-G et al (2013)
Improved sphingolipidomic approach based Extensive diversity in circadian regulation of
on ultra-high performance liquid chromatog- plasma lipids and evidence for different circa-
raphy and multiple mass spectrometries with dian metabolic phenotypes in humans. Proc
application to cellular neurotoxicity. Anal Natl Acad Sci U S A 110(35):14468–14473.
Chem 86(12):5688–5696. https://doi.org/ https://doi.org/10.1073/pnas.1222647110
10.1021/ac5009964 27. Hammad SM, Pierce JS, Soodavar F et al
17. Alshehry ZH, Barlow CK, Weir JM et al (2015) (2010) Blood sphingolipidomics in healthy
An efficient single phase method for the extrac- humans: impact of sample collection method-
tion of plasma lipids. Metabolites 5 ology. J Lipid Res 51(10):3074–3087. https://
(2):389–403. https://doi.org/10.3390/ doi.org/10.1194/jlr.D008532
metabo5020389 28. Denery JR, Nunes AAK, Dickerson TJ (2011)
18. Begum H, Li B, Shui G et al (2016) Discover- Characterization of differences between blood
ing and validating between-subject variations sample matrices in untargeted metabolomics.
in plasma lipids in healthy subjects. Sci Rep Anal Chem 83(3):1040–1047. https://doi.
6:19139. https://doi.org/10.1038/ org/10.1021/ac102806p
srep19139 29. Lima-Oliveira G, Volanski W, Lippi G et al
19. Sales S, Graessler J, Ciucci S et al (2016) Gen- (2017) Pre-analytical phase management: a
der, contraceptives and individual metabolic review of the procedures from patient
162 Bo Burla et al.

preparation to laboratory analysis. Scand J Clin 31. Frej C, Andersson A, Larsson B et al (2015)
Lab Invest 77(3):153–163. https://doi.org/ Quantification of sphingosine 1-phosphate by
10.1080/00365513.2017.1295317 validated LC-MS/MS method revealing strong
30. Heiligers-Duckers C, Peters NALR, van Dijck correlation with apolipoprotein M in plasma
JJP et al (2013) Low vacuum and discard tubes but not in serum due to platelet activation
reduce hemolysis in samples drawn from intra- during blood coagulation. Anal Bioanal Chem
venous catheters. Clin Biochem 46 407(28):8533–8542. https://doi.org/10.
(12):1142–1144. https://doi.org/10.1016/j. 1007/s00216-015-9008-4
clinbiochem.2013.04.005
 Chapter 12

Shotgun Lipidomics Approach for Clinical Samples

Lars F. Eggers and Dominik Schwudke

Abstract
Shotgun lipidomics offers fast and reproducible identification and quantification of lipids in clinical samples.
Lipid extraction procedures based on the methyl tert-butyl protocol are well established for performing
shotgun lipidomics in biomedical research. Here, we describe a shotgun lipidomics workflow that is well suited
for the analysis of clinical samples such as tissue samples, blood plasma, and peripheral blood mononuclear cells.

Key words Shotgun lipidomics, High-resolution mass spectrometry, Clinical samples, Lipid extrac-
tion, Lipid identification

Abbreviations

CE Cholesteryl ester
Cer Ceramide
DAG Diacylglycerol
FT-ICR Fourier transform ion cyclotron resonance
HexCer Hexosyl ceramide
LPC Lysophosphatidylcholine
MAG Monoacylglycerol
MFQL Molecular fragmentation query language
MS Mass spectrometry
MS2 Tandem mass spectrometry
MTBE Methyl tert-butyl ether (UPAC, tert-butyl methyl ether)
PC Phosphatidylcholine
PE Phosphatidylethanolamine
PG Phosphatidylglycerol
PI Phosphatidylinositol
PS Phosphatidylserine
TAG Triacylglycerol
TOF Time of flight

Electronic supplementary material: The online version of this chapter (https://doi.org/10.1007/978-1-4939-

7592-1_12) contains supplementary material, which is available to authorized users.

Martin Giera (ed.), Clinical Metabolomics: Methods and Protocols, Methods in Molecular Biology, vol. 1730,
https://doi.org/10.1007/978-1-4939-7592-1_12, © Springer Science+Business Media, LLC 2018

163
164 Lars F. Eggers and Dominik Schwudke

1 Introduction

Shotgun lipidomics is a well-established approach for lipid analysis,

and a number of clinical applications were reported [1–3]. The
term “shotgun lipidomics” was coined by Han and Gross [4] and
refers to direct analysis of complex lipid extracts by mass spectrom-
etry [5, 6]. The shotgun approach avoids time-consuming chro-
matographic separations and takes advantage of the superior
analytical power of modern tandem mass spectrometers [7]. Appli-
cation of high-resolving mass analyzers like Orbitrap, FT-ICR, and
TOF enables unambiguous quantification of a variety of phospho-
lipids, sphingolipids, and neutral lipids. Top-down lipidomics is
based on accurate mass determination in MS1 and allows to deter-
mine chemical sum compositions. The heteroatom content is char-
acteristic for many lipid classes, and lipid species can be assigned
according to the length of aliphatic chains and number of double
bond [6, 8]. The bottom-up approach identifies lipids on the basis of
specific fragments for the lipid class-specific head groups and fatty
acids in MS2 [9–12]. With fast high-resolving tandem mass spectro-
meters, a sufficient number of MS2 spectra can be collected
together with MS1 information. These complex mass spectrometric
datasets are analyzed with the dedicated LipidXplorer software
[13, 14]. As a result, an improved coverage of the lipidome is
achieved because the aliphatic chain compositions can be deter-
mined [14, 15].
Here, we describe a shotgun lipidomics workflow suitable for
clinical analyses. We describe in detail the sample preparation pro-
cedures for tissue homogenization and lipid extraction using the
methyl tert-butyl ether (MTBE) protocol [16]. We provide detailed
information for lipid profiling based on high-resolution MS instru-
mentation and provide a data analysis strategy for lipid identifica-
tion and quantitation based on LipidXplorer.

2 Materials

All solvents and chemicals should be of LC-MS grade or otherwise

with the highest purity grade available. All glass vials and Pasteur
pipets have to be washed three times with each solvent with increas-
ing hydrophobicity, (1) water, (2) methanol, and (3) MTBE, before
use (see Note 1).
 Clinical Lipidomics 165

2.1 Tissue 1. Homogenization buffer: Weigh 1.86 g KCl (74.55 g/mol) in a

Homogenization 500 mL volumetric flask and fill it to the mark with water. Store
the solution at 4
C.
2. Homogenization instrument with sample tubes (20 mL) con-
taining stainless steel beads for tissue amounts below 200 mg;
otherwise tubes with rotor-stator element (see Note 2).
3. Prepare solution of butylhydroxytoluene (BHT, M ¼
220.35 g/mol) in methanol: Solve 10 mg BHT in 10 mL
methanol and store the solution at 4
C (see Note 3).

2.2 Lipid Extraction 1. Prepare Pasteur pipets, at least two per sample, and Eppendorf
tubes (Eppendorf Safe-Lock tubes). Per sample is needed: two
tubes, size 2 mL, and three tubes, size 0.5 mL (see Note 4).
2. Centrifuge and vacuum centrifuge (SpeedVac).
3. Shaker (e.g., Eppendorf MixMate).
4. Methanol solution (containing 3% acetic acid): Mix 20 mL
methanol and 600 μL acetic acid in a 50 mL glass bottle.
Store the solution at room temperature (see Note 5).
5. Re-extraction solvent: Mix 20 mL MTBE with 6 mL methanol
(with 3% acetic acid) and 5 mL water. From the resulting
biphasic system, the upper phase is used for the second extrac-
tion step (see Note 6).
6. Internal standard solution: We use SPLASH LipidoMix mass
spectrometry standard from Avanti Polar Lipids (No. 330707)
and mix it with ceramide (d18:1/25:0) quantitative mass spec-
trometry standard (Avanti Polar Lipids No. LM-2225). Pur-
chased vials of SPLASH and ceramide (d18:1/25:0) contain
1 mL standard solution. This solution is quantitatively trans-
ferred into a 10 mL volumetric flask and filled with methanol to
the ring mark. From this stock, aliquots of 500 μL volume were
prepared and stored at 20
C (see Note 7) (Table 1).
7. Storage solution: Mix 30 mL of chloroform with 15 mL meth-
anol and 2.25 mL water.
8. For adding internal standard solution, use precision glass capil-
laries with calibrated volumes (Fig. 1).

2.3 Shotgun 1. 96-well plate or 384-well plate (e.g., Eppendorf Twin Tec).
Lipidomics 2. Spray solution: Mix 5 mL chloroform with 10 mL methanol
(containing 0.1% w/v ¼ 1 g/L ammonium acetate,
M ¼ 77.08 g/mol) and 20 mL 2-propanol. The resulting
ammonium acetate concentration is 3.7 mM (see Note 8).
3. Mass spectrometer with high-resolving power (Rmin@700 ¼
100,000); this protocol refers to usage of a Q Exactive Plus
(Thermo, Bremen, Germany).
 166

Table 1
Composition of the internal standards. Concentrations are given for this protocol. Note that exact concentrations may vary between batches of
SPLASH

Molecular
weight Concentration in
Mixture components (g/mol) Chemical formula m/z (positive) m/z (negative) mixture (pmol/μL)
15:0–18:1(d7) PC 753.11 C41H73D7NO8P 753.6134 [M þ H]+ 811.6199 [M þ CH3COO] 21.34
+ 
15:0–18:1(d7) PE 711.03 C38H67D7NO8P 711.5664 [M þ H] 709.5519 [M  H] 0.80
15:0–18:1(d7) PS(Na-Salt) 777.02 C39H66D7NNaO10P – 753.5417 [M  H] 0.54

15:0–18:1(d7) PG(Na-Salt) 764.02 C39H67D7NaO10P – 740.5464 [M  H] 3.81

Lars F. Eggers and Dominik Schwudke

15:0–18:1(d7) PI(NH4-Salt) 847.13 C42H75D7NO13P – 828.5625 [M  H] 1.07


15:0–18:1(d7) PA(Na-Salt) 689.94 C36H61D7NaO8P – 666.5097 [M  H] 1.07
18:1(d7) LPC 528.72 C26H45D7NO7P 529.3994 [M þ H]+ 587.4059 [M þ CH3COO] 4.82
+ 
18:1(d7) LPE 486.64 C23H39D7NO7P 487.3524 [M þ H] 485.3378 [M  H] 1.08
+
18:1(d7) Chol ester 658.16 C45H71D7O2 675.6779 [M þ NH4] – 54.11
+
18:1(d7) MAG 363.59 C21H33D7O4 381.3704 [M þ NH4] – 0.55
15:0–18:1(d7) DAG 587.98 C36H61D7O5 605.5844 [M þ NH4]+ – 1.60
+
15:0–18:1(d7)-18:1 TAG 812.37 C51H89D7O6 829.7985 [M þ NH4] – 7.05
+ 
18:1(d9) SM 738.12 C41H72D9N2O6P 738.6470 [M þ H] 796.6529 [M þ CH3COO] 4.19
+
Cholesterol(d7) 393.71 C27H39D7O 411.4326 [M þ NH4] – 24.99
Cer(d18:1/25:0) 664.14 C43H85NO3 664.6602 [M þ H]+ 2.44
 Clinical Lipidomics 167

Fig. 1 Workflow of MTBE-based lipid extraction. MeOH methanol, ISD internal standard, RT room temperature

4. Nano-electrospray ion source: This protocol refers to the Tri-

Versa NanoMate (Advion, Ithaca, USA). Flow injection strate-
gies are another possibility to automatize shotgun
lipidomics [12].
168 Lars F. Eggers and Dominik Schwudke

3 Methods

3.1 Homogenization 1. Tissue samples, stored at 80
C and thawed on ice, are
of Tissue Samples weighted and transferred to the homogenization tubes.
2. Homogenization buffer is added in 20-fold volumetric excess
to the weight of the tissue sample.
3. Add BHT solution (1 mg/mL) to the tissue samples: 200 μL
per 1 g tissue.
4. Homogenize tissues for 2 min at 6000 rpm and reverse rota-
tion every 30 s. Repeat this step two to five times until the
tissue is sufficiently disintegrated.
5. Collect aliquots of 20 μL in collection tubes. Take one aliquot
for lipid extraction, and store remaining homogenates at
80
C as reserve samples.

3.2 Lipid Extraction 1. Thaw tissue homogenates on ice and keep them cold until
of Tissue methanol is added. Add 20 μL homogenization buffer to the
Homogenates homogenate aliquots.
2. Add 290 μL methanol (3% acetic acid) and 10 μL of the internal
standard mix. Keep samples at room temperature.
3. Add 1 mL MTBE and incubate for 1 h at room temperature
under continuous shaking (600 rpm).
4. After 1 h, add 250 μL of water to the samples and incubate for
another 10 min at 1300 rpm.
5. Centrifuge samples for 10 min at 15,000  g (see Note 9).
6. Transfer 900 μL of the upper phase into a separate 2 mL
collection tube (see Notes 10 and 11).
7. Add 400 μL of the re-extraction solvent to the remaining water
phases.
8. Incubate samples for 20 min at room temperature under con-
tinuous shaking (1300 rpm).
9. Centrifuge samples for 10 min at 15,000  g.
10. Transfer 450 μL of the upper phase into the collection tube.
11. Dry samples in the vacuum centrifuge (see Note 12).
12. Resolve dried lipid extracts in 50 μL storage solution (see Note
13).
13. Pipet three to four aliquots with 10 μL in 500 μL sample tubes.

3.3 Lipid Extraction 1. Dilute EDTA plasma tenfold with water.

of Blood Plasma 2. 50 μL are transferred to a new tube.
3. 20 μL of the internal standard mix are added and the mixture is
well mixed.
 Clinical Lipidomics 169

4. Add 250 μL methanol (3% acetic acid) and agitate the

suspension.
5. Add 1 mL MTBE and incubate the mixture for 1 h with
constant stirring.
6. For induction of phase separation, add 250 μL water. Incubate
the extraction mixture for another 5 min on a shaker.
7. Afterward, the extraction mixture is centrifuged at 15,000  g
for 2 min to improve phase separation.
8. 800 μL of the upper organic phase are collected into a new
sample tube. Optional, the remaining phase is extracted a
second time as described above (Subheading 3.2). The organic
phase should be clear and free of any residual aqueous phase
and insoluble material.
9. The organic phase is dried down with the help of a vacuum
concentrator. Optional, the aqueous phase can be stored for
further determination of the protein content (Subheading
3.4).
10. The dried phase is taken up in 100 μL storage solution and can
be stored until mass spectrometric analysis.

3.4 Lipid Extraction 1. Cell pellets of 1E06 PBMCs are carefully resuspended in 50 μL
of Peripheral Blood water.
Mononuclear Cells 2. Add 4 μL of the internal standard mix.
(PBMCs) 3. Add 270 μL methanol (3% acetic acid) and mix the
suspension well.
4. Add 1 mL MTBE and incubate the mixture for 1 h with
constant stirring.
5. Add 250 μL water to induce phase separation and incubate the
suspension for another 5 min.
6. Centrifuge the extraction mixture at 15,000  g to improve
phase separation.
7. Collect 800 μL of the upper organic phase and transfer it to a
new sample tube. Optional, the remaining phase is extracted a
second time as described above (Subheading 3.2). The organic
phase should be clear and free of any residual aqueous phase
and insoluble material.
8. The combined organic phases are dried down and dissolved in
50 μL storage solution.

3.5 Shotgun 1. Add 190 μL of the spray solution to the 10 μL sample aliquots
Lipidomics directly before MS analysis.
2. Vortex each sample and centrifuge for at least 5 min at
15,000  g to prevent that particles clog nanoESI nozzles.
170 Lars F. Eggers and Dominik Schwudke

3. Pipet 20 μL of each sample onto a 96-well or 384-well plate.

We recommend analyzing each sample twice. Consecutively,
two wells are needed per sample (see Note 14).
4. Set spray voltage to 1.1 kV and back pressure (only for TriVersa
NanoMate) at 1.1 psi.
5. Whole analysis duration is 10 min. Ion source switches polarity
at 5 min.
6. Instrument acquires MS1 spectra in a range of m/z 350–1000
in both ion modes.
7. MS2 spectra are acquired in data-independent acquisitions
(DIA). An MS2 spectra are acquired every m/z 1.001 starting
at m/z 350.2 in the positive ion mode and at m/z 350.1 in the
negative ion mode.
8. When a Q Exactive Plus is utilized, set resolving power of MS1
scans to 270,000 and of MS2 scans to 70,000.

3.6 Data Analysis 1. Install LipidXplorer Version 1.2.7 according to the documenta-
with LipidXplorer tion available on https://wiki.mpi-cbg.de/lipidx/Main_Page.
2. Convert raw data files into *.mzML file by MS convert from
ProteoWizard [17]: http://proteowizard.sourceforge.net/
downloads.shtml
3. Data files are imported into LipidXplorer using the import
settings lpdxImportSettings_LE4.ini as available in the online
supplement (see Note 15).
4. In the online supplement, we provide MFQL (molecular frag-
mentation query language) scripts for lipid identification based
on data acquired on a Q Exactive Plus using both positive and
negative ion mode. Detailed information about LipidXplorer
and MFQL can be found in the tutorial by Herzog et al. [14]
(see Note 16). Specific MFQLs for SPLASH lipid standard
components are also provided.
5. After import of the data and lipid identification with MFQL
scripts for each ion mode separately, *.csv files are written
providing intensities for each lipid species as well as for the
internal standards. In consecutive steps, quantities can be cal-
culated based on this data.
6. We recommend eliminating background ions from the dataset
by deleting all identified lipids whose average abundance is
lower than ten times their abundance in blank extracts.
 Clinical Lipidomics 171

4 Notes

1. Industrial cleaned glass surfaces are in our experience not suffi-

ciently cleaned for lipid analysis. To avoid ionization suppres-
sion of lipids, we employ for all utilized glassware the described
cleaning procedure with LC-MS grade solvents.
2. The selection of the homogenization device depends on the
tissue amounts that should be homogenized. We recommend,
if available, to use higher tissue amounts about 500 mg to cover
the inhomogeneity within the tissue. As described in Subhead-
ing 3, a 20-fold excess of buffer is given to the tissue samples
during homogenization. When used tissues amount below
200 mg, the volume of buffer is too low to yield sufficient
homogenization in rotor-stator system. In this case, use of
steel beads is beneficial.
3. Always prepare a fresh BHT solution.
4. The polypropylene of Eppendorf Safe-Lock tubes has a well-
defined chemical background [18]. We recommend to deter-
mine the chemical background pattern in conjunction of any
mass spectrometric lipid analysis, especially when changing the
manufacturer of the extraction tubes. Some of the known
background signals can be used for mass recalibration.
5. The acidified methanol solution has to be prepared freshly.
Acetic acid reacts with methanol to form methyl acetate. We
observed impaired extraction efficiency and reproducibility if
acidified methanol was not freshly prepared.
6. This biphasic system has an overall composition of 10:3:2.5
(MTBE, methanol, water; v/v/v), which is exact the same
composition in the first extraction step. The goal is to keep
the volumetric ratios of MTBE, water, and methanol constant
during the second extraction step. If only MTBE is added, the
volumetric ratios shift and the extraction efficiency could be
affected.
7. The choice of internal standards has to be customized for the
applied screening experiment. Lipids with dialkyl-bound side
chains, for example, are commercially available and do not
naturally occur in clinical samples. They form the same adducts
than natural lipids of the same class, although ionization effi-
ciencies differ from diacyl species. Unfortunately, these species
undergo different fragmentation mechanisms than diacyl lipids
which makes them only suitable for top-down lipidomics screen.
Another strategy would be to choose lipids with odd numbers
of carbon atoms in their fatty acid residues (e.g., 17:0). These
lipids are not available by mammalian biosynthetic pathways.
From our experience, “odd-chain” lipid species are detected
172 Lars F. Eggers and Dominik Schwudke

with low abundance in human lipidomes. Potentially, these

lipids were taken up by nutrition. The SPLASH mix contains
lipids with deuterated acyl chains (d7 18:1) making this mix-
ture ideal for top-down as well as bottom-up strategies. Addi-
tional lipid standards might be required depending on the
experimental question. Here, we added a ceramide standard.
8. To yield higher sensitivities for anionic phospholipids in nega-
tive ion mode, 0.05 mM ammonium chloride in the spray
solution can be used.
9. The sample tube contains a biphasic system: The upper phase is
composed mainly by MTBE and contains extracted lipids. The
lower phase is aqueous and contains small polar molecules
(metabolites). On the bottom of the tube, precipitated proteins
and biomacromolecules are located.
10. This step has to be carried out with care. Make sure that only
the organic phase is collected.
11. To save time to dry down samples, the extracts in the collection
tubes can be concentrated in a SpeedVac.
12. This step could be time-consuming. Dry sample at first at room
temperature until the pressure notably drops in the SpeedVac
(~45 min). Then MTBE is completely evaporated, and the
remaining sample contains predominantly water. At this point
methanol can be added to the samples to speed up drying. One
can increase the temperature to 45
C. It takes about 1.5 h
until complete dryness. As long as there is water in the tube,
the samples stay cooled. Make sure that dried samples do not
remain an unnecessary long time in the heated vacuum
chamber.
13. Care has to be taken when dried samples are resolved in the
storage solution. Vortex the samples carefully and extensively.
14. We recommend to keep the temperature of the autosampler
between 15 and 18
C for long sequences.
15. The LipidXplorer import settings were written for data
acquired on a Q Exactive Plus instrument and acquisitions as
described under Subheading 3.3.
16. The MFQL scripts report quantifier ions in columns with
prefixes “INT.” The quantifiers were chosen to be specific for
each individual lipid species. In negative ion mode, we quantify
based on the sum intensity of both fatty acid fragments in MS2.
The bases for quantification are signals of the deuterated fatty
acids of the SPLASH internal standard mix. Cardiolipin is
quantified as doubly charged precursor ion in negative ion
mode based on the first isotope signal as proposed by Han
et al. [19]. No internal standard for cardiolipin was utilized; we
utilize the response of the PG internal standard as reference.
 Clinical Lipidomics 173

Quantification of cholesterol esters is based on the fragment

ion at m/z 369.35. For ceramides the C18 long-chain base
signal at m/z 264.26 is used. For DAG and TAG species, the
specific neutral loss of fatty acids is used for quantitation. MAG
species are exclusively identified and quantified by their accu-
rate mass in MS1. SM species are also quantified based on their
precursor ion. We recommend to use the negative ion mode for
PE and PC quantitation.
All provided MFQL scripts contain empirical determined
rules for filtering initial identification results. Intensity ratios of
precursor ions and fragment ions were determined from
internal standards and are used as quality check to avoid false
assignment due to in-source fragmentation and adducts.
These ratios are dependent on the instrument platform.
MFQL script for all glycerophospholipids contains “(avg(PR.
intensity) / (avg(FA1.intensity) þ avg(FA2.intensity)) &gt;¼
X AND avg(PR.intensity) / (avg(FA1.intensity) þ avg(FA2.
intensity)) &lt;¼ Y)” where X represent the lower and Y
higher limit for this ratio for each class.

Acknowledgments

The research was supported by funds of the EXC 306 Inflammation

at Interfaces, German Centre for Infection Research (DZIF
TTU-TB), German Centre for Lung Research, Airway Research
Center North (ARCN), and Lipidomics Informatics for Life-
Sciences (LIFS) of the German Network for Bioinformatics Infra-
structure (de.NBI).

References
1. Griese M, Kirmeier HG, Liebisch G, Rauch D, predisposition shape a healthy plasma lipi-
St€uckler F, Schmitz G, Zarbock R, ILD-BAL dome. Sci Rep 6:27710. https://doi.org/10.
working group of the Kids-Lung-Register 1038/srep27710
(2015) Surfactant lipidomics in healthy chil- 4. Han X, Gross RW (2003) Global analyses of
dren and childhood interstitial lung disease. cellular lipidomes directly from crude extracts
PLoS One 10(2):e0117985. https://doi.org/ of biological samples by ESI mass spectrometry
10.1371/journal.pone.0117985 a bridge to lipidomics. J Lipid Res 44
2. Graessler J, Schwudke D, Schwarz PEH, (6):1071–1079
Herzog R, Shevchenko A, Bornstein SR 5. Shevchenko A, Simons K (2010) Lipidomics:
(2009) Top-down lipidomics reveals ether coming to grips with lipid diversity. Nat Rev
lipid deficiency in blood plasma of hypertensive Mol Cell Biol 11(8):593–598. https://doi.
patients. PLoS One 4(7):e6261. https://doi. org/10.1038/nrm2934
org/10.1371/journal.pone.0006261 6. Schwudke D, Schuhmann K, Herzog R, Born-
3. Sales S, Graessler J, Ciucci S, Al-Atrib R, stein SR, Shevchenko A (2011) Shotgun lipi-
Vihervaara T, Schuhmann K, Kauhanen D, domics on high resolution mass spectrometers.
Sysi-Aho M, Bornstein SR, Bickle M, Cannis- Cold Spring Harb Perspect Biol 3(9). https://
traci CV, Ekroos K, Shevchenko A (2016) Gen- doi.org/10.1101/cshperspect.a004614
der, contraceptives and individual metabolic
174 Lars F. Eggers and Dominik Schwudke

7. Heiskanen LA, Suoniemi M, Ta HX, spectrometry platforms. Curr Protoc Bioinfor-

Tarasov K, Ekroos K (2013) Long-term per- matics 43:14 12 11–30. doi:https://doi.org/
formance and stability of molecular shotgun 10.1002/0471250953.bi1412s43
lipidomic analysis of human plasma samples. 15. Liebisch G, Vizcaino JA, Kofeler H,
Anal Chem 85(18):8757–8763. https://doi. Trotzmuller M, Griffiths WJ, Schmitz G,
org/10.1021/ac401857a Spener F, Wakelam MJ (2013) Shorthand
8. Schwudke D, Hannich JT, Surendranath V, notation for lipid structures derived from
Grimard V, Moehring T, Burton L, mass spectrometry. J Lipid Res 54
Kurzchalia T, Shevchenko A (2007) (6):1523–1530. https://doi.org/10.1194/jlr.
Top-down lipidomic screens by multivariate M033506
analysis of high-resolution survey mass spectra. 16. Matyash V, Liebisch G, Kurzchalia TV,
Anal Chem 79(11):4083–4093. https://doi. Shevchenko A, Schwudke D (2008) Lipid
org/10.1021/ac062455y extraction by methyl-tert-butyl ether for high-
9. Br€ugger B, Erben G, Sandhoff R, Wieland FT, throughput lipidomics. J Lipid Res 49
Lehmann WD (1997) Quantitative analysis of (5):1137–1146. https://doi.org/10.1194/jlr.
biological membrane lipids at the low picomole D700041-JLR200
level by nano-electrospray ionization tandem 17. Chambers MC, Maclean B, Burke R,
mass spectrometry. Proc Natl Acad Sci U S A Amodei D, Ruderman DL, Neumann S,
94(6):2339–2344 Gatto L, Fischer B, Pratt B, Egertson J,
10. Schwudke D, Oegema J, Burton L, Entchev E, Hoff K, Kessner D, Tasman N, Shulman N,
Hannich JT, Ejsing CS, Kurzchalia T, Shev- Frewen B, Baker TA, Brusniak M-Y, Paulse C,
chenko A (2006) Lipid profiling by multiple Creasy D, Flashner L, Kani K, Moulding C,
precursor and neutral loss scanning driven by Seymour SL, Nuwaysir LM, Lefebvre B,
the data-dependent acquisition. Anal Chem 78 Kuhlmann F, Roark J, Rainer P, Detlev S,
(2):585–595. https://doi.org/10.1021/ Hemenway T, Huhmer A, Langridge J,
ac051605m Connolly B, Chadick T, Holly K, Eckels J,
11. Schuhmann K, Herzog R, Schwudke D, Deutsch EW, Moritz RL, Katz JE, Agus DB,
Metelmann-Strupat W, Bornstein SR, Shev- MacCoss M, Tabb DL, Mallick P (2012) A
chenko A (2011) Bottom-up shotgun lipido- cross-platform toolkit for mass spectrometry
mics by higher energy collisional dissociation and proteomics. Nat Biotechnol 30
on LTQ Orbitrap mass spectrometers. Anal (10):918–920. https://doi.org/10.1038/
Chem 83(14):5480–5487. https://doi.org/ nbt.2377. http://www.nature.com/nbt/jour
10.1021/ac102505f nal/v30/n10/abs/nbt.2377.
12. Leidl K, Liebisch G, Richter D, Schmitz G html#supplementary-information
(2008) Mass spectrometric analysis of lipid spe- 18. Schuhmann K, Almeida R, Baumert M,
cies of human circulating blood cells. Biochim Herzog R, Bornstein SR, Shevchenko A
Biophys Acta 1781(10):655–664. https://doi. (2012) Shotgun lipidomics on a LTQ Orbitrap
org/10.1016/j.bbalip.2008.07.008 mass spectrometer by successive switching
13. Herzog R, Schwudke D, Schuhmann K, between acquisition polarity modes. J Mass
Sampaio J, Bornstein S, Schroeder M, Shev- Spectrom 47(1):96–104. https://doi.org/10.
chenko A (2011) A novel informatics concept 1002/jms.2031
for high-throughput shotgun lipidomics based 19. Han X, Yang K, Yang J, Cheng H, Gross RW
on the molecular fragmentation query lan- (2006) Shotgun lipidomics of cardiolipin
guage. Genome Biol 12(1):R8 molecular species in lipid extracts of biological
14. Herzog R, Schwudke D, Shevchenko A (2013) samples. J Lipid Res 47(4):864–879. https://
LipidXplorer: software for quantitative shot- doi.org/10.1194/jlr.D500044-JLR200
gun lipidomics compatible with multiple mass
 Chapter 13

Establishing and Performing Targeted Multi-residue

Analysis for Lipid Mediators and Fatty Acids in Small
Clinical Plasma Samples
Theresa L. Pedersen and John W. Newman

Abstract
LC-MS/MS- and GC-MS-based targeted metabolomics is typically conducted by analyzing and quantify-
ing a cascade of metabolites with methods specifically developed for the metabolite class. Here we describe
an approach for the development of multi-residue analytical profiles, calibration standards, and internal
standard solutions in support of a fast, simple, and low-cost plasma sample preparation that captures and
quantitates a range of metabolite cascades.

Key words Oxylipins, Endocannabinoids, Bile acids, Nonesterified fatty acids, Targeted metabolo-
mics, LC-MS/MS, Metabolic profiling

1 Introduction

Targeted metabolomics provides sensitive, quantitative data sets

that can be used to investigate the metabolic consequences of
disease, drug treatment effects, and outcomes of interventions, as
well as establishing clinical reference ranges and reference materials
for determining accuracy and method validation [1, 2]. The manual
optimization of analyte ionization and fragmentation parameters
for large multi-residue methods is an overwhelming prospect, espe-
cially when transferring a method to a new hardware platform. Over
the past 10 years, we have developed an approach that has allowed
us to successfully transfer complex multi-residue analyses to labora-
tories around the world [3–6]. Here we describe a semiautomated
alternative chromatographic “gradient optimization” approach to
transfer large profiles between LC-MS/MS platforms, where the
acquisition method is used to optimize all analytes simultaneously

Electronic supplementary material: The online version of this chapter (https://doi.org/10.1007/978-1-4939-

7592-1_13) contains supplementary material, which is available to authorized users.

Martin Giera (ed.), Clinical Metabolomics: Methods and Protocols, Methods in Molecular Biology, vol. 1730,
https://doi.org/10.1007/978-1-4939-7592-1_13, © Springer Science+Business Media, LLC 2018

175
176 Theresa L. Pedersen and John W. Newman

and provide a record of the optimization, in a timely manner. This

includes establishing global MS parameters, determining analyte
retention times, and making a series of acquisition methods to
simultaneously optimize analyte-specific parameters under the
LC-MS/MS acquisition method conditions.
Additionally, increased dynamic linear ranges of 5–6 orders of
magnitude allow for the simultaneous observation and quantifica-
tion of many metabolites in one analysis, and we have taken advan-
tage of this to fuse the analysis of multiple lipid classes [7, 8]. Here
we share an approach for quantifying oxygenated lipids, endocan-
nabinoids and like compounds, bile acids, and free fatty acids,
captured in two LC-MS/MS and one GC-MS acquisition methods
from a single small plasma extract. Included are detailed instruc-
tions for the systematic development of the multi-residue calibra-
tion standards for these assays. This includes a worksheet for en
masse calculations to make up analyte mixtures and calibration
standard solutions, which can be expanded to include additional
metabolic cascades of interest.
Classical sample preparation and cleanup methods for multi-
residue quantitative analysis are labor intensive and expensive.
Increased sensitivity in modern MS/MS allows for substantial
reductions in sample volume, eliminating the need for extensive
cleanup procedures, reducing matrix effects, and resulting in more
robust methods. We present a small-volume protein precipitation
with filtration as the plasma sample preparation procedure prior to
LC-MS/MS and GC-MS analysis. Included is a summary instruc-
tion for quality control procedures to assess method robustness.

2 Materials

The Materials section includes procedures needed to perform the

described extraction and quantitative analysis. All LC-MS assay
solvents are Fisher LC-MS grade, 0.2 μm filtered. GC-MS assay
solvents are Fisher Optima grade or equivalent. All liquid chroma-
tography mobile phases are maintained in glass Pyrex bottles, and
the phases and solutions containing organic solvents are stored with
Teflon-lined caps. Aqueous phase solutions are stored at 4
C until
use and not longer than 2 weeks (see Note 1). Prior to use, prerinse
all volumetric flasks, vials, tubes, plates, and glassware with metha-
nol (MeOH), followed by hexane for GC glassware, and allow to
dry in a glass beaker loosely covered with foil, in a dust-free chemi-
cal exhaust hood. Use analytical volumetric glass barrel syringes and
flasks to prepare all solutions containing analytical standards. Glass
barrel syringes should be cleaned in MeOH and dried by house
vacuum (use a funnel on top of a sidearm flask) between measure-
ments. Standard stock solutions, optimization solutions, and work-
ing solutions are stored in amber vials topped with nitrogen gas,
 Targeted Analysis of Lipid Mediators and Fatty Acids in Plasma 177

with Teflon-lined caps. Use thin ½ in. Teflon tape to seal the
outside of caps to prevent solvent evaporation, and mark the menis-
cus to assess volume stability during storage. It is recommended
that all analyte mixtures, calibration standards, surrogate spike
solutions, and internal standard sample reconstitution solutions
are ampouled under nitrogen gas and stored at 20
C. Analyte
stock solutions are stored at 80
C. Use a calibrated analytical
balance with accuracy to 0.1 mg to weigh neat materials.

2.1 Materials 1. Volumetric flasks: A minimum of two 1.0 mL, nine 5 mL, two
for Calibration 10 mL, two 25 mL, and one 100 mL volumetric flasks are
Standard Preparation needed.
2. Glass syringes: Each of the following volumes 10, 50, 100, 250,
and 500 μL of gastight syringes with Teflon-tipped plungers
and blunt-tipped needles is needed.
3. Analytes: Purchase analytes for oxylipin (Table 1), endocanna-
binoid (Table 2), bile acid (Table 3), and fatty acid (Tables 4
and 5) assays. Be sure to purchase fatty acid methyl esters for
GC-MS calibration standard preparation (see Note 2).
4. 900 Pasteur pipettes and bulbs are used for non-volumetric solu-
tion and extract transfer.
5. Calibrated pipettors are used for sample transfer.

2.2 Liquid 1. LC: Use an ultra-performance liquid chromatograph capable of

Chromatography- a running pressure of >8000 psi.
Tandem Mass 2. MS/MS: An atmospheric pressure ionization triple-quadrupole
Spectrometry mass spectrometer with an electrospray probe with positive-
negative mode switching capabilities and fast scanning is pre-
ferred. Otherwise, the profiles can be developed as indepen-
dent positive and negative mode methods, without ESI
switching.
3. Inserts, caps, and vials: Waters Corp autosampler-compatible
2 mL amber vials, slit-top caps, and 150 μL glass polyspring
autosampler inserts are used for analysis.

2.3 LC-MS/MS 1. Liquid chromatography column: 150 mm  2.1 mm i.d.,

of Oxylipin/ 1.7 μm Acquity C18 BEH column, and a 0.2 μm stainless
Endocannabinoid steel frit as a guard column (Waters Corp, Inc. Milford, MA).
Profile 2. Solvent A (0.1% acetic acid): 1 L of water, 1.0 mL of glacial
acetic acid. Measure 1 L of 0.2 μm filtered LC-MS grade water
in a 1 L graduated cylinder, and add 1 mL of LC-MS grade
glacial acetic using a 1 mL pipette. Transfer to a 1 L LC solvent
bottle.
3. Solvent B (10:90 isopropanol/acetonitrile): 900 mL of acetoni-
trile (ACN), 100 mL of LC-MS grade isopropanol (IPA).
178 Theresa L. Pedersen and John W. Newman

Table 1
Oxylipin assay chemical identifiers and quality

# Abbreviation InChIKey Class Quality

1 CUDA HPTJABJPZMULFH-UHFFFAOYSA-N LC ISTD +++
2 PUHA VUPFPVOJLPTJQS-UHFFFAOYSA-N LC ISTD +++
3 6-Keto PGF1a-d4 OXY SSTD +++
4 TXB2-d4 OXY SSTD +++
5 PGF2a-d4 OXY SSTD +++
6 PGD2-d4 OXY SSTD +++
7 LTB4-d4 OXY SSTD +++
8 14,15-DiHETrE-d11 OXY SSTD +++
9 9-HODE-d4 OXY SSTD +++
10 20-HETE-d6 OXY SSTD +++
11 12-HETE-d8 OXY SSTD +++
12 5-HETE-d8 OXY SSTD +++
13 12(13)-EpOME-d4 OXY SSTD +++
14 10-Nitrooleate-d17 OXY SSTD ++
15 AA-d8 OXY SSTD +++
16 6-keto-PGF1a KFGOFTHODYBSGM-ZUNNJUQCSA-N PG +++
17 PGF3a SAKGBZWJAIABSY-SAMSIYEGSA-N PG +++
18 PGE3 CBOMORHDRONZRN-QLOYDKTKSA-N PG +++
19 TXB2 XNRNNGPBEPRNAR-JQBLCGNGSA-N TX +++
20 9,12,13-TriHOME MDIUMSLCYIJBQC-MVFSOIOZSA-N Triol +++
21 PGF2a PXGPLTODNUVGFL-UAAPODJFSA-N PG +++
22 PGE2 XEYBRNLFEZDVAW-ARSRFYASSA-N PG +++
23 PGE1 GMVPRGQOIOIIMI-DWKJAMRDSA-N PG +++
24 PGD2 BHMBVRSPMRCCGG-OUTUXVNYSA-N PG +++
25 15-keto-PGE2 YRTJDWROBKPZNV-KMXMBPPJSA-N PG +++
26 RvD1 OIWTWACQMDFHJG-NJIQAZPPSA-N Triol +++
27 Lipoxin A4 IXAQOQZEOGMIQS-SSQFXEBMSA-N Diol +++
28 LTB5 BISQPGCQOHLHQK-HDNPQISLSA-N LT +++
29 15,16-DiHODE LKLLJYJTYPVCID-OHPMOLHNSA-N Diol +++
30 8,15-DiHETE NNPWRKSGORGTIM-RCDCWWQHSA-N Diol +++
31 12,13-DiHODE RGRKFKRAFZJQMS-OOHFSOINSA-N Diol +++
(continued)
 Targeted Analysis of Lipid Mediators and Fatty Acids in Plasma 179

Table 1
(continued)

# Abbreviation InChIKey Class Quality

32 9,10-DiHODE QRHSEDZBZMZPOA-ZJSQCTGTSA-N Diol +++
33 17,18-DiHETE XYDVGNAQQFWZEF-JPURVOHMSA-N Diol +++
34 5,15-DiHETE UXGXCGPWGSUMNI-BVHTXILBSA-N Diol +++
35 6-trans-LTB4 VNYSSYRCGWBHLG-UKNWISKWSA-N Diol +++
36 14,15-DiHETE BLWCDFIELVFRJY-QXBXTPPVSA-N Diol +++
37 LTB4 KFGOFTHODYBSGM-ZUNNJUQCSA-N Diol +++
38 12,13-DiHOME CQSLTKIXAJTQGA-FLIBITNWSA-N Diol +++
39 9,10-DiHOME XEBKSQSGNGRGDW-YFHOEESVSA-N Diol +++
40 19,20-DiHDoPE FFXKPSNQCPNORO-MBYQGORISA-N Diol +++
41 14,15-DiHETrE SYAWGTIVOGUZMM-ILYOTBPNSA-N Diol +++
42 11,12-DiHETrE LRPPQRCHCPFBPE-KROJNAHFSA-N Diol +++
43 9,10-e-DiHO VACHUYIREGFMSP-SJORKVTESA-N Diol +++
44 9-HOTE YUPHIKSLGBATJK-OBKPXJAFSA-N R-OH +++
45 12(13)Ep-9-KODE RCMABBHQYMBYKV-BUHFOSPRSA-N Epox,R ¼ O ++
46 13-HOTE KLLGGGQNRTVBSU-JDTPQGGVSA-N R-OH +++
47 8,9-DiHETrE DCJBINATHQHPKO-TYAUOURKSA-N Diol +++
48 15-Deoxy PGJ2 VHRUMKCAEVRUBK-GODQJPCRSA-N PG +++
49 15-HEPE UDXLGBLAJBYLSZ-XBCQTNLFSA-N R-OH +++
50 20-HETE NNDIXBJHNLFJJP-DTLRTWKJSA-N R-OH +++
51 12-HEPE MCRJLMXYVFDXLS-QGQBRVLBSA-N R-OH +++
52 5,6-DiHETrE GFNYAPAJUNPMGH-QNEBEIHSSA-N Diol +++
53 9-HEPE OXOPDAZWPWFJEW-FPRWAWDYSA-N R-OH +++
54 13-HODE HNICUWMFWZBIFP-IRQZEAMPSA-N R-OH +++
55 5-HEPE FTAGQROYQYQRHF-FCWZHQICSA-N R-OH +++
56 9-HODE NPDSHTNEKLQQIJ-SIGMCMEVSA-N R-OH +++
57 15(16)-EpODE HKSDVVJONLXYKL-OHPMOLHNSA-N Epox +++
58 17(18)-EpETE GPQVVJQEBXAKBJ-JPURVOHMSA-N Epox +++
59 15-HETE JSFATNQSLKRBCI-VAEKSGALSA-N R-OH +++
60 13-KODE JHXAZBBVQSRKJR-BSZOFBHHSA-N R¼O ++
61 9(10)-EpODE JTEGNNHWOIJBJZ-ZJSQCTGTSA-N Epox +++
62 17-HDoHE SWTYBBUBEPPYCX-VIIQGJSXSA-N R-OH +++
(continued)
180 Theresa L. Pedersen and John W. Newman

Table 1
(continued)

# Abbreviation InChIKey Class Quality

63 12(13)-EpODE BKKGUKSHPCTUGE-OOHFSOINSA-N Epox +++
64 15-KETE YGJTUEISKATQSM-USWFWKISSA-N R¼O ++
65 14-HDoHE ZNEBXONKCYFJAF-BGKMTWLOSA-N R-OH +++
66 11-HETE GCZRCCHPLVMMJE-RSPKXIRXSA-N R-OH +++
67 14(15)-EpETE RGZIXZYRGZWDMI-QXBXTPPVSA-N Epox +++
68 9-KODE LUZSWWYKKLTDHU-ZJHFMPGASA-N R¼O ++
69 11(12)-EpETE QHOKDYBJJBDJGY-BVILWSOJSA-N Epox +++
70 12-HETE ZNHVWPKMFKADKW-FYMOKONMSA-N R-OH +++
71 8-HETE NLUNAYAEIJYXRB-VYOQERLCSA-N R-OH +++
72 9-HETE KATOYYZUTNAWSA-DLJQHUEDSA-N R-OH +++
73 19(20)-EpDoPE OSXOPUBJJDUAOJ-MBYQGORISA-N Epox +++
74 5-HETE KGIJOOYOSFUGPC-JGKLHWIESA-N R-OH +++
75 12(13)-EpOME CCPPLLJZDQAOHD-FLIBITNWSA-N Epox +++
76 14(15)-EpETrE WLMZMBKVRPUYIG-LTCHCNGXSA-N Epox +++
77 4-HDoHE IFRKCNPQVIJFAQ-JGDWKEERSA-N R-OH +++
78 16(17)-EpDoPE BCTXZWCPBLWCRV-ZYADFMMDSA-N Epox +++
79 9(10)-EpOME XEBKSQSGNGRGDW-YFHOEESVSA-N Epox +++
80 5-KETE MEASLHGILYBXFO-XTDASVJISA-N R¼O ++
81 11(12)-EpETrE DXOYQVHGIODESM-IQCOFVSKSA-N Epox +++
82 8(9)-EpETrE DBWQSCSXHFNTMO-TYAUOURKSA-N Epox +++
83 10-Nitrolinoleate LELVHAQTWXTCLY-XYWKCAQWSA-N Nitrolipid ++
84 9(10)-EpO IMYZYCNQZDBZBQ-UHFFFAOYSA-N Epox +++
85 10-Nitrooleate WRADPCFZZWXOTI-UHFFFAOYSA-N Nitrolipid ++
86 9-Nitrooleate CQOAKBVRRVHWKV-UHFFFAOYSA-M Nitrolipid ++
Quality: poor (); fair (þ); good (þþ); excellent (þþþ)

Measure 900 mL of 0.2 μm filtered LC-MS grade ACN in a 1 L

graduated cylinder, and bring to a 1 L final volume with
100 mL of IPA. Transfer to a 1 L LC solvent bottle with a
Teflon-lined cap.

2.4 LC-MS/MS of Bile 1. Liquid chromatography column: 100 mm  2.1 mm i.d.,

Acid Profile 1.7 μm Acquity C18 BEH column, and a 0.2 μm stainless
steel frit as a guard column (Waters Corp, Inc., Milford, MA).
 Targeted Analysis of Lipid Mediators and Fatty Acids in Plasma 181

Table 2
Endocannabinoid target identifiers and assay quality

# Abbreviation InChIKey Class Quality

1 CUDA HPTJABJPZMULFH-UHFFFAOYSA-N LC ISTD +++
2 PUHA VUPFPVOJLPTJQS-UHFFFAOYSA-N LC ISTD +++
3 NA-Gly-d8 Endo SSTD +++
4 2-AG-d5 Endo SSTD ++
5 PGF2a EA-d4 Endo SSTD +++
6 P-EA-d4 Endo SSTD +++
7 A-EA-d8 Endo SSTD +++
8 PGF2a-EA XCVCLIRZZCGEMU-FPLRWIMGSA-N Acyl-amides +++
9 PGE2-EA GKKWUSPPIQURFM-IGDGGSTLSA-N Acyl-amides +++
10 PGD2-EA KEYDJKSQFDUAGF-YIRKRNQHSA-N Acyl-amides +++
11 15-HETE-EA XZQKRCUYLKDPEK-BPVVGZHASA-N Acyl-amides +++
12 11(12)-EET TYRRSRADDAROSO-KROJNAHFSA-N Acyl-amides +++
13 aL-EA HBJXRRXWHSHZPU-PDBXOOCHSA-N Acyl-amides +++
14 DH-EA CXWASNUDKUTFPQ-KUBAVDMBSA-N Acyl-amides +++
15 A-EA LGEQQWMQCRIYKG-DOFZRALJSA-N Acyl-amides +++
16 L-EA KQXDGUVSAAQARU-HZJYTTRNSA-N Acyl-amides +++
17 NA-Gly YLEARPUNMCCKMP-DOFZRALJSA-N Acyl-amides +++
18 DGL-EA ULQWKETUACYZLI-QNEBEIHSSA-N Acyl-amides +++
19 P-EA HXYVTAGFYLMHSO-UHFFFAOYSA-N Acyl-amides +++
20 NO-Gly HPFXACZRFJDURI-KTKRTIGZSA-N Acyl-amides +++
21 O-EA BOWVQLFMWHZBEF-KTKRTIGZSA-N Acyl-amides +++
22 D-EA FMVHVRYFQIXOAF-DOFZRALJSA-N Acyl-amides +++
23 S-EA OTGQIQQTPXJQRG-UHFFFAOYSA-N Acyl-amides +++
24 PGF2a-1G NWKPOVHSHWJQNI-OMVDPNNKSA-N Glyceryl ester ++
25 PGE2-1G RJXVYMMSQBYEHN-SDTVLRMPSA-N Glyceryl ester ++
26 2-AG RCRCTBLIHCHWDZ-DOFZRALJSA-N Glyceryl ester ++
27 2-LG IEPGNWMPIFDNSD-HZJYTTRNSA-N Glyceryl ester ++
28 1-AG DCPCOKIYJYGMDN-HUDVFFLJSA-N Glyceryl ester ++
29 1-LG WECGLUPZRHILCT-GSNKCQISSA-N Glyceryl ester ++
31 2-OG UPWGQKDVAURUGE-KTKRTIGZSA-N Glyceryl ester ++
32 1-OG RZRNAYUHWVFMIP-QJRAZLAKSA-N Glyceryl ester ++
Quality: poor (); fair (þ); good (þþ); excellent (þþþ)
182 Theresa L. Pedersen and John W. Newman

Table 3
Bile acid target identifiers and assay quality

# Abbreviation InChIKey Class Quality

1 CUDA HPTJABJPZMULFH-UHFFFAOYSA-N LC ISTD +++
2 PUHA VUPFPVOJLPTJQS-UHFFFAOYSA-N LC ISTD +++
3 CA-d6 BA SSTD +++
4 CDCA-d4 BA SSTD +++
5 DCA-d4 BA SSTD +++
6 GCA-d4 BA SSTD +++
7 GCDCA-d4 BA SSTD +++
8 LCA-d4 BA SSTD +++
9 TCDCA-d4 BA SSTD +++
10 TDHCA UBDJSBRKNHQFPD-PYGYYAGESA-N 2
BA Conj +++
11 T-ω-MCA m-BA Conj +++
12 T-α-MCA XSOLDPYUICCHJX-QQXJNSDFSA-N m-BA Conj +++
13 T-β-MCA XSOLDPYUICCHJX-UZUDEGBHSA-N m-BA Conj +++
14 TUDCA HMXPOCDLAFAFNT-BHYUGXBJSA-N 2
BA Conj +++
15 TCA WBWWGRHZICKQGZ-HZAMXZRMSA-N 1
BA Conj +++
16 ω-MCA DKPMWHFRUGMUKF-NTPBNISXSA-N m-BA +++


17 GUDCA GHCZAUBVMUEKKP-XROMFQGDSA-N 2 BA Conj +++
18 α-MCA DKPMWHFRUGMUKF-JDDNAIEOSA-N m-BA +++


19 GHDCA SPOIYSFQOFYOFZ-BRDORRHWSA-N 2 BA Conj +++


20 GCA RFDAIACWWDREDC-FRVQLJSFSA-N 1 BA Conj +++
21 β-MCA DKPMWHFRUGMUKF-CRKPLTDNSA-N m-BA +++
22 TCDCA BHTRKEVKTKCXOH-BJLOMENOSA-N 1
BA Conj +++


23 TDCA AWDRATDZQPNJFN-VAYUFCLWSA-N 2 BA Conj +++


24 CA BHQCQFFYRZLCQQ-OELDTZBJSA-N 1 BA +++
25 UDCA RUDATBOHQWOJDD-UZVSRGJWSA-N 2
BA +++
26 GCDCA GHCZAUBVMUEKKP-GYPHWSFCSA-N 1
BA Conj +++


27 GDCA WVULKSPCQVQLCU-BUXLTGKBSA-N 2 BA Conj +++


28 TLCA QBYUNVOYXHFVKC-GBURMNQMSA-N 2 BA Conj +++
29 CDCA RUDATBOHQWOJDD-BSWAIDMHSA-N 1
BA +++
30 DCA KXGVEGMKQFWNSR-LLQZFEROSA-N 2
BA +++


31 GLCA XBSQTYHEGZTYJE-OETIFKLTSA-N 2 BA Conj +++


32 LCA SMEROWZSTRWXGI-HVATVPOCSA-N 2 BA ++
Quality: poor (); fair (þ); good (þþ); excellent (þþþ)
 Targeted Analysis of Lipid Mediators and Fatty Acids in Plasma 183

Table 4
Fatty acid target identifiers and assay quality

# Abbreviation InChIKey (fatty acids) Class Quality

1 C23:0-me VORKGRIRMPBCCZ-UHFFFAOYSA-N GC ISTD +++
2 C6:0-me FUZZWVXGSFPDMH-UHFFFAOYSA-N SFA 
3 C8:0-me WWZKQHOCKIZLMA-UHFFFAOYSA-N SFA 
4 C10:0-me GHVNFZFCNZKVNT-UHFFFAOYSA-N SFA +
5 C12:0-me POULHZVOKOAJMA-UHFFFAOYSA-N SFA ++
6 C14:0-me TUNFSRHWOTWDNC-UHFFFAOYSA-N SFA +++
7 C14:1n5-me YWWVWXASSLXJHU-WAYWQWQTSA-N MUFA +
8 C15:0-me WQEPLUUGTLDZJY-UHFFFAOYSA-N SFA +++
9 C15:1n5-me ISTD +++
10 C16:0-d31-me MQOCIYICOGDBSG-HXKBIXQXSA-M SSTD +++
11 C16:0-me IPCSVZSSVZVIGE-UHFFFAOYSA-N SFA +++
12 C16:1n7t-me SECPZKHBENQXJG-BQYQJAHWSA-N Trans-fat +++
13 C16:1n7-me SECPZKHBENQXJG-FPLPWBNLSA-N MUFA +++
14 C17:0-me KEMQGTRYUADPNZ-UHFFFAOYSA-N SFA +++
15 C17:1n7-me MUFA +++
16 C18:0-d35-me QIQXTHQIDYTFRH-KNAXIHRDSA-N SSTD +++
17 C18:0-me QIQXTHQIDYTFRH-UHFFFAOYSA-N SFA +++
18 C18:1n9-me ZQPPMHVWECSIRJ-KTKRTIGZSA-N MUFA +++
19 C18:1n7-me UWHZIFQPPBDJPM-FPLPWBNLSA-N MUFA +++
20 C18:2n6-me OYHQOLUKZRVURQ-HZJYTTRNSA-N PUFA-n6 +++
21 C18:3n6-me VZCCETWTMQHEPK-QNEBEIHSSA-N PUFA-n6 +++
22 C18:3n3-me DTOSIQBPPRVQHS-PDBXOOCHSA-N PUFA-n3 +++
23 C18:4n3-me JIWBIWFOSCKQMA-LTKCOYKYSA-N PUFA-n3 +++
24 C19:0-me ISYWECDDZWTKFF-UHFFFAOYSA-N SFA +++
25 C19:1n9-me BBOWBNGUEWHNQZ-KTKRTIGZSA-N MUFA +++
26 9c,11 t–CLA-me JBYXPOFIGCOSSB-GOJKSUSPSA-N Trans-fat +++
27 10 t,12c–CLA-me GKJZMAHZJGSBKD-BLHCBFLLSA-N Trans-fat +++
28 C20:0-me VKOBVWXKNCXXDE-UHFFFAOYSA-N SFA +++
29 C20:1n9-me BITHHVVYSMSWAG-KTKRTIGZSA-N MUFA +++
30 C20:2n6-me XSXIVVZCUAHUJO-HZJYTTRNSA-N PUFA-n6 +++
31 C20:3n6-me HOBAELRKJCKHQD-QNEBEIHSSA-N PUFA-n6 +++
(continued)
184 Theresa L. Pedersen and John W. Newman

Table 4
(continued)

# Abbreviation InChIKey (fatty acids) Class Quality

32 C20:4n6-me YZXBAPSDXZZRGB-DOFZRALJSA-N PUFA-n6 +++
33 C20:3n3-me AHANXAKGNAKFSK-PDBXOOCHSA-N PUFA-n3 +++
34 C21:0-me CKDDRHZIAZRDBW-UHFFFAOYSA-N SFA +++
35 C20:4n3-me YNVYKJQCWARJFA-ZKWNWVNESA-N PUFA-n3 +++
36 C20:5n3-me JAZBEHYOTPTENJ-JLNKQSITSA-N PUFA-n3 +++
37 C22:0-me UKMSUNONTOPOIO-UHFFFAOYSA-N SFA +++
38 C22:1n9-me DPUOLQHDNGRHBS-KTKRTIGZSA-M SSTD +++
39 C22:2n6-me HVGRZDASOHMCSK-HZJYTTRNSA-N PUFA-n6 +++
40 C22:4n6-me TWSWSIQAPQLDBP-DOFZRALJSA-N PUFA-n6 +++
41 C22:5n6-me AVKOENOBFIYBSA-WMPRHZDHSA-N PUFA-n6 +++
42 C22:3n3-me WBBQTNCISCKUMU-PDBXOOCHSA-N SSTD +++
43 C22:5n3-me YUFFSWGQGVEMMI-RCHUDCCISA-N PUFA-n3 +++
44 C22:6n3-me MBMBGCFOFBJSGT-KUBAVDMBSA-N PUFA-n3 +++
45 C24:0-me QZZGJDVWLFXDLK-UHFFFAOYSA-N SFA +++
46 C24:1n9-me GWHCXVQVJPWHRF-KTKRTIGZSA-M MUFA +++
Quality: poor (); fair (þ); good (þþ); excellent (þþþ)

Table 5
Internal standards, lipid class surrogates, and FAME derivatization controls

# Abbreviation InChIKey Class Use

1 TAG-tri(16:0)-d93 PVNIQBQSYATKKL-JGBASCRPSA-N TG TG SSTD
2 PC-di(18:0)-d70) NRJAVPSFFCBXDT-MNVUKWGGSA-N PL PL SSTD
3 CE-C22:1n9 SQHUGNAFKZZXOT-QXAJUEOOSA-N CE CE SSTD
4 C22:3n3 WBBQTNCISCKUMU-PDBXOOCHSA-N NEFA FAME SSTD
5 C15:1n5 NEFA FAME ISTD

2. Solvent A (0.1% formic acid): 1 L of water, 1.0 mL of formic

acid. Measure 1 L of water in a 1 L graduated cylinder, and add
1 mL of LC-MS grade formic acid using a 1 mL pipette.
Transfer to a 1 L LC solvent bottle.
3. Solvent B (0.1% formic acid in ACN): 1 L of ACN, 1 mL formic
acid. Measure 1 L of ACN in a 1 L graduated cylinder, and add
1 mL of LC-MS grade formic acid using a 1 mL pipette.
Transfer to a 1 L LC solvent bottle with a Teflon-lined cap.
 Targeted Analysis of Lipid Mediators and Fatty Acids in Plasma 185

2.5 Preparing Prepare the individual analyte optimization solutions and the “all
LC-MS/MS Analyte analyte” optimization solution described below at the same time
Stocks while the stock solutions are being handled. Clean one 2 mL amber
and Optimization vial for each analyte optimization solution and one vial for the all
Solutions analyte optimization solution being prepared for each assay.
1. Vials: Per assay, ~200 2 mL and ~50 5 mL amber borosilicate
vials with solid Teflon-lined caps for solution storage.
2. LC-MS analyte stock solutions: 0.10–10.0 mg/mL of analyte in
MeOH. One mg/mL is ideal for the volumetric transfers used
to make the LC-MS/MS calibration solutions. Stock solutions
of each analyte are prepared or purchased in a concentration
range from 0.10 to 10.0 mg/mL. For neat compounds, weigh
materials, record weight, transfer to an appropriate volumetric
flask, and solubilize with nitrogen or argon purged MeOH. If
necessary, dilute concentrated solutions by transferring the
appropriate volume with an analytical syringe to an appropriate
volumetric flask and bring to volume in purged MeOH. Vortex
each new solution 2 s to mix and transfer to a glass vial with a
Pasteur pipette. Top the analyte stock solutions with nitrogen
gas, cap, seal, mark the meniscus, and label vial. Store analyte
stocks at 80
C.
3. LC-MS analyte optimization solutions: 0.10–10.0 μg/mL of
analyte in MeOH. Using one vial per analyte, transfer 1.0 μL
of each analyte stock solution into a 2 mL amber glass vial
containing 1 mL purged 1:1 MeOH/ACN (v/v). Label the
vial with the analyte name, seal, and store at 20
C.
4. LC-MS all analyte optimization solution: 0.10–10.0 μg/mL of
all analytes in MeOH. Transfer 1 μL of each analyte stock
solution into a 2 mL amber vial containing 1 mL purged 1:1
MeOH/ACN (v/v). Label as “All Analyte Opt” and store at
20
C.

2.6 Calibration Generating a spreadsheet containing target weights with all volu-
Standard Preparation metric deliveries (and volumetric constraints) provides an excellent
Worksheet format for both projecting approximate target concentrations,
recording actual weights and volumes, and developing the volu-
metrics for making a complex suite of analytes (see Note 3). Sup-
plementary File 1 provides an organized approach for making the
complex mixtures needed for this protocol. Simple mathematical
functions are embedded in the sheet, where the adjustment of
delivery and final volumes will change the solution and standard
concentrations, en masse. This worksheet can be expanded with the
positive mode endocannabinoids (Table 6) and/or used to build
other multi-residue profiles, such as the bile acids (Table 7) and
fatty acids (Table 8). If your mass spectrometer software quantifies
internal standards on a curve, you will need to adjust your internal
186 Theresa L. Pedersen and John W. Newman

Table 6
Representative oxy/endo fusion calibrations for the 6500 QTRAP (nM)

# Mixture Compound Mix target (μM) Range (nM)

1 IS PHAU 50 100
2 IS CUDA 50 100
3 Oxy SSTD 14,15-DiHETrE-d11 13 25
4 Oxy SSTD 6-keto PGF1a-d4 12 25
5 Oxy SSTD TXB2-d4 12 25
6 Oxy SSTD PGD2-d4 13 25
7 Oxy SSTD LTB4-d4 13 25
8 Oxy SSTD 12-HETE-d8 12 25
9 Oxy SSTD 9-HODE-d4 12 25
10 Oxy SSTD 5-HETE-d8 13 25
11 Oxy SSTD 20-HETE-d6 12 25
12 Oxy SSTD AA-d8 13 25
13 Oxy SSTD 10-Nitrooleate-d17 13 25
14 Oxy SSTD PGF2a-d4 13 25
15 Oxy SSTD 12(13)-EpOME-d4 12 25
16 Endo SSTD NA-Gly-d8 6.2 22
17 Endo SSTD 2-AG-d5 6.3 22
18 Endo SSTD PGF2a EA-d4 6.2 22
19 Endo SSTD P-EA-d4 6.3 22
20 Endo SSTD A-EA-d8 6.2 22
21 Oxy mix 1 9,12,13-TriHOME 15 [0.0116–1210]
22 Oxy mix 1 12,13-DiHOME 16 [0.0122–1270]
23 Oxy mix 1 9,10-DiHOME 16 [0.0122–1270]
24 Oxy mix 1 EKODE 16 [0.0123–1280]
25 Oxy mix 1 13-HODE 16 [0.0124–1300]
26 Oxy mix 1 9-HODE 16 [0.0124–1300]
27 Oxy mix 1 13 KODE 17 [0.013–1360]
28 Oxy mix 1 9-KODE 17 [0.013–1360]
29 Oxy mix 1 12(13)-EpOME 17 [0.013–1350]
30 Oxy mix 1 9(10)-EpOME 17 [0.013–1350]
31 Oxy mix 1 9,10-DiHODE 16 [0.0123–1280]
(continued)
 Targeted Analysis of Lipid Mediators and Fatty Acids in Plasma 187

Table 6
(continued)

# Mixture Compound Mix target (μM) Range (nM)

32 Oxy mix 1 12,13-DiHODE 15 [0.0118–1230]
33 Oxy mix 1 15,16-DiHODE 16 [0.0123–1280]
34 Oxy mix 1 13-HOTE 16 [0.0125–1300]
35 Oxy mix 1 9-HOTE 16 [0.0125–1300]
36 Oxy mix 1 9(10)-EpODE 16 [0.0123–1280]
37 Oxy mix 1 12(13)-EpODE 16 [0.0125–1300]
38 Oxy mix 1 15(16)-EpODE 16 [0.0123–1280]
39 Oxy mix 1 6-keto-PGF1a 15 [0.0116–1210]
40 Oxy mix 1 TXB2 15 [0.0116–1210]
41 Oxy mix 1 PGF2a 16 [0.0121–1260]
42 Oxy mix 1 PGE1 16 [0.0121–1260]
43 Oxy mix 1 PGE2 16 [0.0122–1270]
44 Oxy mix 1 PGE3 16 [0.0123–1280]
45 Oxy mix 1 PGD2 16 [0.0122–1270]
46 Oxy mix 1 PGJ2 15 [0.0115–1200]
47 Oxy mix 1 15-deoxy PGJ2 16 [0.0121–1260]
48 Oxy mix 1 LTB5 6.0 [0.00482–502]
49 Oxy mix 1 6-trans-LTB4 15 [0.0114–1190]
50 Oxy mix 1 LTB4 15 [0.0114–1190]
51 Oxy mix 1 Lipoxin a4 14 [0.0109–1130]
52 Oxy mix 1 LTE4 11 [0.00874–910]
53 Oxy mix 2 14,15-DiHETrE 15 [0.0113–1180]
54 Oxy mix 2 11,12-DiHETrE 15 [0.0113–1180]
55 Oxy mix 2 8,9-DiHETrE 15 [0.0113–1180]
56 Oxy mix 2 5,6-DiHETrE 15 [0.0113–1180]
57 Oxy mix 2 20-HETE 16 [0.012–1250]
58 Oxy mix 2 15-HETE 16 [0.012–1250]
59 Oxy mix 2 12-HETE 16 [0.012–1250]
60 Oxy mix 2 11-HETE 16 [0.012–1250]
61 Oxy mix 2 9-HETE 16 [0.0121–1260]
62 Oxy mix 2 8-HETE 16 [0.012–1250]
(continued)
188 Theresa L. Pedersen and John W. Newman

Table 6
(continued)

# Mixture Compound Mix target (μM) Range (nM)

63 Oxy mix 2 5-HETE 16 [0.012–1250]
64 Oxy mix 2 15-KETE 16 [0.0121–1260]
65 Oxy mix 2 5-KETE 16 [0.0121–1260]
66 Oxy mix 2 14(15)-EpETrE 16 [0.012–1250]
67 Oxy mix 2 11(12)-EpETrE 16 [0.012–1250]
68 Oxy mix 2 8(9)-EpETrE 16 [0.012–1250]
69 Oxy mix 2 5(6)-EpETrE 16 [0.012–1250]
70 Oxy mix 3 14,15-DiHETE 15 [0.0114–1190]
71 Oxy mix 3 17,18-DiHETE 15 [0.0114–1190]
72 Oxy mix 3 5,15-DiHETE 15 [0.0114–1190]
69 Oxy mix 3 8,15-DiHETE 15 [0.0114–1190]
73 Oxy mix 3 15(S)-HEPE 16 [0.0121–1260]
74 Oxy mix 3 12(S)-HEPE 16 [0.0121–1260]
75 Oxy mix 3 5(S)-HEPE 16 [0.0121–1260]
76 Oxy mix 3 17(R)-HDoHE 15 [0.0111–1160]
77 Oxy mix 3 14(15)-EpETE 16 [0.0121–1260]
78 Oxy mix 3 17(18)-EpETE 16 [0.0121–1260]
79 Oxy mix 3 16(17)-EpDoPE 15 [0.0111–1160]
80 Oxy mix 3 19(20)-EpDoPE 15 [0.0111–1160]
81 Oxy mix 3 19,20-DiHDoPE 14 [0.0106–1100]
82 Oxy mix 3 Resolvin D1 13 [0.0102–1060]
83 Oxy mix 3 11(12)-EpETE 16 [0.0121–1260]
84 Oxy mix 3 9-HEPE 16 [0.0121–1260]
85 Oxy mix 3 4-HDoHE 15 [0.0111–1160]
86 Oxy mix 3 14-HDoHE 15 [0.0111–1160]
87 Oxy mix 3 PGF3a 14 [0.0109–1130]
88 Oxy mix 3 10-Nitrolinoleate 15 [0.0118–1230]
89 Oxy mix 3 10-Nitrooleate 15 [0.0117–1220]
90 Oxy mix 3 9-Nitrooleate 15 [0.0117–1220]
91 Oxy mix 3 9,10-EpO 13 [0.0103–1070]
92 Oxy mix 3 9,10-e-DiHO 15 [0.0116–1200]
(continued)
 Targeted Analysis of Lipid Mediators and Fatty Acids in Plasma 189

Table 6
(continued)

# Mixture Compound Mix target (μM) Range (nM)

98 Oxy mix 3 9,10-t-DiHHex 17 [0.0128–1330]
99 Oxy mix 3 15-keto PGE2 16 [0.0123–1280]
100 Endo mix P-EA 13 [0.0128–1340]
101 Endo mix S-EA 12 [0.0117–1220]
102 Endo mix O-EA 15 [0.0147–1540]
103 Endo mix L-EA 15 [0.0148–1550]
104 Endo mix aL-EA 16 [0.0149–1560]
105 Endo mix DGL- EA 14 [0.0137–1430]
106 Endo mix A-EA 14 [0.0138–1440]
107 Endo mix D-EA 13 [0.0128–1330]
108 Endo mix DH-EA 13 [0.0129–1350]
109 Endo mix NA-Gly 14 [0.0133–1380]
110 Endo mix NO-Gly 12 [0.0113–1180]
111 Endo mix 1-LG 23 [0.0217–2260]
112 Endo mix 2-LG 42 [0.0406–4230]
113 Endo mix 1-AG 21 [0.0203–2110]
114 Endo mix 2-AG 21 [0.0203–2110]
115 Endo mix 1-OG 42 [0.0404–4210]
113 Endo mix 2-OG 42 [0.0404–4210]
114 Endo mix 15-HETE-EA 3.0 [0.00264–275]
115 Endo mix PGE2-EA 10 [0.00971–1010]
116 Endo mix PGD2-EA 13 [0.0121–1260]
117 Endo mix PGF2a-EA 13 [0.0121–1260]
118 Endo mix PGE2 1G 12 [0.0113–1170]
119 Endo mix PGF2a 1G 12 [0.0112–1170]
120 Endo mix 11(12)-EpETrE-EA 3.0 [0.00264–275]

standard volumetric deliveries accordingly to create appropriate

concentrations for your standards. The worksheet will guide the
following steps: (1) record analyte stock information, (2) establish
volumetric transfers, (3) prepare internal standard (ISTD) and
surrogate (SSTD) reconstitution solution used to make calibration
standards, (4) prepare calibration solutions, (5) prepare sample
ISTD reconstitution solution, and (6) prepare the surrogate spike
solution.
190 Theresa L. Pedersen and John W. Newman

Table 7
Representative bile acid calibration ranges (nM) for the 4000 QTRAP

# Mixture Compound Mix target (μM) Range (nM)

1 ISTD PUHA 10 100
2 ISTD CUDA 10 100
3 BA-SSTD CA-d6 10 100
4 BA-SSTD CDCA-d4 10 100
5 BA-SSTD DCA-d4 10 100
6 BA-SSTD GCA-d4 10 100
7 BA-SSTD GCDCA-d4 10 100
8 BA-SSTD LCA-d4 10 100
9 BA-SSTD TCDCA-d4 10 100
10 BA mix 1 CA 10 [0.245–5510]
11 BA mix 1 CDCA 10 [0.201–4520]
12 BA mix 1 DCA 10 [0.183–4120]
13 BA mix 1 GCA 10 [0.214–4810]
14 BA mix 1 GCDCA 10 [0.212–4770]
15 BA mix 1 GUDCA 10 [0.149–3360]
16 BA mix 1 TCA 10 [0.209–4710]
17 BA mix 2 GDCA 10 [0.176–1320]
18 BA mix 2 GHDCA 10 [0.159–1190]
19 BA mix 2 GLCA 10 [0.188–1410]
20 BA mix 2 HDCA 10 [0.168–1260]
21 BA mix 2 LCA 10 [0.18–1350]
22 BA mix 2 TCDCA 10 [0.194–1460]
23 BA mix 2 TDCA 10 [0.176–1320]
24 BA mix 2 TDHCA 10 [0.188–1410]
25 BA mix 2 TLCA 10 [0.165–1240]
26 BA mix 2 TUDCA 10 [0.229–1720]
27 BA mix 2 UDCA 10 [0.298–2240]
28 BA mix 3 α-MCA 10 [0.245–1840]
29 BA mix 3 β-MCA 10 [0.245–1840]
30 BA mix 3 ω-MCA 10 [0.196–1470]
31 BA mix 3 T-α-MCA 10 [0.186–1390]
32 BA mix 3 T-β-MCA 10 [0.186–1390]
33 BA mix 3 T-ω-MCA 10 [0.186–1390]
 Targeted Analysis of Lipid Mediators and Fatty Acids in Plasma 191

2.7 Gas The gas chromatography mass spectrometry hardware specification

Chromatography Mass described below is for the analysis of fatty acid methyl esters
Spectrometry (FAMEs) on an Agilent 6890 with the second electronic-controlled
inlet configured as a backflush unit to facilitate the removal of high-
boiling components (e.g., cholesterol) and shorten run times from
46 to 25 min. A modern GC with factory-implemented backflush
capabilities can replace this configuration.
1. GC-MS: Agilent 6890 Gas Chromatograph with dual split/
splitless injection ports under electronic pressure control, cou-
pled to an Agilent 5975N Mass Selective Detector with simul-
taneous selected ion monitoring/full-scan spectral acquisition
capabilities; Agilent 7683B Series Injector Autosampler (Agi-
lent Technologies, Santa Clara, CA, USA).
2. Gas chromatography column: 30 m  0.25 mm, 0.25 μm
DB-225 ms (Agilent Technologies, Santa Clara, CA, USA).
3. Pre-column, backflush column, and post-column transfer line:
5 m  0.25 mm deactivated fused silica (Agilent Technologies,
Santa Clara, CA, USA).
4. Column connections: Columns are connected using straight
(pre-column) or T- (post-column and backflush) unions.
5. Inserts, caps, and vials: Agilent autosampler-compatible 2 mL
vials, caps, and 150 μL glass polyspring autosampler inserts
are used.
6. Carrier gas: High purity 99.9990% helium is used as the chro-
matography carrier gas.
7. Syringe rinse solvents: Optima grade n-hexane and isooctane (see
Note 4).

2.8 Gas 1. FAME stocks and mixes: Using a standard preparation work-
Chromatography Stock sheet as described in Subheading 2.7, prepare the fatty acid
Mixtures methyl ester (FAME) mixes defined in Table 8. Weigh each of
the materials into the appropriate volumetric flask and dissolve
in Fisher Optima grade or equivalent n-hexane to achieve the
final solutions (see Note 5).
2. FAME ISTD stock: Target concentration ~1.5 mM C23:0
methyl ester in hexane.
3. FAME ISTD reconstitution solution: Target concentration
4 μM C23:0 methyl ester in hexane.
4. 100 μM FAME SSTD spike solution: 100 μM C22:3n3, 100 μM
CE-22:1n9, 100 μM PC-18:0-d70, TAG-16:0-d93, in 1:1
MeOH/toluene. Using the standard preparation worksheet
described in Subheading 2.7, prepare a 100 μM solution in
1:1 MeOH/toluene (v/v) of deuterated tri-palmitoyl-
192 Theresa L. Pedersen and John W. Newman

Table 8
Representative FAME calibration concentrations (μM)

# Mixture Compound Mix target (μM) Range (μM)

1 FAME ISTD stock C23:0-me 1400 4
2 FAME mix 1 C6:0-me 200 [0.0154–51.2]
3 FAME mix 1 C8:0-me 300 [0.0253–84.3]
4 FAME mix 1 C10:0-me 400 [0.0429–143]
5 FAME mix 1 C12:0-me 400 [0.0373–124]
6 FAME mix 1 C14:0-me 500 [0.0495–165]
7 FAME mix 1 C14:1n5-me 80 [0.00832–27.7]
8 FAME mix 1 C15:0-me 80 [0.00781–26]
9 FAME mix 1 C16:0-me 600 [0.0592–197]
10 FAME mix 1 C16:1n7-me 70 [0.00745–24.8]
11 FAME mix 1 C17:0-me 70 [0.00703–23.4]
12 FAME mix 1 C17:1n7-me 70 [0.00708–23.6]
13 FAME mix 1 C18:0-me 400 [0.0402–134]
14 FAME mix 1 C18:1n9-me 300 [0.027–89.9]
15 FAME mix 1 C18:2n6-me 200 [0.0204–67.9]
16 FAME mix 1 C18:3n3-me 100 [0.0109–36.2]
17 FAME mix 1 C19:0-me 60 [0.0064–21.3]
18 FAME mix 1 C20:0-me 60 [0.00613–20.4]
19 FAME mix 1 C20:1n9-me 100 [0.0123–41.1]
20 FAME mix 1 C22:0-me 60 [0.00564–18.8]
21 FAME mix 1 C24:0-me 50 [0.00523–17.4]
22 FAME mix 2 C18:4n3-me 10 [0.00124–4.12]
23 FAME mix 2 C20:3n6-me 40 [0.00359–12]
24 FAME mix 2 C20:4n6-me 300 [0.0344–115]
25 FAME mix 2 C20:4n3-me 30 [0.00299–9.95]
26 FAME mix 2 C22:2n6-me 70 [0.00698–23.3]
27 FAME mix 2 C22:5n6-me 80 [0.0078–26]
28 FAME mix 3 C16:1n7t-me 80 [0.00837–27.9]
29 FAME mix 3 C19:1n9-me 90 [0.00941–31.4]
30 FAME mix 3 9c,11 t–CLA-me 70 [0.00735–24.5]
31 FAME mix 3 C20:2n6-me 70 [0.00703–23.4]
(continued)
 Targeted Analysis of Lipid Mediators and Fatty Acids in Plasma 193

Table 8
(continued)

# Mixture Compound Mix target (μM) Range (μM)

32 FAME mix 3 C21:0-me 30 [0.00309–10.3]
33 FAME mix 3 C20:3n3-me 60 [0.0055–18.3]
34 FAME mix 3 C20:5n3-me 50 [0.0051–17]
35 FAME mix 3 C22:4n6-me 80 [0.0081–27]
36 FAME mix 3 C22:6n3-me 200 [0.0188–62.7]
37 FAME mix 4 C18:1n7-me 30 [0.00295–9.85]
38 FAME mix 4 C18:3n6-me 40 [0.00403–13.4]
39 FAME mix 4 10 t,12c–CLA-me 30 [0.00336–11.2]
40 FAME mix 4 C22:5n3-me 70 [0.00681–22.7]
41 FAME mix 4 C24:1n9-me 80 [0.00836–27.9]
42 FAME mix 5 C15:1n5-me 70 [0.00659–22]
43 FAME mix 5 C16:0-d31-me 100 [0.0108–36]
44 FAME mix 5 C18:0-d35-me 100 [0.00227–7.56]
45 FAME mix 5 C22:1n9-me 80 [0.00753–25.1]
46 FAME mix 5 C22:3n3-me 40 [0.00397–13.2]

glycerides (TAG-tri(16:0)-d93), deuterated di-stearoyl-phos-

phatidylcholine (PC-di(18:0)-d70), the rare cholesteryl ester
(CE-22:1n9), and the rare NEFA (C22:3n3) (see Note 6).

2.9 Sample 1. Tubes (or plates): 2.0 mL Eppendorf Safe-Lock micro-centri-

Preparation Materials fuge tube, natural, round bottom.
2. Plates: Thermo Fisher Scientific Nunc™ 96-well, 500 μL
U-bottom plate, natural, non-treated, mfr. No. 267245.
3. Non-slit 96-well cap mats: Thermo Fisher Scientific Nunc™
96-well non-slit cap mats, natural, mfr. No. 276002.
4. Spin filters: Ultrafree-MC VV Centrifugal Filters Durapore
(R) PVDF 0.1 μm EMD Millipore Corporation #
UFC30VV00.
5. Extraction antioxidant solution: 0.2 mg/mL butylated hydro-
xytoluene, 0.2 mg/mL EDTA in MeOH. Dissolve 10 mg
butylated hydroxytoluene (BHT) in 10 mL of Optima grade
MeOH. Dissolve 10 mg ethylenediaminetetraacetic acid
(EDTA) in 10 mL of LC-MS grade water. Mix solutions 1:1
to prepare a 0.2 mg/mL BHT/EDTA solution. Sub-aliquot
1 mL proportions into clean vials, seal, and store at 20
C.
194 Theresa L. Pedersen and John W. Newman

6. Centrifugal evaporator: Genevac EZ-2 (SP Scientific, Warmin-

ster, PA, USA).
7. Centrifuge: A refrigerated centrifuge capable of holding 2 mL
Eppendorf tubes and maintaining 6
C (e.g., Eppendorf
5430 R, model # 5428).
8. Plasma reference material: Prepare or purchase a large homo-
geneous volume of plasma, and sub-aliquot into 110 μL
volumes in polypropylene Eppendorf tubes and store at
80
C (see Note 7).
9. NEFA extraction solvents: LC-MS grade water, Optima grade
cyclohexane, and 1 M ultrahigh purity ammonium acetate.

2.10 FAME Free fatty acids are derivatized to produce methyl ester for GC-MS
Derivatization analysis. To control for derivatization efficiency, a fatty acid deriva-
Reagents tization surrogate control is introduced after sample extraction and
prior to derivatization.
1. Methylation regent: (Trimethylsilyl)diazomethane solution
2.0 M in hexane.
2. Derivatization SSTD: 62.5 μM pentadecenoic acid (15:1n5) in
MeOH. In a 10 mL volumetric flask, dissolve 15 mg (15:1n5
Nu-Chek Prep #U-38-A) in MeOH to yield a 6.25 mM solu-
tion. In a second 10 mL volumetric flask, dilute 100 μL of the
6.25 mM solution with 10 mL MeOH to the
derivatization SSTD.

3 Methods

The Methods section contains instruction for building the

LC-MS/MS and GC-MS instrument acquisition methods for anal-
ysis of plasma sample extracts, developing the multi-residue calibra-
tion standards, the plasma sample preparation protocol, and a
review of post-acquisition QC.

3.1 Building 1. Establishing global MS/MS parameters: The analyte optimiza-

the Multi-residue tion solutions (see Subheading 2.6) are required to build the
LC-MS/MS Instrument LC-MS/MS acquisition method(s). To optimize the ESI
Acquisition Method oxylipin MRM global ion source parameters and collision gas
Using the Gradient pressures (Table 9), infuse the 14,15-DiHETrE-d11 optimiza-
Optimization Approach tion solution while collecting data for its defined precursor to
product ion (i.e., Q1 > Q3) mass transition listed in Table 10
(see Note 8). Transfer the optimized global parameters into the
table of global parameters (Table 9). Establish the þESI endo-
cannabinoid MRM experiment by repeating the global optimi-
zation steps by infusing the A-EA-d8 optimization solution,
while collecting data for its defined Q1 > Q3 mass transition
listed in Table 11. The þESI and ESI source and collision gas
 Targeted Analysis of Lipid Mediators and Fatty Acids in Plasma 195

Table 9
Sciex QTRAP global MS parameters for oxylipin, endocannabinoid, and bile acid assays

Assay Mode IS (V) TEM (
C) CUR (L/min) GS1 (L/min) GS2 (L/min) CAD (pressure)
Oxylipin ESI 4500 525 35 60 50 Med
Endo þESI 5500 525 35 60 50 Med
Bile acid ESI 4500 600 35 60 50 Med
CUR curtain gas flow, CAD collision gas pressure setting, IS ion source voltage, TEM source temperature, GS1 nebulizer
gas, GS2 heater gas

Table 10
Oxy/endo ESI analyte-specific parameters for Sciex 6500 QTRAP

# Analyte tR Q1 Q3 DCP EP CE CXP SSTD

1 PUHA esi 3.06 249 130 30 10 18 13 –
2 6-Keto PGF1α-d4 3.54 373 167 70 10 36 13 6-Keto PGF1α-d4
3 6-Keto PGF1α 3.51 369 163 70 10 33 13 6-Keto PGF1α-d4
4 PGF3α 4.12 351 307 60 10 24 13 PGF2α-d4
5 PGE3 4.25 349 269 35 10 21 13 PGD2-d4
6 TXB2-d4 4.55 373 173 50 10 45 13 PUHA esi
7 TXB2 4.61 369 169 50 10 21 13 TXB2-d4
8 9_12_13-TriHOME 4.82 329 211 60 10 30 13 PGF2α-d4
9 PGF2α-d4 4.87 357 197 70 10 33 13 CUDA esi
10 PGF2α 4.89 353 193 70 10 33 13 PGF2α-d4
11 PGE2 4.99 351 271 40 10 21 13 PGD2-d4
12 PGE1 5.15 353 317 40 10 18 13 PGD2-d4
13 PGD2-d4 5.16 355 275 40 10 27 13 CUDA esi
14 PGD2 5.26 351 271 40 10 24 13 PGD2-d4
15 15-Keto PGE2 5.3 349 331 50 10 15 13 PGD2-d4
16 Resolvin D1 5.78 375 121 60 10 39 13 PGD2-d4
17 Lipoxin A4 5.89 351 217 45 10 27 13 PGD2-d4
18 LTB5 6.99 333 195 50 10 21 13 LTB4-d4
19 15_16-DiHODE 7.49 311 235 55 10 27 13 14_15-DiHETrE-d11
20 8_15-DiHETE 7.58 335 235 65 10 21 13 14_15-DiHETrE-d11
21 12_13-DiHODE 7.59 311 183 60 10 27 13 14_15-DiHETrE-d11
22 9_10-DiHODE 7.64 311 201 65 10 27 13 14_15-DiHETrE-d11
(continued)
196 Theresa L. Pedersen and John W. Newman

Table 10
(continued)

# Analyte tR Q1 Q3 DCP EP CE CXP SSTD

23 17_18-DiHETE 7.8 335 247 70 10 24 13 14_15-DiHETrE-d11
24 5_15-DiHETE 7.81 335 173 40 10 18 13 14_15-DiHETrE-d11
25 6-trans-LTB4 7.86 335 195 85 10 21 13 LTB4-d4
26 CUDA esi 8.7 339 214 70 10 30 13 –
27 LTB4-d4 8.48 339 163 70 10 33 13 CUDA esi
28 14_15-DiHETE 8.51 335 207 40 10 24 13 14_15-DiHETrE-d11
29 LTB4 8.58 335 195 50 10 24 13 LTB4-d4
30 12_13-DiHOME 8.9 313 183 70 10 30 13 14_15-DiHETrE-d11
31 9_10-DiHOME 9.34 313 201 65 10 27 13 14_15-DiHETrE-d11
32 14_15-DiHETrE-d11 9.5 348 207 65 10 24 13 CUDA esi
33 19_20-DiHDoPE 9.55 361 273 80 10 24 13 14_15-DiHETrE-d11
34 14_15-DiHETrE 9.6 337 207 60 10 24 13 14_15-DiHETrE-d11
35 11_12-DiHETrE 10.2 337 167 55 10 27 13 14_15-DiHETrE-d11
36 9_10-e-DiHO 10.4 315 297 110 10 30 13 14_15-DiHETrE-d11
37 12(13)-Ep-9-KODE 10.4 309 291 90 10 21 13 9-HODE-d4
38 13-HOTrE 10.6 293 195 70 10 24 13 9-HODE-d4
39 8_9-DiHETrE 10.8 337 127 65 10 27 13 14_15-DiHETrE-d11
40 15-Deoxy PGJ2 10.8 315 271 50 10 21 13 PGD2-d4
41 20-HETE-d6 11 325 281 80 10 21 13 CUDA esi
42 15-HEPE 11 317 219 55 10 18 13 12-HETE-d8
43 20-HETE 11 319 275 95 10 24 13 20-HETE-d6
44 12-HEPE 11.4 317 179 45 10 18 13 12-HETE-d8
45 5,6-DiHETrE 11.5 337 145 55 10 24 13 14_15-DiHETrE-d11
46 9-HEPE 11.6 317 167 45 10 18 13 12-HETE-d8
47 13-HODE 11.9 295 195 90 10 24 13 9-HODE-d4
48 5-HEPE 12 317 115 40 10 21 13 5—HETE-d8
49 9-HODE-d4 12 299 172 85 10 27 13 CUDA esi
50 9-HODE 12.1 295 171 70 10 24 13 9-HODE-d4
51 15(16)-EpODE 12.2 293 275 75 10 18 13 12(13)-EpOME-d4
52 17(18)-EpETE 12.3 317 259 55 10 15 13 12(13)-EpOME-d4
53 15-HETE 12.3 319 219 55 10 18 13 12-HETE-d8
(continued)
 Targeted Analysis of Lipid Mediators and Fatty Acids in Plasma 197

Table 10
(continued)

# Analyte tR Q1 Q3 DCP EP CE CXP SSTD

54 13-KODE 12.3 293 179 80 10 27 13 9-HODE-d4
55 9(10)-EpODE 12.4 293 275 65 10 18 13 12(13)-EpOME-d4
56 9-HOTrE 10.4 293 171 60 10 21 13 9-HODE-d4
57 17-HDoHE 12.4 343 281 55 10 18 13 12-HETE-d8
58 12(13)-EpODE 12.5 293 183 50 10 24 13 12(13)-EpOME-d4
59 15-KETE 12.6 317 273 60 10 18 13 12-HETE-d8
60 14-HDoHE 12.7 343 281 60 10 18 13 12-HETE-d8
61 11-HETE 12.7 319 167 45 10 21 13 12-HETE-d8
62 14(15)-EpETE 12.7 317 247 35 10 18 13 12(13)-EpOME-d4
63 9-KODE 12.8 293 185 100 10 27 13 9-HODE-d4
64 12-HETE-d8 12.8 327 184 60 10 21 13 CUDA esi
65 11(12)-EpETE 12.8 317 167 40 10 18 13 12(13)-EpOME-d4
66 12-HETE 12.9 319 179 60 10 21 13 12-HETE-d8
67 8-HETE 13.1 319 155 45 10 21 13 12-HETE-d8
68 9-HETE 13.2 319 167 60 10 18 13 12-HETE-d8
69 5-HETE-d8 13.5 327 116 70 10 18 13 CUDA esi
70 19(20)-EpDoPE 13.5 343 281 60 10 18 13 12(13)-EpOME-d4
71 5-HETE 13.5 319 115 50 10 18 13 5-HETE-d8
72 12(13)-EpOME-d4 13.6 299 198 90 10 24 13 CUDA esi
73 12(13)-EpOME 13.7 295 195 85 10 21 13 12(13)-EpOME-d4
74 14(15)-EpETrE 13.8 319 219 50 10 18 13 12(13)-EpOME-d4
75 4-HDoHE 13.8 343 281 60 10 18 13 12-HETE-d8
76 16(17)-EpDoPE 13.9 343 274 45 10 15 13 12(13)-EpOME-d4
77 9(10)-EpOME 13.9 295 171 75 10 21 13 12(13)-EpOME-d4
78 5-KETE 14.2 317 203 75 10 27 13 5-HETE-d8
79 11(12)-EpETrE 14.2 319 167 40 10 21 13 12(13)-EpOME-d4
80 8(9)-EpETrE 14.4 319 155 40 10 18 13 12(13)-EpOME-d4
81 10-Nitrolinoleate 14.6 324 277 40 10 18 13 10-Nitrooleate-d17
82 9_10-EpO 15.1 297 279 105 10 27 13 12(13)-EpOME-d4
83 10-Nitrooleate-d17 15.3 343 308 65 10 18 13 CUDA esi
84 10-Nitrooleate 15.4 326 280 40 10 24 13 10-Nitrooleate-d17
(continued)
198 Theresa L. Pedersen and John W. Newman

Table 10
(continued)

# Analyte tR Q1 Q3 DCP EP CE CXP SSTD

85 9-Nitrooleate 15.5 326 308 50 10 18 13 10-Nitrooleate-d17
86 C20:5n3 15.5 301 257 60 10 15 13 C20:4n6-d8
87 C18:3n3 15.6 277 259 115 10 24 13 C20:4n6-d8
88 C22:6n3 16.2 327 283 45 10 15 13 C20:4n6-d8
89 C20:4n6-d8 16.5 311 267 60 10 18 13 C20:4n6-d8
90 C20:4n6 16.6 303 259 40 10 18 13 C20:4n6-d8
91 C18:2n6 16.9 279 261 185 10 38 13 C20:4n6-d8
CUR curtain gas flow, CAD collision gas pressure setting, IS ion source voltage, TEM source temperature, GS1 nebulizer
gas, GS2 heater gas

Table 11
Oxy/endo þESI analyte-specific parameters for Sciex 6500 QTRAP

# Analyte tR Q1 Q3 DCP EP CE CXP SSTD

1 PUHA þesi 3.06 251.2 114.1 65 10 21 12 –
2 PGF2α-EA-d4 3.62 384.3 62.1 50 10 42 12 PUHA þesi
3 PGF2a-EA 3.63 380.3 62.1 45 10 39 12 PGF2α-EA-d4
4 PGE2-EA 3.65 378.301 62.1 65 10 39 12 PGF2α-EA-d4
5 PGD2-EA 3.92 378.302 62.1 65 10 42 12 PGF2α-EA-d4
6 PGF2a 1G 4.29 411.3 301.2 40 10 21 12 PGF2α-EA-d4
7 PGE2 1G 4.33 409.3 317.2 75 10 21 12 PGF2α-EA-d4
8 CUDA þesi 8.7 341.3 216.2 50 10 24 12 –
9 15-HETE-EA 9.62 346.3 62.1 75 10 21 12 A-EA-d8
10 11(12)-EpETre-EA 11.73 364.3 62.1 75 10 45 12 A-EA-d8
11 aL-EA 13.46 322.2 62.1 60 10 21 12 A-EA-d8
12 DH-EA 14.61 372.3 62.1 55 10 45 12 A-EA-d8
13 A-EA-d8 14.72 356.3 63.1 50 10 45 12 CUDA þesi
14 A-EA 14.78 348.3 62.1 70 10 39 12 A-EA-d8
15 L-EA 14.81 324.2 62.1 70 10 21 12 A-EA-d8
16 NA-Gly-d8 14.86 370.3 76.1 45 10 21 12 CUDA þesi
17 NA-Gly 14.91 362.3 76.1 65 10 21 12 NA-Gly-d8
18 2-AG-d5 15.41 384.3 287.2 60 10 21 12 CUDA þesi
(continued)
 Targeted Analysis of Lipid Mediators and Fatty Acids in Plasma 199

Table 11
(continued)

# Analyte tR Q1 Q3 DCP EP CE CXP SSTD

19 2-AG 15.43 379.3 287.2 110 10 24 12 2-AG-d5
20 DGL-EA 15.45 350.3 62.1 35 10 36 12 A-EA-d8
21 2-LG 15.57 355.3 263.2 25 10 12 12 2-AG-d5
22 P-EA-d4 15.62 304.2 62.1 90 10 18 12 CUDA þesi
23 1-AG 15.64 379.3 287.2 95 10 21 12 2-AG-d5
24 P-EA 15.64 300.2 62.1 70 10 18 12 P-EA-d4
25 1-LG 15.82 355.3 263.2 40 10 12 12 2-AG-d5
26 O-EA 16.1 326.2 62.1 105 10 21 12 A-EA-d8
27 D-EA 16.12 376.3 62.1 105 10 45 12 A-EA-d8
28 NO-Gly 16 340.2 76.2 60 10 21 12 NA-Gly-d8
29 2-OG 16.97 357.3 265.2 50 10 15 12 2-AG-d5
30 1-OG 17.31 357.3 265.2 55 10 15 12 2-AG-d5
31 S-EA 17.79 328.2 62.1 100 10 21 12 P-EA-d4
CUR curtain gas flow, CAD collision gas pressure setting, IS ion source voltage, TEM source temperature, GS1 nebulizer
gas, GS2 heater gas

flows and temperatures must be kept constant due to the

inability to adjust with the speed of þESI/ESI switching
(see Note 9).
2. Creation of single MRM methods to establish retention times:
Using the following instructions, build a series of methods,
each with ~20 analytes. (1) Create a method for the first
20 analytes by importing, pasting, or manually programing
the analyte names and values for precursor (Q1) and product
(Q3) ions from Table 10; (2) set analyte source voltage and
collision energy (CE) at the midrange, based on your mass
spectrometer user manual; (3) enter the ESI global para-
meters from the infusions in Subheading 3.1, step 1; (4) pro-
gram the liquid chromatography parameters from Table 12;
(5) name and save the method.
Repeat these five steps for all oxylipin analytes, being sure
to save as new files for each set of ~20 analytes. Repeat for the
þESI endocannabinoid targets. For bile and acids, refer to
Tables 13 and 14.
3. Establishing retention times and creating the “all analyte”
method: Inject 5 μL of the “all analyte” solution for each of
the methods to establish analyte retention times for each of the
groups of 20 analytes (see Notes 10 and 11). Update Table 10
200 Theresa L. Pedersen and John W. Newman

Table 12
Oxy/endo UPLC solvent gradient

Time %B
0.01 25
1.00 40
2.50 42
4.50 50
10.50 65
12.50 75
13.25 80
17.25 85
18.25 95
18.75 100
19.00 100
19.10 25
20.00 Stop
Solvent A: 0.1% acetic acid in water
Solvent B: 10:90 isopropanol/acetonitrile
Flow: 250 μL/min

Table 13
Bile acid positive mode analyte-specific parameters for Sciex 4000 QTRAP

# Analyte tR Q1 Q3 DCP EP CE CXP SSTD

1 PHAU 2.6 249.2 130 65 10 20 5 –
2 TDHCA 2.82 508.2 80 155 10 110 7 TCDCA-d4
3 T-ω-MCA 3.33 514.3 80 155 10 110 2 TCDCA-d4
4 T-α-MCA 3.47 514.3 80 155 10 110 4 TCDCA-d4
5 T-β-MCA 3.56 514.3 80 155 10 110 4 TCDCA-d4
6 TUDCA 5.48 498.3 80 155 10 110 9 TCDCA-d4
7 TCA 5.91 514.3 80 185 10 115 9 TCDCA-d4
8 ω-MCA 6.96 407.3 407 115 10 30 10 CA-d4
9 GUDCA 7.12 448.3 74 115 10 70 4 GCA-d4
10 α-MCA 7.3 407.3 407 115 10 30 4 CA-d4
11 GHDCA 7.32 448.3 74 120 10 70 4 GCA-d4
12 GCA-d4 7.4 468.3 74 125 10 70 4 PHAU
(continued)
 Targeted Analysis of Lipid Mediators and Fatty Acids in Plasma 201

Table 13
(continued)

# Analyte tR Q1 Q3 DCP EP CE CXP SSTD

13 GCA 7.41 464.3 74 125 10 70 4 GCA-d4
14 β-MCA 7.71 407.3 407 115 10 30 9 CA-d4
15 TCDCA-d4 8.63 502.3 80 175 10 110 9 PHAU
16 TCDCA 8.66 498.3 80 145 10 110 9 TCDCA-d4
17 TDCA 9.4 498.3 80 140 10 110 9 TCDCA-d4
18 CA-d4 10.1 411.3 411 120 10 30 4 CUDA
19 CA 10.2 407.3 407 125 10 30 4 CA-d4
20 UDCA 10.3 391.3 391 125 10 30 4 DCA-d4
21 GCDCA-d4 10.8 452.3 74 120 10 65 9 CUDA
22 GCDCA 10.8 448.3 74 125 10 65 4 GCDCA-d4
23 GDCA 11.5 448.3 74 125 10 65 3 GCDCA-d4
24 CUDA 12.1 339.3 214 65 10 35 9 –
25 TLCA 12.3 482.3 80 150 10 110 9 LCA-d4
26 CDCA-d4 12.4 395.3 395 125 10 25 9 CUDA
27 CDCA 12.4 391.3 391 130 10 30 9 CDCA-d4
28 DCA-d4 12.5 395.3 395 125 10 30 4 CUDA
29 DCA 12.5 391.3 391 130 10 30 4 DCA-d4
30 GLCA 12.6 432.3 74 120 10 65 10 GCDCA-d4
31 LCA-d5 13 380.3 380 135 10 30 8 CUDA
32 LCA 13 375.3 375 130 10 35 10 LCA-d5
CUR curtain gas flow, CAD collision gas pressure setting, IS ion source voltage, TEM source temperature, GS1 nebulizer
gas, GS2 heater gas

with your retention times and import or paste into your MS

method, as a scheduled MRM method or as defined MRM
periods. Scheduled MRM allows for more scans across peaks
and higher sensitivity. For a scheduled MRM method, use a
60-s retention time window for all analytes, or define in
advanced settings. For defined MRM period methods, set
dwell times at ~10–20 ms as appropriate for your instrument.
Import the updated ESI and þESI MRM experiments into a
single method, and save method as the “all analyte Tr” method
(see Note 12).
4. Regio-isomeric purity check for analyte solutions: Before con-
ducting analyte-specific source and collision energy
202 Theresa L. Pedersen and John W. Newman

Table 14
Bile acid UPLC solvent gradient

Time %B
1.01 10
0.50 10
1.00 25
11.00 40
12.50 95
14.00 95
14.50 10
16.00 Stop
Solvent A: 0.1% formic acid in water
Solvent B: 0.1% formic acid in acetonitrile
Flow: 250 μL/min

optimizations or making the calibration solutions, each analyte

optimization solution should be analyzed by retention time for
regio-isomeric purity to allow for adjusting calibration concen-
trations. Inject each individual analyte optimization solution
using your all analyte Tr method. Export area responses to
check for substantial contamination and lack of analyte signal.
Occasionally (sometimes often), the contents of manufacturer
stock can contain substantial amounts of other regio-isomers
or analytes. We have observed up to 99% impurity
(or mislabeled manufacturers stock)! In this case the analyte
should be returned to the manufacturer and purchased from an
alternative source, if possible.
5. “Gradient optimization” of analyte-specific source and collision
energies: Check the user manual for your instrument’s range of
source voltages and collision energies. In your all analyte Tr
method, adjust the source voltage in both þESI and ESI
MRM to the lowest voltage in the range (e.g., 15/15). This
can be set as a global parameter or analyte-specific parameter.
Table 10 can be updated and imported or pasted into the MS
method, for this purpose. Save this method (e.g., Oxy SV15).
Now adjust the source voltage to 20/20 and save as Oxy
SV20. Repeat for the entire range in increments of 5. Collect
data in a series of 5 μL injections of your all analyte optimiza-
tion solution (or the dilution) using each method. Export
integrated area responses to Excel, and using the MAX func-
tion, determine the voltage producing the maximum response
for each analyte (see Note 13). Replace the estimated source
voltages with these optimized values in Table 10. Import the
 Targeted Analysis of Lipid Mediators and Fatty Acids in Plasma 203

updated Table 10 into the acquisition method and save (e.g.,

Oxy SV-OPT). Using the Oxy SV-Opt method, repeat this
process for collision energies in increments of 3. You now
have an optimized method! Fine-tuning of global mass spec-
trometer parameters can be done to boost specific analyte
responses.

3.2 Preparing Before making the analyte standard mixtures described below, build
Calibration Standards, your acquisition method, run each optimization solution for iso-
Surrogate Spike, meric purity, and note concentration adjustments to your stock
and Internal Standard solutions. Refer to the worksheet in Supplementary File 1. The
Solutions procedures described below can be followed to prepare all solutions
for the Analytical and calibration standards for the assay, using your analyte stock
Assay mixture concentrations and the concentration range defined in
Tables 6–8. These concentration ranges are for general guidance
and can be updated to suit the linear range of your instrument.
1. In the worksheet, record analyte stock information (refer to work-
sheet, step 1): Use the Name Box pulldown in the top left
corner of the Excel worksheet to locate each of the following
steps. Record the amount (mg) of analyte and the μL volume of
solution in your analyte stock solutions.
2. Establish desired calibration standard range (refer to step 2): In
the sheet, there are four adjustments that can be made to
obtain your desired Cal Standard concentration values: the
volume of analyte stock solution delivered to make the mix-
tures, the final volume of the mixtures, the volume of mixture
used to make the Cal Standard, and the final volume of the Cal
Standard (see Note 14). Adjust these values to assure reason-
able volumetric deliveries of stock solutions and mixtures to
reach your targets. For example, as shown in the Supplemen-
tary File 1 worksheet, our internal standard mixture (ISTD
Mix) contains 50 μM of each of the two internal standards,
with a target of 100 nM in the calibration standards. To make
this mixture, use a glass barrel analytical syringe to deliver the
following volumes into a 10 mL volumetric flask: 170 μL of a
1 mg/mL solution of cyclohexyl ureido dodecanoic acid
(CUDA) and 25 μL of 5 mg/mL phenyl ureido hexanoic
acid (PUHA). Bring to volume in purged MeOH. Make the
remaining mixtures with the analyte stock solutions in your
sheet.
3. Prepare an IS SSTD dilution solution for Cal Standards that are
serial diluted (refer to step 3): Make sure the concentration of
your IS SSTD dilution solution matches the concentrations of
the calculated values for the higher-concentration Cal Stan-
dards that are not serial diluted; see Cal 5–9 in the oxy sheet.
Special Instructions: If you quantify surrogates (SSTDs) on a
204 Theresa L. Pedersen and John W. Newman

standard curve (e.g., using Micromass MassLynx or Agilent

MassHunter software), calculate their delivery volumes like
the analytes, and do not include SSTD in the dilution solution!
Instead, use purged MeOH to bring all Cal Standards to final
volume. Adjust the worksheet accordingly.
4. Prepare calibration standards (refer to step 4): Make calibra-
tion standards (5–9) by delivering volumes of the analyte mix-
tures to the indicated Cal Standard volumetric flasks. Assure
you do not add more than the volume of the flask! Bring to
final volume in purged MeOH. If you are maintaining constant
internal standard and surrogate concentrations across your
calibration standards, make Cal Standards (1–4) as a serial
dilution, by delivering the indicated volume of Cal Standard
6 to the Cal 1–4 volumetric flasks, and bring to final volume
with the IS SSTD dilution solution for Cal Standards. Cal
Standard 0 contains no analytes, is comprised of the remaining
IS SSTD dilution solution for Cal Standards, and allows real-
time corrections for the addition of non-isotopically labeled
impurities into samples at extraction.
5. Prepare LC ISTD sample reconstitution solution (refer to step
5): Refer to worksheet, for example, deliver 200 μL of the LC
ISTD mixture into a 100 mL volumetric flask. Bring to volume
in purged 1:1 MeOH/ACN for a final concentration of
100 nM. If you are maintaining constant concentration of
LC-ISTD across all Cal Standards, the concentration of this
solution should be the same concentration of your Cal
Standards.
6. Prepare surrogate spike solution (refer to step 6): As shown on
the sheet, we make surrogate spike solution by delivering 2 mL
of the SSTD mixture into a 25 mL volumetric flask and bring to
a final volume in purged MeOH. Use the sample spike calcula-
tor to assure that the final concentration of SSTD spike solu-
tion in samples corresponds to the final concentration in Cal
Standards.

3.3 The Shake This protocol captures the free oxylipins, endocannabinoids, bile
and Shoot Sample acids, and free fatty acids listed in Tables 6–8.
Prep Protocol 1. Sample extraction: Label prerinsed 2 mL Eppendorf tubes or
map out a rinsed 1 mL polypropylene plate. Thaw plasma
samples on wet ice (see Note 15). Per ~44 samples, include a
sample replicate, a plasma reference material, and a blank
(50 μL LC-MS grade water). If you do not have enough sample
to make a replicate, then use two aliquots of the plasma refer-
ence material. For small studies include one replicate and one
blank. Centrifuge plasma samples at 15,000 rcf (relative cen-
trifugal force, g force), at 4
C for 5 min to concentrate solids
on the bottom of the tube. Spike newly labeled clean tubes or
Targeted Analysis of Lipid Mediators and Fatty Acids in Plasma 205

wells with 5 μL of extraction antioxidant solution, 5 μL of

1000 nM oxy/endo surrogate spike solution (SSTD), 5 μL of
1000 nM bile acid SSTD, and 5 μL 100 μM FAME SSTD (see
Note 16). Transfer a 50 μL aliquot of either plasma superna-
tant or water blank into the labeled tubes/plate, cap, and
gently vortex 2 s to mix. Add 200 μL LC internal standard
solution (LC ISTD). The total sample volume at this stage
equals 270 μL of an 81% alcohol solution. Cap and vortex at
medium speed for 30 s to mix-denature-extract the samples.
Centrifuge plasma extracts at 15,000 rcf, at 4
C for 5 min to
concentrate solids on the bottom of the tube.
2. Preparation for NEFA analysis: Aliquot 150 μL of sample
extract to a 2 mL amber vial, add 289 μL of isopropanol to
bring the total alcohol amount in the extract to 410 μL, and
mix for 30 s. Add 520 μL cyclohexane and mix for 2 min. Add
484 μL of LC-MS grade water followed by 57 μL of ultrahigh
purity 1 M ammonium acetate, and mix for 2 min (see Note
17). The total water content of the extract, including water
from the sample itself, must equal a total of 570 μL at this
point. Centrifuge for 5 min at 15,000 rcf to break emulsion,
and transfer approximately 400 μL of upper organic phase to a
second labeled 2 mL glass vial being careful not to collect any of
the aqueous phase. Add 520 μL of cyclohexane to the remain-
ing aqueous sample, vortex 2 min to mix, and transfer ~600 μL
of the upper phase and combine with the first extract. Again, it
is critical not to collect water at this step! Centrifuge extracts
for 5 min at 15,000 rcf. Inspect for the appearance of any water
droplets, and if observed, remove with a pasture pipette.
Remove solvent by centrifugal evaporation in a speed vac. If
using a Genevac system, use the low boiling point (i.e., low BP)
setting with no heat for ~30 min. It is critical that the dry
extract is absolutely free from water to prevent hydrolysis of
esterified fatty acids during derivatization. Confirm complete
solvent removal and reconstitute dry residues in 100 μL tolu-
ene and vortex 30 s to dissolve residues. Create a derivatization
blank by placing 100 μL of toluene into a separate vial. Dilute
samples and blanks with 100 μL of MeOH and 20 μL of
derivatization SSTD (62.5 μM 15:1n5). Add 45 μL of (tri-
methylsilyl)diazomethane, cap, and vortex to mix for 10 s.
React for 30 min at room temperature, and then, evaporate
to dryness on a centrifugal vacuum evaporator, at low BP
setting. Reconstitute the sample in 100 μL of FAME ISTD
reconstitution solution. Vortex 10 s to mix. Store at 20
C
until GC-MS analysis.
3. Preparation for oxy/endo and bile acid analysis: Maintain sam-
ple extract on wet ice for 15 min. In batches of 8, aliquot
100 μL of extract to a Millipore Durapore PVDF 0.1 μm
206 Theresa L. Pedersen and John W. Newman

centrifugal filter, and centrifuge at 7400 rcf for 3 min, at 4
C

(see Note 18). Transfer filtrate with a Pasteur pipette to an
autosampler insert, cap, and vortex 2 (s) to remove bubbles.
Store at 20
C until analysis. Use slit-top caps for analysis, and
analyze 5 μL injections by LC-MS/MS. Build your sample list
starting with two MeOH shots, followed by a set of concentra-
tion standards, two MeOH (see Note 19), 20–30 samples, a
methanol, three calibration standards, a methanol, 20–30 sam-
ples, etc. End your acquisition with a set of calibration stan-
dards (see Note 20).

3.4 Gas Fatty acid methyl esters are quantified in 1 μL splitless injections
Chromatography Mass against an eight-point calibration curve of authentic standards.
Spectrometry Analysis
1. FAME GC instrument configuration: The injector is configured
of Fatty Acid Methyl for fast injection of 1 μL aliquots after three sample pumps, and
Esters post-injection syringe cleaning is conducted with three rinses of
hexane and isooctane. The front injection port, which receives
the sample, has a temperature of 230
C and is operated in
splitless mode with an initial pressure of 14.75 psi and a 1-min
purge delay and 18.8 mL total flow. The rear injection port has
a 2 psi initial pressure, 0 min purge delay, and a 35 mL/min
total flow. The initial column temperature of 64
C is held for
1 min, then increased at 32
C/min to 224
C, held for 8 min,
increased at 16
C/min to 240
C, and held for 8 min. The
chromatographic column carrier gas is regulated in a ramped
flow mode. Initial flow is 1.1 mL/min held for 12 min,
increased at 0.2 mL/min/min to 2.1 mL/min, held for
4 min, ramped at 2 mL/min to 0.1 mL/min, and held for
8.00 min, with outlet pressure 2 psi. In contrast, the rear
backflush column is operated in a ramped pressure mode.
Initial pressure is 2 psi, held for 21 min, ramped at 30 psi/
min to 32 psi, held for 8.00 min, and then ramped at 30 mL/
min to 2.0 psi, outlet pressure ambient (see Note 21).
2. FAME MS instrument configuration: FAMEs are detected with
a 5973N mass spectral detector with simultaneous selected ion
monitoring/full-scan (SIM/SCAN) spectral acquisition cap-
abilities. Zone temperatures are transfer line: 240
C, source:
230
C, quadrupole: 180
C. Mass spectrometer parameters:
filament delay of 3.00 min; scan parameters (50–400 m/z);
selected ion monitoring parameters are m/z 55.10, 67.10,
69.10, 74.10, 77.10, 79.10, and 368.40. Analytes are quanti-
fied using the SIM ion data and identified based on retention
time and characteristic mass fragmentation as described in
Table 15.

3.5 Quantitation Use internal standard methodologies with ratio response to quan-
titate surrogate recoveries and analyte concentrations. With this in
mind, organize your quantitation method to fit the constraints of
 Targeted Analysis of Lipid Mediators and Fatty Acids in Plasma 207

Table 15
FAME GC-MS quantitation and identification parameters

Analyte tR (min) Target ion (m/z) Q1 ion (m/z) Q2 ion (m/z)

C23:0 17.55 368.4 74.1 79.1
C6:0 4.30 74.1 55.1
C8:0 4.70 74.1 55.1
C10:0 5.20 74.1 55.1
C12:0 6.06 74.1 55.1
C14:0 7.12 74.1 55.1
C14:1n5 7.33 74.1 67.1 55.1
C15:0 7.79 74.1 55.1
C15:1n5 8.05 74.1 67.1 55.1
C16:0-d31 8.32 77.1 74.1 79.1
C16:0 8.59 74.1 55.1
C16:1n7t 8.74 74.1 67.1 55.1
C16:1n7 8.80 74.1 67.1 55.1
C17:0 9.51 74.1 55.1
C17:1n7 9.76 74.1 67.1 55.1
C18:0-d35 10.17 77.1 74.1 79.1
C18:0 10.58 74.1 55.1
C18:1n9 10.79 74.1 67.1 55.1
C18:1n7 10.86 74.1 67.1 55.1
C18:2n6 11.22 67.1 55.1 79.1
C18:3n6 11.45 79.1 67.1 55.1
C18:3n3 11.76 79.1 67.1 55.1
C19:0 11.81 74.1 55.1
C19:1n9 11.99 74.1 55.1 55.1
C18:4n3 12.05 79.1 67.1 55.1
9c,11t–CLA 12.22 67.1 55.1 79.1
10 t,12c–CLA 12.36 67.1 55.1 79.1
C20:0 13.02 74.1 55.1
C20:1n9 13.26 74.1 67.1 55.1
C20:2n6 13.76 67.1 55.1 79.1
C20:3n6 14.02 79.1 67.1 55.1
(continued)
208 Theresa L. Pedersen and John W. Newman

Table 15
(continued)

Analyte tR (min) Target ion (m/z) Q1 ion (m/z) Q2 ion (m/z)

C20:4n6 14.15 79.1 67.1 55.1
C21:0 14.33 74.1 55.1
C20:3n3 14.43 79.1 67.1 55.1
C20:4n3 14.70 79.1 67.1 55.1
C20:5n3 14.84 79.1 67.1 55.1
C22:0 15.82 74.1 55.1
C22:1n9 16.14 74.1 67.1 55.1
C22:2n6 16.82 67.1 55.1 79.1
C22:4n6 17.44 79.1 67.1 55.1
C22:5n6 17.59 79.1 67.1 55.1
C22:3n3 17.74 108.1 67.1 79.1
C22:5n3 18.43 79.1 67.1 55.1
C22:6n3 18.61 79.1 67.1 55.1
C24:0 19.74 74.1 55.1
C24:1n9 20.26 74.1 67.1 55.1

your quantitation software. Use the ratio of ISTD/SSTD to deter-

mine surrogate recoveries. If possible use the SSTDs as internal
standards for the analytes to correct for losses directly in your
quantitation software. Otherwise, treat your analytical surrogates
as analytes, determine the percentage recovered, and apply these
corrections to your measured target analyte concentrations. Surro-
gate to analyte assignments are indicated for LC-MS analyses in the
analyte-specific parameter tables (Tables 10, 11, and 13). Calibra-
tion curves should be weighted 1/x, and linear by default, but
should be inspected, and modest quadratic fits can be used (see
Note 22). Review all integrations, modify integration parameters
to maximize integration quality, and manually adjust integrations if
needed.

3.6 Assessing There are five basic quality controls used for these quantitative
Quality Controls assays.
1. Internal standard response (ISTD): The internal standard area
response can be graphed chronologically, shot to shot, over the
entire acquisition. This display can be used to determine if any
injections were not successful, as shown by a severe lack of
signal. Common causes are from not removing air bubbles
 Targeted Analysis of Lipid Mediators and Fatty Acids in Plasma 209

from inserts prior to injection (LC-MS) or a lack of extract in

the insert due to evaporation or multiple injections (LC-MS
and GC-MS). The sample should be reshot as soon as possible.
The ISTD response will also show if there has been substantial
instrumental drift, displayed as an increase or decrease in signal
over time. This is typically due to progressive ion source con-
tamination (LC-MS and GC-MS) or post-cleaning system sta-
bilization (GC-MS), but increases can also be due to sample
concentration in the vial. Replace slit-top caps between assays.
The CUDA and PUHA LC ISTDs are observed in positive and
negative mode (see Note 23).
2. Relative retention time: Plot analyte/ISTD relative retention
times, by analyte over the course of the run to assure correct
peak picking and integration.
3. Surrogate recoveries: The surrogates can be compared as a ratio
response against theoretical and measured values to determine
performance (% recovery) of the sample preparation.
4. Replicate analysis: Replicates can be compared to determine
overall method variance for the analytes. Variance (%RSD) can
be plotted against sample concentration. Variance will increase
as concentration decreases, and this plot provides a simple
means to determine the concentration at which your variance
is no longer constant. For ESI-LC-MS applications, replicate
precision of analytes above the concentration providing con-
stant variance is generally considered acceptable if less than
30%. For GC-MS applications, this cutoff is generally less
than 20%.
5. Method blank: Use to determine if the reagents are contami-
nated (see Note 24).

4 Notes

1. There is a risk of bacterial growth in solutions with <10%

organic, and these solutions should be discarded regularly.
Bacterial overgrowth in an idle LC can cause serious problems
with the hardware. Maintain idle solvent lines in organic sol-
vent to avoid growth.
2. We have successfully obtained these compounds using a variety
of vendors including Avanti Polar Lipids (Birmingham, AL,
USA), Cayman Chemical (Ann Arbor Michigan, USA), Laro-
dan Fine Chemicals (Malmo, Sweden), and Sigma-Aldrich
(Saint Louis, MO, USA). Due to inconsistent naming of chem-
ical structures, the compound International Chemical Identi-
fier (InChI) keys are provided to aid in compound searches.
However, a complete set of InChIKeys for the fatty acid methyl
esters (FAMEs) were not available while preparing this
210 Theresa L. Pedersen and John W. Newman

protocol, and we provide the free fatty acid structures to locate

these methyl esters, use the InChIKey to locate the compound
in PubChem, find a chemical name for the acid, append this
name with methyl ester, and search the worldwide web.
3. We have found this worksheet a necessary tool to assure rea-
sonable volumetric deliveries and final volumes for the concen-
tration targets of the analytes in the assay. It also functions as a
complete record of the analyte stock solutions, mixtures, cali-
bration standards, spike solutions, and the internal standard
reconstitution solutions. See Supplementary File 1 for a work-
ing example of this spreadsheet as a Microsoft Excel workbook.
4. High-boiling solvent as second rinse prevents syringe from
drying while over hot GC injection port.
5. Preparing separate mixtures allows for the confirmation of
spectral purity for chromatographically co-eluting analytes.
When weighing viscous oils into a mixture, we have found it
easiest to deliver materials sequentially to volumetric flask using
a pasture pipette, depositing the material below the volumetric
demarcation line, and accepting measured weights within 20%
of the targeted weight. To accomplish this, it is critical that you
confirm and document that your analytical balance can make
accurate measures of 1 mg in the 10 mg range. To test this, place
a volumetric flask of appropriate size on the balance, tare, and
measure the accuracy of weighing a certified 10 mg weight and a
10 mg þ 1 mg weight. Repeat this measure no less than three
times and allow time for drift between measures of at least 5 min.
6. Prior to extraction, samples are spiked with the FAME SSTD
which contains a suite of complex lipid esters. If we observe the
fatty acid products from the SSTD triglycerides, glyceropho-
spholipid, and cholesteryl ester, it indicates hydrolysis during
the NEFA derivatization, likely due to water, and the assay has
failed. In addition, with the increased availability of isotopically
labeled fatty acids, the authors would encourage replacement
of the rare compounds used here with deuterated compounds.
In particular, the C22:3n3 can have problematic resolution
from the C22:5n6 on some DB-225ms and replacement with
d8 C20:4n6 should enhance method robustness.
7. We have found that Cryovials designed for 80
C storage
often have high fatty acid backgrounds and should not be used.
8. If you have an AB Sciex 6500 QTRAP, the DCP and CE voltages
of the oxylipin/endocannabinoid analytes from Table 1 and the
global parameters from Table 2 can be used to build the acquisi-
tion method. These values also approximate the AB Sciex 4000
and can be used to capture signal for determining analyte reten-
tion time prior to analyte-specific optimization; see Subheading
3.2, step 5. Bile acid tables are for the Sciex 4000 QTRAP.
9. After the method is built, choose the parameter values that
have the least effect on overall signal.
 Targeted Analysis of Lipid Mediators and Fatty Acids in Plasma 211

10. The analytes in Table 10 are in retention time order for the
Waters Corp. 150 mm C18 BEH Acquity column, as described
in Subheading 2. Some analytes are regio-isomers and have
identical precursor and product ions, so the retention time
order is very important to maintain, so as to not switch their
identities.
11. If an analyte signal is not observed, a manual optimization by
infusion of the analyte’s optimization solution should be con-
ducted, according to the MS user manual. Optimize source
voltage by MS scan and collision energy by MS/MS or MRM
scan. The optimized analyte source voltage and collision
energy parameters can now be entered into the appropriate
20 analyte method. Rerun the method with the “all analyte”
solution to determine the analyte retention times.
12. If any of your peaks are above 1e6, dilute a portion of the all
analyte optimization solution, inject, and check peak. Assure
other analyte peaks are observed, if not consider conducting a
manual optimization to observe.
13. Use the value at the highest source voltage setting, where there
is typically less background noise.
14. In this sheet, the Cal Standard range reflects in general, the
dynamic linear range of our instrument.
15. Transfer samples from 80 to 20
C overnight to reduce
time to thaw.
16. Please use the spike calculator from step 6 to develop your
spiking solution.
17. Developing this protocol around a 1 M ammonium acetate
addition with a separate addition of water allows flexibility to
increase or decrease the amount of sample delivered to the
assay, such that the final molar concentration in the extract
can be maintained at 0.1 M. If such adjustments are made, be
sure to maintain the alcohol/cyclohexane/0.1 M ammonium
acetate ratio at 8:10:11 v/v/v.
18. If anybody can find a solvent-compatible hydrophilic polypro-
pylene filter plate, at 0.2 μm, please contact the corresponding
author:)
19. Determine if there is carryover from your high standards to a
methanol injection.
20. It is critical to track the response of the ISTD over the run to
assure there is no substantial loss of signal.
21. For calibration solutions, methods should be created that stop
rather than execute the backflush routine, thus reducing the
time required to calibrate the instrument.
22. Level zero should only be used in the calibration curves when a
significant background is detected to avoid inaccuracies in the
curves.
212 Theresa L. Pedersen and John W. Newman

23. The chromatographic behavior of both CUDA and PUHA is

sensitive to pH, and therefore shifts in this behavior are good
indications of changes in mobile phase pH. These can be
included in a broad array of assays as a measure to check
instrument sensitivity.
24. Surrogate behavior in method blanks generally does not track
those observed in biological samples, particularly in LC-MS
applications. Therefore, the quantitated values are often inap-
propriate to directly correct for any observed contamination
and should only be used to qualify data that may be affected.
When present, contamination associated with derivatization
steps in the GC-MS applications should be removed from
reported values.

Acknowledgment

The authors would like to thank Ira J. Gray, Michael R. La Frano,

and William R. Keyes for their technical support on the develop-
ment and implementation of these assays over the past 10 years.
The methods reported in this chapter were developed as part of
research projects funded by the United States Department of Agri-
culture (5306-51530-016-00D, 2032-51530-019-00-D, 2032-
51530-022-00-D) and DK/NIDDK NIH HHS Grant U24
DK097154/United States

References

1. Griffiths WJ, Koal T, Wang Y, Kohl M, Enot DP, JZ, Grunewald J, Svartengren M, Hammock
Deigner HP (2010) Targeted metabolomics for BD, Larsson BM, Eklund A, Wheelock AM,
biomarker discovery. Angew Chem Int Ed Engl Wheelock CE (2011) Asthmatics exhibit altered
49(32):5426–5445 oxylipin profiles compared to healthy individuals
2. Su LJ, Fiehn O, Maruvada P, Moore SC, after subway air exposure. PLoS One 6(8):
O’Keefe SJ, Wishart DS, Zanetti KA (2014) e23864
The use of metabolomics in population-based 6. Sun Y, Koh HW, Choi H, Koh WP, Yuan JM,
research. Adv Nutr 5(6):785–788 Newman JW, Su J, Fang J, Ong CN, van Dam
3. Strassburg K, Huijbrechts AM, Kortekaas KA, RM (2016) Plasma fatty acids, oxylipins, and risk
Lindeman JH, Pedersen TL, Dane A, Berger R, of myocardial infarction: the Singapore Chinese
Brenkman A, Hankemeier T, van Duynhoven J, Health Study. J Lipid Res 57(7):1300–1307
Kalkhoven E, Newman JW, Vreeken RJ (2012) 7. Grapov D, Adams SH, Pedersen TL, Garvey
Quantitative profiling of oxylipins through com- WT, Newman JW (2012) Type 2 diabetes asso-
prehensive LC-MS/MS analysis: application in ciated changes in the plasma non-esterified fatty
cardiac surgery. Anal Bioanal Chem 404 acids, oxylipins and endocannabinoids. PLoS
(5):1413–1426 One 7(11):e48852
4. Yang J, Schmelzer K, Georgi K, Hammock BD 8. Agrawal K, Hassoun LA, Foolad N, Pedersen
(2009) Quantitative profiling method for oxyli- TL, Sivamani RK, Newman JW (2017) Sweat
pin metabolome by liquid chromatography elec- lipid mediator profiling: a noninvasive approach
trospray ionization tandem mass spectrometry. for cutaneous research. J Lipid Res 58
Anal Chem 81(19):8085–8093 (1):188–195
5. Lundstrom SL, Levanen B, Nording M,
Klepczynska-Nystrom A, Skold M, Haeggstrom
 Chapter 14

Chemical Isotope Labeling LC-MS for Human Blood

Metabolome Analysis
Wei Han and Liang Li

Abstract
Blood is a widely used biofluid in discovery metabolomic research to search for clinical metabolite
biomarkers of diseases. Analyzing the entire human blood metabolome is a major analytical challenge, as
blood, after being processed into serum or plasma, contains thousands of metabolites with diverse chemical
and physical properties as well as a wide range of concentrations. We describe an enabling method based on
high-performance chemical isotope labeling (CIL) liquid chromatography-mass spectrometry (LC-MS) for
in-depth quantification of the metabolomic differences in comparative blood samples with high accuracy
and precision.

Key words Chemical isotope labeling, Dansylation, DmPA, LC-MS, Metabolomics, Blood

1 Introduction

Blood samples are being extensively used for discovery metabolo-

mics with the goal of finding sensitive and specific biomarkers that
are indicative of healthy and diseased states. Despite great advances
in blood metabolomic profiling techniques in the past decades,
in-depth quantitative analysis of the blood metabolome is still a
major analytical challenge. Because of diverse chemical and physical
properties of blood metabolites, a conventional strategy of increas-
ing the metabolomic coverage is to use several analytical tools with
different metabolite detectability to analyze the same sample. For
example, a widely used metabolomic profiling platform includes the
use of reversed-phase (RP) liquid chromatography-mass spectrom-
etry (LC-MS) to analyze the hydrophobic metabolites and hydro-
philic interaction (HILIC) LC-MS to analyze the relatively polar
metabolites. For each separation, positive and negative ion modes
of MS detection are separately carried out in order to ionize as
many metabolites as possible. This platform is relatively easy to
implement. However, the overall metabolomic coverage from the
combined LC-MS analyses is still limited. In addition, untargeted

Martin Giera (ed.), Clinical Metabolomics: Methods and Protocols, Methods in Molecular Biology, vol. 1730,
https://doi.org/10.1007/978-1-4939-7592-1_14, © Springer Science+Business Media, LLC 2018

213
214 Wei Han and Liang Li

metabolomic analysis using LC-MS without standards provides

limited quantification accuracy and precision due to matrix and
ion suppression effects.
An alternative strategy of performing metabolomic analysis is
to classify the metabolites into several subgroups based on the
presence of common functional moieties and then perform
in-depth analysis of the individual chemical-group-submetabo-
lomes [1]. The combined data from the submetabolomes would
allow the analysis of the entire metabolome with high coverage.
This divide-and-conquer strategy does not require many analytical
measurements of the same sample. For example, when we examined
the chemical structures of the endogenous human metabolites in
the Human Metabolome Database (HMDB), we found that more
than 95% of these metabolites contain one or more of the four
functional groups: amine, carboxyl, hydroxyl, and carbonyl. Thus,
in principle, if we could analyze all the metabolites within these four
submetabolomes, a near-complete metabolomic profile could be
generated.
Our laboratory has been developing a high-performance chem-
ical isotope labeling (CIL) LC-MS approach to analyze the individ-
ual chemical-group-submetabolomes with very high coverage
[1–3]. The idea is to use a rationally designed chemical labeling
reagent to react with a common functional group within a subme-
tabolome (e.g., amines) to form metabolite derivatives, followed by
MS detection. The properties of the labeled metabolites are altered,
compared to the original unlabeled metabolites, to such an extent
that all the labeled metabolites can be retained on RPLC for effi-
cient separation without the need of switching to another mode of
separation column, ionized with high efficiency to improve overall
detection sensitivity, and detected in positive ion mode only to
avoid the need of running negative ion mode. For accurate relative
quantification, a differential isotope labeling strategy using heavy
and light isotope reagents is used. To this end, we have reported the
use of 12C-/13C-dansyl chloride (DnsCl) for profiling the amine/
phenol submetabolome [1], 12C-/13C-dimethylaminophenacyl
(DmPA) bromide for profiling the carboxylic acid submetabolome
[2], base-activated 12C-/13C-dansyl chloride (b-DnsCl) labeling
for profiling the hydroxyl submetabolome [3], and 12C-/13C-dan-
syl hydrazine labeling for profiling the ketone/aldehyde submeta-
bolome including many sugars [4]. Each labeling method generates
a submetabolome profile with very high metabolic coverage. For
example, dansyl labeling of the amine/phenol submetabolome
detects over 2000 metabolites routinely from a plasma sample.
Figure 1 shows the high-performance CIL LC-MS workflow
for quantitative metabolomics. A control sample is prepared by
mixing small aliquots of individual samples to form a pool, followed
by heavy (e.g., 13C-dansyl) labeling. This 13C-labeled control is
 CIL LC-MS Analysis of the Human Blood Metabolome 215

Individual blood samples Pooled sample

Protein precipitation

12C-chemical isotope 13C-chemical isotope

labeling labeling

LC-UV quantification and pre-

acquisition sample normalization

Mixing equal amounts of 12C-

individual/13C-pooled samples

Mixed samples

LC-MS Analysis

Data processing and statistical studies

Fig. 1 CIL LC-MS workflow

spiked into all the light or 12C-labeled individual samples and thus
serves as a global internal standard [5]. The 12C-labeled metabolite
and its corresponding 13C-labeled counterpart in a 12C-/13C-mix-
ture are detected as a peak pair in MS, and their peak area ratio
reflects the concentration difference of the metabolite in an indi-
vidual sample vs. the control. Since the same 13C-labeled control is
used for preparing all individual 12C-/13C-mixtures, the peak ratio
values can be used for relative quantification of individual metabo-
lites in different samples. We have developed the required data
processing software for analyzing the CIL LC-MS data [6]. In
addition, over 650 metabolite standards have been individually
labeled with the proper labeling reagents for positive metabolite
identification [7]. In the library of standards, each labeled
216 Wei Han and Liang Li

metabolite contains triplet parameters: molecular mass, LC reten-

tion time, and MS/MS spectrum. Library search program has been
developed for fully automated search for rapid metabolite identifi-
cation [7]. For putative metabolite identification or structure
match, accurate mass search is used to match the detected labeled
metabolite masses to the metabolites in metabolome databases.
In this chapter, we describe the updated protocols for
performing dansyl and DmPA labeling of metabolites and LC-MS
analysis of labeled metabolites for human blood metabolomics.

2 Materials

All chemicals and reagents were purchased from Sigma-Aldrich

Canada (Markham, ON, Canada) unless otherwise noted. The
chemical isotope labeling reagents, including 12C- and 13C-dansyl
chloride and 12C- and 13C-DmPA bromide, are available on http://
mcid.chem.ualberta.ca. All solutions and LC mobile phases were
prepared using LC-MS grade solvents (see Note 1), and the solu-
tions were stored at room temperature. Blood samples were col-
lected from participants with standard operating procedures after
they reviewed and signed the informed consents. All human sam-
ples were processed in a Containment Level II laboratory and
stored in a Level I or regular chemistry laboratory after being
processed. The waste was disposed following the waste disposal
regulations. All the studies involving human subject have been
reviewed and approved by the University of Alberta Health
Research Ethics Board.

2.1 Human Blood 1. We recommend BD Vacutainer 10 mL serum collection tubes

Sample Collection (see Note 2).
2. We recommend VACUETTE 5 mL plasma collection tubes.

2.2 Dansylation 1. 250 mM NaHCO3/Na2CO3 buffer solution: Weigh 26.5 g of

Labeling anhydrous Na2CO3 and 21.0 g of anhydrous NaHCO3 in a
clean 1 L glass bottle. Measure 1 L of water in a 1 L volumetric
flask, and transfer the water to the glass bottle to dissolve the
solid (see Note 3).
2. 250 mM NaOH solution: Weigh 1.0 g of NaOH and transfer it
to a Nalgene lab quality bottle. Measure 100 mL of water in
100 mL volumetric flask, and transfer the water to the plastic
bottle. Dissolve the NaOH.
3. 425 mM formic acid solution: Dilute 1.60 mL of formic acid to
100 mL with (50/50, v/v) acetonitrile/water.
4. 20 mg/mL dansyl chloride solution: Before the chemical iso-
tope labeling experiment, 20 mg/mL 12C-dansyl chloride
 CIL LC-MS Analysis of the Human Blood Metabolome 217

(12C-DnsCl) solution is prepared by dissolving 20 mg of 12C-

DnsCl in 1 mL of acetonitrile, and 20 mg/mL 13C-dansyl
chloride (13C-DnsCl) solution is prepared by dissolving
20 mg of 13C-DnsCl in 1 mL of acetonitrile.

2.3 DmPA Labeling 1. 0.5 M triethanolamine solution: Dissolve 7.46 g of triethano-

lamine in 100 mL of acetonitrile.
2. 10 mg/mL DmPA bromide solution: Before the chemical
isotope labeling experiment, 10 mg/mL 12C-DmPA bromide
(12C-DmPABr) solution is prepared by dissolving 10 mg of
12
C-DmPABr in 1 mL of acetonitrile, and 10 mg/mL 13C-
DmPA bromide (13C-DmPABr) solution is prepared by dissol-
ving 10 mg of 13C-DmPABr in 1 mL of acetonitrile.

2.4 LC-MS Analysis 1. Mobile phase A: 1 mL of formic acid and 50 mL of acetonitrile

are transferred into a 1 L volumetric flask and then diluted to
1 L with water.
2. Mobile phase B: 1 mL of formic acid is transferred into a 1 L
volumetric flask and then diluted to 1 L with acetonitrile.
3. HPLC column: Agilent reversed-phase Eclipse plus C18 col-
umn (2.1 mm
100 mm, 1.8 μm particle size, 95 Å pore size)
for LC-MS analysis.

3 Methods

3.1 Human Serum Ten milliliters of venipuncture blood are collected into a BD Vacu-
Samples tainer 10 mL serum collection tube. An individual sample is
allowed to clot spontaneously at room temperature for 1 h and
then centrifuged at 1,500
g for 15 min to separate serum and
cells. The supernatant (serum) is divided into multiple 250 μL
aliquots in 1.5 mL microcentrifuge tubes for analysis or storage in
a 80  C freezer. Also, a pooled serum sample is prepared by
mixing equal volumes of individual serum samples. This pooled
sample is processed with 13C-labeling reagents to serve as the
internal standard during the quantitative analyses (see Note 4).

3.2 Human Plasma Blood should be collected by venipuncture into a VACUETTE

Samples 5 mL plasma collection tube with lithium heparin (see Note 5).
The tube is then inverted for ten times to ensure blood mixing with
the inner tube coating material. The sample is left on the bench for
30 min and then centrifuged at 2,100
g for 30 min at 4  C. The
plasma sample is divided into multiple 250 μL aliquots in 1.5 mL
microcentrifuge tubes for analysis or storage in a 80  C freezer.
Also, a pooled plasma sample is prepared by mixing equal volumes
of individual plasma samples.
218 Wei Han and Liang Li

30 mL of blood sample
Mix with 90 mL of methanol
Centrifuge

90 mL of supernatant
Dry down

Dried sample
Redissolve with 75 mL of 2:1 water/acetonitrile
Add 25 mL of 250 mM HCO3-/CO32- buffer
Add 50 mL of 20 mg/mL 12C- or 13C-DnsCl
Incubate at 40 °C for 45 min.

Add 15 mL of 250 mM NaOH solution

Incubate at 40 °C for 10 min.

Cool down in an ice bath.

Add 50 mL of 425 mM formic acid solution

Labeled sample for LC-MS

analysis

Fig. 2 Dansylation labeling scheme

3.3 Dansylation The workflow of dansylation labeling is shown in Fig. 2.

Labeling Reaction
1. Frozen serum or plasma samples are thawed (see Note 6) in an
ice bath and then centrifuged at 15,000
g for 15 min.
2. Thirty microliters of supernatant are transferred into a micro-
centrifuge tube and mixed with 90 μL of methanol.
3. Store the mixture at 20  C for 2 h to precipitate the proteins.
After this, the mixture is centrifuged at 15,000
g for 15 min.
4. Take 90 μL of supernatant and dry using a Speed-Vac centrifu-
gal evaporator.
5. Redissolve to 75 μL with 2:1 water/ACN. Then add 25 μL of
250 mM sodium carbonate/sodium bicarbonate buffer to the
sample to generate a basic environment for the dansylation
reaction.
6. Vortex, spin down, and mix with 50 μL of freshly prepared 12C-
DnsCl solution (20 mg/mL) (for light labeling) or 13C-DnsCl
solution (20 mg/mL) (for heavy labeling).
 CIL LC-MS Analysis of the Human Blood Metabolome 219

30 mL of blood sample
Mix with 90mL of acetonitrile
Centrifuge

90 mL of supernatant

Add 20 mL of 0.5 M triethanolamine

Add 50 mL of 10 mg/mL 12C- or 13C-DmPABr

Incubate at 85 °C for 1 hour

Cool down in an ice bath.

Labeled sample for LC-MS

analysis

Fig. 3 DmPA labeling scheme

7. After 45 min incubation at 40  C, 10 μL of 250 mM NaOH are

added to the reaction mixture to quench the excess dansyl
chloride. The solution is then incubated at 40  C for another
10 min.
8. Finally, add formic acid (425 mM) in 1:1 ACN/H2O to con-
sume excess NaOH and to make the solution acidic. The
labeled samples are now ready for LC-MS analysis or stored at
a –80  C freezer (see Note 7).

3.4 DmPA Labeling The workflow is shown in Fig. 3.

Reaction
1. Frozen serum or plasma samples are thawed in an ice bath and
then centrifuged at 15,000
g for 15 min.
2. Transfer 30 μL of supernatant into a microcentrifuge tube and
mix with 90 μL of acetonitrile.
3. Store the mixture at 20  C for 2 h to precipitate the proteins.
After this, the mixture is centrifuged at 15,000
g for 15 min.
4. Take 90 μL of supernatant, and mix with 20 μL of 0.5 M
triethanolamine and 50 μL of freshly prepared 12C-DmPA
bromide solution (10 mg/mL) (for light labeling) or 13C-
DmPA bromide solution (10 mg/mL) (for heavy labeling).
5. Incubate the mixture at 85  C for 1 h (see Note 8).
6. Finally, the samples are cooled down in an ice bath and are
ready for LC-MS analysis. The labeled samples can also be
stored in a –80  C freezer for future use.
220 Wei Han and Liang Li

3.5 LC-UV Variations in total sample amount in different samples must be

Quantification minimized in order to detect the concentration differences of
and Pre-acquisition individual metabolites caused by the biological or clinical factors
Sample Normalization being studied. An LC-UV method [8] can be applied to determine
the total concentration of dansylated amine/phenol submetabo-
lome based on the UV absorption of the dansyl group (see Note 9).
1. For the LC-UV setup, a Waters ACQUITY UPLC system with
a photodiode array (PDA) detector can be used for the quanti-
fication of dansyl-labeled metabolites for sample amount
normalization.
2. Four microliters of the labeled serum or plasma are injected
onto a Phenomenex Kinetex C18 column (2.1 mm
5 cm,
1.7 μm particle size) for a fast step-gradient run.
3. Solvent A is 0.1% (v/v) formic acid in 5% (v/v) ACN/H2O, and
solvent B is 0.1% (v/v) formic acid in ACN. The gradient starts
with 0% B for 1 min and is increased to 95% B within 0.01 min
and held at 95% B for 1 min to ensure complete elution of all
labeled metabolites. The flow rate used is 0.45 mL/min.
4. The peak area, which can represent the total labeled metabolite
concentration in the sample, is integrated using the Empower
software (6.00.2154.003).
5. Based on the quantification results, the 12C- and 13C-labeled
samples are mixed in equal mole amounts for LC-MS analysis.
When a 12C-labeled individual sample is mixed with the 13C-
labeled pooled sample, each peak pair ratio in the LC-MS result
represents the relative concentration of one specific metabolite
in the individual sample, with respect to the average concentra-
tion of the sample set.

3.6 LC-FTICR-MS 1. The LC-Fourier transform ion cyclotron resonance (FTICR)-

MS analysis is performed using an Agilent 1100 series binary
system (Agilent, Palo Alto, CA) connected to a 9.4 T Apex-Qe
FTICR-MS (Bruker, Billerica, MA). The MS data are acquired
in the positive ion mode with an electrospray ionization (ESI)
source. An Agilent reversed-phase Eclipse plus C18 column
(2.1 mm
100 mm, 1.8 μm particle size, 95 Å pore size) is
used for chromatographic separation (see Note 10). The
mobile phase A is 0.1% formic acid in ACN/H2O (5/95,
v/v), and the mobile phase B is 0.1% formic acid in ACN.
2. All dansyl-labeled samples are analyzed with a 32-min gradient:
0 min (20% B), 0–3.5 min (20–35% B), 3.5–18 min (45–65%
B), 18–21 min (65–95% B), 21–24 min (95–99% B), and
24–32 min (99% B). The column is re-equilibrated with the
initial mobile phase conditions for 15 min before injecting the
next sample. The flow rate is 180 μL/min, and the injection
volume is 5.0 μL.
 CIL LC-MS Analysis of the Human Blood Metabolome 221

(A) Intens. (B) Intens.

x10 8 x10 8
3
2.0

2 1.5

1.0
1
0.5

0 0.0
0 5 10 15 20 25 Time [min] 0 5 10 15 20 25 Time [min]

(C) Intens. (D) Intens.

x108 x108
2.5 2.5

2.0 2.0

1.5 1.5

1.0 1.0

0.5 0.5

0.0 0.0
0 10 20 30 Time [min] 0 10 20 30 Time [min]

Fig. 4 Representative LC-MS chromatograms from (a) a dansyl-labeled human serum sample, (b) a dansyl-
labeled human plasma sample, (c) a DmPA-labeled human serum sample, and (d) a DmPA-labeled human
plasma sample (acquired by LC-FTICR-MS)

3. The 40-min gradient for all DmPA-labeled samples is 0 min

(20% B), 0–9 min (20–50% B), 9–22 min (50–65% B),
22–26 min (65–80% B), 26–28 min (80–98% B), and
28–40 min (98% B). The column is re-equilibrated with the
initial mobile phase conditions for 15 min before injecting the
next sample. The flow rate is 180 μL/min, and the injection
volume is 2.5 μL.
4. The MS conditions used for FTICR-MS are as follows: nitro-
gen nebulizer gas, 2.3 L/min; dry gas flow, 7.0 L/min; dry
temperature, 190  C; capillary voltage, 4,200 V; spray shield,
3,700 V; acquisition size, 256 k; scan range, 200–1100; and
ion accumulation time, 1 s.
5. Figure 4 shows the representative base peak chromatograms of
(A) a dansyl-labeled serum, (B) a dansyl-labeled plasma, (C) a
DmPA-labeled serum, and (D) a DmPA-labeled plasma. In
Fig. 5, the peak pair of dansyl-serine is given as an example
for a representative mass spectrum. The peak with m/z of
339.0957 is the 12C-dansyl-labeled serine from the individual
sample, and the peak with m/z of 341.1032 is the 13C-dansyl-
labeled serine from the pooled sample.

3.7 LC-QTOF-MS 1. For LC-quadrupole time-of-flight (QTOF)-MS, an Agilent

1100 series binary system (Agilent, Palo Alto, CA) and an Agilent
reversed-phase Eclipse plus C18 column (2.1 mm
100 mm,
222 Wei Han and Liang Li

8
2x10 341.1032
Serine

Intensity
8 339.0957
1x10

339.0957 342.1053
0
338 339 340 341 342 343
m/z

Fig. 5 Molecular ion region of a peak pair (labeled serine)

1.8 μm particle size, 95 Å pore size) are used. The flow is loaded
to the electrospray ionization (ESI) source of a Bruker maXis
impact high-resolution QTOF mass spectrometer (Bruker, Bill-
erica, MA). All MS spectra are obtained in the positive ion mode.
2. LC solvent A is 0.1% (v/v) formic acid in 5% (v/v) ACN/H2O,
and solvent B is 0.1% (v/v) formic acid in ACN. The gradient
elution profile for dansyl-labeled samples is as follows: 0 min
(20% B), 0–3.5 min (20–35% B), 3.5–18 min (45–65% B),
18–21 min (65–95% B), 21–24 min (95–99% B), and
24–32 min (99% B). The column is re-equilibrated with the
initial mobile phase conditions for 15 min before injecting the
next sample. The flow rate is 180 μL/min and the injection
volume is 10.0 μL.
3. The 40-min gradient for running the DmPA-labeled samples
is: 0 min (20% B), 0–9 min (20–50% B), 9–22 min (50–65% B),
22–26 min (65–80% B), 26–28 min (80–98% B), and
28–40 min (98% B). The column is re-equilibrated with the
initial mobile phase conditions for 15 min before injecting the
next sample. The flow rate is 180 μL/min and the injection
volume is 5.0 μL.
4. The MS conditions used for QTOF-MS are as follows: end
plate offset, 500 V; capillary voltage, 4,500 V; nebulizer,
1.0 bar; dry gas, 8.0 L/min; dry temperature, 230  C; transfer
time, 40 μs; and prepulse storage, 10 μs.

3.8 Data Processing The data processing workflow is shown in Fig. 6 (see Note 11). A
and Statistical software tool, IsoMS, is used to process the raw data generated
Analysis from multiple LC-MS runs by peak picking, peak pairing, peak-pair
filtering, and peak-pair intensity ratio calculation [6].
1. Align peak pairs detected from multiple samples using IsoMS-
Align.
2. Fill missing ratio values using the zero-fill program [9].
 CIL LC-MS Analysis of the Human Blood Metabolome 223

Fig. 6 IsoMS data processing workflow

3. Finally, use IsoMS-Quant to determine the chromatography-

peak-intensity ratio of a 12C-/13C-pair [10]. The final metabo-
lite relative concentration file can be exported to SIMCA-Pþ
12.0 software (Umetrics, Umeå, Sweden) for multivariate sta-
tistical analysis (see Note 12).

3.9 Metabolite Positive metabolite identification is performed based on mass and

Identification retention time matching to a labeled standard library [3, 7]. Putative
identification is done based on accurate mass matches to the meta-
bolites in the Human Metabolome Database (HMDB) (8,021
known human endogenous metabolites) and in the evidence-
based metabolome library (EML) (375,809 predicted human
metabolites with 1 reaction) using MyCompoundID (www.
MyCompoundID.org) [11]. The mass accuracy tolerance window
is set at 10 ppm for database search.

4 Notes

1. HPLC grade organic solvents or purified deionized water are

found to be adequate for preparing solutions for dansylation.
224 Wei Han and Liang Li

2. Serum tubes without clot activators are recommended for bet-

ter reproducibility of the analysis results.
3. Ultra-sonication may be needed to dissolve the solid.
4. Normally the pooled sample is made by mixing equal
mole amounts of all the individual samples from the sample
set being studied. However, if some of the samples are in very
limited amount, they can be skipped from contributing to the
pooled sample; the pooled sample works as an internal refer-
ence for relative quantification, and thus not including a few
samples within a group in the pooled sample does not affect the
results. In addition, a preprepared universal serum/plasma
sample can also be used as the internal standard, which enables
inter-study comparisons.
5. EDTA plasma and citrate plasma can also be used for the CIL
LC-MS analysis without significant matrix effect.
6. Samples used in one study should experience the same number
of freeze-thaw cycles. This rule also applies to the samples after
labeling.
7. The labeled samples can be stored at 4  C for 1 week without
significant degradation. However, 80  C freezer is recom-
mended for long-term storage.
8. The leftover amount of unreacted DmPA is very small, and thus
it does not affect the metabolomic profiling of blood. If it
interferes with the detection of a specific compound with simi-
lar retention time, triphenylacetic acid can be used to quench
the excess amount of DmPA.
9. To reduce the effect of real samples on the lifetime of a C18
column, all samples should be centrifuged at 15,000
g for
10 min before analysis. A pre-column filter should also be
helpful, particularly for running blood samples.
10. Generally, the standard deviation of the distribution of total
amine-/phenol-containing metabolite amounts among human
blood samples is less than 15%. If a variance of as large as 50% is
allowed in a specific study, post-acquisition normalization
methods can be used, instead of the pre-acquisition normaliza-
tion. Nevertheless, the LC-UV quantification is recommended
for profiling blood samples for increased accuracy.
11. The data processing packages, as well as metabolite identifica-
tion libraries, are available on the MyCompoundID website:
http://www.mycompoundid.org.
12. Other metabolomic statistical tools, such as MetaboAnalyst,
can also be used for the analysis.
 CIL LC-MS Analysis of the Human Blood Metabolome 225

Acknowledgment

This work was supported by Genome Canada, the Natural Sciences

and Engineering Research Council of Canada (NSERC), and
Canada Research Chairs (CRC) programs.

References
1. Guo K, Li L (2009) Differential C-12/C-13- (10):4675–4679. https://doi.org/10.1021/
isotope dansylation labeling and fast liquid ac5009089
chromatography/mass spectrometry for abso- 7. Huan T, YM W, Tang CQ, Lin GH, Li L
lute and relative quantification of the metabo- (2015) DnsID in MyCompoundID for rapid
lome. Anal Chem 81(10):3919–3932. https:// identification of dansylated amine- and phenol-
doi.org/10.1021/ac900166a containing metabolites in LC-MS-based meta-
2. Guo K, Li L (2010) High-performance isotope bolomics. Anal Chem 87(19):9838–9845.
labeling for profiling carboxylic acid- https://doi.org/10.1021/acs.analchem.
containing metabolites in biofluids by mass 5b02282
spectrometry. Anal Chem 82(21):8789–8793. 8. YM W, Li L (2012) Determination of total
https://doi.org/10.1021/ac102146g concentration of chemically labeled metabo-
3. Zhao S, Luo X, Li L (2016) Chemical isotope lites as a means of metabolome sample normal-
labeling LC-MS for high coverage and quanti- ization and sample loading optimization in
tative profiling of the hydroxyl submetabolome mass spectrometry-based metabolomics. Anal
in metabolomics. Anal Chem 88 Chem 84(24):10723–10731. https://doi.
(21):10617–10623. https://doi.org/10. org/10.1021/ac3025625
1021/acs.analchem.6b02967 9. Huan T, Li L (2015) Counting missing values
4. Zhao S, Dawe M, Guo K, Li L (2017) Devel- in a metabolite-intensity data set for measuring
opment of high-performance chemical isotope the analytical performance of a metabolomics
labeling LC-MS for profiling the carbonyl sub- platform. Anal Chem 87(2):1306–1313.
metabolome. Anal Chem 89:6758–6765. https://doi.org/10.1021/ac5039994
https://doi.org/10.1021/acs.analchem. 10. Huan T, Li L (2015) Quantitative metabolome
7b01098 analysis based on chromatographic peak recon-
5. Peng J, Chen YT, Chen CL, Li L (2014) struction in chemical isotope labeling liquid
Development of a universal metabolome- chromatography mass spectrometry. Anal
standard method for long-term LC-MS meta- Chem 87(14):7011–7016. https://doi.org/
bolome profiling and its application for bladder 10.1021/acs.analchem.5b01434
cancer urine-metabolite-biomarker discovery. 11. Li L, Li RH, Zhou JJ, Zuniga A, Stanislaus AE,
Anal Chem 86(13):6540–6547. https://doi. YM W, Huan T, Zheng JM, Shi Y, Wishart DS,
org/10.1021/ac5011684 Lin GH (2013) MyCompoundID: using an
6. Zhou R, Tseng CL, Huan T, Li L (2014) evidence-based metabolome library for metab-
IsoMS: automated processing of LC-MS data olite identification. Anal Chem 85
generated by a chemical isotope labeling meta- (6):3401–3408. https://doi.org/10.1021/
bolomics platform. Anal Chem 86 ac400099b
 Chapter 15

Direct Infusion-Tandem Mass Spectrometry (DI-MS/MS)

Analysis of Complex Lipids in Human Plasma and Serum
Using the Lipidyzer™ Platform
Baljit K. Ubhi

Abstract
Lipids play a key role in the signaling pathways of cancer, cardiovascular, diabetic, and inflammatory
diseases. A major challenge in the analysis of lipids is the many isobaric interferences present in highly
complex samples that confound identification and accurate quantitation. After obtaining the total lipid
extract from a sample, differential mobility separation has proven to be a powerful tool for gas-phase
fractionation of lipid classes. When combined with mass spectrometry, this allows the unambiguous
identification and thus quantification of lipid molecular species. These components, sample extraction,
gas-phase separation, and mass spectrometry, form the basis of a novel integrated quantitative lipid analysis
platform.

Key words DI-MS/MS, Lipidyzer, Lipidomics, Plasma, Serum

1 Introduction

The metabolome is an important class of molecules to study as it

can lend insight into the health and well-being of an organism and
the causative elements associated with disease. It integrates the
effects of our genes with the impact of our environment by being
the product of both internal factors (our genome and proteome)
and external factors (our lifestyle and our environment). In addi-
tion, the metabolome is extremely dynamic and always changing.
By identifying and quantifying changes in specific metabolites, any
changes that are observed can be mapped back to specific pathways
for detailed interpretation. Thus, quantitative analysis of the meta-
bolome can provide important information that can help predict
the onset of a disease or disorder or characterize its progression.
Lipids are a key part of the metabolome and are involved in the
formation of important biological elements such as membranes,
lipid droplets, and lipoproteins. The study of lipids as a class of

Martin Giera (ed.), Clinical Metabolomics: Methods and Protocols, Methods in Molecular Biology, vol. 1730,
https://doi.org/10.1007/978-1-4939-7592-1_15, © Springer Science+Business Media, LLC 2018

227
228 Baljit K. Ubhi

molecules within the metabolome is important as they act as signal-

ing and inflammatory molecules and are known to be significant
players in a number of metabolic disorders, cardiovascular disease,
oncology, and other disease areas. However, lipids are an extremely
complex and diverse group of compounds. Lipid molecules are
polymers that are formed by combining metabolites from different
metabolic pathways and result from the combinatorial products of
fatty acids and head groups that can often have the same mass.
Thus, using phosphatidylcholines (a phospholipid class) as an
example, for 40 different fatty acids and five main head groups,
these 45 components could account for 8000 unique molecular
species.
In order to simplify their analysis and improve the accuracy of
identification and quantification, the first step in analyzing lipids by
mass spectrometry is efficient extraction of the entire class of lipid
compounds from the biological sample. There are several methods
which allow the extraction and recovery of total lipid from any kind
of organism or matrix. For isolating lipids from a biological sample,
there are really only three methods that are readily employed in the
field of lipidomics: Folch [1], Bligh and Dyer [2], and the methyl
tert-butyl ether (MTBE) method [3].
While extraction creates a clean sample of virtually only lipid
compounds, the remaining lipid fraction is highly complex. The
extracted lipids exhibit tremendous isobaric overlap with many
lipids per class, all within a narrow mass range. Separation prior to
analysis is essential. Liquid chromatography (LC) has traditionally
been used to separate and simplify compounds prior to mass spec-
trometry. However, because of the extreme complexity of the lipid
fraction, one LC method alone doesn’t supply the necessary sepa-
ration power to efficiently resolve all of the lipid compounds for
accurate quantification. Traditionally, several methods must be
used or a sub-fractionation technique must be added in order to
resolve each lipid class and achieve accurate qualitative and quanti-
tative data on the molecular species. This can be done by using
solid-phase extraction and separating the polar from the neutral
lipid components; however, it is time- and labor-intensive.
Differential mobility spectrometry (DMS) has been used as an
effective separation tool to couple to mass spectrometry, used most
often to reduce interferences and improve signal-to-noise ratio and
therefore sensitivity of detection. For lipid analysis, it has proven
very successful as an orthogonal separation technique by allowing
lipids to be fractioned in the gas phase by the dipole moment on
their head groups. As the device is located at the front of the mass
spectrometer, it helps eliminate isobaric overlap often encountered
during any lipid analysis, whether applying accurate mass or nomi-
nal mass spectrometry. In DMS, an ionized sample enters a mobility
cell that is located at atmospheric pressure in front of the mass
spectrometer orifice. An RF voltage is applied across the cell that
 DI-MS/MS Lipidomics 229

is cycled between high and low fields. As the ions transit the cell,
they are separated based on the difference in their mobility between
high and low field, which is impacted by numerous molecular
properties. A second voltage (a compensation voltage known as
COV) is applied to steer analytes through the cell into the mass
spectrometer. Now mass isolation, fragmentation, and analysis can
be performed on the separated classes yielding simplified, more
reliable data for qualitative and quantitative analysis. The DMS is
a small planar cell that is added between the source region and the
high vacuum region of the instrument.
When DMS is used for lipid analysis, each lipid class will have a
different COV that can be used to separate each individual class by
head group and allow them to enter into the MS, one class at a time
(Fig. 1). This is because each lipid class possesses a different head
group with differing dipole moments that are affected by the
voltages they encounter within the mobility cell. In a paper by
Lintonen et al. [4], a linear relationship between the dipole
moment of each lipid class and the COV value was found. By
applying a specific COV, each lipid class could be directed through
the mobility cell into the MS, and all other classes (and species) will
not be transmitted. Alternatively, by ramping the COV, different
classes of lipids could, in turn, become stable and transit success-
fully through the cell, thereby allowing sequential analysis of all
lipid classes and respective molecular species. In Fig. 1, the separa-
tion of the lipids by class is demonstrated by analyzing the lysopho-
sphatidylcholine/lysophosphatidylethanolamine/phosphatidyl-
choline/phosphatidylethanolamine (LPC/LPE/PC/PE) classes in
the negative ion mode and sphingomyelin (SM) class in the positive
ion mode.
Separating lipid classes by DMS allows us to overcome the
confounding factor of isobaric overlap of these molecules. When
coupled with targeted multiple reaction monitoring (MRM) on a
triple quadrupole or QTRAP® mass spectrometer, this simple sepa-
ration of lipid classes eliminates the need for any up-front liquid
chromatography (LC), greatly simplifying the workflow. By select-
ing specific COV values for lipid class transmission that provide the
least amount of lipid class overlap, isobaric interference, inherent in
lipid analysis, is significantly reduced.
A new LC-MS/MS solution (Lipidyzer™ Platform, SCIEX)
simplifies and automates the high-throughput analysis of lipids by
infusion, leveraging the optimized sample preparation procedure
and DMS separation power described above. Using flow injection
sample introduction, the platform can analyze up to 45 samples per
day in a fully automated fashion, including system optimization and
data processing. This platform allows for the quantitation of over
1100 lipid molecular species across 13 lipid classes from complex
lipid metabolism readily found in human plasma and serum.
230 Baljit K. Ubhi

Fig. 1 A differential mobility spectrometry (DMS) ionogram. A compensation voltage (COV) ramp of a mixture of
standards infused into the DMS cell. COV ramped from 25 to 10 V separates lipid classes by their head
groups in the gas phase. This figure highlights the specificity of the DMS cell to separate lipid classes from
complex mixtures. The COV per class is added to the MRM tables in the acquisition method and at any 1 V
across the DMS, only that specific lipid class is selected and allowed and passed through to Q1 for subsequent
MRM analysis. In this figure, the COV tuning mixture is infused at 7 μL/min, and the Lipidomics Workflow
Manager (LWM) software automatically tunes the COV value per class by collecting lysophosphatidylcholine
(LPC)/lysophosphatidylethanolamine (LPE)/phosphatidylcholine (PC)/phosphatidylethanolamine (PE) MRMs in
the negative ion mode and sphingomyelin (SM) MRMs in the positive ion mode

The platform employs the Bligh and Dyer extraction protocol

(modified to use dichloromethane) because of the acceptance in the
field and its simplicity. It had readily been shown to extract the
larger proportion (~98%) of lipids and is more efficient in its
recoveries. The full extraction can be completed in one test tube
without any complicated filtration steps. The optimized protocols
are included with the platform. Finally, a novel internal standard
strategy has been developed for the Lipidyzer Platform that enables
more accurate quantification with reduced bias [5]. The chemical
kits include over 50 labeled internal standards that cover 13 lipid
classes across complex lipid metabolism.
In this chapter, we describe the methods for identification and
quantification of complex lipids in plasma and serum using
the direct infusion-tandem mass spectrometric (DI-MS/MS)
approach employed by the Lipidyzer Platform.

2 Materials

Please be sure to use proper PPE (personal protective equipment).

Only HPLC or LC-MS grade solvents should be used.
 DI-MS/MS Lipidomics 231

2.1 Standards 1. SelexION tuning mixture (Part Number 5040141, SCIEX,

MA, USA): Aliquot 0.1 mL of the SelexION tuning mixture
into a vial. Add 0.9 mL of the 10 mM ammonium acetate in
dichloromethane/methanol (50:50) to the vial. Cap the mix-
ture and vortex gently for 5 s.
2. QC Spike Kit (Part Number 5040408, SCIEX, MA, USA).
3. QC control plasma (SCIEX, MA, USA).
4. Lipidyzer internal standards (Part Number 5040156, SCIEX,
MA, USA).

2.2 Sample 1. 10 mM ammonium acetate solution in (50:50) dichloro-

Extraction methane/methanol: Dissolve 770 mg ammonium acetate in
1 L of methanol, and mix 500 mL of this solution with 500 mL
dichloromethane.
2. Extraction solvents: 100% methanol, 100% water, and 100%
dichloromethane (HPLC grade).
3. For the Bligh and Dyer extraction, borosilicate glass culture
test tubes (16  125 mm, 10 mL) are used.

2.3 DI-MS/MS 1. All tubing used for the FIA setup is PEEKsil which allows for
minimal carry-over between injections (< 0.5%).
2. 10 mM ammonium acetate solution in (50:50) dichloro-
methane/methanol; see Subheading 2.1, item 4 for
preparation.
3. Rinse solvents for lines R0 and R1: (50:50) dichloromethane/
methanol; see Subheading 2.1, item 1 for preparation.
4. Rinse solvents for lines R2 and R3: 100% 2-propanol
(isopropanol).
5. SelexION chemical modifier: 100% 1-propanol.
6. For sample analysis, 12  32 mm 2 mL glass vials are used with
250 μL pulled conical point glass inserts and 9 mm vial caps
with pre-slit septa.

3 Methods

3.1 SelexION® 1. Prepare the SelexION tuning standard by warming one vial
Tuning from the SelexION Tuning Kit to ambient temperature before
opening. If lipids precipitate out of solution during storage,
gentle warming followed by vortexing will redissolve lipids.
2. Fill a 1 mL syringe with the SelexION tuning mixture.
3. Prepare a “blank” sample by filling a 2 mL vial with 10 mM
ammonium acetate in (50:50) dichloromethane/methanol
and cap.
232 Baljit K. Ubhi

4. Place in the autosampler according to the tray layout

(as detailed by the Lipidomics Workflow Manager (LWM)
software).
5. Attach the syringe to the PEEK adaptor and set in the MS
syringe holder.
6. Attach the tubing from the probe to the PEEK adaptor.
7. In the LWM software, in SelexION tuning step 2, click Next.
8. Click Start Run.
9. A window, “Purge Modifier,” will automatically pop up. Click
on the Purge button. Purging will automatically start and will
finish after 4 min. A 30 min equilibration will automatically
start after the purging.
10. While the system is purging/equilibrating, press the MS
syringe button to start infusing the standard.
11. Complete tuning and save the updated compensation voltages
to the MRM tables.

3.2 System For a Quick Test, the following steps are necessary:
Suitability Testing
1. Preparation of the SST LOD sample for the SST Quick Test.
(SST)
2. Warm one vial from the System Suitability Test Kit to ambient
temperature before opening. If lipids precipitate out of solu-
tion during storage, gentle warming followed by vortexing will
redissolve lipids.
3. Aliquot 0.01 mL of the System Suitability Mixture into a vial.
4. Add 0.99 mL of 10 mM ammonium acetate (50:50) dichlor-
omethane/methanol to the vial.
5. Cap the mixture and vortex gently for 5 s.
6. Transfer 0.25 mL of the mixture into an insert.
7. Place the insert in a vial and cap.
8. Place in the autosampler according to the tray layout
(as detailed by the LWM software).
9. The remaining SST LOD mixture can be stored in the 20  C
freezer for future use.
10. Preparation of a “blank” sample.
11. Fill a 2 mL vial with the 10 mM ammonium acetate (50:50)
dichloromethane/methanol and cap.
12. Place in the autosampler according to the tray layout (in LWM
software).
For a Comprehensive Test, the following steps are necessary:
1. Purge the LC system.
 DI-MS/MS Lipidomics 233

2. Preparation of the SST LOD sample for the SST

Comprehensive Test.
3. Warm one vial of System Suitability Test Kit to ambient tem-
perature before opening. If lipids precipitate out of solution
during storage, gentle warming followed by vortexing will
redissolve lipids.
4. Aliquot 0.01 mL of the system suitability standard into a vial.
5. Add 0.99 mL of the 10 mM ammonium acetate dichloro-
methane/methanol (50:50) to the vial.
6. Cap the mixture and vortex gently for 5 s.
7. Transfer 0.25 mL of the mixture into an insert.
8. Place the insert in a vial and cap.
9. Place in the autosampler according to the tray layout (in LWM
software).
10. The remaining SST LOD mixture can be stored in the 20  C
freezer for future use.
11. Preparation of the two SST RSD samples for the SST
Comprehensive Test.
12. Warm a vial of the System Suitability Test Kit to ambient
temperature before opening, if not done so already. If lipids
precipitate out of solution during storage, gentle warming
followed by vortexing will redissolve lipids.
13. Place two inserts into two separate vials.
14. Aliquot 0.05 mL of the System Suitability Mixture into each
insert.
15. Add 0.2 mL of the 10 mM ammonium acetate dichloro-
methane/methanol (50:50) to each insert.
16. Cap both vials and vortex gently for 5 s.
17. Place in the autosampler according to the tray layout.
18. Preparation of a “blank” sample.
19. Fill a 2 mL vial with the 10 mM ammonium acetate dichlor-
omethane/methanol (50:50) and cap.
20. Place in the autosampler according to the tray layout.

3.3 Internal Standard 1. Warm each of the internal standard vials to ambient tempera-
Mixture Preparation ture prior to opening. If lipids precipitate out of solution
during storage, gentle warming followed by vortexing will
redissolve lipids.
2. Before using each syringe, rinse the syringe three times with
methanol and then three times with dichloromethane.
234 Baljit K. Ubhi

3. Vortex the first internal standard lot vial and uncap the vial.
Add the appropriate amount of volume according to the Inter-
nal Standard Mixture Page in step 4 in the LWM software.
4. Cap the internal standard lot vial.
5. Repeat steps 2–4 for each internal standard class you wish to
analyze.
6. Concentrate the internal standard mixture solution under
nitrogen.
7. Add the calculated final volume (mL) of (50:50) dichloro-
methane/methanol from the Internal Standard Mixture Page
in step 4 (LWM software) to the vial. Cap the vial.
8. Portion the internal standard mixture into separate vials as
needed according to the Internal Standard Mixture Page in
step 4 (in LWM software). Cap the vials.
9. If the mixture will not be used immediately, wrap the capped
vial with Teflon tape, and store in a 20  C freezer.
10. Store the internal standard lot vial(s) in a 20  C freezer.

3.4 Sample 1. Thaw frozen serum or plasma experimental samples and QC

Extraction control plasma for 30 min at room temperature.
2. Thaw the internal standard mixture (made in the previous
section) and a vial of the QC Spike Kit for 30 min at ambient
temperature.
3. Aliquot 0.1 mL of the serum or plasma samples to glass culture
tubes.
4. To make the QC samples, aliquot 0.1 mL of the QC control
plasma to glass culture tubes.
5. To make the QC Spike samples, aliquot 0.1 mL of the QC
control plasma to glass culture tubes. Add 0.05 mL of the QC
Spike mixture to each QC Spike sample.
6. To all samples, add 0.9 mL of water, 2 mL of methanol, and
0.9 mL of dichloromethane.
7. The extracts are gently vortexed for 5 s.
8. Add 0.1 mL of the internal standards mixture to the extracts.
9. The extracts are then set on the benchtop at room temperature
for 30 min.
10. An additional 1 mL of water and 0.9 mL of dichloromethane
are added to the extracts.
11. The extracts are gently vortexed for 5 s and then centrifuged on
high speed for 10 min or until the extracts are separated into a
bilayer. Note: Over-vortexing samples at this stage will lead to
poor-phase partitioning.
 DI-MS/MS Lipidomics 235

12. The bottom organic layer is transferred to a new test tube for
each extract.
13. Another 1.8 mL of dichloromethane is added to the original
extract test tubes.
14. The original extracts are gently vortexed for 5 s and centrifuged
on high speed for 10 min or until the extracts are separated into
a bilayer. Note: Over-vortexing samples at this stage will lead to
poor-phase partitioning.
15. The bottom layers are taken again and added to the previous
aliquots.
16. The combined bottom layers for each sample are concentrated
under nitrogen and reconstituted in 0.25 mL of 10 mM
ammonium acetate in (50:50) dichloromethane/methanol.
17. The extracts are transferred to inserts and placed in vials for
analysis on the Lipidyzer Platform.

3.5 DI-MS/MS 1. A QTRAP® system with SelexION Technology (SCIEX) is

Analysis used for targeted profiling (SCIEX, MA, USA).
2. Two methods are used covering 13 lipid classes using a flow
injection analysis (FIA): one injection with the SelexION vol-
tages turned ON and another with the SelexION voltages
turned OFF.
3. The lipid molecular species are measured using multiple reac-
tion monitoring (MRM) and positive/negative switching. Pos-
itive ion mode detected the following lipid classes: SM/DAG/
CE/CER/TAG. Negative ion mode detected the following
lipid classes: LPE/LPC/PC/PE/FFA.
4. A flow injection analysis (FIA) setup is employed by using the
LC to flow at an isocratic rate of 7 μL/min with a ramp up to
30 μL/min for the last 2 min of the experiment to allow for
washing.
5. Data acquisition is around 20 min per sample, and 50 μL of the
reconstituted sample is infused and the area under the flat
infusion line reported and corrected to the appropriate internal
standard.
6. Samples are quantified using the LWM software which reports
all the detected lipids in nmol/g.

4 Notes

1. Proper storage of lipid standards is crucial to accurate lipid

quantitation. It is recommended to store lipid standards at
20  C.
236 Baljit K. Ubhi

2. All Lipidyzer kits and standards are stored in (50:50) dichlor-

omethane/methanol which is a highly volatile solvent.
Extreme caution should be used to prevent any evaporation
during storage. Evaporation will cause the internal standards to
degrade. Best practice is to store the standards after opened in
amber glass vials wrapped with Teflon tape preferably under
argon/nitrogen.
3. Standards should never be stored in plastic or polymer material
containers as this will leach impurities from the container.
4. If lipids precipitate out during storage, then gentle warming at
ambient temperature followed by mixing on a vortex mixer will
redissolve them.
5. Ammonium acetate is hygroscopic, meaning it absorbs water
over time. This can lead to insufficient acetate adduct ions
being formed and thus impacting accurate quantitation. Store
ammonium acetate in a desiccator.
6. The PEEKsil tubing can become blocked; make sure to care-
fully extract the organic (bottom) layer free from any protein
precipitate. Also spinning of the reconstituted sample (in the
LC-MS vial) eliminates any particulates in the sample to the
bottom of the insert, thus eliminating this problem.

Acknowledgments

The author would like to thank the Lipidyzer development team at

Metabolon (Raleigh, NC) and the Lipidyzer development team at
SCIEX (Framingham, MA) including Avanti Polar Lipids (Alabas-
ter, AL) for manufacturing the novel internal standards accompa-
nying the Lipidyzer Platform.

References
1. Folch J, Lees M, Stanley GHS (1957) A simple throughput lipidomics. J Lipid Res 49
method for the isolation and purification of total (5):1137–1146
lipids from animal tissues. J Biol Chem 226: 4. Lintonen TPI, Baker PRS, Suoniemi M, Ubhi
497–509 BK, Koistinen KM, Duchoslav E, Campbell JL,
2. Bligh EG, Dyer WJ (1959) A rapid method of Ekroos K (2014) Differential mobility
total lipid extraction and purification. Can J Bio- spectrometry-driven shotgun lipidomics. Anal
chem Physiol 37(8):911–917 Chem 86(19):9662–9669
3. Matyash V, Liebisch G, Kurzchalia TV, 5. Ubhi BK, Watkins S (2015) High quality, repro-
Shevchenko A, Schwudke D (2008) Lipid ducible, lipidomics: making lipid research rou-
extraction by methyl-tert-butyl ether for high- tinely accessible. Chromatography Today
 Part III

GC-MS-Based Metabolomics
 Chapter 16

Exploratory GC/MS-Based Metabolomics of Body Fluids

Carole Migné, Stéphanie Durand, and Estelle Pujos-Guillot

Abstract
GC/MS-based metabolomics is a powerful tool for metabolic phenotyping and biomarker discovery from
body biofluids. In this chapter, we describe an untargeted metabolomic approach for plasma/serum and
fecal water sample profiling. It describes a multistep procedure, from sample preparation, oximation/
silylation derivatization, and data acquisition using GC/QToF to data processing consisting in data
extraction and identification of metabolites.

Key words Untargeted metabolomics, Gas chromatography, Mass spectrometry, Data processing

1 Introduction

Mass spectrometry (MS) coupled with advanced chromatographic

techniques, such as liquid and gas chromatography (GC), has
become a powerful metabolomic tool to identify subtle changes
in metabolite profiles in biological systems. In particular, GC/MS is
a widely used analytical platform in discovery-phase studies for
metabolic profiling of various physiological biofluids, such as
urine, blood, as well as fecal water samples [1–4]. It has very high
separation efficiency, reproducibility, and high sensitivity. More-
over, it provides characteristic, reproducible, and standardized
mass spectra, which allow identification when searching against
databases. The major disadvantages of GC/MS are a limited ana-
lytical coverage (to a set of small volatile biological molecules,
thermally stable, or to those that can be derivatized) and a long
sample preparation requirement. In fact, in order to provide
non-biased data, representative of sample metabolic complexity,
GC/MS profiling requires adequate extraction of metabolites
from the biological matrix and derivatization for nonvolatile com-
pounds, often necessary to reduce polarity, to increase thermal
stability and volatility [5], as well as to improve compound ioniza-
tion. Because of its ability to derivatize a wide range of metabolites,
the most commonly used derivatization method is based on a

Martin Giera (ed.), Clinical Metabolomics: Methods and Protocols, Methods in Molecular Biology, vol. 1730,
https://doi.org/10.1007/978-1-4939-7592-1_16, © Springer Science+Business Media, LLC 2018

239
240 Carole Migné et al.

two-step protocol involving a methoximation reaction followed by

a silylation [6], but it requires a nonaqueous environment for the
reaction and thereby leads to a more complex sample pretreatment.
In most applications, 0.5–2 μL of TMS-derivatized samples are
introduced into a heated injector (200–250  C), and separation
of metabolites is typically performed using a fused silica capillary
column with a 5% diphenyl cross-linked 95% dimethylpolysiloxane
stationary phase (0.25 μm film thickness) using helium as carrier gas
[2]. Metabolomics for high-throughput experiments requires fast
acquisition rates, which are now available, thanks to recent prog-
ress. GC is often coupled to single-quadrupole analyzers, which
offer high sensitivity and good dynamic range but operate with
slower scan rates and lower resolution compared to time-of-flight
(ToF) systems. GC coupled with ToF mass spectrometry is increas-
ingly used for metabolic profiling [7] because of fast acquisition
rates, particularly useful for an accurate deconvolution of overlap-
ping peaks obtained from complex mixtures.
The nontargeted approach is based on both annotated and
unannotated peak information and therefore requires an important
data processing step to handle the large amount of data. In partic-
ular, it consists in the alignment of chromatographic peaks along
large sample sets, the annotation to keep only one entry for each
metabolite [8], and finally metabolite identification. Highly diverse
and specialized software solutions for GC/MS data processing have
been published. Both automated peak extraction and automated
deconvolution of mass spectra are necessary for the comprehensive
analysis of GC/MS experiments [9]. The conventional approach in
GC/MS is based on mass spectral deconvolution by processing the
information present within single chromatograms. One of the most
widely used tools is the automated mass spectral deconvolution and
identification system (AMDIS; http://chemdata.nist.gov/mass-
spc/amdis/overview.html). For GC-ToF data, commercial pro-
grams are also commonly used. However, to be able to compare
large numbers of samples and to improve data processing through-
put, software projects, based on the comprehensive extraction of
mass selective peak apex intensities, were initiated by academia, for
example, XCMS [10].
Finally, metabolite identification is performed by retention
time comparison with pure standard compounds and/or compari-
son with mass spectral library databases (i.e., NIST). Although
these libraries are extensive, they do not yet contain a large number
of metabolites that are found in biological metabolic pathways.
Public repositories of mass spectra of metabolites (i.e., MassBank
[11]) are also used as complementary tools.
In this chapter, we focused on two important human biofluids,
plasma/serum and fecal water. The objective is to provide a full
description of untargeted metabolomics for biomarker discovery,
 GC-MS Metabolomics of Body Fluids 241

from sample treatment and analysis to data processing and metab-

olite identification.

2 Materials

Prepare all solutions using ultrapure water (prepared by purifying

deionized water to 18 MΩ-cm at 25  C) and analytical grade
reagents. All solvents should be analytical grade. Prepare and
store all reagents and solvents at room temperature (unless indi-
cated otherwise).

2.1 Standards 1. [13C1]-L-valine (99%), a stable isotope-labeled internal stan-

and Preservatives dard (IS), is obtained from Cambridge Isotope Laboratories
(Andover, MA). Two solutions in ultrapure water are prepared
at 100 μg/mL for fecal water extraction and 200 μg/mL for
plasma or serum extraction. Dissolve 2 mg of [13C1]-L-valine
in 1 mL of ultrapure water. Dilute this solution by 10 or
20 according to the biofluid extracted.
2. Sodium azide (NaN3  99.0%) 100 mg/mL in ultrapure
water: dissolve 0.5 g of NaN3 in 5 mL of ultrapure water
(see Note 1).

2.2 Extraction 1. Heptane and methanol (HPLC Grade) are used for liquid
extraction.
2. An aliquot of 200 mL of methanol is Stored at 20  C during
24 h.

2.3 Derivatization 1. N,O-bis(trimethylsilyl)trifluoroacetamide (BSTFA) with 1%

Reagents for Gas trimethylchlorosilane (TMCS) is purchased from Thermo
Chromatography Mass Fisher Scientific and stored at 4  C.
Spectrometry 2. Pyridine is stored in a fume hood in amber glass vials in
the dark.
3. Methoxylamine solution (15 mg/mL) in pyridine: dissolve
15 mg of methoxylamine in 1 mL of pyridine (see Note 2).

2.4 Gas 1. GC-QToF: use an Agilent 7890B Gas Chromatograph coupled

Chromatography Mass to an Agilent Accurate Mass QTOF 7200, 7693A Injector
Spectrometry (SSL) auto-sampler (Agilent Technologies, Inc.).
2. Gas chromatography column: HP-5MS UI 30 m  0.25 mm i.
d.  0.25 μm film thickness (Agilent J&W Scientific, Folsom, CA,
USA) with a deactivated fused silica precolumn 5 m  0.32 mm
i.d. (Agilent J&W Scientific, Folsom, CA, USA).
3. Use N11 crimp caps natural rubber/TEF 1.0 mm (Macherey-
Nagel) and micro-vials with crimp neck N 11, amber, conical with
a round pedestal glass plate 1.1 mL, 32  11.6 mm (Macherey-
Nagel).
242 Carole Migné et al.

3 Methods

3.1 Fecal Water, 1. For fecal water samples, homogenized stool samples (5–8 g)
Plasma, and Serum are ultracentrifuged at 4  C and 171,500  g using a Beckman
Sample Collection Ti 70.1 rotor for 2 h. 2 μL of NaN3 as an antimicrobial agent
and Storage per gram of fecal water is added, and supernatants are aliquoted
as 1 mL samples and kept at 80  C for a maximum period of
6 months.
2. Plasma is either collected on EDTA, heparin, or citrate. Then,
250 μL aliquots of collected plasma or serum are stored at
80  C until analysis.

3.2 Processing 1. Thaw fecal water samples at room temperature.

of Fecal Water 2. Add 100 μL ice-cold methanol and subsequently 50 μL fecal
Samples water to an Eppendorf centrifuge tube (1.5 mL or smaller),
vortex, keep at 20  C for 30 min, and centrifuge (Sigma
3-16PK, Fischer Bioblok Scientific) at 15,493  g at 4  C for
10 min.
3. Transfer 90 μL supernatant to high recovery vial with
crimp cap.
4. Spike 10 μL of 100 μg/mL IS (13C1-L-valine) to each vial and
close. Store the sample vial (in a suitable container) at  20  C
for more than 30 min.
5. Remove the screw cap, and transfer vial to a freeze drier (pre-
cooling) for lyophilization (see Note 3).
6. Ensure rapid transfer into a drying oven and keep at 50  C
(about 1 min) to avoid absorbing moisture and close
sample vial.
7. At the same time, a derivatization control sample (fecal water
substituted by Milli-Q water) is prepared in order to determine
the background noise produced during sample preparation,
derivatization, and GC/MS analysis.

3.3 Processing Extraction

of Plasma/Serum
1. Thaw blood samples at 4  C overnight.
Samples
2. Add 200 μL of ice-cold methanol ( 20  C) to 100 μL plasma
sample in an Eppendorf tube and vortex, or add 400 μL of
ice-cold methanol ( 20  C) to 100 μL serum sample in an
Eppendorf tube and vortex.
3. After precipitation of the proteins, store at 20  C for 30 min.
4. Centrifuge the mixture at 15,493  g for 10 min at 4  C.
5. Transfer 200 μL supernatant in a brown glass vial of 2 mL, add
10 μL of [13C1]-L-valine (200 μg/mL), and evaporate to
dryness using a SpeedVac.
 GC-MS Metabolomics of Body Fluids 243

6. At the same time, a derivatization control sample (plasma/

serum is substituted by Milli-Q water) is prepared in order to
determine the background noise produced during sample
preparation, derivatization, and GC/MS analysis.

3.4 Derivatization Process to the derivatization step straight away after the extraction.
1. Dissolve the dry residue by adding 80 μL of methoxylamine
solution to each vial.
2. Vortex vigorously for 1 min and incubate at 37  C for 24 h
(in order to inhibit the cyclization of reducing sugars and the
decarboxylation of α-keto acids).
3. Add 80 μL of BSTFA (1%TMCS) to the mixture and derivatize
at 70  C for 60 min.
4. Cool down to room temperature, and transfer 50 μL of deri-
vatized mixture to a glass vial containing 100 μL of heptane,
prior to injection.
5. In the same way, a pool sample is prepared from each extracted
sample and derivatized in order to monitor the drift of the
spectrometer during GC/MS analysis (see Note 4).
6. Transfer 50 μL of sample pool in a glass vial containing 100 μL
of heptane prior to injection; repeat this step eight times (pre-
processing analysis) (see Note 5).

3.5 Gas 1. Analytical sequence: to avoid possible differences between sam-

Chromatography Mass ple batches, randomize the analytical sequence using a Latin
Spectrometry Analysis square.
2. Four heptane blanks are injected at the beginning of each
sequence, followed by four pool samples, and then one pool
sample (see Note 4) and one derivatization control sample after
each set of 10 samples.
3. Chromatography method.
Use an Agilent HP-5MS column with helium as the carrier
gas at a flow rate of 1 mL/min (ultrahigh purity helium gas
(99.9990%) with the following conditions: injector tempera-
ture, 250  C; transfer line heater, 280  C; and injection vol-
ume, 0.5–2 μL in splitless mode (see Note 6).
The initial oven temperature is 60  C for 2 min, ramped to
140  C at a rate of 10  C/min, to 240  C at a rate of 4  C/min,
and to 300  C at a rate of 10  C/min and finally held at 300  C
for 8 min. The total run time is 49 min. The solvent delay is
5.5 min (see Note 7).
Agilent “retention time locking” (RTL) is applied to control
the reproducibility of retention times (RT). [13C1]-L-valine
(IS) is used to lock the GC method.
244 Carole Migné et al.

4. MS method.
Initially tune and calibrate the system using PFTBA. Subse-
quently, in the analytical sequence, a calibration is done
between each sample.
Use the following conditions: electron ionization source
temperature, 230  C; quadrupole temperature, 150  C; elec-
tron energy, 70 eV; ToF settings, acquisition 2GHzEDR with
N2 (1.5 mL/min); mass range, m/z 50–800; acquisition rate,
5 spectra/s; acquisition time, 200 ms/spectrum; limits for
average PPM error, 3.0, and maximum error, 8.0; and resolu-
tion, 8500 (fwhm) at m/z 501.9706.
Raw data files are transformed to mzData files in MassHun-
ter B.07.01 software (Agilent).

3.6 Data Processing: 1. Process the mzData files using XCMS [10, 12] to yield a data
Extraction matrix containing retention times, accurate masses, and nor-
and Alignment malized peak intensities. Set XCMS settings (matchfilter algo-
rithm) as follows: fwmh 10–12, step 0.01, Snthreshold 5, bw
10, and mzwid 0.1 (see Note 8).
2. Import the obtained peak list under the Galaxy web-based
platform Worflow4metabolomics [13] (W4M), for quality
checks and signal drift correction according to the algorithm
described by Van der Kloet et al. [14], to correct for batch
effects.

3.7 Metabolite 1. Perform database queries, using the resulting peak list and
Identification W4M. If pure compounds (standards) are previously analyzed
on the same instrument, use the “bank in-house” tool with a
mass error of 0.005 Da and a retention time difference of
0.1 min. If not, tentative identifications can be obtained from
comparison of exact masses to those registered in MassBank
(https://metlin.scripps.edu).
2. Deconvolute the raw data files using Agilent MassHunter soft-
ware package (Unknowns Analysis and Qualitative Analysis
B.07.01) (or AMDIS, see Note 9) and compare the mass
spectral data against a personal compound database library
(Agilent MassHunter PCDL version B.07.00) containing spec-
tra of pure compounds (standards) previously analyzed on the
same instrument. If not compare to NIST library.
3. In the last step, integrate the results from both approaches.

4 Notes

1. NaN3 is toxic and should be handled with caution. NaN3 is

fatal if swallowed or in contact with the skin: wear protective
gloves, protective clothing, and eye protection to prepare this
solution.
 GC-MS Metabolomics of Body Fluids 245

2. Use BSTFA with 1% TMS and pyridine in a fume hood.

3. Alternatively, it is possible to evaporate the sample to dryness
using a SpeedVac.
4. A minimum of 400 μL of pool sample has to be prepared to be
able to aliquot it in 8 vials to be analyzed for batch correction.
If more pool samples or backup samples are needed, adjust the
volume consequently.
5. Heptane should be used in a fume hood to avoid vapors. Care
should be taken to avoid contact with the skin.
6. Adjust the injection volume to obtain a good signal-to-noise
ratio.
7. Adjust the solvent delay to detect small volatile compounds if
necessary.
8. Adjust the XCMS parameters to have a good detection of
chromatographic peaks without duplicates.
9. In case of low-resolution data (from GC/MS simple quadru-
pole), AMDIS can be used alternatively for deconvolution.

Acknowledgments

This work was in part funded by the French National Research

Agency within the MetaboHUB infrastructure (ANR-INBS-
0010).

References
1. A J, Trygg J, Gullberg J, Johansson AI, mass spectrometry-based metabolomics plat-
Jonsson P, Antti H, Marklund SL, Moritz T form for metabolic phenotype analysis. J Biosci
(2005) Extraction and GC/MS analysis of the Bioeng 112(3):292–298. https://doi.org/10.
human blood plasma metabolome. Anal Chem 1016/j.jbiosc.2011.05.001
77(24):8086–8094. https://doi.org/10. 5. Dettmer K, Aronov PA, Hammock BD (2007)
1021/ac051211v Mass spectrometry-based metabolomics. Mass
2. Pasikanti KK, Ho PC, Chan EC (2008) Gas Spectrom Rev 26(1):51–78. https://doi.org/
chromatography/mass spectrometry in meta- 10.1002/mas.20108
bolic profiling of biological fluids. J Chroma- 6. Kanani H, Chrysanthopoulos PK, Klapa MI
togr B Analyt Technol Biomed Life Sci 871 (2008) Standardizing GC-MS metabolomics.
(2):202–211. https://doi.org/10.1016/j. J Chromatogr B Analyt Technol Biomed Life
jchromb.2008.04.033 Sci 871(2):191–201. https://doi.org/10.
3. Gao X, Pujos-Guillot E, Sebedio JL (2010) 1016/j.jchromb.2008.04.049
Development of a quantitative metabolomic 7. Almstetter MF, Oefner PJ, Dettmer K (2012)
approach to study clinical human fecal water Comprehensive two-dimensional gas chroma-
metabolome based on trimethylsilylation deriv- tography in metabolomics. Anal Bioanal Chem
atization and GC/MS analysis. Anal Chem 82 402(6):1993–2013. https://doi.org/10.
(15):6447–6456. https://doi.org/10.1021/ 1007/s00216-011-5630-y
ac1006552 8. Koek MM, Jellema RH, van der Greef J, Tas
4. Tsugawa H, Bamba T, Shinohara M, AC, Hankemeier T (2011) Quantitative meta-
Nishiumi S, Yoshida M, Fukusaki E (2011) bolomics based on gas chromatography mass
Practical non-targeted gas chromatography/ spectrometry: status and perspectives.
246 Carole Migné et al.

Metabolomics 7(3):307–328. https://doi. sciences. J Mass Spectrom: JMS 45

org/10.1007/s11306-010-0254-3 (7):703–714. https://doi.org/10.1002/jms.
9. Mastrangelo A, Ferrarini A, Rey-Stolle F, 1777
Garcia A, Barbas C (2015) From sample treat- 12. Benton HP, Wong DM, Trauger SA, Siuzdak G
ment to biomarker discovery: a tutorial for (2008) XCMS2: processing tandem mass spec-
untargeted metabolomics based on GC-(EI)- trometry data for metabolite identification and
Q-MS. Anal Chim Acta 900:21–35. https:// structural characterization. Anal Chem 80
doi.org/10.1016/j.aca.2015.10.001 (16):6382–6389. https://doi.org/10.1021/
10. Smith CA, Want EJ, O’Maille G, Abagyan R, ac800795f
Siuzdak G (2006) XCMS: processing mass 13. Giacomoni F, Le Corguille G, Monsoor M,
spectrometry data for metabolite profiling Landi M, Pericard P, Petera M, Duperier C,
using nonlinear peak alignment, matching, Tremblay-Franco M, Martin JF, Jacob D,
and identification. Anal Chem 78 Goulitquer S, Thevenot EA, Caron C (2015)
(3):779–787. https://doi.org/10.1021/ Workflow4Metabolomics: a collaborative
ac051437y research infrastructure for computational
11. Horai H, Arita M, Kanaya S, Nihei Y, Ikeda T, metabolomics. Bioinformatics 31
Suwa K, Ojima Y, Tanaka K, Tanaka S, (9):1493–1495. https://doi.org/10.1093/
Aoshima K, Oda Y, Kakazu Y, Kusano M, bioinformatics/btu813
Tohge T, Matsuda F, Sawada Y, Hirai MY, 14. van der Kloet FM, Bobeldijk I, Verheij ER,
Nakanishi H, Ikeda K, Akimoto N, Maoka T, Jellema RH (2009) Analytical error reduction
Takahashi H, Ara T, Sakurai N, Suzuki H, using single point calibration for accurate and
Shibata D, Neumann S, Iida T, Tanaka K, precise metabolomic phenotyping. J Proteome
Funatsu K, Matsuura F, Soga T, Taguchi R, Res 8(11):5132–5141. https://doi.org/10.
Saito K, Nishioka T (2010) MassBank: a public 1021/pr900499r
repository for sharing mass spectral data for life
 Chapter 17

GC-MS Analysis of Short-Chain Fatty Acids in Feces, Cecum

Content, and Blood Samples
Lisa R. Hoving, Marieke Heijink, Vanessa van Harmelen,
Ko Willems van Dijk, and Martin Giera

Abstract
Short-chain fatty acids, the end products of fermentation of dietary fibers by the gut microbiota, have been
shown to exert multiple effects on mammalian metabolism. For the analysis of short-chain fatty acids, gas
chromatography–mass spectrometry is a very powerful and reliable method. Here, we describe a fast,
reliable, and reproducible method for the separation and quantification of short-chain fatty acids in
mouse feces, cecum content, and blood samples (i.e., plasma or serum) using gas chromatography–mass
spectrometry. The short-chain fatty acids analyzed include acetic acid, propionic acid, butyric acid, valeric
acid, hexanoic acid, and heptanoic acid.

Key words Short-chain fatty acids, Gas chromatography–mass spectrometry, GC-MS, Targeted
metabolomics, PFBBr

1 Introduction

Lipids and fatty acids (FAs) are essential molecules in the regulation
and control of various biological functions and play a role in the
onset and progression of disease [1]. FAs with less than 8 carbon
atoms are considered short-chain fatty acids (SCFAs) [2]. SCFAs
(predominantly acetic acid, propionic acid, and butyric acid with
respectively 2, 3, and 4 carbon atoms) are mainly produced in the
colonic lumen after anaerobic fermentation of indigestible carbo-
hydrates by saccharolytic gut bacteria [3]. The link between diet,
the gut microbiota, the production of SCFAs and their role in
human health and disease is an active area of research [4]. This
requires suitable analytical techniques for sensitive and accurate
quantification of SCFAs. One technique traditionally used for the
analysis of small, volatile molecules is gas chromatography–mass
spectrometry (GC-MS). Here, were describe step by step the quan-
titative analysis of the SCFAs acetic acid, propionic acid, butyric
acid, valeric acid, hexanoic acid, and heptanoic acid using GC-MS

Martin Giera (ed.), Clinical Metabolomics: Methods and Protocols, Methods in Molecular Biology, vol. 1730,
https://doi.org/10.1007/978-1-4939-7592-1_17, © Springer Science+Business Media, LLC 2018

247
248 Lisa R. Hoving et al.

in feces, cecum content, as well as in blood samples (i.e., plasma or

serum).
GC-MS is an analytical technique, well suited for the analysis of
SCFAs and other (longer) FAs [5]. However, one critical step in the
GC-MS analysis of FAs is their conversion into suitable volatile
derivatives by derivatization (e.g., by alkylation or silylation)
[6]. Traditionally, FAs are being transformed into their methyl
ester, or trimethylsilyl ester derivatives in GC-MS analysis
[7]. While both approaches work well for longer chain FAs, the
intrinsically low boiling point of the SCFA methyl ester or tri-
methylsilyl ester derivatives results in some issues with their
GC-MS based analysis. For example, the trimethylsilyl ester of
acetic acid roughly presents the same boiling point as the com-
monly used derivatization reagents, leading to severe signal overlap.
Alternatively, for SCFA analysis FAs can be derivatized by the
alkylation reagent pentafluorobenzyl bromide (PFBBr) [8, 9]. The
benzyl bromide group reacts with the carboxylic acid group to form
an ester, allowing analysis as pentafluorobenzyl ester. Additionally,
this so-formed ester presents ideal properties for electron-capture
negative ionization (ECNI), which is a highly selective and sensitive
ionization technique. ECNI allows analysis of the negatively
charged molecular ions, usually detected in the single ion monitor-
ing mode (SIM) on quadrupole based mass spectrometers.
For targeted analysis of SCFAs, ideally isotopically labeled
internal standards (IS) should be used. The use of IS enables
quantitative analysis of biological samples and greatly improves
specificity [7, 9].
Apart from GC-MS, Nuclear magnetic resonance (NMR) spec-
troscopy can be used to analyze SCFAs in feces or cecum content.
An interplatform comparison performed in our lab, comparing
GC-MS and NMR spectroscopy, showed good correlations for
the measurements of SCFA concentrations. However, the advan-
tage of GC-MS over NMR is its higher sensitivity, which is essential
for the analysis of SCFAs at low concentrations such as SCFAs
present in blood.

2 Materials

Use only high purity solvents (preferably LC-MS grade) in order to

prevent elevated background signals (see Note 1). An overview of
the amounts of materials and chemicals is provided in Table 1. If
vendors of different materials are specifically mentioned in this
section, the use of these materials are recommended based on our
previous experiences.
 GC-MS Analysis of Short-chain Fatty Acids 249

Table 1
Chemicals and materials required per sample for the quantification of SCFAs. For every type of
biological matrix, take along three blank samples. Blank samples should be processed in exactly the
same way as biological samples

Biological sample

Chemical/material Calibration series sample Feces Cecum content Plasma/serum

Biological sample – 50 mg 10 mg 10 μL
Water 250 μL 550 μL 690 μL 250 μL
Acetone 250 μL 250 μL 250 μL 250 μL
1 μg/mL IS solution 10 μL 10 μL 10 μL 10 μL
Standards in EtOH 10 μL a
– – –
EtOH – 10 μL 10 μL 10 μL
172 mM PFBBr 100 μL 100 μL 100 μL 100 μL
n-hexane 500 μL 500 μL 500 μL 500 μL
1.5 mL plastic tubes – 2 3 –
Clean steel beads – 2 2 –
Glass autosampler vials 2 2 2 2
Glass autosampler inserts 1 1 1 1
Glass autosampler caps 2 2 2 2
a
For every individual sample of the calibration series, a specific concentration of standards in EtOH is used

2.1 Materials 1. Glass autosampler vials, inserts, and caps. It is important to use
for Sample Preparation the highest quality glass ware. We recommend to use the
following items: Agilent certified 2 mL vials with screw top;
Agilent certified 250 μL inserts with polymer feet; Agilent
screw caps with PTFE/red silicone septum.
2. 1 μg/mL IS solution in ethanol (EtOH) (see Note 2): mix
acetic acid-d4, propionic acid-d6 and butyric acid-d8 and dis-
solve in EtOH to a final concentration of 1 μg/mL. Store at

80  C. Apart from butyric acid, butyric acid-d8 is also used as
the IS for valeric acid, hexanoic acid, and heptanoic acid.
3. Concentration series of SCFA standards in EtOH: use the
SCFA standard mixture (Sigma-Aldrich) and dilute with
EtOH. Prepare concentrations ranging from the lower limit
of quantification (LLOQ) (see Table 2) to 1000 μM. Store at

80  C.
4. 172 mM PFBBr in acetone: add 26.8 μL PFBBr to 1 mL
acetone. Prepare fresh daily.
250 Lisa R. Hoving et al.

Table 2
Overview of FAs. For each SCFA, an indication of the retention time (RT), the m/z-value, an indication
of the LLOQ, and the IS to be used are shown

FA Name RT (min) Monitored m/z (M
) LLOQ (μM)a IS

FA 2:0-d4 Acetic acid-d4 7.19 62 N/A N/A
FA 2:0 Acetic acid 7.22 59 20 FA 2:0-d4
FA 3:0-d6 Propionic acid-d6 7.86 78 N/A N/A
FA 3:0 Propionic acid 7.89 73 5 FA 3:0-d6
FA 4:0-d8 Butyric acid-d8 8.41 94 N/A N/A
FA 4:0 Butyric acid 8.44 87 2 FA 4:0-d8
FA 5:0 Valeric acid 8.99 101 1 FA 4:0-d8
FA 6:0 Hexanoic acid 9.51 115 5 FA 4:0-d8
FA 7:0 Heptanoic acid 9.99 129 1 FA 4:0-d8
a
An indication of the lowest concentration to be included in the calibration series. This LLOQ is determined for every
individual experiment. The calibration series samples are measured twice. A specific concentration is included if signal/
noise >10 and if the accuracy based on the calibration obtained 80 and 120% for both measurements
N/A not applicable

5. Clean steel beads (only for fibrous biological matrices): rinse

3.2 mm stainless steel beads with methanol and dry at room
temperature.

2.2 Materials for GC- 1. GC with split/splitless injector, coupled to a quadrupole mass
MS spectrometer with chemical ionization source.
2. Injection: autosampler (recommended).
3. GC column: use an Agilent VF-5 ms column (5% phenyl-
methyl; 25 m  0.25 mm internal diameter; 0.25 μm film
thickness).
4. Pure helium (99.9990%) and methane (99.9995%) should be
used as carrier and as chemical ionization gas, respectively.

3 Methods

SCFAs are ubiquitous. Hence, extra care has to be taken to prevent

sample contamination. Possible sources of contamination include
environmental air, pipettes, pipette tips, low quality plastics, and
(low purity) solvents (see Note 1). Interday and intraday repeatabil-
ity of the method, validated in fetal calf serum (FCS), is provided in
Table 3.
 GC-MS Analysis of Short-chain Fatty Acids 251

Table 3
Interday and intraday repeatability data in FCS. FCS was spiked with 5 μM and 100 μM SCFA. Acetic
acid, propionic acid and butyric acid in these samples were quantified in triplicate on three different
days using the described method

FCS FCS þ 5 μM SCFA FCS þ 100 μM SCFA

FA Day Mean (μM) RSD Mean (μM) RSD Mean (μM) RSD REa
Acetic acid Intraday 1 108.5 12% 113.4 12% 204.4 3%
4%
Intraday 2 126.5 5% 120.8 4% 180.2 3%
46%
Intraday 3 117.8 11% 121.4 6% 192.5 6%
25%
Interday 117.6 8% 118.5 4% 192.4 6%
25%
Propionic acid Intraday 1 <LLOQ N/A 10.5 18% 101.3 2%
4%
Intraday 2 <LLOQ N/A 9.3 8% 89.4 4%
16%
Intraday 3 8.1 24% 13.5 8% 96.5 2%
13%
Interday N/A N/A 11.1 19% 95.7 6%
11%
Butyric acid Intraday 1 <LLOQ N/A 7.3 24% 96.7 5%
6%
Intraday 2 <LLOQ N/A 5.7 9% 87.9 7%
14%
Intraday 3 <LLOQ N/A 6.7 6% 80.8 10%
22%
Interday <LLOQ N/A 6.6 12% 88.5 9%
14%
a
RE, based on the difference in the determined concentrations. For acetic acid, the RE is determined based on the
difference between the FCS and FCS þ 100 μM SCFA sample. For propionic acid and butyric acid, the RE is determined
based on the difference between the FCS þ 5 μM SCFA and FCS þ 100 μM SCFA sample
N/A not applicable, RE relative error, RSD relative standard deviation

3.1 Sample 1. Facilitate rapid sampling. Store the samples at
80  C upon
Preparation of Feces, collection if the samples are not prepared immediately (see
Cecum Content, Note 3).
and Blood 2. Matrix dependent preprocessing of feces:
Prepare an aqueous extract of feces. Weigh feces (approxi-
mately 50 mg mouse feces) (see Note 4) in a 1.5 mL plastic
tube with 0.1 mg accuracy and add 300 μL water. Homogenize
the sample using a bullet blender: add two clean 3.2 mm steel
beads and blend the sample for 5 min. Centrifuge at 1400  g
for 10 min. Transfer the supernatant to a fresh 1.5 mL plastic
tube.
3. Matrix dependent preprocessing of cecum content:
Weigh cecum content (approximately 10 mg mouse cecum
content) (see Note 4) in a 1.5 mL plastic tube with 0.1 mg
accuracy and add 400 μL water. Homogenize by vortexing.
Use a bullet blender if the material is fibrous: add two clean
3.2 mm steel beads and blend the sample for 5 min. Centrifuge
at 1400  g for 10 min. Transfer the supernatant to a fresh
1.5 mL plastic tube. Dilute the supernatant 1:5 with water in a
total volume of 50 μL using a fresh 1.5 mL plastic tube.
252 Lisa R. Hoving et al.

4. Matrix dependent preprocessing of blood:

Obtain plasma and/or serum. No further preprocessing is
needed.
5. Prepare a glass autosampler vial for every sample:
6. For calibration samples, add 250 μL acetone, 10 μL 1 μg/mL
IS solution, and 10 μL of the calibration series SCFA standards
at the desired concentration. In case of feces or cecum content
analysis, add 10 μL water which is preprocessed exactly the
same as the biological samples. This includes bullet blending
if necessary.
7. For biological samples, add 250 μL acetone (see Note 5), 10 μL
1 μg/mL IS solution in EtOH (see Note 2), 10 μL EtOH (see
Note 6), and 10 μL aqueous feces, 10 μL cecum content
extract, or 10 μL plasma/serum into a glass autosampler vial.
8. For blank samples, add 250 μL acetone, 10 μL 1 μg/mL IS
solution, and 10 μL EtOH into a glass autosampler vial. In case
of feces or cecum content analysis, 10 μL water should be
added which is preprocessed in exact the same way as the
biological samples. This includes bullet blending if necessary.
For every type of biological matrix used in an experiment, three
blank samples should be included.
9. Vortex all samples.
10. Add 100 μL 172 mM PFBBr in acetone (see Note 7). Vortex all
samples.
11. Heat the samples at 60  C for 30 min in a laboratory stove. Let
the samples cool down to room temperature (approximately
15 min) (see Note 8).
12. Add 500 μL n-hexane and 250 μL water to the samples. Shake
the vial in vertical direction for approximately 10 s. Let the
samples rest for 1 min at room temperature.
13. Prepare a new empty glass autosampler vial with a glass insert
for every sample. Transfer 250 μL of the n-hexane (upper layer)
into the glass insert.

3.2 GC-MS Analysis 1. Inject 1 μL in the GC-MS, splitless at 280  C.

2. Use helium as carrier gas at a constant flow rate of 1.20 mL/
min.
3. Use the following temperature gradient: 1 min at 40  C, linear
increase at 40  C/min to 60  C, held for 3 min at 60  C, linear
increase at 25  C/min to 210  C, linear increase at 40  C/min
to 315  C, and held for 3 min at 315  C.
4. Set the transfer line temperature at 280  C.
 GC-MS Analysis of Short-chain Fatty Acids 253

5. Keep the ionization source temperature at 280  C.

6. Use methane as chemical ionization gas at approximately
15 psi.
7. Detect ions obtained in the negative mode using SIM (see
Notes 9 and 10). Table 2 provides the m/z-values to be moni-
tored and an indication of retention times (RT). As a conse-
quence of small chromatographic differences (e.g., GC column
length), the exact RT varies between various GC systems.
Hence, calibration using external standards is mandatory.

3.3 Data Analysis 1. Integrate the obtained signal (see Note 11).
2. Calculate the relative retention time (RRT) and area ratios
using the respective IS (see Table 2) (see Notes 12 and 13).
3. Determine the slope and LLOQ for every SCFA by performing
linear regression. It is recommended to use a weighing factor of
1/x2 [10].
4. Calculate the SCFA concentrations by using the area ratios
obtained from the biological samples, average signal of the
blank samples as intercept (see Note 14), and the slopes
obtained from the analysis of the calibration series samples.
Take into account the sample dilution for feces and cecum
content.

4 Notes

1. SCFAs (especially acetic acid and propionic acid) usually show

high background signals, resulting in a relatively high LLOQ
(see Fig. 1). SCFA background signals can be diminished by
using high purity solvents (preferably LC-MS grade). Addi-
tionally, use glass vials for organic solvents. For plastic tubes,
we strongly recommend to use Eppendorf polypropylene
tubes.
2. The IS signal should be present in every sample. The IS is used
to correct for differences in sample preparation between the
samples. Use exactly the same batch of 1 μg/mL IS solution in
EtOH for the entire experiment, as minor differences in IS
composition potentially translate into systematic under- or
overestimation of SCFAs in samples.
3. Collect the biological samples as quick as possible and store the
samples at
80  C. Levels of SCFAs within biological samples
are vulnerable for change, especially when the collection is
performed slowly or when samples are improperly stored.
SCFAs can evaporate from the samples or SCFAs from the air
can contaminate the samples.
254 Lisa R. Hoving et al.

Fig. 1 Background signal of SCFAs. Several SCFAs usually show a high

background signal. As a consequence, the LLOQ for these SCFAs is dependent
on the background obtained. In our experience, the background signal of acetic
acid is higher than that of propionic acid, which in turn is higher than the
background signal of butyric acid. The dashed lines in the graph show the
extent of the background signals

4. The amount of sample that is required for the analysis of SCFAs

might vary between biological samples from different species.
5. Acetone facilitates the precipitation of proteins.
6. The addition of 10 μL EtOH to the samples ensures that the
solvents of the biological samples are matched to the solvents in
the calibration samples.
7. Within this protocol no base is added to catalyze the derivati-
zation reaction, since the addition of base can severely increase
SCFA background signal [8].
8. n-Hexane is added after the samples have been cooled down in
order to prevent evaporation and spilling.
9. Sensitivity is higher when the MS is operated in SIM mode as
compared to the full scan mode. However, the full scan mode
can be useful to detect FAs that are not incorporated in the SIM
method, or to determine the RT of a specific SCFA. If one
decides to operate in full scan mode, an m/z-range of 50–150 is
recommended to be used.
10. For isotopolog analysis, either m/z-values corresponding to
isotopologs can be added to the SIM method (e.g., M0, M1,
M2, etc. for every SCFA) [8], or the MS can be operated in
scheduled scan mode (e.g., scan window including m/z-values
corresponding to M0, M1, M2, etc., for every SCFA).
11. Pyruvic acid has the same mass and almost the same RT as
butyric acid and both acids are derivatized by PFBBr. Consec-
utively, care has to be taken when both analytes are simulta-
neously present in the sample (see Fig. 2). Particularly in plasma
 GC-MS Analysis of Short-chain Fatty Acids 255

Fig. 2 Signal interference between butyric acid (RT ¼ 8.22 min) and pyruvic acid (RT ¼ 8.25 min). Butyric
acid and pyruvic acid are eluting closely while being monitored in the same SIM trace. Particularly in plasma
(a) and serum samples, pyruvic acid and butyric acid are simultaneously present. In feces and cecum content
(b), pyruvic acid is usually not detected

and serum samples, pyruvic acid and butyric acid are simulta-
neously present. In feces and cecum content, pyruvic acid is
usually not detected.

retention time analyte

12. RRT ¼
retention time I S
area analyte
13. Area ratio ¼
area IS
14. The blank samples reflect the background signal of the
biological samples more accurately than the intercept obtained
from the linear regression of the calibration series samples.
Therefore, use the average area ratio of the blank samples as
background signal/intercept to calculate the concentrations of
the biological samples. Use the following formula:

area ratio
average area ratio blank samples

Concentration ¼
slope
References
1. Siri-Tarino PW, Chiu S, Bergeron N et al 4. O’Keefe SJD (2016) Diet, microorganisms and
(2015) Saturated fats versus polyunsaturated their metabolites, and colon cancer. Nat Rev
fats versus carbohydrates for cardiovascular dis- Gastroenterol Hepatol 13:691–706
ease prevention and treatment. Annu Rev Nutr 5. Dołowy M, Pyka A (2015) Chromatographic
35:517–543 methods in the separation of long-chain mono-
2. Desbois AP, Smith VJ (2010) Antibacterial free and polyunsaturated fatty acids. J Chem
fatty acids: activities, mechanisms of action and 2015:20
biotechnological potential. Appl Microbiol 6. Gao P, Xu G (2015) Mass-spectrometry-based
Biotechnol 85(6):1629–1642 microbial metabolomics: recent developments
3. Cummings JH, Pomare EW, Branch WJ et al and applications. Anal Bioanal Chem
(1987) Short chain fatty acids in human large 407:669–680
intestine, portal, hepatic and venous blood. 7. Kloos D-P, Gay E, Lingeman H et al (2014)
Gut 28:1221–1227 Comprehensive gas chromatography-electron
ionisation mass spectrometric analysis of fatty
256 Lisa R. Hoving et al.

acids and sterols using sequential one-pot sily- analysis of fatty acids in biological samples by
lation: quantification and isotopologue analy- gas chromatography-mass spectrometry. Bio-
sis. Rapid Commun Mass Spectrom chim Biophys Acta 1811:648–656
28:1507–1514 10. Gu H, Liu G, Wang J et al (2014) Selecting the
8. Tomcik K, Ibarra RA, Sadhukhan S et al (2011) correct weighting factors for linear and qua-
Isotopomer enrichment assay for very short dratic calibration curves with least-squares
chain fatty acids and its metabolic applications. regression algorithm in bioanalytical LC-MS/
Anal Biochem 410:110–117 MS assays and impacts of using incorrect
9. Quehenberger O, Armando AM, Dennis EA weighting factors on curve stability, data qual-
(2011) High sensitivity quantitative lipidomics ity, and assay perfo. Anal Chem 86:8959–8966
 Chapter 18

GC-MS Analysis of Medium- and Long-Chain Fatty Acids

in Blood Samples
Lisa R. Hoving, Marieke Heijink, Vanessa van Harmelen,
Ko Willems van Dijk, and Martin Giera

Abstract
Our body contains a wide variety of fatty acids that differ in chain length, the degree of unsaturation, and
location of the double bonds. As the various fatty acids play distinct roles in health and disease, methods
that can specifically determine the fatty acid profile are needed for fundamental and clinical studies. Here we
describe a method for the separation and quantification of fatty acids ranging from 8 to 24 carbon chain
lengths in blood samples using gas chromatography-mass spectrometry following derivatization using
pentafluorobenzyl bromide. This method quantitatively monitors fatty acid composition in a manner that
satisfies the requirements for comprehensiveness, sensitivity, and accuracy.

Key words Medium-chain fatty acids, Long-chain fatty acids, Gas chromatography-mass spectrome-
try (GC-MS), Targeted metabolomics, PFBBr

1 Introduction

Lipids and fatty acids (FAs) are present in all organisms and consti-
tute essential structural elements of biological membranes, regulate
and control cellular function, and are involved in the onset and
progression of various diseases [1]. The majority of FAs are present
as esters in lipids, such as triacylglycerols, sterol esters, and phos-
pholipids. Only a small fraction is nonesterified, generally termed
free FAs (FFAs) [2]. The role of FAs in health and disease has
gained extensive interest. FAs differ in chain length, the degree of
unsaturation, and location of the double bond(s). As various types
of FAs have different associations with disease outcomes, the assess-
ment of the FA composition in biological samples may provide
suitable information. Therefore, great effort has been put in the
development of comprehensive, sensitive, and reliable methodolo-
gies to quantify the FA composition. Here we describe step-by-step
the quantitative analysis of the FA profile in plasma using GC-MS,
measuring FAs with a chain length ranging from 8 to 24 carbons.

Martin Giera (ed.), Clinical Metabolomics: Methods and Protocols, Methods in Molecular Biology, vol. 1730,
https://doi.org/10.1007/978-1-4939-7592-1_18, © Springer Science+Business Media, LLC 2018

257
258 Lisa R. Hoving et al.

GC-MS is an analytical technique that is well-suited for the

analysis of the total amount of FAs as well as the FA composition
within a sample [3]. As blood contains both esterified FAs and
FFAs, a separate hydrolysis step is required during sample prepara-
tion to determine the total FA composition. For the analysis of FAs
by GC-MS, it is of importance to convert FAs into suitable volatile
derivatives by derivatization (e.g., alkylation or silylation) [4]. Tra-
ditionally in GC-MS analysis, FAs were being transformed into
their methyl ester or trimethylsilyl ester derivatives [5]. Alterna-
tively, pentafluorobenzyl bromide (PFBBr) can be used to deriva-
tize FAs. It has been successfully applied for the analysis of FAs with
different chain lengths [6–9]. The benzyl bromide group reacts
with the carboxylic acid group to form a pentafluorobenzyl ester. In
addition, this pentafluorobenzyl ester contains ideal properties for
electron capture negative ionization (ECNI), a highly selective and
sensitive ionization technique. The combination of PFBBr deriva-
tization and ECNI ionization allows for the analysis of negatively
charged molecular ions. These ions are usually detected in the
single ion monitoring (SIM) mode on quadrupole-based mass
spectrometers. Isotopically labeled internal standards (IS) have to
be used for quantitative analysis of FAs by GC-MS. The addition of
IS in GC-MS analysis enables quantitative analysis of biological
samples and greatly improves detection specificity [10].

2 Materials

Use only high-purity solvents (preferably LC-MS grade) in order to

prevent increased background signals (see Note 1). If vendors from
different materials are mentioned in this method section, the use of
these chemicals is recommended based on our previous experi-
ences. The use of 10 M NaOH forms an exception. In order for
the method to succeed, it is urged to use the specific items men-
tioned in the Materials section. An overview of the amounts of
materials and chemicals is provided in Table 1.

2.1 Materials 1. Glass autosampler vials, inserts, and caps. It is recommended to

for Sample Preparation use Agilent certified 2 mL vials with screw top; Agilent certified
250 μL inserts with polymer feet; and Agilent screw caps with
PTFE/red silicone septum.
2. 1 μg/mL IS solution in ethanol (EtOH) (see Note 2): accurately
weigh decanoic acid-d19, palmitic acid-d31, and arachidonic
acid-d8, and dissolve in EtOH to a final concentration of
1 μg/mL. Store at
80  C.
3. Concentration series of FA standards in EtOH: use GLC refer-
ence standard 85 mix (Nu-Chek Prep), eicosapentaenoic acid
(Cayman), docosapentaenoic acid (Cayman), and docosahexae-
noic acid (Cayman), and serially dilute using EtOH. Prepare
 GC-MS Analysis of Fatty Acids 259

Table 1
Chemicals and materials needed per sample for the quantification of FAs

Chemical/material Calibration series sample Plasma/serum sample

Plasma/serum – 10 μL
Acetone 250 μL 250 μL
10 M NaOH – 10 μL
1 μg/mL IS solution 10 μL 10 μL
Standards in EtOH 10 μL a
–
EtOH – 10 μL
172 mM PFBBr 100 μL 100 μL
n-Hexane 500 μL 500 μL
Water 250 μL 250 μL
Glass autosampler vials 2 2
Glass autosampler inserts 1 1
Glass autosampler caps 2 2
For every batch of samples, take along three blank samples. Blank samples should be processed in exactly the same way as
biological samples
a
For every individual sample of the calibration series, a specific concentration of standards in EtOH is used

concentrations ranging from the lower limit of quantification

(LLOQ) (see Table 2) to 50 μg/mL. Store at
80  C.
4. 172 mM PFBBr in acetone: add 26.8 μL PFBBr to 1 mL ace-
tone. Prepare fresh daily.
5. 10 M NaOH in water. In order for the method to succeed, it is
urged to use a prepared solution from Sigma-Aldrich (Art.
No. 72068) (see Note 3).

2.2 Materials 1. GC with split/splitless injector, coupled to a quadrupole mass

for GC-MS spectrometer with chemical ionization source.
2. Injection: autosampler (recommended).
3. GC column: use an Agilent VF-5 ms column (5% phenylmethyl;
25 m  0.25 mm internal diameter; 0.25 μm film thickness).
4. Pure helium (99.9990%) and methane (99.9995%) are used as
carrier and chemical ionization gas, respectively.

3 Methods

Palmitic acid and stearic acid are ubiquitous. Hence, extra care has
to be taken to prevent sample contamination. Sources of contami-
nation include low-quality plastics and (low-purity) solvents (see
Note 1).
260 Lisa R. Hoving et al.

Table 2
Overview of FAs

RT Monitored m/z LLOQ

FA Name (min) (M
) (ng/mL)a IS
FA 08:0 Octanoic acid 10.14 143.1 200 FA 10:0-d19
FA 10:0 Decanoic acid 11.01 171.1 50 FA 10:0-d19
FA 10:0-d19 Decanoic acid-d19 10.92 190.3 N/A N/A
FA 11:0 Undecanoic acid 11.42 185.2 50 FA 10:0-d19
FA 12:0 Lauric acid 11.87 199.2 100 FA 10:0-d19
FA 13:0 Tridecanoic acid 12.40 213.2 50 FA 10:0-d19
FA 14:0 Myristic acid 13.03 227.2 50 FA 16:0-d31
FA 14:1 (n-5) Myristoleic acid 12.97 225.2 20 FA 16:0-d31
FA 15:0 Pentadecanoic acid 13.78 241.2 50 FA 16:0-d31
FA 15:1 (n-5) 10-Pentadecenoic acid 13.72 239.2 10 FA 16:0-d31
FA 16:0 Palmitic acid 14.68 255.2 500 FA 16:0-d31
FA 16:0-d31 Palmitic acid-d31 14.43 286.4 N/A N/A
FA 16:1 (n-7) Palmitoleic acid 14.50 253.2 50 FA 16:0-d31
FA 17:0 Heptadecanoic acid 15.72 269.3 20 FA 16:0-d31
FA 17:1 (n-7) 10-Heptadecenoic acid 15.53 267.2 10 FA 16:0-d31
FA 18:0 Stearic acid 16.54 283.3 500 FA 16:0-d31
FA 18:1 (n-9) Oleic acid 16.37 281.3 100 FA 16:0-d31
cis
FA 18:1 (n-9) Elaidic acid 16.41 281.3 50 FA 16:0-d31
trans
FA 18:2 (n-6) Linoleic acid 16.34 279.2 50 FA 16:0-d31
FA 18:3 (n-6) Gamma-linolenic acid (GLA) 16.15 277.2 50 FA 16:0-d31
FA 18:3 (n-3) Alpha-linolenic acid (ALA) 16.40 277.2 50 FA 16:0-d31
FA 20:0 Arachidic acid 17.66 311.3 50 FA 20:4-d8
FA 20:1 (n-9) 11-Eicosenoic acid 17.55 309.3 20 FA 20:4-d8
FA 20:2 (n-6) 11,14-Eicosadienoic acid 17.54 307.3 10 FA 20:4-d8
FA 20:3 (n-6) Homo-Gamma linolenic acid 17.43 305.3 10 FA 20:4-d8
(DGLA)
FA 20:3 (n-3) 11,14,17-Eicosatrienoic acid 17.58 305.3 10 FA 20:4-d8
FA 20:4 (n-6) Arachidonic acid (AA) 17.28 303.2 10 FA 20:4-d8
FA 20:4-d8 Arachidonic acid-d8 (AA-d8) 17.26 311.3 N/A N/A
FA 20:5 (n-3) Eicosapentaenoic acid (EPA) 17.33 301.2 10 FA 20:4-d8
(continued)
 GC-MS Analysis of Fatty Acids 261

Table 2
(continued)

RT Monitored m/z LLOQ

FA Name (min) (M
) (ng/mL)a IS
FA 22:0 Behenic acid 18.48 339.3 50 FA 20:4-d8
FA 22:1 (n-9) Erucic acid 18.40 337.3 20 FA 20:4-d8
FA 22:2 (n-6) 13,16-Docosadienoic acid 18.39 335.3 10 FA 20:4-d8
FA 22:4 (n-6) Adrenic acid (AdA) 18.21 331.3 10 FA 20:4-d8
FA 22:5 (n-3) Docosapentaenoic acid (DPA) 18.25 329.3 10 FA 20:4-d8
FA 22:6 (n-3) Docosahexaenoic acid (DHA) 18.15 327.2 20 FA 20:4-d8
FA 24:1 (n-9) Nervonic acid 19.28 365.4 20 FA 20:4-d8
For each FA, an indication of the retention time (RT), the m/z value, an indication of the LLOQ, and the IS to be used
are shown. N/A not applicable
a
An indication of the lowest concentration to be included in the calibration series. This LLOQ is determined for every
individual experiment. The calibration series samples are measured twice. A specific concentration is included if signal/
noise >10 and if the accuracy based on the calibration obtained 80 and 120% for both measurements

3.1 Sample 1. Facilitate rapid sampling. Store samples at
80  C upon collec-
Preparation tion if the samples are not prepared immediately (see Note 4).
2. Prepare a glass autosampler vial for every sample: for calibration
samples, add 250 μL acetone and 10 μL of your calibration series
FA standards at the desired concentration. An indication of the
expected LLOQ for every FA is provided in Table 2. For
biological samples, add 250 μL acetone (see Note 5), 10 μL
EtOH (see Note 6), and 10 μL plasma or serum into a glass
autosampler vial. For blank samples, add 250 μL acetone and
10 μL EtOH into a glass autosampler vial. For every experiment,
three blank samples should be included.
3. Hydrolyze the biological and blank samples (see Note 7): add
10 μL 10 M NaOH to the biological and blank samples. Vortex
all samples. Heat the biological and blank samples at 60  C for
30 min in a laboratory stove. Let the samples cool down to room
temperature (approximately 15 min).
4. Add 10 μL 1 μg/mL IS solution (see Notes 2 and 8) to every
sample and vortex all samples.
5. Add 100 μL 172 mM PFBBr in acetone (see Note 9). Vortex the
samples.
6. Heat the samples at 60  C for 30 min in a laboratory stove. Let
the samples cool down to room temperature (approximately
15 min) (see Note 10).
7. Add 500 μL n-hexane and 250 μL water to the samples. Shake
vigorously in vertical direction of the vial for 10 s. Let the
samples rest for 1 min at room temperature.
262 Lisa R. Hoving et al.

8. Prepare a new empty glass autosampler vial with a glass insert for
every sample. Transfer 250 μL of the n-hexane (upper layer) into
the glass insert.

3.2 GC-MS Analysis 1. Inject 1 μL in the GC-MS, splitless at 280  C.

2. Use helium as carrier gas at a constant flow rate of 1.20 mL/
min.
3. Use the following temperature gradient: 1 min at 50  C, linear
increase at 40  C/min to 60  C, held for 3 min at 60  C, linear
increase at 25  C/min to 237  C, linear increase at 3  C/min to
250  C, linear increase at 25  C/min to 315  C, held for
1.55 min at 315  C.
4. Set the transfer line temperature at 280  C.
5. Keep the ionization source temperature at 280  C.
6. Use methane as chemical ionization gas at approximately 15 psi.
7. Detect ions obtained in the negative mode using SIM analysis
(see Notes 11 and 12). Table 2 provides the m/z values to be
monitored and an indication of retention times (RT). As a
consequence of small chromatographic differences (e.g., GC
column length), the exact RT varies between various GC sys-
tems. Hence, calibration using external standards is mandatory.

3.3 Data Analysis 1. Integrate the obtained signal (see Note 13).
2. Calculate the relative retention times (RRT) and area ratios
using the respective IS (see Table 2) (see Notes 14 and 15).
3. Determine the slope and LLOQ for every FA by performing
linear regression. It is recommended to use a weighing factor of
1/x2 [11].
4. Calculate the FA concentrations by using the area ratios
obtained from the biological samples, average signal of the
blank samples as intercept (see Note 16), and the slopes obtained
from the analysis of the calibration series samples.

4 Notes

1. Palmitic acid and stearic acid usually give high background

signals, resulting in a relatively high LLOQ (Fig. 1). Sources
of these FAs include solvents and plastic containers of inferior
quality. Background signals of these FAs can be diminished by
using high-purity solvents (preferably LC-MS grade). Addi-
tionally, use glass vials for organic solvents.
2. The IS signal should be present in every sample. The IS are
used to correct for differences in sample preparation between
the samples. Use exactly the same batch of 1 μg/mL IS
 GC-MS Analysis of Fatty Acids 263

Fig. 1 Background signal of palmitic acid and stearic acid. Palmitic acid and
stearic acid usually show a high background signal. As a consequence, the LLOQ
for these FAs is higher than for FAs which do not show a high background signal
like docosahexaenoic acid (DHA). The dashed lines in the graph show the size of
the background signals

solution in EtOH for the entire experiment, as minor differ-

ences in IS composition might translate into systematic under-
or overestimation of FAs in samples.
3. Sodium hydroxide pellets can be heavily contaminated by FAs.
4. Collect the biological samples as quickly as possible, and store
the samples at
80  C. Levels of FAs can change upon sample
collection if the collection is performed slowly or when samples
are stored improperly by auto-oxidation of polyunsaturated
FAs and enzymatic hydroxylation.
5. Acetone facilitates the precipitation of proteins.
6. The addition of 10 μL EtOH to the samples ensures that the
solvents of the biological samples are matched to the solvents in
the calibration samples.
7. No esterified FAs are present in the calibration series samples.
Therefore, no hydrolyzation step is needed.
8. Under highly alkaline conditions, risk of hydrogen–deuterium
exchange exists [12]. Therefore, the IS need to be added after
hydrolysis.
9. Within this protocol no base is added to catalyze the derivati-
zation reaction, since the addition of base can severely increase
FA background [7].
10. n-Hexane is added after the samples have been cooled down in
order to prevent evaporation and spilling.
11. Sensitivity is higher when the mass spectrometer is operated in
SIM mode as compared to the full scan mode. However, the
full scan mode can be very useful to detect FAs that are not
incorporated in the SIM method or to determine the RT of a
264 Lisa R. Hoving et al.

specific FA. If one decides to operate in full scan mode, an m/z

range of 100–400 can be used.
12. For isotopologue analysis, either m/z values corresponding to
isotopologues can be added to the SIM method (e.g., M0, M1,
M2, etc. for every FA) [7] or the MS can be operated in
scheduled scan mode (e.g., scan window including m/z values
corresponding to M0, M1, M2, etc., for every FA).
13. Oleic and elaidic acids have equal masses and are not baseline
separated. In some cases, it is therefore not possible to accu-
rately and precisely quantify one or both of these two FAs
within the same sample. FA 20:1 (n-9) and FA 22:1 (n-9),
co-elute with, respectively, FA 20:2 (n-6) and FA 22:2 (n-6).
As a consequence, the M2 isotopes (containing 2 13C instead
of 12C) of FA 20:2 (n-6) and FA 22:2 (n-6) might contaminate
the FA 20:1 (n-9) and FA 22:1 (n-9) signal.
retention time analyte
14. RRT ¼ .
retention time IS
area analyte
15. Area ratio ¼ :
area IS
16. The blank samples reflect the background signal of the
biological samples more accurately than the intercept obtained
from the linear regression of the calibration series samples. There-
fore, use the average area ratio of the blank samples as background
signal/intercept to calculate the concentrations of the biological
samples. Use the following formula: concentration ¼
area ratio
average area ratio blank samples
:
slope

References
1. Siri-Tarino PW, Chiu S, Bergeron N et al acids and sterols using sequential one-pot sily-
(2015) Saturated fats versus polyunsaturated lation: quantification and isotopologue analy-
fats versus carbohydrates for cardiovascular dis- sis. Rapid Commun Mass Spectrom
ease prevention and treatment. Annu Rev Nutr 28:1507–1514
35:517–543 6. Quehenberger O, Armando AM, Dennis EA
2. van Meer G, Voelker DR, Feigenson GW (2011) High sensitivity quantitative lipidomics
(2008) Membrane lipids: where they are and analysis of fatty acids in biological samples by
how they behave. Nat Rev Mol Cell Biol gas chromatography-mass spectrometry. Bio-
9:112–124 chim Biophys Acta Mol Cell Biol Lipids
3. Dołowy M, Pyka A (2015) Chromatographic 1811:648–656
methods in the separation of long-chain mono- 7. Tomcik K, Ibarra RA, Sadhukhan S et al (2011)
and polyunsaturated fatty acids. J Chem 2015 Isotopomer enrichment assay for very short
4. Gao P, Xu G (2015) Mass-spectrometry-based chain fatty acids and its metabolic applications.
microbial metabolomics: recent developments Anal Biochem 410:110–117
and applications. Anal Bioanal Chem 8. Quehenberger O, Armando A, Dumlao D et al
407:669–680 (2008) Lipidomics analysis of essential fatty
5. Kloos D-P, Gay E, Lingeman H et al (2014) acids in macrophages. Prostaglandins Leukot
Comprehensive gas chromatography-electron Essent Fat Acids 79:123–129
ionisation mass spectrometric analysis of fatty
 GC-MS Analysis of Fatty Acids 265

9. Pawlosky RJ, Sprecher HW, Salem N (1992) quadratic calibration curves with least-squares
High sensitivity negative ion GC-MS method regression algorithm in bioanalytical LC-MS/
for detection of desaturated and chain- MS assays and impacts of using incorrect
elongated products of deuterated linoleic and weighting factors on curve stability, data qual-
linolenic acids. J Lipid Res 33:1711–1717 ity, and assay perfo. Anal Chem 86:8959–8966
10. Kloos D, Lingeman H, Mayboroda OA et al 12. Lee J, Jang E-S, Kim B (2013) Development of
(2014) Analysis of biologically-active, endoge- isotope dilution-liquid chromatography/mass
nous carboxylic acids based on chromatography- spectrometry combined with standard addition
mass spectrometry. TrAC Trends Anal Chem techniques for the accurate determination of
61:17–28 tocopherols in infant formula. Anal Chim
11. Gu H, Liu G, Wang J et al (2014) Selecting the Acta 787:132–139
correct weighting factors for linear and
 Chapter 19

Analysis of Oxysterols
Fabien Riols and Justine Bertrand-Michel

Abstract
Oxysterols are oxygenated derivatives of cholesterol formed in the human body or ingested in the diet. By
modulating the activity of many proteins (for instance, liver X receptors, oxysterol-binding proteins, some
ATP-binding cassette transporters), oxysterols can affect many cellular functions and influence various
physiological processes (e.g., cholesterol metabolism, membrane fluidity regulation, intracellular signaling
pathways). Due to their crucial role, it is important to be able to quantify them in pathological conditions.
The method described here permits to measure the content of oxysterol in plasma, cell, or media using
GC-MS.

Key words Oxysterol, Plasma, Gas chromatography mass spectrometry

1 Introduction

Cholesterol has many functions; for instance, it affects biophysical

properties of membranes and is a precursor to hormone synthesis.
These actions are governed by enzymatic pathways that modify the
sterol nucleus or the isooctyl tail [1]. The addition of oxygen to the
cholesterol backbone produces its derivatives known as oxysterols.
In addition to having an enzymatic origin, oxysterols can be formed
in the absence of enzymatic catalysis in a pathway usually termed
“autoxidation,” [2] which is known for almost a century and
observed under various experimental conditions. These metabolic
intermediates are involved in cholesterol homeostasis where they
play a role as ligands to nuclear and G protein-coupled receptors.
These molecular classes are implicated in the etiology of a diverse
array of diseases including autoimmune disease, Parkinson’s dis-
ease, motor neuron disease, breast cancer, the lysosomal storage
disease Niemann-Pick type C, and the autosomal recessive disorder
Smith-Lemli-Opitz syndrome. For example, the cholesterol meta-
bolites 25-hydroxycholesterol and 7α, 25-dihydroxycholesterol
have recently been shown to play important signaling roles in the
immune system [3, 4]. 24-hydroxycholesterol has been shown to

Martin Giera (ed.), Clinical Metabolomics: Methods and Protocols, Methods in Molecular Biology, vol. 1730,
https://doi.org/10.1007/978-1-4939-7592-1_19, © Springer Science+Business Media, LLC 2018

267
268 Fabien Riols and Justine Bertrand-Michel

be involved in cholesterol turnover in the brain, and it plays a role in

memory [5] and glucose metabolism [6]. In Niemann-Pick type
C1 (NPC1) disease which is a rare progressive neurodegenerative
disorder characterized by accumulation of cholesterol in the endo-
lysosomes, oxysterols are accumulated during the disease progres-
sion [7]. A similar case is observed in Smith-Lemli-Opitz
syndrome, another inborn error of sterol metabolism [8]. Because
of their large implication in physiopathology, they are good poten-
tial biomarkers, so they need to be readily measureable in biological
fluids, in organs, and in cells. Gas chromatography mass spectrom-
etry (GC-MS) with the use of stable isotope-labeled standards is the
gold standard methodology for the analysis of neutral cholesterol
metabolites [9] and for acidic oxysterols after appropriate derivati-
zation [10]. As these molecules are often isobaric species and also
their steroid nucleus is quite hard to fragment, separation is key.
The esterification rate of oxysterols is approximately 70%; in turn a
saponification step is needed to measure total oxysterol levels. Due
to the possibility of autoxidative oxysterol production, samples
should be stored at low temperature (
80  C), under inert atmo-
sphere, ideally with argon or nitrogen gas. Samples will then be
submitted to organic liquid-liquid extraction and to saponification
(or not) depending on the pool of oxysterols targeted. Oxysterols
are in minority compared to cholesterol; hence, it is important to
remove cholesterol before analysis, in order to minimize matrix
effects. Extracts are pre-purified by solid-phase extraction (SPE)
and finally submitted to derivatization to improve GC-MS separa-
tion and detection. Two types of detectors can be used: a single
quadrupole where the SIM (single ion monitoring) mode will be
chosen to have the best sensitivity or a triple quadrupole where the
SRM (single reaction monitoring) mode will be preferred due to a
better signal-to-noise ratio.

2 Materials

All solvents have to be prepared with ultrapure water (18 MΩ-cm at

25  C) and analytical grade reagents. All solvents used are at least of
HPLC grade.

2.1 Cell and Media 1. Directly after collection, snap freeze cell or medium samples
Collection with liquid nitrogen (see Note 1), and store at
80  C for a
maximum period of 1 month.
2. 5 mM ethylene glycol tetraacetic acid (EgTA) solution: dissolve
9.51 mg of EgTA in 10 mL of NaOH 10 M, add 10 mL of
water, adjust the pH to 7.4 with concentrated hydrochloric
acid (HCl), and complete the volume to 50 mL with water.
 Analysis of Oxysterols 269

Take 10 mL of this solution, and dilute it with 990 mL of water

to obtain the 5 mM solution of EgTA.
3. Methanol (MeOH)/EgTA solution (2:1; v/v): 10 mL of
EgTA 5 mM is diluted with 20 mL of methanol (see Note 2).

2.2 Blood Collection 1. Collect blood sample into commercial blood collection tube
containing ethylenediaminetetraacetic acid (EDTA). Centri-
fuge immediately at 1000  g for 10 min at 4  C.
2. Collect plasma and snap freeze it with liquid nitrogen (see Note 1),
and store at
80  C for a maximum of 3 months.

2.3 Standards 1. Primary standards: the following standard materials are

needed—7α-hydroxycholesterol, 7β-hydroxycholesterol,
4β-hydroxycholesterol, 5α,6α-epoxycholestanol, 24(S)-
hydroxycholesterol, 25-hydroxycholesterol, 7-ketocholesterol,
and 27-hydroxycholesterol.
2. Internal standard: 19-hydroxycholesterol serves as the internal
standard.
3. Standard stock solutions: dilute each standard in a glass vial to
1000 ng/μL using MeOH, and store at
80  C after flushing
with nitrogen gas.
4. Standard mix solution for calibration: pipet 50 μL of each
standard stock solution to a glass vial, and adjust the final
volume to 1000 μL with MeOH in order to obtain a final
concentration of 50 ng/μL per standard.
5. Internal standard solution for extractions: pipet (or transfer)
50 μL of internal standard stock solution to a glass vial, and
adjust the final volume to 1000 μL with MeOH obtaining a
final concentration of 50 ng/μL.

2.4 Extraction 1. Extraction tube: Pyrex™ Round Bottom Threaded Culture

Material Tubes (13  100 mm) with cap PTFE faced rubber insert
should be used.
2. Brine (NaCl) solution: prepare a saturated solution of NaCl in
water.

2.5 Solid-Phase 1. Typically use a Waters Positive Pressure 96-SPE Processor.

Extraction 2. The following solid-phase extraction cartridges should be used
(SPE): CHROMABOND® Multi 96 SiOH (96  100 mg).
3. For sample collection, use a 96-well 2 mL square collection
plate.
4. 0.5% (v/v) 2-propanol in n-heptane: measure 199 mL
n-heptane into a 250 mL graduated cylinder, then add 1 mL
2-propanol, and store at ambient temperature in a 250 mL
reagent bottle.
270 Fabien Riols and Justine Bertrand-Michel

5. 30% (v/v) 2-propanol in n-heptane: measure 60 mL n-heptane

and 140 mL 2-propanol in a 250 mL graduated cylinder, and
store at ambient temperature in a 250 mL reagent bottle.

2.6 Alkaline 1. 0.5 M potassium hydroxide (KOH) methanolic solution: dis-

Hydrolysis solve 14.03 g of KOH in 500 mL of MeOH.
and Derivatization 2. 0.5 M phosphoric acid solution: add 339 μL of phosphoric acid
85% with 9.661 mL of water.
3. 50% (v/v) BSTFA solution: mix 25 μL N,O-Bis(trimethylsilyl)
trifluoroacetamide þ TMCS (99:1) with 25 μL of acetonitrile
(ACN).

2.7 Calibration 1. Serially dilute the calibration stock solution 1:1, starting with
10 μL of standard mix solution (50 ng/μL) mixed with 10 μL
of MeOH, and continue seven times.
2. Dry samples under a constant stream of nitrogen, and dissolve
in 20 μL of n-heptane; this will result in the concentrations
given in Table 1.

2.8 Gas 1. GC-MS single quadrupole: exemplary a Thermo Fisher Trace

Chromatography Mass Gas Chromatograph coupled to a Thermo Fisher ISQ mass
Spectrometry selective detector; AI 3000 Injector Autosampler (Thermo
Fisher Scientific, Waltham, MA, USA) can be used.
2. GC-MS triple quadrupole: exemplary a Thermo Fisher Trace
Gas Chromatograph coupled to a Thermo Fisher TSQ 8000
mass selective; AI 1310 Injector Autosampler (Thermo Fisher
Scientific, Waltham, MA, USA) can be used.
3. Gas chromatography column: use a HP-5MS capillary column
(30 m, 0.25 mm, 0.25 μm phase thickness) (see Note 2).

Table 1
Calibration standards

Calibration point [ng/μL] of each oxysterol Final volume (μL)

8 12.5 20
7 6.25 20
6 3.12 20
5 1.563 20
4 0.781 20
3 0.391 20
2 0.195 20
1 0.098 20
 Analysis of Oxysterols 271

4. Glass screw-top vials: 2 mL, 32  12 mm, and 9 mm clear glass

screw thread vials purchased from Thermo Fisher Scientific can
be used.
5. Glass inserts for screw-top vials: use 0.1 mL micro insert
29  5.7 mm clear glass with attached plastic.
6. Caps for screw-top vials: use 9 mm combination seal PP short
thread caps.
7. Ultrahigh purity helium gas (99.9990%) and ultrahigh purity
ammoniac gas (99.990%) are used as carrier and chemical
ionization gas, respectively.

3 Methods

3.1 Sample 1. In case of cells, add 1 mL of MeOH/EgTA solution to 5  106

Processing cells (see Note 3), and transfer the suspension to a glass culture
tube (13 mm  100 mm). In case of media samples, transfer
200 μL of cell media (see Note 3) into a glass culture tube. In
case of plasma, transfer 400 μL of human plasma into a glass
culture tube.
2. Add 2.5 mL of dichloromethane, 2.5 mL of MeOH, and 2 mL
of water.
3. Add 10 μL of internal standard solution (500 ng).
4. Flush with nitrogen gas (see Note 4).
5. Vortex for 10 min in capped glass culture tube.
6. Centrifuge the sample for 10 min at 1095  g.
7. Collect the lower organic phase, and transfer to a new glass
culture tube (see Note 5).
8. Dry lipid extract under a constant stream of nitrogen gas.
For free oxysterol extraction, go directly to step 17.
9. Add 2.5 mL of dichloromethane and 1.75 mL of 0.5 M KOH
methanolic solution.
10. Flush with nitrogen gas, and keep for 2 h under stirring at
room temperature in capped glass culture tube (see Note 4).
11. Neutralize with 1.3 mL of 0.5 M phosphoric acid solution (see
Note 6).
12. Add 0.75 mL of MeOH and 0.7 mL of water.
13. Flush with nitrogen gas (see Note 4).
14. Vortex and then centrifuge for 5 min at 1095  g in capped
glass culture tube.
15. Collect the lower organic phase and transfer to a new tube.
16. Dry lipid extract under a constant stream of nitrogen gas.
272 Fabien Riols and Justine Bertrand-Michel

17. Condition the 100 mg SiOH SPE plate with 2 mL of

n-heptane.
18. Dissolve sample in 1 mL of n-heptane and load onto SPE plate.
19. Wash the plate with 4 mL of 0.5% (v/v) 2-propanol in n-heptane.
20. Elute oxysterols with 2 mL of 30% (v/v) 2-propanol in
n-heptane.
21. Dry lipid extract under nitrogen gas.
22. Resolubilize the lipid extract in 2  80 μL of n-heptane, and
transfer to a glass vial with insert.
23. Dry lipid extract under nitrogen gas (see Notes 7 and 8).
24. Add 50% (v/v) BSTFA solution in ACN, and heat for 1 h at
55  C in a heating block in the capped glass vial (see Note 9).
25. Cool and dry lipid extract under nitrogen gas.
26. Dissolve the extract in 20 μL of n-heptane and vortex.
27. Immediately analyze sample extract by GC-MS (see Fig. 1 for
exemplary data).

3.2 Quantification 1. The oven temperature program is as follows: 180  C for 1 min,
Using SIM or SRM 20  C/min to 250  C, 5  C/min to 300  C where the temper-
Scan Mode ature is kept for 8 min, and then 35  C/min to 325  C.
2. The carrier gas is kept at a constant flow rate of 0.8 mL/min.
3. The samples are injected in splitless mode with an injection
volume of 1 μL.
4. The injector, transfer line, and source temperature are 270  C,
280  C, and 250  C, respectively.
5. For SIM analysis, the mass spectrometer is operated in electron
ionization mode (70 eV).
6. For SRM analysis, the mass spectrometer is operated in positive
chemical ionization mode using ammonia as ionization gas at a
flow rate of 1.5 mL/min (47.5 cm/s).
7. Ions used are as follows (Table 2):
8. For quantitative analysis construct calibration linear lines (2.7)
plotting nominal concentrations against the area ratio obtained
for each analyte. Use the weighting equal and ignore the origin.
Calculate analyte concentrations using the obtained calibration
functions.

4 Notes

1. Liquid nitrogen should be manipulated under natural or

mechanical ventilation to prevent oxygen-deficient atmo-
spheres below 19.5%. Use loose fitting thermal insulated or
 Analysis of Oxysterols 273

90
85
A
80
75
70
65 7-ketocholesterol
15.77
60
Relative Abundance

55 7β-hydroxycholesterol 25-hydroxycholesterol
50 4β-hydroxycholesterol
45 16.55

40
35 19-hydroxycholesterol 24(S)-hydroxycholesterol 27-hydroxycholesterol
15.24
30
18.27 19.90
25
14.13
20 5α,6α-epoxycholestanol
7α-hydroxycholesterol 16.14
15
18.69
10

5 15.08
16.77
14.23 15.39 18.52 19.27
14.01
0
12.5 13.0 13.5 14.0 14.5 15.0 15.5 16.0 16.5 17.0 17.5 18.0 18.5 19.0 19.5 20.0
Time (min)
RT:12.99 - 20.07
100
95
90
B
85
80
75
70
Relative Abundance

65 7-ketocholesterol
60
19-hydroxycholesterol 7β-hydroxycholesterol
55
50
45
40
35
30
4β-hydroxycholesterol
25
27-hydroxycholesterol
20
15 5α,6α-epoxycholestanol
10
5
0
13.0 13.5 14.0 14.5 15.0 15.5 16.0 16.5 17.0 17.5 18.0 18.5 19.0 19.5 20.0
Time (min)

Fig. 1 Oxysterol profile using GC-MS analysis with positive chemical ionization in SRM mode: (Panel a) a
mixture of nine standards is shown; (Panel b) an example analysis of human multipotent adipose-derived stem
cells (hMADS) is depicted
274 Fabien Riols and Justine Bertrand-Michel

Table 2
Monitored ions and retention times

Ions used for SIM Mass transitions used for SRM Retention time
Substance analysis analysis (min)
7α-hydroxycholesterol 456; 367 457.4 ¼ > 367.4 14.1
19-hydroxycholesterol 456; 353 457.4 ¼ > 367.4 15.2
7β-hydroxycholesterol 456; 367 457.4 ¼ > 367.4 15.7
4β-hydroxycholesterol 456; 417 457.4 ¼ > 159.1 16.1
5α,6α-epoxycholestanol 474; 384 385.4 ¼ > 367.4 16.5
24(S)- 546; 456 474.4 ¼ > 367.4 18.2
hydroxycholesterol
25-hydroxycholesterol 546; 456 474.4 ¼ > 367.4 18.7
7-ketocholesterol 472; 382 473.4 ¼ > 383.4 18.9
27-hydroxycholesterol 456; 417 474.4 ¼ > 367.4 19.9

leather gloves for skin protection and tongs to immerge and

withdraw objects in liquid nitrogen. Be careful to minimize
boiling and splashing by handling the liquid nitrogen slowly.
2. Dichloromethane and MeOH should be used in a fume hood
to avoid vapors. Don’t allow contact with the skin or eyes.
3. The quantity must be adjusted according to the cell type.
4. Be extremely careful not to disturb and/or aspirate the inter-
face which contains proteins.
5. Phosphoric acid should be used in a fume hood. Wear gloves,
and add the acid to water.
6. BSTFA should be used in extractor fume hood. Don’t allow
contact with the skin or eyes.
7. Flush the culture tubes with a stream of nitrogen gas, and cap
immediately to avoid autoxidation during the extraction
process.
8. To ensure complete derivatization, ensure complete dryness.
9. Dried samples flushed with nitrogen gas can be stored at

80  C overnight prior to derivatization.
10. The method described is optimized on a 5MS capillary
column.
 Analysis of Oxysterols 275

Acknowledgments

This work was in part funded by National Health and Medical

Research Council of France (Inserm) and the French National
Infrastructure MetaboHUB-ANR-11-INBS-0010.

References
1. Russell DW (2003) The enzymes, regulation, Diabetes and insulin in regulation of brain cho-
and genetics of bile acid synthesis. Annu Rev lesterol metabolism. Cell Metab 12
Biochem 72:137–174. https://doi.org/10. (6):567–579. https://doi.org/10.1016/j.
1146/annurev.biochem.72.121801.161712 cmet.2010.11.006
2. Iuliano L (2011) Pathways of cholesterol oxi- 7. Porter FD, Scherrer DE, Lanier MH, Lang-
dation via non-enzymatic mechanisms. Chem made SJ, Molugu V, Gale SE, Olzeski D,
Phys Lipids 164(6):457–468. https://doi. Sidhu R, Dietzen DJ, Fu R, Wassif CA, Yanja-
org/10.1016/j.chemphyslip.2011.06.006 nin NM, Marso SP, House J, Vite C, Schaffer
3. Bauman DR, Bitmansour AD, McDonald JG, JE, Ory DS (2010) Cholesterol oxidation pro-
Thompson BM, Liang G, Russell DW (2009) ducts are sensitive and specific blood-based
25-hydroxycholesterol secreted by macro- biomarkers for Niemann-pick C1 disease. Sci
phages in response to toll-like receptor activa- Transl Med 2(56):56ra81. https://doi.org/
tion suppresses immunoglobulin a production. 10.1126/scitranslmed.3001417
Proc Natl Acad Sci U S A 106 8. Griffiths WJ, Wang Y, Karu K, Samuel E,
(39):16764–16769. https://doi.org/10. McDonnell S, Hornshaw M, Shackleton C
1073/pnas.0909142106 (2008) Potential of sterol analysis by liquid
4. Hannedouche S, Zhang J, Yi T, Shen W, chromatography-tandem mass spectrometry
Nguyen D, Pereira JP, Guerini D, Baumgarten for the prenatal diagnosis of Smith-Lemli-
BU, Roggo S, Wen B, Knochenmuss R, Noel S, Opitz syndrome. Clin Chem 54
Gessier F, Kelly LM, Vanek M, Laurent S, (8):1317–1324. https://doi.org/10.1373/
Preuss I, Miault C, Christen I, Karuna R, clinchem.2007.100644
Li W, Koo DI, Suply T, Schmedt C, Peters 9. Matysik S, Klunemann HH, Schmitz G (2012)
EC, Falchetto R, Katopodis A, Spanka C, Roy Gas chromatography-tandem mass spectrome-
MO, Detheux M, Chen YA, Schultz PG, Cho try method for the simultaneous determination
CY, Seuwen K, Cyster JG, Sailer AW (2011) of oxysterols, plant sterols, and cholesterol pre-
Oxysterols direct immune cell migration via cursors. Clin Chem 58(11):1557–1564.
EBI2. Nature 475(7357):524–527. https:// https://doi.org/10.1373/clinchem.2012.
doi.org/10.1038/nature10280 189605
5. Kotti TJ, Ramirez DM, Pfeiffer BE, Huber 10. Saeed A, Floris F, Andersson U, Pikuleva I,
KM, Russell DW (2006) Brain cholesterol Lovgren-Sandblom A, Bjerke M, Paucar M,
turnover required for geranylgeraniol produc- Wallin A, Svenningsson P, Bjorkhem I (2014)
tion and learning in mice. Proc Natl Acad Sci U 7alpha-hydroxy-3-oxo-4-cholestenoic acid in
S A 103(10):3869–3874. https://doi.org/10. cerebrospinal fluid reflects the integrity of the
1073/pnas.0600316103 blood-brain barrier. J Lipid Res 55
6. Suzuki R, Lee K, Jing E, Biddinger SB, McDo- (2):313–318. https://doi.org/10.1194/jlr.
nald JG, Montine TJ, Craft S, Kahn CR (2010) P044982
 Chapter 20

Analysis of Metabolites from the Tricarboxylic Acid Cycle

for Yeast and Bacteria Samples Using Gas Chromatography
Mass Spectrometry
Reza Maleki Seifar, Angela ten Pierick, and Patricia T.N. van Dam

Abstract
We here explain step by step the implementation of gas chromatography coupled with tandem mass
spectrometry for the quantitative analysis of intracellular metabolites from the tricarboxylic acid (TCA)
cycle such as citrate, isocitrate, alpha-ketoglutarate, succinate, malate, and fumarate. Isotope dilution is
used to correct for potential metabolite losses during sample processing, matrix effects, incomplete
derivatization, and liner contamination. All measurements are performed in selected reaction monitoring
(SRM) mode. Standards and samples are first diluted with a fixed volume of a mixture of fully 13C-labeled
internal standards and then derivatized to give trimethylsilyl-methoxylamine derivatives prior GC-MS/MS
analysis.

Key words TCA cycle, Gas chromatography, Mass spectrometry, TMS-MOX derivatives,
Metabolomics

1 Introduction

The tricarboxylic acid (TCA) cycle is central to energy production

by the oxidation of pyruvic acid to carbon dioxide. The cycle is also
important for biosynthesis of amino acids, nucleotide bases, and
other metabolites. Alterations in metabolite levels caused by defects
in TCA enzymes have been described as underlying hallmarks of
cancer [1]. Furthermore, studying the pyruvate metabolism can
also help in the construction of important industrial strains and
provide tools for controlled cultivation [2]. In this chapter, we
describe in detail the application of gas chromatography (GC)
coupled with tandem mass spectrometry (MS/MS) for the quanti-
tative analysis of intracellular citrate, isocitrate, alpha-ketoglutarate,
succinate, malate, fumarate, and oxaloacetate levels. A global
13
C-labeled Saccharomyces cerevisiae extract (containing all targeted
metabolites as 13C-labeled compounds) is added as internal stan-
dard to the sample solutions and standard calibration mixtures,

Martin Giera (ed.), Clinical Metabolomics: Methods and Protocols, Methods in Molecular Biology, vol. 1730,
https://doi.org/10.1007/978-1-4939-7592-1_20, © Springer Science+Business Media, LLC 2018

277
278 Reza Maleki Seifar et al.

helping to overcome analytical limitations. This extract is produced

by cultivation of S. cerevisiae on fully labeled 13C-glucose as the
exclusive carbon source [3]. Internal standards are added to the
samples immediately after quenching or right before extraction of
the metabolites.

2 Materials

All chemicals should be of p.a. or higher purity.

2.1 Chemicals 1. Succinic acid, alpha-ketoglutaric acid, malic acid, fumaric acid,
citric acid, and isocitric acid are needed as calibration standards
(see Note 1).
2. 13
C-labeled S. cerevisiae cell extract is obtained by growing
S. cerevisiae on fully labeled 13C-glucose as the exclusive carbon
source [3].
3. Silylation reagent mix: trimethylchlorosilane (TMCS) and N-
methyl-N-trimethylsilyltrifuoroacteamide (MSTFA) are used
as silylation reagents. Mix 1 mL of MSTFA with 50 μL
TMCS (see Note 2).
4. Methoxyamine hydrochloride (MOX): prepare a solution of
(20 mg/mL) MOX in anhydrous pyridine. Vortex and prefera-
bly heat for 1 min at the heating block to let MOX completely
dissolve (prepare freshly every day).
5. Isooctane and ethyl acetate are used for syringe cleaning.

2.2 Equipment A freeze dryer, a heating module (e.g., a Reacti-Therm III Heating
Module (Thermo Scientific)), a 80  C freezer, and a centrifuge
are needed for sample preparation.

2.3 GC-MS/MS 1. GC system: 7890 A (Agilent Technologies) or equivalent.

Instrumentation 2. MS: Triple Quad 7000 (Agilent Technologies) or equivalent.
3. A Zebron ZB-50 capillary column (30 m  250 μm internal
diameter, 0.25 μm film (Phenomenex)) should be used.
4. A straight glass liner, Ultra Inert, single taper with glass wool
(Agilent) is recommended (see Note 3).
5. 10 μL gastight syringe with Teflon-tipped plunger.
6. 2 mL deactivated, silanized vial.
7. Cap screw with PTFE/silicone/PTFE septa (see Note 4).
8. Short thread screw caps (transparent) with PTFE septa only
(Grace).
9. Snap top shell vial, 1 mL, 8 mm, clear (Grace).
10. 250 μL glass inserts.
 Analysis of TCA Cycle Intermediates by GC-MS/MS 279

3 Methods

3.1 Preparation In the case of yeast and bacteria samples, 1.2 mL broth is rapidly (less
of Intracellular than a second) taken from a fermentor into a pre-weighted 50 mL
Samples Greiner tube containing 5 mL cold methanol ( 40  C) to freeze
whole metabolism of the cells. After weighting the tube once more,
the cell pellet is collected by centrifugation for 3 min at 1000  g at
6  C. Rotors are kept in a freezer ( 20  C) and installed before the
start of centrifugation. Subsequently, the cell pellet is washed with
another 5 mL of cold methanol ( 40  C) followed by centrifugation
(3 min at 1000  g) at 6  C. After discarding the supernatant,
120 μL of labeled internal standard is added to the cell pellet (see
Note 5). Extraction of intracellular metabolites is done using hot
ethanol [3]. Briefly, 5 mL of preheated (95  C) ethanol 75% (v/v) is
added to the cell pellet and heated for 3.5 min at 95  C. Then the
extract is immediately cooled down using an ice bath followed by
another centrifugation step (3 min 1000  g) to remove cell debris.
In the next step, the extract is dried using a SpeedVac. The dry
matter is reconstituted in 600 μL of deionized pure water and is
stored at 80  C or immediately analyzed. In the case of mammalian
cells, the quenching and extraction protocols described by Kostidis
et al. [4] can be applied.

3.2 Preparation Stock solutions from each targeted compound are prepared. From
of Standards for GC- these stock solutions, a standard mix containing all targeted meta-
MS bolites is prepared. Aliquots from this standard mix are prepared
and stored at 80  C. For each set of samples, one aliquot is taken
for preparation of standard calibration dilution series. Each calibra-
tion set includes ten dilution series. A typical standard calibration
range for each metabolite is 0.25 μM to 100 μM. The standard
calibration range should cover the concentration of the targeted
compounds in the sample solutions. A fixed volume of internal
standard (13C-labeled extract) is required to be added to the sam-
ples and each of the standard calibration mixtures (see Notes 6 and
7). It should be mentioned that the concentration of 13C-labeled
internal standards needs to be in the calibration range of the 12C
standards.
1. Pipette the following volumes (Table 1) to GC vials and close
them with caps containing Teflon septa.
2. Pipette 100 μL sample (containing 20 μL internal standard) to
a GC vial, and close it with a cap containing a Teflon septum.
3. Make two holes in each Teflon septum using a needle.

3.3 Freeze-Drying 1. Freeze samples in a 80 C freezer.

2. Place the frozen samples in a freeze dryer chamber.
3. Start the vacuum pump.
4. Freeze dry overnight.
280 Reza Maleki Seifar et al.

Table 1
Dilution scheme for preparation of standard calibration mixtures

a
Dilution Standard Ultrapure Internal standard
series mix μL water μL mix μL
CAL 1 1 200 20
CAL 2 2 200 20
CAL 3 5 200 20
CAL 4 10 200 20
CAL 5 20 200 20
CAL 6 40 200 20
CAL 7 60 200 20
CAL 8 100 20
CAL 9 200 20
CAL 10 400 20
a
200 μL is added to the low standards; otherwise, the required volume for freeze-drying
is too little, and samples might be defrosted

3.4 Derivatization 1. Work in a fume hood and wear nitrile gloves.

2. Switch on the heating block (70  C).
3. Add 50 μL of the MOX solution to each GC vial (see Note 2).
4. Incubate samples for 50 min at 70  C using a heating block.
5. Remove from the heating block and let cool down to room
temperature.
6. Add 80 μL of the silylation reagent mix to each sample.
7. Incubate samples for 50 min at 70  C using a heating block.
8. Remove vials from the heating block and let them cool down to
room temperature.
9. Transfer samples into the centrifuge glass tubes (1 mL shell
vials).
10. Centrifuge for 1 min at 8000  g.
11. Transfer 70 μL of supernatant of the sample from the centri-
fuge tube into a glass insert, and place it back into the original
GC vial (remove air bubbles if needed).

3.5 GC-MS Settings 1. Set the carrier gas flow (helium 99.9990%) to 1 mL/min.
2. Program the GC oven temperature as follows: keep constant
for 1 min at 70  C followed by a ramp of 10  C per min up to
220  C.
3. Set the temperature of the transfer line to 250  C.
 Analysis of TCA Cycle Intermediates by GC-MS/MS 281

Table 2
GC-MS/MS parameters for TMS-MOX derivatives

13 a
Retention Precursor C Precursor Fragment Collision
Metabolite time (min) ion ion ion energy
Fumaric acid 9.80 245 249 147 10
Succinic acid 9.87 247 251 147 5
Malic acid 11.46 335 339 147 15
alpha-Ketoglutaric 13.65 304 309 147 5
acid
Citric acid 14.97 465 471 147 35
Isocitric acid 15.14 465 471 147 35
a
Fragment ions for labeled metabolites were the same as unlabeled metabolites

4. Set the injection volume to 1 μL (see Note 8).

5. If using a multimode inlet (MMI) (Agilent), use the following
settings: split-less mode for metabolites with low concentration
and in split mode 1:5 for metabolites with high concentrations.
6. Set the MMI temperature: 70  C (with a split-less time of 1 min
if it is needed), after injection temperature is increased to
220  C with a ramp of 720  C/min and held for 5 min. Then
temperature is increased to 300  C to clean injector.
7. At the end of the run, backflush the column with five column
volumes at 220  C (see Notes 9 and 10).
8. Set the MS transfer line to 250  C.
9. Keep the MS source at 230  C.
10. Set the helium gas flow of the collision cell to 2.25 mL/min.
11. Set the N2 collision gas flow to 1.5 mL/min.
12. Use electron ionization operated at 70 eV.
13. The MS is operated in SRM mode.
14. Mass resolution is set to 0.7 mass unit for both quadrupoles.
15. Use the settings shown in Table 2 for GC-MS/MS analysis.
16. Mass Hunter quantitative analysis software (version B.07.00;
Agilent) can be used for data processing.
17. Integrate signals and calculate concentrations based on the cali-
bration line as established according to Table 1 (see Note 7).

4 Notes

1. Use certificate of analysis from chemical providers for prepara-

tion of stock solutions.
282 Reza Maleki Seifar et al.

2. Derivatizing reagents should be stored in a desiccator in a

fridge at 4  C.
3. The liner in GC might need to be replaced based on the sample
type after different numbers of injections.
4. For GC vials, silicone rubber should be avoided because of
possible sample contamination and reactivity with the derivati-
zation reagents.
5. Internal standards were added to the samples immediately after
quenching or right before the extraction step.
6. Fixed volumes of 13C-labeled internal standard (20 μL) were
added to samples and each one of the calibration mixtures.
7. Calibration graphs are constructed based on peak height ratios
of natural 12C peak height of targeted metabolites to 13C-
labeled peak height of internal standards versus concentration
of natural isotope 12C standards. 1/x2 weighting is used for
construction of calibration curves.
8. For multiple injections from the same sample vial, renew the
cap after each two injections if it is possible.
9. After about 100 injections, cut 10 cm of the column (injection
side) and at the same time replace the liner and inlet septum
(gas flow needs to be modified accordingly using retention
time locking (RTL), to have reproducible retention times).
10. Run the alkane mix for RTL and system check about every ten
samples.

References
1. Cardaci S, Ciriolo MR (2012) TCA cycle defects metabolome by isotope dilution mass spectrom-
and cancer: when metabolism tunes redox state. etry using uniformly C-13-labeled cell extracts
Int J Cell Biol. https://doi.org/10.1155/ as internal standards. Anal Biochem
2012/161837 336:164–171
2. Pronk JT, YDE Steensma H, van Dijken JP 4. Kostidis S, Addie RD, Morreau H et al (2017)
(1996) Pyruvate metabolism in Saccharomyces Quantitative NMR analysis of intra- and extra-
cerevisiae. Yeast 12:1607–1633 cellular metabolism of mammalian cells: a tuto-
3. Wu L, Mashego MR, van Dam JC et al (2005) rial. Anal Chim Acta 980:1–24
Quantitative analysis of the microbial
 Chapter 21

GC-MS Analysis of Lipid Oxidation Products in Blood,

Urine, and Tissue Samples
Anne Barden and Trevor A. Mori

Abstract
Oxidant stress has been identified as important in the pathology of many diseases. Oxidation products of
polyunsaturated fatty acids collectively termed isoprostanes, neuroprostanes, and isofurans are considered
the most reliable measures of in vivo lipid oxidation, and they are widely used to assess oxidant stress in
various diseases. Here we describe the measurement of these lipid oxidation products using gas chroma-
tography mass spectrometry with electron capture negative ionization.

Key words Lipid oxidation, F2-isoprostanes, Neuroprostanes, Isofurans, Gas chromatography mass
spectrometry

1 Introduction

Free radicals have been linked to a wide range of human diseases

and are known to oxidize lipids, proteins, and DNA [1]. Although a
number of measures of lipid peroxidation are available, most
involve assays directed against nonspecific or unstable metabolites
such as short-chain alkanes, malondialdehyde, lipid hydroperox-
ides, or ex vivo induced lipoprotein oxidizability [1]. Polyunsatu-
rated fatty acids (PUFAs) react with free radicals to form
oxygenated metabolites collectively termed isoprostanes, neuro-
prostanes, and isofurans [2]. The F2-isoprostanes (F2-IsoPs) first
reported by Morrow et al. [3] are formed in vivo predominantly by
free radical nonenzymatic oxidation of arachidonic acid (AA, C20:4
n-6). The F2-IsoPs are widely acknowledged as the most reliable
in vivo biomarkers of oxidative stress in animals [4] and humans
[5, 6]. Importantly, a number of F2-IsoPs are also biologically
active [6]. The presence and quantity of oxygen can alter the lipid
peroxidation products formed. For example, under conditions of
high oxygen tension, free radical-induced peroxidation of AA leads
to formation of the isofurans (IsoFs) [7, 8].

Martin Giera (ed.), Clinical Metabolomics: Methods and Protocols, Methods in Molecular Biology, vol. 1730,
https://doi.org/10.1007/978-1-4939-7592-1_21, © Springer Science+Business Media, LLC 2018

283
284 Anne Barden and Trevor A. Mori

The n-3 fatty acids eicosapentaenoic acid (EPA, C20:5 n
3)

and docosahexaenoic acid (DHA, C22:6 n-3) are oxidized to
F3-isoprostanes (F3-IsoPs) [9] and F4-isoprostanes (F4-IsoPs)
[10], respectively. The F4-IsoPs are commonly referred to as neu-
roprostanes (NeuroPs). Measurement of lipid oxidation products
of EPA and DHA is relevant in the setting of n-3 fatty acid supple-
mentation and in organs that have a high n-3 fatty acid content
such as the heart, retina, and brain.
The measurement of in vivo oxidative stress requires samples be
collected and stored in a manner that avoids ex vivo autoxidation.
This is particularly important for the measurement of lipid oxida-
tion markers in plasma and tissue samples. We have shown that the
collection of blood into collection tubes containing dipotassium
ethylenediaminetetraacetic acid (EDTA), reduced glutathione
(GSH), and butylated hydroxytoluene (BHT) with storage at

80  C is necessary to minimize ex vivo autoxidation of AA that
can lead to artifactual elevation of plasma F2-IsoP concentrations
[11]. Our data show that F2-IsoP concentrations in samples col-
lected in EDTA alone were significantly elevated compared with
those collected into EDTA with the addition of antioxidants
[11]. We also showed plasma stored at
20  C resulted in a marked
elevation of plasma F2-IsoPs irrespective of whether blood collec-
tion included EDTA/GSH/BHT.
Measurement of lipid oxidation products in biological samples
is complex due to the number of products that can be formed. To
date enzyme-linked immunoassays (EIA) are available only for
F2-IsoP, but we have shown these have poor agreement with mass
spectrometry [12]. Therefore, mass spectrometry remains the
“gold standard” for measurement of lipid oxidation products.
In this chapter, we describe the methods for measurement of
IsoPs, NeuroPs, and IsoFs in plasma, urine, and tissue using gas
chromatography mass spectrometry (GC-MS) with electron cap-
ture negative ionization.

2 Materials

Prepare all solutions using ultrapure water (prepared by purifying

deionized water to 18 MΩ-cm at 25  C) and analytical grade
reagents. All solvents should be analytical grade. Prepare and
store all reagents and solvents at room temperature (unless indi-
cated otherwise).

2.1 Blood Collection 1. Dissolve 1 g of dipotassium ethylenediaminetetraacetic acid

Tubes (EDTA) into 6 mL 0.9% saline. Add 1 g of reduced glutathi-
one, and adjust to pH 7.4 using 5 M NaOH. Make up to a final
volume of 10 mL with 0.9% saline. To make 5 M NaOH,
dissolve 20 g NaOH in 100 mL water (see Note 1).
 GC-MS Analysis of Lipid Autoxidation Products in Biological Samples 285

2. Dissolve 200 mg butylated hydroxytoluene (BHT) in 50 mL of

ethanol.
3. To each 2.5 mL blood collection tube, add 25 μL of EDTA/
reduced glutathione and 10 μL of BHT in ethanol to prevent
ex vivo oxidation.
4. Collection tubes can be prepared in advance and kept at 4  C
for a period of months.

2.2 Urine Collection 1. Two 500 μL aliquots of urine from a 24-h collection are stored
at
80  C until analysis. The total volume of the 24-h urine
collection and the creatinine measurement are required.
2. If a 24-h collection is not available, a spot urine can be used,
and the creatinine concentration should be determined.

2.3 Tissue Collection 1. Liquid nitrogen and porcelain mortar and pestle for processing
tissue samples (see Note 2).
2. Chloroform (CCl3)/methanol (CH3OH) (2:1 vol/vol, con-
taining 500 mg/L BHT): Measure 660 mL CCl3 and 330 mL
CH3OH into a 1 L graduated cylinder. Add 500 mg BHT and
store in a 1 L reagent bottle (see Note 3).

2.4 Standards 1. 8-iso-prostaglandin F2α-d4 (8-iso-PGF2α-d4) (item № 316,350,

Cayman Chemicals, Ann Arbor, MI, USA): Make up as 5 ng/
50 μL in CH3OH, and store at
20  C.
2. 8,12-iso-isoprostane F2α-VI-d11 (8,12-iso-iPF2α-VI-d11) (item
№ 10006878, Cayman Chemicals, Ann Arbor, MI, USA):
Make up as 5 ng/50 μL in CH3OH, and store at
20  C.
3. F2-IsoPs: Use 8-iso-prostaglandin F2α (8-iso-PGF2α) (item
№ 16350, Cayman Chemicals, Ann Arbor, MI, USA) as the
standard. Make up as 5 ng/50 μL in CH3OH, and store at

20  C.
4. F3-IsoPs: Use 8-iso-prostaglandin F3α (8-iso-PGF3α) (item
№ 16,992, Cayman Chemicals, Ann Arbor, MI, USA) as the
standard. Make up as 5 ng/50 μL in CH3OH, and store at

20  C.
5. IsoF standard prepared as 5 ng/50 mL in CH3OH and stored
at
20  C was made available as a gift from Professor
L.J. Roberts II (Department of Pharmacology, Vanderbilt Uni-
versity, Nashville, TN, USA).
6. NeuroPs: Use 4(RS)-F4t-neuroprostane as the standard. 4
(RS)-F4t-neuroprostane was made available as a gift from
Dr. Thierry Durand (Faculty of Pharmacy, Institut des Biomo-
lécules Max Mousseron, Montpellier, France) and is prepared
as 5 ng/50 mL in CH3OH and stored at
20  C.
286 Anne Barden and Trevor A. Mori

2.5 Solid-Phase 1. Bond Elut™ Certify II cartridges (200 mg, 3 mL) (Agilent
Chromatography Technologies, Santa Clara, CA, USA).
2. Glass culture tubes: Borosilicate disposable culture yubes
13  100 mm and caps are purchased from Kimble Chase
Life Science and Research Products, Vineland, NJ, USA.
3. 1 M potassium hydroxide (KOH) methanolic: Dissolve
14.03 g of KOH in 250 mL CH3OH.
4. 100 mM sodium acetate (CH3COONa), pH 4.6: Dissolve
8.2 g of CH3COONa (anhydrous) in 1000 mL of water, and
adjust to pH 4.6 with concentrated acetic acid (see Note 4).
5. 100 mM sodium acetate with 5% CH3OH pH 7.0: Dissolve
8.2 g of CH3COONa in 950 mL of water. Add 50 mL
CH3OH, mix, and adjust to pH 7.0 with concentrated acetic
acid (see Note 4).
6. 1 M hydrochloric acid (HCl): Fill a 1 L graduated cylinder with
900 mL pure water, and carefully add 100 mL concentrated
HCl (10 M) (see Note 4).
7. CH3OH/water (1:1): Measure 500 mL CH3OH and 500 mL
water in a 1 L graduated cylinder, and store in a 1 L reagent
bottle.
8. Ethyl acetate/n-hexane (1:3): Measure 200 mL ethyl acetate
and 600 mL hexane in a 1 L graduated cylinder, and store in a
1 L reagent bottle.
9. Ethyl acetate/CH3OH (9:1): Measure 900 mL ethyl acetate
and 100 mL CH3OH in a 1 L graduated cylinder, and store in a
1 L reagent bottle.
10. Solid-phase extraction (SPE) VacElut 20 Manifold (Agilent
Technologies, Santa Clara, CA, USA).

2.6 Derivatization 1. 10% pentafluorobenzyl bromide (PFBBr) in acetonitrile:

Reagents for Gas PFBBr (0.5 mL) (Sigma, St. Louis, MO, USA; Cat# 101052)
Chromatography Mass is added to 4.5 mL acetonitrile and stored in a dark glass bottle
Spectrometry protected from light at 4  C (see Note 5).
2. 10% N,N-diisopropylethylamine (DIPEA) in acetonitrile:
DIPEA (0.5 mL) (Sigma, St. Louis, MO, USA; Cat#
D125806) is added to 4.5 mL acetonitrile and stored in a
dark glass bottle protected from light in a fume hood (see
Note 5).
3. N,O-bis(trimethylsilyl)trifluoroacetamide with 1% trimethyl-
chlorosilane (BSTFA with 1% TMS): Purchased from Sigma,
St. Louis, MO, USA (Cat# T6381), and is stored at 4  C (see
Note 5).
4. Pyridine: Purchased from Sigma, St. Louis, MO, USA, and is
stored in a fume hood in a glass vial in the dark (see Note 5).
 GC-MS Analysis of Lipid Autoxidation Products in Biological Samples 287

5. Isooctane (2,2,4-trimethylpentane): Purchased from Sigma,

St. Louis, MO, USA.
6. High purity nitrogen gas.

2.7 Gas 1. GC-MS: Uses an Agilent 6890 Gas Chromatograph coupled to

Chromatography Mass an Agilent 5975 Mass Selective Detector and Agilent 6890
Spectrometry 7683B Series Injector Autosampler (Agilent Technologies,
Santa Clara, CA, USA).
2. Gas chromatography column: DB-5MS 25 m, 0.20 mm,
0.33 μm film thickness (P/N 128–5522) (Agilent Technolo-
gies, Santa Clara, CA, USA).
3. Glass screw-top vials: 2 mL, 32  12 mm, 9 mm clear glass
screw thread vials purchased from Thermo Fisher Scientific
(Waltham, MA, USA).
4. Glass inserts for screw-top vials: 0.1 mL micro insert
29  5.7 mm clear glass with attached plastic spring purchased
from Thermo Fisher Scientific (Waltham, MA, USA).
5. Caps for screw-top vials: 9 mm combination seal PP short
thread caps purchased from Thermo Fisher Scientific (Wal-
tham, MA, USA).
6. Ultrahigh purity helium gas (99.999%) and ultrahigh purity
methane gas (99.995%).

3 Methods

3.1 Processing 1. Blood samples are collected into chilled 2.5 mL blood collec-
of Blood Samples tion tubes and centrifuged immediately at 4  C and 1000  g
for 10 min.
2. Two 500 μL plasma aliquots are removed and stored at
80  C
until analysis.
3. Plasma (200 μL) þ 50 μL (5 ng) 8-iso-PGF2α-d4 (internal
standard) are subjected to base hydrolysis using 1 mL of 1 M
KOH in CH3OH in capped glass culture tubes under nitrogen
(see Note 6). Samples are heated for 40 min at 40  C in a water
bath or a heating block.
4. The samples are cooled to room temperature, 2 mL of 100 mM
sodium acetate pH 4.6 is added, and samples are adjusted to
pH 4.0 with 1 M HCl (see Note 7).
5. Samples are centrifuged at 1500  g for 10 min at 4  C, and the
supernatant is removed for solid-phase chromatography.
6. Bond Elut™ Certify II cartridges (200 mg, 3 mL) applied to a
VacElut 20 Manifold under vacuum are preconditioned with
2 mL CH3OH, followed by 2 mL 100 mM sodium acetate
with 5% CH3OH pH 7 (see Note 8).
288 Anne Barden and Trevor A. Mori

7. The supernatant is applied to the Bond Elut™ Certify II car-

tridge (see Note 8).
8. The cartridge is washed sequentially with 2 mL CH3OH/
water (1:1) and 2 mL ethyl acetate/hexane (1:3) (see Note 8).
9. Oxidation products are eluted with 2 mL ethyl acetate/
CH3OH (9:1) into a glass culture tube (see Note 8).
10. Samples are evaporated to dryness under a stream of nitrogen
(see Note 9).
11. 10% PFBBr in acetonitrile (40 μL) and 10% DIPEA in acetoni-
trile (20 μL) are added to each glass culture tube and left at
room temperature for 30 min.
12. Samples are evaporated to dryness under nitrogen (see Note 10).
13. Add 20 μL BSTFA with 1% TMS and 10 μL pyridine to each
sample and then heat for 20 min at 45  C in a heating block
(see Note 10).
14. Samples are evaporated to dryness under nitrogen (see Note 10).
15. Reconstitute in 30 μL isooctane in a 0.1 mL glass micro insert
in a 2 mL screw-top vial and cap. Samples are ready for GC-MS
analysis.

3.2 Processing 1. To 200 μL urine, add 50 μL (5 ng) 8,12-iso-iPF2α-VI-d11

of Urine Samples (internal standard) (see Note 11) and 2 mL of 100 mM sodium
acetate pH 4.6 buffer, and adjust to pH 4.0 with 1 M HCl.
2. Subject samples to solid-phase chromatography and derivatiza-
tion as described above for plasma (Subheading 3.1, steps
6–15).

3.3 Processing 1. A weighed amount of tissue sample (approx. 50 mg) is col-

of Tissue Samples lected into liquid nitrogen (see Note 2) and ground to a fine
powder with a porcelain mortar and pestle.
2. Tissue samples are placed in glass culture tubes, extracted using
the Folch method with 2 mL CCl3/CH3OH (2:1) (containing
500 mg/L BHT), and evaporated to dryness in a centrifugal
evaporator under vacuum at 40  C.
3. To each sample, add 50 μL (5 ng) 8-iso-PGF2α-d4 (internal
standard) and 200 μL water.
4. Samples are hydrolyzed using 1 mL of 1 M KOH in CH3OH in
capped glass culture tubes under nitrogen (see Note 6) with
heating for 40 min at 40  C in a water bath or a heating block.
5. Samples are extracted and subjected to solid-phase chromatog-
raphy and derivatization as described above for plasma (Sub-
heading 3.1, steps 4–15).
 GC-MS Analysis of Lipid Autoxidation Products in Biological Samples 289

3.4 Gas 1. Samples are analyzed by GC-MS. Chromatography uses an

Chromatography Mass Agilent DB-5MS column with helium as the carrier gas at a
Spectrometry Analysis flow rate of 0.9 mL/min (40 cm/s). The mass spectrometer is
operated in electron capture negative ionization mode using
methane as the ionizing gas (ion source pressure of 1.8 torr).
The injector temperature is 250  C, the transfer line is 280  C,
and the ion source and quadrupole temperatures are both
150  C. Injections (2 μL) are made using an Agilent 6890
7683B Series Injector Autosampler in splitless mode for the
first 2 min. The initial column temperature is 180  C, pro-
grammed at 20  C/min from 180  C to 270  C, then 5  C/
min from 270  C to 300  C, then 2  C/min from 300  C to
310  C, and then 25  C/min from 310  C to 320  C and held
at 320  C for 4 min. The total run time is 20.80 min.
2. Lipid oxidation products are detected by retention time
(RT) comparison with authentic standards and selected ion
monitoring (SIM) as follows: F2-IsoPs (8-iso-PGF2α) at RT
12.66 min and mass-to-charge (m/z) 569; 8-iso-PGF2α-d4 at
RT 12.68 min and m/z 573; 8,12-iso-iPF2α-VI-d11 at RT
12.85 min and m/z 580; F3-IsoPs (8-iso-PGF3α) at RT
12.76 min and m/z 567; NeuroPs at RT 14.50 min and m/z
593; and IsoFs at RT 14.90 min and m/z 585.
3. Calculating the concentration of lipid oxidation products:
Fig. 1 provides an example for calculating the concentration
of 8-iso-PGF2α in urine.
4. Identify the peak for 8-iso-PGF2α at RT 12.66 min (Peak A) by
comparison with the standard for 8-iso-PGF2α-d4 at RT
12.68 min (Peak B). Integrate the area of Peak A.
5. Identify and integrate the area of the internal standard (Peak C)
8,12-iso-iPF2α-VI-d11 at RT 12.85 min and m/z 580.
6. (a) To quantify F2-IsoP concentration in a 24-h urine collec-
tion, use the formula:
Equation 1: Area Peak A/Area Peak C  5 ng  1000 μL/
200 μL  1000  1000  1/354.5 ¼ pM.
(b) To calculate the total F2-IsoP over a 24-h period, use the
formula:
Equation 2: Eq. 1  24-h urine volume (L) ¼ pmol/24 h.
(c) To calculate the total F2-IsoP in a spot urine, use the
formula:
Equation 3: Eq. 1  1/urine creatinine (mmol/L) ¼ pmol/
mmol creatinine.
290 Anne Barden and Trevor A. Mori

A
A 8-iso-PGF2α
Abundance
m/z 569
12000
12.66
10000

8000

6000

4000

2000

11.20 11.40 11.60 11.80 12.00 12.20 12.40 12.60 12.80 13.00 13.20 13.40 13.60 13.80 Time

B
Abundance 12.68
140000
12.85
12000
120000

100000
C 8,12-iso-iPF2α-VI-d11
80000 B 8-iso-PGF2α-d4
m/z 580
m/z 573
60000

40000

20000

11.20 11.40 11.60 11.80 12.00 12.20 12.40 12.60 12.80 13.00 13.20 13.40 13.60 13.80 Time

Fig. 1 GC-MS analysis with electron capture negative ionization of (Panel a) a urine sample showing the peak
at RT ¼ 12.66 min with m/z ¼ 569 corresponding to F2-IsoPs (8-iso-PGF2α) and (Panel b) the internal
standards 8-iso-PGF2α-d4 at RT ¼ 12.68 and m/z ¼ 573 and 8,12-iso-iPF2α-VI-d11 at RT ¼ 12.85 and m/
z ¼ 580

4 Notes

1. NaOH is caustic and should be handled with caution. Special

care is required to prepare a solution of NaOH because consid-
erable heat is liberated by the exothermic reaction. Add NaOH
to water a little at a time with stirring, and then dilute the
solution to make up to a final volume of 100 mL.
2. Liquid nitrogen should be handled in well-ventilated areas.
Handle the liquid slowly to minimize boiling and splashing.
Use tongs to withdraw objects immersed in liquid nitrogen.
 GC-MS Analysis of Lipid Autoxidation Products in Biological Samples 291

3. Chloroform should be used in a fume hood to avoid vapors.

Care should be taken to avoid contact with the skin.
4. Concentrated acetic acid and HCl should be handled with care
in a fume hood. Wear gloves when handling, and always add the
acid to water not the reverse.
5. Use PFBBr, DIPEA, BSTFA with 1% TMS, and pyridine in a
fume hood.
6. Flush the culture tubes with a stream of nitrogen gas and cap
immediately.
7. Samples can also be cooled by placing them in ice.
8. Ensure all the solvent has been eluted from the cartridge, and
leave the vacuum on between steps.
9. Ensure the glass culture tube is completely dry of solvents.
Samples can be stored at
80  C overnight prior to
derivatization.
10. Ensure all solvents are removed prior to commencing the
next step.
11. Urine samples use 8,12-iso-iPF2α-VI-d11 as internal standard
because of an interfering peak in the gas chromatography
chromatogram at RT 12.68 min that co-elutes with the 8-iso-
PGF2α-d4 internal standard.

Acknowledgments

This work was in part funded by the National Health and Medical
Research Council of Australia, the National Heart Foundation of
Australia, and the Royal Perth Hospital Medical Research
Foundation.

References

1. Halliwell B, Gutteridge JMC (2015) Free radi- free radical-catalyzed mechanism. Proc Natl
cals in biology and medicine. Fifth edition edn. Acad Sci U S A 87(23):9383–9387. https://
Oxford University Press, Oxford doi.org/10.1073/pnas.87.23.9383
2. Galano JM, Mas E, Barden A, Mori TA, 4. Kadiiska MB, Gladen BC, Baird DD,
Signorini C, De Felice C, Barrett A, Opere C, Germolec D, Graham LB, Parker CE,
Pinot E, Schwedhelm E, Benndorf R, Roy J, Le Nyska A, Wachsman JT, Ames BN, Basu S,
Guennec JY, Oger C, Durand T (2013) Iso- Brot N, FitzGerald GA, Floyd RA, George M,
prostanes and neuroprostanes: total synthesis, Heinecke JW, Hatch GE, Hensley K, Lawson
biological activity and biomarkers of oxidative JA, Marnett LJ, Morrow JD, Murray DM,
stress in humans. Prostaglandins Other Lipid Plastaras J, Roberts LJ, Rokach J, Shigenaga
Mediat 107:95–102. https://doi.org/10. MK, Sohal RS, Sun J, Tice RR, Van Thiel
1016/j.prostaglandins.2013.04.003 DH, Wellner D, Walter PB, Tomer KB,
3. Morrow JD, Hill KE, Burk RF, Nammour TM, Mason RP, Barrett JC (2005) Biomarkers of
Badr KF, Roberts LJ (1990) A Series of prosta- oxidative stress study II. Are oxidation pro-
glandin-F2-like compounds are produced ducts of lipids, proteins, and DNA markers of
in vivo in humans by a noncyclooxygenase, CCl4 poisoning? Free Radic Biol Med 38
292 Anne Barden and Trevor A. Mori

(6):698–710. https://doi.org/10.1016/j.fre 9. Barden A, Mas E, Henry P, Durand T, Galano

eradbiomed.2004.09.017 JM, Roberts LJ, Croft KD, Mori TA (2011)
5. Galano JM, Lee YY, Durand T, Lee JCY The effects of oxidation products of arachido-
(2015) Special issue on “analytical methods nic acid and n3 fatty acids on vascular and
for oxidized biomolecules and antioxidants” platelet function. Free Radic Res 45
the use of isoprostanoids as biomarkers of oxi- (4):469–476. https://doi.org/10.3109/
dative damage, and their role in human dietary 10715762.2010.544730
intervention studies. Free Radic Res 49 10. Nourooz-Zadeh J, Liu EHC, Anggard EE,
(5):583–598. https://doi.org/10.3109/ Halliwell B (1998) F-4-isoprostanes: a novel
10715762.2015.1007969 class of prostanoids formed during peroxida-
6. Milne GL, Dai Q, Roberts LJ (2015) The tion of docosahexaenoic acid (DHA). Biochem
isoprostanes-25 years later. BBA-Mol Cell Biophys Res Commun 242(2):338–344.
Biol L 1851(4):433–445. https://doi.org/ https://doi.org/10.1006/bbrc.1997.7883
10.1016/j.bbalip.2014.10.007 11. Barden AE, Mas E, Croft KD, Phillips M, Mori
7. Fessel JP, Jackson Roberts L (2005) Isofurans: TA (2014) Minimizing artifactual elevation of
novel products of lipid peroxidation that define lipid peroxidation products (F-2-isoprostanes)
the occurrence of oxidant injury in settings of in plasma during collection and storage. Anal
elevated oxygen tension. Antioxid Redox Sig- Biochem 449:129–131. https://doi.org/10.
nal 7(1–2):202–209. https://doi.org/10. 1016/J.Ab.2013.12.030
1089/ars.2005.7.202 12. Proudfoot J, Barden A, Mori TA, Burke V,
8. Fessel JP, Porter NA, Moore KP, Sheller JR, Croft KD, Beilin LJ, Puddey IB (1999) Mea-
Roberts LJ 2nd (2002) Discovery of lipid per- surement of urinary F(2)-isoprostanes as mar-
oxidation products formed in vivo with a sub- kers of in vivo lipid peroxidation-a comparison
stituted tetrahydrofuran ring (isofurans) that of enzyme immunoassay with gas chromatog-
are favored by increased oxygen tension. Proc raphy/mass spectrometry. Anal Biochem 272
Natl Acad Sci U S A 99(26):16713–16718. (2):209–215. https://doi.org/10.1006/abio.
https://doi.org/10.1073/pnas.252649099 1999.4187S0003-2697(99)94187-8.
 Part IV

CE-MS-Based Metabolomics
 Chapter 22

Metabolic Profiling of Urine by Capillary Electrophoresis-

Mass Spectrometry Using Non-covalently Coated
Capillaries
Rawi Ramautar

Abstract
In the field of metabolomics, capillary electrophoresis-mass spectrometry (CE-MS) can be considered a
very useful analytical tool for the profiling of polar and charged metabolites. However, variability of
migration time is an important issue in CE. An elegant way to minimize this problem is the use of
non-covalently coated capillaries that is dynamic coating of the bare fused-silica capillary with solutions of
charged polymers. In this protocol, an improved strategy for the profiling of cationic metabolites in urine by
CE-MS using multilayered non-covalent capillary coatings is presented. Capillaries are coated with a bilayer
of polybrene (PB) and poly(vinyl sulfonate) (PVS) or with a triple layer of PB, dextran sulfate (DS), and
PB. The bilayer- and triple-layer-coated capillaries have a negative and positive outside layer, respectively. It
is shown that the use of such capillaries provides very repeatable migration times.

Key words Capillary electrophoresis, Mass spectrometry, Metabolomics, Cationic metabolites, Non-
covalently coated capillaries, Urine

1 Introduction

In metabolomics, CE-MS can be considered a useful analytical

technique for the global profiling of highly polar and charged
metabolites in various biological samples [1–3]. A stable CE-MS
method is crucial to obtain reproducible results. For example, the
stability of analyte migration times is of utmost importance in
metabolomic studies where multiple biological samples have to be
profiled and compared [4]. When conventional bare fused-silica
capillaries are used, the analysis of biological samples with minimal
sample pretreatment may lead to adsorption of matrix components
to the capillary wall causing detrimental changes of the electro-
osmotic flow (EOF) and, therefore, analyte migration times. More-
over, separation efficiencies may be compromised as a result of
adverse analyte-capillary wall interactions.

Martin Giera (ed.), Clinical Metabolomics: Methods and Protocols, Methods in Molecular Biology, vol. 1730,
https://doi.org/10.1007/978-1-4939-7592-1_22, © Springer Science+Business Media, LLC 2018

295
296 Rawi Ramautar

Over the past few years, various strategies have been developed
to correct for migration time-shifts and for aligning electrophero-
grams in CE-MS-based metabolomic studies [5–7]. However, these
procedures are not very effective for aligning metabolites showing
strong migration time-shifts among different samples, notably for
late-migrating compounds. An attractive way to minimize these
issues is the use of non-covalently coated capillaries, i.e., dynamical
coating of bare fused-silica capillaries with charged polymers [8]. So
far, various CE-MS methods employing non-covalently coated capil-
laries have been developed for the highly efficient and repeatable
analysis of proteins, peptides, and metabolites in various matrices
[9–12].
Recently, the utility of CE-MS using non-covalently coated
capillaries with layers of charged polymers has been demonstrated
for the highly repeatable metabolic profiling of urine [13]. Capil-
laries were coated with a bilayer of polybrene (PB) and poly(vinyl
sulfonate) (PVS) or with a triple layer of PB, dextran sulfate (DS),
and PB. The bilayer and triple-layer coatings were evaluated at low
and high pH separation settings, thereby providing separation
conditions for basic and acidic metabolites. In this chapter, atten-
tion is paid to the methodological aspects of the bilayer and triple-
layer capillary coatings in CE-MS for the profiling of cationic
metabolites in urine from rats using minimal sample pretreatment.
It is shown that the use of these easy to produce capillary coatings
of charged polymers significantly improves the performance of
CE-MS for urinary metabolomic studies.

2 Materials

Prepare all solutions using ultrapure water (prepared by purifying

deionized water to obtain a sensitivity of 18 MΩ-cm at 25
C) and
analytical grade reagents.

2.1 Solutions 1. Background electrolyte (BGE) solution: 1 M formic acid,

and Samples pH 2.0. Add 9.6 mL of water into a 10 mL glass vial and add
for Analysis 0.4 mL of concentrated formic acid to the water in a fume
hood. Mix the solution thoroughly using a vortex. Store at
4
C.
2. Metabolite standard mixture: Prepare stock solutions of 1 mg/
mL of creatinine, dopamine, adrenaline, L-phenylalanine, L-tyro-
sine, glutathione, folic acid, guanosine, and hippuric acid by
dissolving appropriate amounts in water. Make aliquots of stock
solutions by dilution with BGE to obtain a working solution of
20 μg/mL for each analyte. Store stock and working solutions at
80
C when not in use.
 Metabolic Profiling by CE-MS using Non-Covalently Coated Capillaries 297

3. 10% (m/v) polybrene (PB) solution: Add 9.0 mL of water into

a 10 mL glass vial and add 1.0 g of PB to the water in a fume
hood. Mix the solution thoroughly using a vortex. Store at
4
C.
4. 3% (m/v) dextran sulfate (DS) solution: Add 9.7 mL of water
into a 10 mL glass vial and add 300 mg of DS to the water in a
fume hood. Mix the solution thoroughly using a vortex mixer.
Store at 4
C.
5. 5% (v/v) poly(vinyl sulfonate) (PVS) solution: Add 9.5 mL of
water into a 10 mL glass vial and add 0.5 mL of PVS to the
water in a fume hood. Mix the solution thoroughly using a
vortex mixer. Store at 4
C.
6. Sheath-liquid solution for CE-MS analysis: Mix 50 mL of water
with 50 mL methanol. Add 100 μL of concentrated formic acid
to this solution. Mix the solution thoroughly using a vortex
mixer.

2.2 Analytical 1. CE-MS: A commercially available CE equipment (Sciex,

Equipment P/ACE ProteomeLab PA 800) is coupled to MS via a coaxial
sheath-liquid interface (Agilent Technologies).
2. CE separation: Commercially available fused-silica capillaries
(dimensions: 50 μm ID  100 cm total length) are coated
with charged polymers for electrophoretic separations.

3 Methods

The protocol described here for CE-MS using non-covalently

coated capillaries for metabolic profiling studies is for laboratory
use only. Prior to using this protocol, consult all relevant material
safety data sheets (MSDS). Please use all appropriate laboratory
safety procedures, including safety glasses, lab coat, and gloves,
when performing the experiments described in this protocol.

3.1 Preparation 1. Prior to CE-MS analysis, mix the urine sample with BGE (1:1,
of Urine Samples v/v) and centrifuge for 10 min at 4
C and 16,100  g. Rat
urine samples used to obtain the here presented results were
kindly provided by AstraZeneca (Department of Drug Metab-
olism and Pharmacokinetics, Macclesfield, UK) and stored at
80
C when not in use.

3.2 Preparation 1. Place a new bare fused-silica capillary in the CE instrument and
of the Bilayer-Coated rinse with water at 1380 mbar for 5 min. Assess whether drop
Capillary formation is observed at the end of the capillary (see Note 1).
2. Rinse the separation capillary with 1 M NaOH at 1380 mbar for
15 min and then with water at 1380 mbar for 15 min (see Note 2).
298 Rawi Ramautar

3. Rinse the separation capillary with 10% (m/v) PB solution at

350 mbar for 15 min and then with water at 1380 mbar for
5 min. Subsequently, flush the capillary with 5% (v/v) PVS
solution at 350 mbar for 30 min and finally with water at
1380 mbar for 5 min. The PB-PVS-coated capillary is now
ready for use (see Note 3).

3.3 Preparation 1. Place a new bare fused-silica capillary in the CE instrument and
of the Triple-Layer- rinse with water at 1380 mbar for 5 min. Assess whether drop
Coated Capillary formation is observed at the end of the capillary (see Note 1).
2. Rinse the separation capillary with 1 M NaOH at 1380 mbar
for 15 min and then with water at 1380 mbar for 15 min.
3. Rinse the separation capillary with 10% (m/v) PB solution at
350 mbar for 15 min and then with water at 1380 mbar for
5 min. Subsequently, flush the capillary with 3% (m/v) DS
solution at 350 mbar for 15 min, followed by water at
1380 mbar for 5 min.
4. Rinse the separation capillary with 10% (m/v) PB solution at
350 mbar for 15 min and finally with water at 1380 mbar for
5 min. The PB-DS-PB-coated capillary is now ready for use.

3.4 Performance 1. Prior to CE-MS analysis, assess first in CE-UV mode using
Assessment absorbance detection at 200 nm the performance of the
of Bilayer- and Triple- bilayer- and triple-layer-coated capillaries with the cationic
Layer-Coated metabolite standards.
Capillaries 2. Add 20 μL of the cationic metabolite standard mixture into an
empty 100 μL microvial (PCR vial) which fits into a CE vial and
put this vial in the inlet sample tray.
3. Rinse the bilayer- or triple-layer-coated capillary with BGE at
1380 mbar for 5 min followed by sample injection at 35 mbar
for 30 s (~15 nL).
4. Apply a voltage of þ30 kV with bilayer or 30 kV with triple-
coated capillary using a ramp time of 1 min and start acquiring
UV absorbance data at 200 nm.
5. For CE analysis with bilayer-coated capillary, assess the
recorded data by determining the migration times and the
plate numbers of the analyzed cationic metabolite mixture.
Check whether the compounds appear in the region between
10 and 18 min (Fig. 1a) and whether the plate numbers are
between 100,000 and 300,000 (see Note 4).
6. For CE analysis with triple-layer-coated capillary, assess the
recorded data by determining the migration times and the
plate numbers of the analyzed cationic metabolite mixture.
Check whether the compounds appear in the region between
 Metabolic Profiling by CE-MS using Non-Covalently Coated Capillaries 299

Fig. 1 CE-UV analysis of a test mixture of cationic metabolites using (a) a bilayer-coated capillary and (b) a
triple-layer-coated capillary. Experimental conditions: BGE, 1 M formic acid (pH 2.0); sample injection,
35 mbar for 30 s; detection wavelength, 200 nm (reproduced from ref. 13 with permission)

20 and 60 min (Fig. 1b) and whether the plate numbers are
between 100,000 and 300,000 (see Note 4).
7. Repeat the CE analysis of the cationic metabolite mixture ten
times by both the bilayer- and triple-layer-coated capillaries and
300 Rawi Ramautar

determine whether the variation for migration times is below

1% for each test compound (see Note 4).
8. For multiple/repeated CE analysis with bilayer-coated capil-
lary, rinse the coated capillary with 5% (v/v) PVS solution at
1380 mbar for 5 min and then with BGE at 1380 mbar for
5 min between runs (see Note 5).
9. For multiple/repeated CE analysis with triple-layer-coated cap-
illary, rinse the coated capillary with 1% (m/v) PB solution at
1380 mbar for 5 min and then with BGE at 1380 mbar for
5 min between runs (see Note 5).

3.5 CE-MS Analysis 1. Prior to CE-MS analysis, ensure that the height of the BGE
of Metabolite vials in the CE instrument matches the height of the coaxial
Standards sheath-liquid sprayer tip.
and Biological 2. Insert the outlet part of the coated CE capillary into the coaxial
Samples sheath-liquid interface in such a way that less than 1 mm of the
capillary is protruding from the electrospray needle (see Note
6). Add the sheath liquid at a flow rate of 4 μL/min.
3. Rinse the coated separation capillary with BGE at 1380 mbar
for 10 min in the forward direction.
4. Analyze the cationic metabolite mixture by CE-MS with the
bilayer- and triple-layer-coated capillaries using an injection
volume of circa 15 nL (35 mbar for 30 s).
5. Apply a voltage of þ30 kV with bilayer or 30 kV with triple-
layer-coated capillary using a ramp time of 1 min and start
acquiring MS data in the m/z range from 50 to 1000 for
metabolic profiling using an ESI voltage of 4.5 kV.
6. Assess whether the results obtained by CE-MS are comparable
to the results obtained by CE-UV for migration times and plate
numbers (see Note 7).
7. Between sample injections by CE-MS, rinse the coated capil-
lary with water, 1% (m/v) PB or 5% (v/v) PVS solution, and
BGE, each at 1380 mbar for 5 min. During these rinsing steps,
ensure that the end-plate voltage, capillary voltage, and the
nebulizer gas of the MS instrument are set to 0 (see Note 8).
8. Apply the same procedure used for CE-MS analysis of the
metabolite standards to cationic metabolic profiling of (rat)
urine samples with both coated capillaries. A typical profile
obtained for cationic metabolites in rat urine with the bilayer-
and triple-layer-coated capillaries is shown in Fig. 2.
9. Repeat the CE-MS analysis of the (rat) urine sample ten times
using both the bilayer- and triple-layer-coated capillaries and
determine whether the variation for migration times is below
1% for the following endogenous compounds: creatinine, phe-
nylalanine, and tyrosine (see Note 9).
 Metabolic Profiling by CE-MS using Non-Covalently Coated Capillaries 301

Fig. 2 Metabolic profiles (base peak electropherograms) obtained during CE-MS analysis of rat urine using (a)
a bilayer-coated capillary and (b) a triple-layer-coated capillary. Experimental conditions: BGE, 1 M formic
acid (pH 2.0); sample injection, 35 mbar for 30 s; data acquired for mass range from 50 to 1000 m/z
(reproduced from ref. 13 with permission)
302 Rawi Ramautar

10. After analysis of the urine samples, analyze the cationic metab-
olite standard mixture to determine whether the performance
of the CE-MS method using coated capillaries is still adequate
in terms of expected migration times, plate numbers, and
detection sensitivity (see Note 7).
11. After the analyses or when not in use, rinse the coated capil-
laries with water at 1380 mbar for 15 min and store both the
inlet and outlet part of the capillary in a vial containing water.

4 Notes

1. If no drop formation is observed at the end of the capillary,

repeat this step at a pressure of 3500 mbar. If no drop is
observed under these conditions, then remove a small piece
of the capillary at the inlet and outlet using a capillary cutter. If
drop formation is still not observed after this procedure, a new
capillary needs to be installed and repeat generation of the
coatings.
2. Rinsing with 1 M NaOH solution is needed to ensure that the
inner wall of the fused-silica capillary is fully negatively charged.
Only then, the positively charged polybrene polymers will
effectively attach electrostatically to the negatively charged
inner wall.
3. For practical reasons, when rinsing the capillary with solutions
of charged polymers, a lower pressure is applied to ensure
proper attachment of the second or third polymer layer to the
previous layer via electrostatic interactions.
4. The following analytical performance data need to be obtained
by CE using coated capillaries for cationic metabolite standards
(each present at 20 μg/mL): migration time variation below 1%
for ten repeated analyses using an injection volume of circa
15 nL, and plate numbers ranging between 100,000 and
300,000. In case these data are not obtained for the metabolite
standards, then regeneration/renewing of the coated capillary
is needed.
5. In order to obtain consistent migration times for metabolite
standards and for urinary metabolic profiling using multiple
sample injections, it is crucial that between runs the outer
capillary coating is regenerated/renewed by flushing with the
charged polymer solution.
6. If the coated capillary is not properly aligned into the coaxial
sheath-liquid interface, then an instable MS signal may be
observed or electrophoretic current drops.
 Metabolic Profiling by CE-MS using Non-Covalently Coated Capillaries 303

7. In case no comparable data is obtained, then the MS instru-

ment needs to be tuned and re-calibrated or the CE capillary
needs to be renewed.
8. It is important to switch off these MS parameters during the
rinsing procedure in order to prevent that the charged polymer
solution is entering the vacuum part of the MS instrument. The
ion source of the MS instrument needs to be cleaned after 24 h
of analysis.
9. In case such data is not obtained, then the MS instrument
needs to be tuned and re-calibrated or the CE capillary needs
to be renewed.

Acknowledgment

Dr. Rawi Ramautar would like to acknowledge the financial sup-

port of the Veni and Vidi grant scheme of the Netherlands Organi-
zation for Scientific Research (NWO Veni 722.013.008 and Vidi
723.016.003).

References
1. Ramautar R, Somsen GW, de Jong GJ (2013) rats by capillary electrophoresis-mass spec-
The role of CE–MS in metabolomics. In: trometry. Anal Bioanal Chem 405
Metabolomics in practice. Wiley-VCH Verlag (8):2585–2594. https://doi.org/10.1007/
GmbH & Co. KGaA, Weinheim, Germany, pp s00216-012-6701-4
177–208. https://doi.org/10.1002/ 7. Garcia-Perez I, Whitfield P, Bartlett A,
9783527655861.ch8 Angulo S, Legido-Quigley C, Hanna-Brown-
2. Kuehnbaum NL, Britz-McKibbin P (2013) M, Barbas C (2008) Metabolic fingerprinting
New advances in separation science for meta- of Schistosoma Mansoni infection in mice
bolomics: resolving chemical diversity in a urine with capillary electrophoresis. Electro-
post-genomic era. Chem Rev 113 phoresis 29(15):3201–3206. https://doi.
(4):2437–2468. https://doi.org/10.1021/ org/10.1002/elps.200800031
cr300484s 8. Huhn C, Ramautar R, Wuhrer M, Somsen GW
3. Hirayama A, Wakayama M, Soga T (2014) (2010) Relevance and use of capillary coatings
Metabolome analysis based on capillary in capillary electrophoresis-mass spectrometry.
electrophoresis-mass spectrometry. TrAC Anal Bioanal Chem 396(1):297–314. https://
Trends Anal Chem 61:215–222. https://doi. doi.org/10.1007/s00216-009-3193-y
org/10.1016/j.trac.2014.05.005 9. Dominguez-Vega E, Haselberg R, Somsen GW
4. Ramautar R (2016) Capillary electrophoresis- (2016) Capillary zone electrophoresis-mass
mass spectrometry for clinical metabolomics. spectrometry of intact proteins. Methods Mol
Adv Clin Chem 74:1–34. https://doi.org/10. Biol 1466:25–41. https://doi.org/10.1007/
1016/bs.acc.2015.12.002 978-1-4939-4014-1_3
5. Nevedomskaya E, Derks R, Deelder AM, May- 10. Katayama H, Ishihama Y, Asakawa N (1998)
boroda OA, Palmblad M (2009) Alignment of Stable cationic capillary coating with successive
capillary electrophoresis-mass spectrometry multiple ionic polymer layers for capillary elec-
datasets using accurate mass information. Anal trophoresis. Anal Chem 70(24):5272–5277
Bioanal Chem 395(8):2527–2533. https:// 11. Catai JR, Torano JS, de Jong GJ, Somsen GW
doi.org/10.1007/s00216-009-3166-1 (2006) Efficient and highly reproducible capil-
6. Kok MG, Ruijken MM, Swann JR, Wilson ID, lary electrophoresis-mass spectrometry of pep-
Somsen GW, de Jong GJ (2013) Anionic met- tides using Polybrene-poly(vinyl sulfonate)-
abolic profiling of urine from antibiotic-treated coated capillaries. Electrophoresis 27
304 Rawi Ramautar

(11):2091–2099. https://doi.org/10.1002/ mass spectrometry. Anal Chem 74

elps.200500915 (10):2233–2239
12. Soga T, Ueno Y, Naraoka H, Ohashi Y, 13. Ramautar R, Torano JS, Somsen GW, de Jong
Tomita M, Nishioka T (2002) Simultaneous GJ (2010) Evaluation of CE methods for
determination of anionic intermediates for global metabolic profiling of urine. Electro-
Bacillus Subtilis metabolic pathways by capil- phoresis 31(14):2319–2327. https://doi.
lary electrophoresis electrospray ionization org/10.1002/elps.200900750
 Chapter 23

CE-MS for the Analysis of Amino Acids

Karina Trevisan Rodrigues, Marina Franco Maggi Tavares,
and Ann Van Schepdael

Abstract
Amino acids play an important role in clinical analysis. Capillary electrophoresis-electrospray ionization-
mass spectrometry (CE-ESI-MS) has proven to possess several characteristics that make it a powerful and
useful tool for the analysis of amino acids in clinical studies. Here we present a method for the separation
and quantitative analysis of 27 amino acids in urine based on CE-ESI-MS. The method presents an
improved resolution between the isomers Leu, Ile, and aIle, in comparison to other CE-ESI-MS methods
in the literature. This method is fast, selective, and simple and has improved sensitivity by applying a
pH-mediated stacking strategy, showing that it can be successfully used for amino acid analysis and probably
for other small cationic metabolites.

Key words Capillary electrophoresis, Mass spectrometry, Electrospray ionization, Amino acids analy-
sis, Targeted metabolomics, Urine

1 Introduction

It is already well established that amino acids play essential roles in

many different fields, including human nutrition, food science,
synthesis of drugs, and cosmetics, and play an essential role in
metabolomics to name a few. In clinical analysis, the screening of
amino acids in biological fluids can help to diagnose and to monitor
treatment of diseases. The understanding of the concentration of
amino acids is crucial for a precise diagnosis of metabolic disorders
[1–3].
Chromatography-based methods (ion chromatography,
reversed-phase liquid chromatography, hydrophilic interaction liq-
uid chromatography, or liquid chromatography-mass spectrometry
(LC-MS)) have been the most commonly applied techniques for
the analysis of amino acids in biological fluids; however, methods
reporting quantitative determination of these compounds in urine
still have drawbacks. Different column chemistries and derivatiza-
tion strategies are necessary, and some methods are limited to only a

Martin Giera (ed.), Clinical Metabolomics: Methods and Protocols, Methods in Molecular Biology, vol. 1730,
https://doi.org/10.1007/978-1-4939-7592-1_23, © Springer Science+Business Media, LLC 2018

305
306 Karina Trevisan Rodrigues et al.

few amino acids. In addition, some chromatographic modes are not

able to provide an appropriate retention time for all amino acids.
Although LC-MS has shown a great performance for amino acid
analysis, ion suppression can often affect quantification accuracy
and sensitivity [4–6]. Capillary electrophoresis (CE) has already
proven to be a valuable analytical technique for clinical analysis
[7]. Capillary electrophoresis coupled to mass spectrometry with
electrospray ionization (CE-ESI-MS) has been demonstrated to be
a powerful and attractive analytical tool for the determination and
quantification of small charged compounds in biological samples
[8]. Considering that CE offers a great mass selectivity and sensi-
tivity and MS provides the possibility of structural information, the
coupling of CE-MS increases analysis reliability without the need of
extensive and complicated sample pretreatment or the use of a
specific amino acid analyzer. Several authors have already demon-
strated the possibility of amino acid analysis by CE-MS; good
separations and solid method development have been reported
[1, 9–12]. Recently, our group has also developed a quantitative
amino acid analytical method using ion trap mass spectrometry
[13]. In CE, the choice of background electrolyte (BGE) strongly
affects the separation system. It is well known that the resolving
power for charged compounds is the difference in charge to size
ratio; thus, at a medium pH range, basic and acidic amino acids
migrate in different directions. Thus, in order to achieve simulta-
neous separation of all amino acids under investigation by CE-MS,
either low or high pH BGE can be attempted since the isoelectric
points of the amino acids under investigation range from 2.77
(aspartic acid) to 10.76 (arginine). When applying a pH value
below 2.77, each amino acid will be positively charged, and conse-
quently they will migrate toward the cathode, which in this work is
the terminal pole coupled to the mass spectrometer, and will then
be detected with high selectivity and sensitivity. To achieve the
simultaneous analysis of amino acids, 0.8 M formic acid was
employed. The choice of appropriate sheath-liquid parameters is
also very important in a CE-ESI-MS method. The optimized con-
ditions are described in Subheading 3. This chapter explains the
technical details of the CE-ESI-MS method for amino acid analysis
using urine as sample.

2 Materials

All reagents should be analytical grade. Methanol and formic acid

are LC-MS grade. All solutions should be prepared using Milli-Q
water by purifying deionized water to attain a resistivity of 18 MΩ
cm at 25
C. Amino acid standard solutions, individual stock solu-
tions of amino acids, and internal standard (methionine sulfone), at
a concentration of 10 or 100 mM, are prepared in water, containing
 CE-MS Analysis of Amino Acids 307

0.1 or 0.2 M hydrochloric acid (HCl). All stock solutions can be

stored at 4
C. However, some of the amino acid solutions should
be prepared fresh daily, due to a lack of stability (see Note 1). The
working standard mixture is prepared by diluting the stock solu-
tions with water prior to use.

2.1 CE-MS 1. Running buffer: 0.80 M formic acid containing 15% (v/v)
methanol. Initially, prepare 50 mL of 8 M formic acid stock
solution by adding 15.094 mL of 99% formic acid reagent to
34.906 mL of water. Mix well and store at 4
C. Subsequently,
prepare the BGE: mix 1 mL of the formic acid stock solution
and 1.5 mL of methanol and complete to 10 mL with water (see
Note 2).
2. Sheath liquid (SHL): 0.5% formic acid in 60% (v/v) methanol-
water. Mix 60 mL of methanol and 40 mL of a 1.25% formic
acid solution prepared in water. Store at 4
C until use.
3. 12.5% ammonium hydroxide solution (NH4OH) is prepared
by diluting 25% ammonium hydroxide, with an equal volume
of water.

2.2 Instrumentation 1. The CE-ESI-MS experiments are performed on a Beckman

P/ACE MDQ capillary electrophoresis system coupled to a
Bruker HCT Esquire 3000 plus ion trap mass spectrometer
system using an Agilent coaxial sheath-liquid CE-MS interface
(G1607A Agilent CE-ESI-MS sprayer kit) (see Note 3).
2. The sheath liquid is delivered through a KD Scientific syringe
pump (see Note 4).
3. The CE system control is performed using the Beckman
32 Karat 8.0 software, while the mass spectrometry control
and data acquisition are performed with Bruker Esquire Con-
trol™ software and the data evaluation with Bruker ESI Com-
pass Data Analysis 4.1 software.

3 Methods

3.1 Sample 1. Following the sample collection, each urine sample is divided in
Collection smaller volumes and added to Eppendorf tubes. The tubes are
and Preparation then centrifuged at 14.103  g for 15 min and immediately
stored at 80
C.
2. Before analysis, the selected Eppendorf containing sample is
thawed and again centrifuged at 14.103  g for 10 min. Then
750 μL of urine is mixed with 250 μL 0.1% formic acid contain-
ing 50 μM methionine sulfone (internal standard) immediately
before CE-MS analysis.
308 Karina Trevisan Rodrigues et al.

3.2 CE-ESI-MS 1. Before the first use, a new capillary has to be conditioned for at
Conditions least 20 min with 1 M NaOH, followed by water and BGE (see
Note 5). The capillary should be conditioned daily by applying
20 psi of water for 4 min and BGE for 5 min. Additionally, in
between analyses, the capillary is flushed with BGE for 3 min
and at the end of the day for 3 min with water and 4 min
with air.
2. Uncoated fused silica capillaries (50 μm i.d.  85 cm total
length) are used in all experiments (see Note 6).
3. For sample injection, a pH-mediated stacking is employed by
hydrodynamic injection of a 12.5% NH4OH solution for 9 s at
0.5 psi, followed by hydrodynamic injection of sample for 20 s
at 0.6 psi (see Note 7). Other CE parameters are 30 kV applied
voltage and a capillary temperature of 20
C.
4. SHL is delivered at a flow rate of 5 μL/min (see Note 8).
5. The ESI source is operated in positive mode by applying a
capillary voltage of 4.5 kV.
6. The drying gas flow is maintained at 5 L/min and the heater
temperature at 200
C. Nitrogen is used as nebulizing gas and
the pressure is maintained at 8 psi.
7. Other MS parameters comprise a scan range of m/z 50-275
with an accumulation time of 15 ms.
8. Figure 1 shows base peak and extracted ion electropherograms
of amino acids analyzed in a pooled urine sample. Table 1
shows linearity results, LOQ, and LOD of each amino acid.

4 Notes

1. All amino acid solutions should be prepared at 100 mM using

0.1 or 0.2 M HCl. Glutamic acid, aspartic acid, and tyrosine are
less soluble; thus, 0.2 M HCl has to be used. The concentration
of methionine sulfone (internal standard) is 10 mM using
0.1 M HCl as solvent. Methionine and glutamine are very
unstable compounds and can suffer from degradation; conse-
quently, the standard solutions of those amino acids should be
prepared fresh daily.
2. We recommend preparing the BGE mixture fresh daily, as
methanol can evaporate throughout the day possibly
compromising method performance.
3. To avoid a siphoning effect, the CE inlet vial has to be posi-
tioned at the same height as the sprayer tip of the mass spec-
trometer. For the same reason, the capillary tip on the inlet
should be at a similar height as the capillary tip in the sprayer.
 CE-MS Analysis of Amino Acids 309

Fig. 1 Selected base peak electropherogram (BPE) of amino acid analysis in a pooled urine sample with
extracted ion electropherograms (EIE) of each amino acid detected with the proposed CE-ESI-MS method.
Reprinted with permission from 13
Table 1
Linearity results, LOD, and LOQ of each individual amino acid analyzed with the proposed CE-ESI-MS method
310

Exact Concentration Linearity Regression p- LOD LOQ

Amino acid mass range mM (R2, n ¼ 5) error Slope Intercept F value (μmol/L) (μmol/L)
Alanine 89.048 0.20–1.0 0.9997 0.0004 0.30  0.001 0.002  0.0005 94,922 0.16 5.7 17
a-isoleucine 131.095 0.020–0.10 0.999 0.0031 5.0  0.071 0.005  0.004 4941 0.39 2.5 7.5
Arginine 174.112 0.020–0.10 0.999 0.004 8.6  0.081 0.010  0.004 11,182 0.26 1.7 5.0
Aspartic acid 133.038 0.050–0.25 0.996 0.004 1.2  0.021 0.003  0.005 3146 0.65 16 47
Asparagine 132.053 0.10–0.50 0.991 0.006 1.8  0.009 0.006  0.007 391 0.55 13 40
b-alanine 89.048 0.20–1.0 0.993 0.0025 0.38  0.006 0.0001  0.003 4467 0.98 27 80
Karina Trevisan Rodrigues et al.

Carnosine 226.107 0.020–0.10 0.999 0.011 7.1  0.11 0.014  0.014 4549 0.50 6.7 20
Citrulline 175.096 0.050–0.25 0.999 0.007 3.5  0.061 0.007  0.008 3224 0.52 7.7 23
Cysteine 121.020 0.20–1.0 0.999 0.004 0.51  0.008 0.015  0.004 3756 0.18 29 86
GABA 103.063 0.10–0.50 0.996 0.001 1.3  0.007 0.004  0.001 28,833 0.25 4.0 12
Glutamic acid 147.053 0.050–0.25 0.998 0.004 2.4  0.040 0.0008  0.005 3483 0.91 7.3 22
Glutamine 146.069 0.050–0.25 0.998 0.002 2.1  0.015 0.005  0.002 19,813 0.26 3.1 9.4
Glycine 75.032 0.20–1.0 0.998 0.0005 0.13  0.0009 0.001  0.0006 21,963 0.33 15 45
4-Hydroxy 131.058 0.050–0.25 0.999 0.002 2.6  0.026 0.011  0.003 9910 0.16 3.7 11
proline
Histidine 155.069 0.020–0.10 0.998 0.0008 4.9  0.018 0.005  0.0009 76,690 0.11 0.63 1.9
Isoleucine 131.095 0.020–0.10 0.997 0.004 3.7  0.10 0.004  0.005 1342 0.57 4.7 14
Leucine 131.095 0.020–0.10 0.996 0.007 4.9  0.17 0.005  0.009 790 0.70 6.3 19
Lysine 146.106 0.020–0.10 0.9996 0.002 3.8  0.042 0.001  0.003 8252 0.76 2.4 7.3
Methionine 149.051 0.050–0.25 0.9998 0.005 3.1  0.050 0.005  0.007 3754 0.60 7.3 22
Ornithine 132.090 0.050–0.25 0.9997 0.002 1.9  0.035 0.002  0.003 2980 0.64 4.8 15
Phenylalanine 165.079 0.050–0.25 0.996 0.001 7.3  0.013 0.015  0.002 306,822 0.074 0.6 2.4
Proline 115.063 0.020–0.10 0.999 0.008 5.0  0.038 0.014  0.010 16,965 0.41 6.7 20
Serine 105.043 0.10–0.50 0.9995 0.003 0.90  0.012 0.004  0.003 5332 0.41 12 36
Threonine 119.058 0.10–0.50 0.996 0.002 0.91  0.010 0.001  0.003 7859 0.78 10 30
Tryptophan 204.090 0.020–0.10 0.998 0.002 5.5  0.045 0.001  0.002 14,975 0.66 1.4 4.3
Tyrosine 181.074 0.020–0.10 0.998 0.002 4.1  0.034 0.007  0.002 14,869 0.19 1.8 5.4
Valine 117.079 0.050–0.25 0.997 0.004 3.8  0.037 0.005  0.005 10,523 0.51 4.3 13
CE-MS Analysis of Amino Acids
311
312 Karina Trevisan Rodrigues et al.

4. If no pulseless HPLC pump is available, we recommend using a

syringe pump instead. A non-pulse-free HPLC pump can cause
instability in the current, affecting the performance of the
CE-MS.
5. We recommend conditioning a new capillary also with sodium
hydroxide before first use, but the conditioning should be
performed with the capillary outside of the needle source to
avoid contamination of the mass spectrometer. This is an
important step for a better performance of the method.
6. The electrospray performance depends on the quality of the
capillary cut. Jagged edges prevent the formation of a uniform
spray and can also act as adsorption sites for sample compo-
nents. We recommend using a diamond blade cutter.
7. In order to obtain a better focusing of the sample zone, urine
samples should be acidified with 0.1% formic acid.
8. Low sheath-liquid flow rates are not recommended when using
a syringe pump because current drop can be observed.

References

1. Mayboroda OA, Neus€ uß C, Pelzing M et al https://doi.org/10.1016/j.jchromb.2011.

(2007) Amino acid profiling in urine by capil- 07.030
lary zone electrophoresis–mass spectrometry. J 7. Jabeen R, Payne D, Wiktorowicz J et al (2006)
Chromatogr A 1159:149–153. https://doi. Capillary electrophoresis and the clinical labo-
org/10.1016/j.chroma.2007.04.055 ratory. Electrophoresis 27:2413–2438.
2. Bouatra S, Aziat F, Mandal R et al (2013) The https://doi.org/10.1002/elps.200500948
human urine metabolome. PLoS One 8: 8. Zhong X, Zhang Z, Jiang S, Li L (2014)
e73076. https://doi.org/10.1371/journal. Recent advances in coupling capillary
pone.0073076 electrophoresis-based separation techniques to
3. De Benedetto GE (2008) Biomedical applica- ESI and MALDI-MS. Electrophoresis
tions of amino acid detection by capillary elec- 35:1214–1225. https://doi.org/10.1002/
trophoresis. In: Capillary Electrophor. elps.201300451
Humana Press, Totowa, NJ, pp 457–481 9. Soga T, Kakazu Y, Robert M et al (2004)
4. Wang C, Zhu H, Pi Z et al (2013) Classifica- Qualitative and quantitative analysis of amino
tion of type 2 diabetes rats based on urine acids by capillary electrophoresis-electrospray
amino acids metabolic profiling by liquid chro- ionization-tandem mass spectrometry. Electro-
matography coupled with tandem mass spec- phoresis 25:1964–1972. https://doi.org/10.
trometry. J Chromatogr B Analyt Technol 1002/elps.200305791
Biomed Life Sci 935:26–31. https://doi.org/ 10. Soga T, Heiger DN (2000) Amino acid analysis
10.1016/j.jchromb.2013.07.016 by capillary electrophoresis electrospray ioniza-
5. Sakaguchi Y, Kinumi T, Yamazaki T, Takatsu A tion mass spectrometry. Anal Chem
(2015) A novel amino acid analysis method 72:1236–1241. https://doi.org/10.1021/
using derivatization of multiple functional ac990976y
groups followed by liquid chromatography/ 11. Ramautar R, Mayboroda OA, Derks RJE et al
tandem mass spectrometry. Analyst (2008) Capillary electrophoresis time of flight
140:1965–1973. https://doi.org/10.1039/ mass spectrometry using noncovalently bilayer
C4AN01672F coated capillaries for the analysis of amino acids
6. Held PK, White L, Pasquali M (2011) Quanti- in human urine. Electrophoresis
tative urine amino acid analysis using liquid 29:2714–2722. https://doi.org/10.1002/
chromatography tandem mass spectrometry elps.200700929
and aTRAQ reagents. J Chromatogr B Analyt 12. Soliman LC, Hui Y, Hewavitharana AK, Chen
Technol Biomed Life Sci 879:2695–2703. DDY (2012) Monitoring potential prostate
 CE-MS Analysis of Amino Acids 313

cancer biomarkers in urine by capillary electro- validation of a CE-MS method for the targeted
phoresis–tandem mass spectrometry. J Chro- assessment of amino acids in urine. Electropho-
matogr A 1267:162–169. https://doi.org/ resis 37:1039–1047. https://doi.org/10.
10.1016/j.chroma.2012.07.021 1002/elps.201500534
13. Rodrigues KT, Mekahli D, Tavares MFM, Van
Schepdael A (2016) Development and
 Part V

NMR-Based Metabolomics
 Chapter 24

NMR Analysis of Fecal Samples

Hye Kyong Kim, Sarantos Kostidis, and Young Hae Choi

Abstract
Fecal analysis can generate data that is relevant for the exploration of gut microbiota and their relationship
with the host. Nuclear magnetic resonance (NMR) spectroscopy is an excellent tool for the profiling of fecal
extracts as it enables the simultaneous detection of various metabolites from a broad range of chemical
classes including, among others, short-chain fatty acids, organic acids, amino acids, bile acids, carbohy-
drates, amines, and alcohols. Compounds present at low μM concentrations can be detected and quantified
with a single measurement. Moreover, NMR-based profiling requires a relatively simple sample preparation.
Here we describe the three main steps of the general workflow for the NMR-based profiling of feces: sample
preparation, NMR data acquisition, and data analysis.

Key words NMR spectroscopy, Feces, Sample preparation, Data analysis, Identification, Gut
microbiota

1 Introduction

Metabolomics has been applied to the study of diverse biological

samples such as urine, plasma, and cerebrospinal fluid [1]. In recent
years, however, the study of feces has gained attention due to the
awareness of the importance of the role of gut microbes in human
health. For example, gut-microbe interaction is now considered to
play a significant role in the regulation of the human immune system
[2]. Feces result from a direct contact with the intestine and there-
fore can be explored to study gut microbial activities and relate these
with the gut health status. Gut diseases are often characterized by
dysregulation of gut microbiota and their activities. Because gut
bacteria are able to break down indigestible food components and
produce critical metabolites which cannot be produced by the host,
the results of their activity can be reflected in the metabolite profile of
the fecal samples. For example, the fecal metabolites of patients with
Crohn’s disease showed a significant decrease in short-chain fatty
acids, methylamine, and trimethylamine [3].

Martin Giera (ed.), Clinical Metabolomics: Methods and Protocols, Methods in Molecular Biology, vol. 1730,
https://doi.org/10.1007/978-1-4939-7592-1_24, © Springer Science+Business Media, LLC 2018

317
318 Hye Kyong Kim et al.

Generally, nuclear magnetic resonance (NMR) spectroscopy

and mass spectrometry (MS) are the most common analytical plat-
forms for metabolomics studies. Each platform possesses unique
advantages, and NMR though less sensitive exhibits greater repro-
ducibility and easier sample preparation than MS. Moreover, NMR
is capable to provide structural and quantitative information of
metabolites across a broad range of chemical classes, with a single
measurement and using only one internal standard. Thus, using
NMR for the profiling of fecal samples has a great potential and has
already been applied to study gastrointestinal diseases such as ulcer-
ative colitis (UC), irritable bowel syndrome (IBS), Crohn’s disease,
and helminth infection [3–6]. The generated NMR data from fecal
extracts is similar in terms of spectral complexity to that of all other
body fluids that have been well studied (e.g., urine, blood serum,
etc.), and thus, it can be processed and analyzed using the same,
univariate, and multivariate statistical methods [7, 8]. It does pres-
ent some challenges particularly in sampling due to its heteroge-
neous nature and the need to define meaningful results due to the
inherent interindividual variability of the fecal metabolome.
In this chapter, we describe the steps for the NMR-based
analysis of fecal samples. Following the sample collection and logis-
tics for human feces, which have both been recently discussed [9],
we will focus on the details of the three subsequent steps of fecal
analysis: fecal sample preparation, data acquisition, and data analy-
sis. The latter covers data processing for multivariate statistical
analysis and metabolites identification. We will not include a com-
prehensive description of the available statistical methods to analyze
the data since it is well beyond the scope of this chapter. Instead, we
will focus on the generation of data using our own experience with
both human [6] and murine specimen, as well as data available from
optimization studies and reviews from other researchers [10–12].

2 Materials

Prepare all solutions with ultrapure deionized water and analytical

grade reagents and solvents. Store all reagents and solutions at
room temperature unless indicated otherwise.

2.1 Homogenization 1. Phosphate buffer 0.15 M (K2HPO4/KH2PO4, pH 7.4): pre-

of Feces Samples pare 100 mL solution of 0.15 M K2HPO4 and 0.2 mM NaN3
(solution 1) and 100 mL of 0.15 M KH2PO4 and 0.2 mM
NaN3 (solution 2) both in deionized water. Add solution 2 to
solution 1, dropwise, until a pH of 7.4 is reached.
2. Mechanical homogenizer: bullet blender with ~ 5
1 mm
zirconium oxide beads per sample (see Note 1).
 NMR Analysis of Fecal Samples 319

3. Centrifuge with temperature control and capable of speeds

>10,000
g.
4. Microcentrifuge tubes (see Note 2).

2.2 NMR Sample 1. Internal standard solution: 4 mM sodium 3-trimethylsilyl

Preparation (2,2,3,3-d4) propionate (TSP) in D2O (99.9% D).
2. 96 Ritter well plates if a robotic liquid handler is available.
3. 5 or 3 mm NMR tubes (see Note 3).

2.3 NMR 1. Autosampler integrated to an NMR instrument and software

Instrumentation for automation.
and System 2. Methanol-d4 (99.98% D) for temperature calibration: prefera-
Optimization bly ampules of 750 μL.
3. Aqueous solution with reference compounds for shimming
optimization: generally provided by the manufacturer. For
Bruker instruments, the standard sucrose sample in 9:1
H2O/D2O, including sodium 3-(trimethylsilyl) propanesulfo-
nic-d6 acid (DSS-d6) as internal standard, can be used.

3 Methods

3.1 Processing 1. Defrost samples at room temperature (about 1 h).

of Fecal Samples 2. Weigh approximately 300 mg into a microcentrifuge tube and
add ~ 5
1 mm zirconium oxide beads (see Note 4).
3. Add sufficient phosphate buffer 0.15 M, pH 7.4, to obtain a ratio
of fecal sample (mg) to volume of buffer (μL) 1:2 (see Note 5).
4. Homogenize the sample by bead beating using a bullet blender
for 30 s. Observe the samples and repeat this step if fecal slurry
is not homogeneous.
5. Centrifuge at 16,000
g for 15 min, at 4  C (see Note 6).
6. Transfer 600 μL of the supernatant (see Note 7) of each sample
to a new microcentrifuge tube and store at 80  C until NMR
analysis. If NMR measurements can be performed immediately,
then proceed to the next section.

3.2 NMR Sample 1. Defrost samples at room temperature.

Preparation 2. Add 60 μL of the internal standard solution (4 mM TSP in
D2O) to each aqueous fecal sample to reach the final concen-
tration of TSP in the NMR sample of 0.4 mM. D2O will be
used for the field lock and TSP as the chemical shift reference
and quantification standard (see Note 8).
3. Vortex for 10 s.
4. Centrifuge at 16,000
g for 10 min, at 4  C.
320 Hye Kyong Kim et al.

5. Transfer 560 μL to a 5 mm NMR tubes using a pipette or a

robotic liquid handler. If 3 mm tubes are used, transfer 165 μL
with either a syringe or a robotic liquid handler (see Note 9).
6. Place tubes in the autosampler for NMR measurements. If
temperature regulation of the autosampler system is not avail-
able, then use batches of a small number of samples (e.g., up to
6–8), and store the remaining samples temporarily at 4  C for
no longer than 24 h (see Note 10).
7. Mix all leftover solutions from step 5 to prepare pool samples.
Ideally, one pool sample per group of samples of the study
should be made. Vortex for 10 s and store all of them at
80  C, keeping one sample that will be used for the NMR
spectrometer optimization.

3.3 NMR Instrument 1. Optimize all axes (3D) shims using the aqueous reference
Setup solution, generally provided by the instrument manufacturer.
The internal standard (TSP or DSS) peak half-height linewidth
without line broadening should be <0.7 Hz for acceptable
resolution.
2. Calibrate probe temperature with a fresh sample of pure meth-
anol-d4 to obtain an actual sample temperature of 300 K (i.e.,
27  C). To achieve this, acquire a short single-pulse 1H NMR
experiment (two scans). Apply a line broadening of 3 Hz and
calculate the actual temperature from the distance between the
two methanol peaks (CH3 and OH) (see Note 11). Repeat the
experiment with appropriate adjustments of the temperature
setting until an exact temperature of 300 K (27  C) is achieved.
Save the temperature setting.
3. Optimize shims for the fecal aqueous samples using a sample of
similar matrix. It is preferable to use a pooled sample made
from the samples of the study. Store the optimized shims file
and use it as the starting point prior to the shimming of each
sample during the automatic acquisition process.
4. Use the 1D 1H NOESY experiment with presaturation to
optimize the frequency in which the presaturation pulse for
the water signal suppression will be applied. For aqueous fecal
samples, a bandwidth of 50 Hz is required to sufficiently sup-
press the water signal. Store the optimum frequency value
(parameter O1 for Bruker instruments) at which the residual
water peak does not affect the spectral baseline outside of
4.65–5.1 ppm when a cryoprobe is used or 4.7–4.9 for the
room-temperature operating inverse probes.
5. Optimize profiling parameters using one of the pooled samples.
Two experiments per sample should be used for fecal extract
profiling, a 1D 1H NOESY (preferably with z-gradients; see
Note 12) and a 2D 1H J-resolved (JRES). Except for the
 NMR Analysis of Fecal Samples 321

water irradiation frequency, described in the previous step, the

experimental time and the receiver gain (RG) should be adjusted
in this step. The optimum RG value is automatically calculated.
The number of scans (NS) should be manually adjusted based
on a number of factors. The most important are the signal-to-
noise ratio (SNR) of the weakest signals observed (e.g., the
singlet of fumarate or aromatic compounds can be used) and
the aim of the study (see Note 13). Typically, for a 600 MHz
NMR instrument equipped with a cryoprobe, 64–128 scans
(7–15 min) can be used for the 1D NOESY experiment. Other
standard parameters are 20 ppm spectral width, 10 ms mixing
time, 4 s relaxation delay, and 64 K FID data points (see Note
14). The 2D JRES experiment should be acquired using the
same pulses as the 1D NOESY (90-degree pulse and presatura-
tion) and the same RG. The relaxation delay should be 2 s,
12,288 data points in the direct dimension (F2), 2 transients
over 40 increments in the indirect dimension (F1), 10,000 Hz
spectral width in F2, and 78 Hz in F1. Store the parameters of
the two experiments so that they can be used for the automatic
profiling.

3.4 NMR Profiling 1. Prior to data acquisition, each sample should be allowed to rest
of Fecal Extracts in the probe for 5 min to reach the required temperature with
and Data Processing less than 0.1 K fluctuation.
2. The following steps should be performed automatically using
the software routines provided by the manufacturer of the
instrument: load the stored shim file settings, tune and match
the probe, field lock, 1D (z-axis) shimming, autophase of the
lock signal, and calculation of the 90-degree pulse length
[13]. The calculated pulse is automatically implemented in
the two saved profiling experiments (1D NOESY and JRES)
and the acquisition starts.
3. Use the automation routines provided by the software to pro-
cess the spectra: Fourier transformation, phase and baseline
correction, calibration of chemical shift scale to TSP signal at
0.00 ppm, and apodization (exponential multiplication of FID)
with line broadening of 0.3 Hz.
4. Acquire 2D NMR spectra, using the pooled samples. Typically,
the homonuclear 1H-1H total correlation spectroscopy
(TOCSY) and the heteronuclear 1H-13C heteronuclear single
quantum correlation (HSQC) experiments are collected in
addition to the profiling experiments used for all samples (see
Note 15).

3.5 NMR Data 1. Observe all spectra and if necessary apply manual baseline
Analysis correction. A typical 1H NMR spectrum of human feces is
shown in Fig. 1.
322 Hye Kyong Kim et al.

Fig. 1 1H NMR spectrum of human feces. In total 60 fecal metabolites were identified as 1, 2-methylbutyrate;
2, valerate; 3, n-butyrate; 4, leucine; 5, isoleucine; 6, valine; 7, propionate; 8, isobutyrate; 9, 3-methyl-2-
oxoisovalerate; 10, 2-oxoisovalerate; 11, ethanol; 12, 3-hydroxybutyrate; 13, threonine; 14, lactate; 15, -
2-hydroxyisobutyrate; 16, 3-hydroxy-2-butanone; 17, alanine; 18, lysine; 19, thymine; 20, acetate; 21, -
5-aminopentanoate; 22, ornithine; 23, proline; 24, glutamate; 25, methionine; 26, glutamine; 27, succinate;
28, 2-oxoglutarate; 29, 3-phenylpropionate; 30, aspartate; 31, methylamine; 32, malate; 33, trimethylamine;
34, tyrosine; 35, malonate; 36, choline; 37, D-glucose; 38, taurine; 39, methanol; 40, glycine; 41, D-xylose;
42, D-galactose; 43, fructose; 44, dihydroxyacetone; 45, uracil; 46, fumarate; 47, urocanate; 48, ethanolamine;
49, xanthine; 50, hypoxanthine; 51, nicotinate; 52, 3-hydroxyphenylacetate; 53, tryptophan; 54, phenylalanine;
55, orotate; 56, UDP-glucuronate; 57, formate; 58, benzoate; 59, 4-aminohippurate; 60, homovanillate;
61, putrescine; 62, asparagine. Adapted from [7] with great acknowledgment

2. Identify the regions that do not contain relevant information,

e.g., the water region.
3. Convert NMR data to a format suitable for multivariate data
analysis using either proprietary (e.g., AMIX from Bruker
BioSpin) or freely available NMR processing software within
MATLAB or the R statistical environment [14]. In this step,
the dimensions of the data will be reduced by dividing it into
small integrated regions (bins). The integral of each bin is
calculated and used as a new variable. The same region, which
might contain one or more NMR peaks, is calculated for all
samples, resulting in a new data set of a number of rows equal
to the number of samples and a number of columns equal to
the number of bins. The bin size ranges from 0.0025 to
 NMR Analysis of Fecal Samples 323

0.04 ppm and can either be constant or of a variable size (see

Note 16).
4. Normalize the data. Since the weight of each fecal sample has
been recorded, the total spectral area, i.e., the sum of bins, can
be normalized to that weight. This can be performed in Excel
or by routines that are implemented with the software used for
statistical analysis (see Note 17).
5. Scale the data so that all variables have an equal importance for
the development of the statistical model. Usually the autoscale
method or Pareto scale is used, i.e., each variable is divided by
its standard deviation or the root of the standard deviation,
respectively. This can be performed in Excel or by routines that
can be implemented with the software used for statistical anal-
ysis, e.g., SIMCA (Umetrics, Umea, Sweden) (see Note 17).
6. Perform principal component analysis (PCA), using available
software such as AMIX, SIMCA, MATLAB, or freely available
packages within the R statistical environment. PCA is an excel-
lent tool to reveal underlying trends of such complex data sets
such as grouping of samples. Possible outliers, e.g., measure-
ment errors or sample outliers, can also be detected with PCA
(see Note 18).
7. Depending on the study design, supervised statistical methods
such as partial least square discriminant analysis (PLS-DA) or
orthogonal PLS-DA (OPLS-DA) can also be performed to get
an overview of the data set (see Note 18).
8. Perform the validation of the statistical model using either a
test set, consisting of un-modeled samples (cross-validation),
or permutation tests (see Note 19).
9. Analyze the score plots and loading plots of the first two to
three components (e.g., those that explain most of the total
explained variance) to identify the bins that contribute to
group separation.

3.6 Identification 1. Select the significant bins emerging from the data analysis.
of Fecal Metabolites 2. Compare the NMR signals included in these bins to available
web databases and the literature (see Notes 20 and 21).
3. Verify the assigned metabolites with 2D–NMR method spectra
of the pooled samples. Use the TOCSY data to connect the
1
H-1H correlations and the HSQC to annotate singlets and
severely overlapped multiplets using the 1H and 13C chemical
shifts of compounds.

3.7 Quantification 1. Quantify the selected metabolites from the statistical analysis
of Metabolites using deconvolution. Some proprietary software packages for
this task are Chenomx NMR suite (Chenomx, Edmonton Inc.)
324 Hye Kyong Kim et al.

or MestreNova (MNova software, Mestrelab Research SL).

Follow the instructions provided by the software companies.
If such software is not available, look for other free options
within the R environment, such as Batman and BQuant
[15–17].
2. Calculate the concentration in each fecal sample by correcting
the measured concentration with the weight of each fecal
sample.
3. If needed, annotate the full spectra using the methods
described in Subheading 3.6 and quantify as many compounds
as possible by repeating steps 1 and 2.

4 Notes

1. We have successfully used the bead-beating method for crude

fecal homogenization. However, other efficient methods are
also available which can be used for this step such as tissue
lysers. Ultrasonication or vortexing can replace the
homogenization step.
2. When the bead-beating homogenization method is used, care-
fully select the microcentrifuge tubes compatible with the bul-
let blender machine. Check this with the manufacturer
instructions to avoid leakage while bead beating.
3. The choice of 5 or 3 mm NMR tubes depends on the probe
equipment in the first place and secondly on the amounts of
available samples. A 5 mm cryoprobe can support the analysis
with 3 mm tubes or even the smaller 1.7 mm tube due to the
reduction in thermal noise. However, the conventional 5 mm
ambient temperature operating probes typically exhibit best
signal-to-noise ratio (SNR) with 5 mm tubes.
The smaller tubes are ideal if a limited amount of sample is
available as they require much less volume, 165 μL and 30 μL
for 3 mm and 1.7 mm tubes, respectively, compared to the
560 μL needed for a 5 mm tube.
4. The weight can be changed depending on the availability of the
sample. If murine stool specimens are to be analyzed, the
weight of available material may be lower. Also, the amount
needed per sample depends on the available instrumentation;
see Note 3.
In general the ratio of weight of fresh fecal sample to buffer
(Wf/Vb) 1:2 is recommended [10].
5. Human fecal samples can be very heterogeneous and their
structure and composition also depends on the diet. Some-
times, the proposed Wf/Vb ratio of 1:2 might not be sufficient
to achieve the homogenization of the fecal slurry and the
 NMR Analysis of Fecal Samples 325

separation of a fecal aqueous extract after centrifugation. In

those cases, the volume of buffers should be increased to Wf/
Vb ratios of 1:3 to 1:10. The same is necessary if there is a
limited quantity of available material (see also Notes 3 and 4).
6. Repeat centrifugation if it is not possible to separate the fecal
water from insoluble particles. If the problem persists, homog-
enization should be repeated with additional buffer.
7. For 3 mm tubes, a volume of 200–250 μL is sufficient for the
NMR sample, with a residual solution that can be used to
prepare the pooled samples.
8. It is important to ensure the accuracy of the volume of internal
standard that is added, since the integral of its peak will be used
for the quantitation of the metabolites. If a robotic liquid
sampler is available, it should be preferred for this step.
9. If a syringe is used, wash it twice with distilled water of analyti-
cal grade between each sample preparation.
10. Fecal aqueous samples are not stable at room temperature for
longer than 5 h [9]. Therefore, depending on the time of
analysis required for each sample, small batches of samples
should be placed in the autosampler. If refrigerated rack posi-
tions are available, like with the Bruker SampleJet system (Bru-
ker BioSpin Ltd.), then sample stability is extended to 24 h.
11. The actual sample temperature can be calculated by measuring
the distance (Δδ, in ppm) between the methanol peaks in the
NMR spectrum of a 100% methanol solution. The measured
Δδ is then used in the equation, T ( C) ¼ 130.00 – 29.53Δδ –
23.87Δδ2, to extract the actual temperature [18]. In Bruker’s
TopSpin, this is automatically done with the “calctemp”
command.
12. Gradients result in better water suppression, and compared to
the non-gradient version of 1D NOESY (pulse program: noe-
sy1dphpr for Bruker systems), the benefit in SNR can be
significant with fecal water samples.
13. In some cases, the aim of studies might be the quantification of
specific metabolites rather than the full profiling of each sam-
ple. In these cases, the total acquisition time is adjusted to the
molecular targets, so that the resulted SNR is sufficient for
accurate quantification.
14. These profiling parameters are based on the pulse sequence
-RD – gz,1–90 - t - 90 - tm – gz,2–90 —ACQ. RD is the
relaxation delay of 4 s, gz,1 and gz,2 are gradients with 1 ms
duration, t is a short delay of 3 μs between pulses, 90 is the 90
radiofrequency pulse (RF), tm is the mixing time of 10 ms, and
ACQ is the acquisition of the free induction decays (FIDs), set
up to 2.72 s. The water resonance is suppressed with a
326 Hye Kyong Kim et al.

continuous RF of 50 Hz, applied during the relaxation delay

and the mixing time.
15. 2D NMR spectroscopy can be applied for further identification
of metabolites. Among various 2D NMR methods, HSQC
(heteronuclear single quantum coherence spectroscopy) is par-
ticularly useful for this purpose. HSQC detects any carbon
atom directly attached to a proton. Consequently, the chemical
shift of a 13C atom connected to a specific proton can be
obtained. Standard Bruker pulse sequences for HSQC mea-
surement (with gradients and water presaturation) are as fol-
lows: HSQC hsqcetgpprsisp2.2 2048
256, 12 ppm
(1H)
165.7 ppm (13C); spectra are processed with zero
filling in both dimensions and standard windows (900 phase-
shifted QSINE for phase-sensitive HSQC). Another useful
method is TOCSY (1H-1H total correlation spectroscopy),
which creates correlations between all protons within a given
spin system. It provides similar information to a COSY (corre-
lation spectroscopy) experiment with regard to directly cou-
pled hydrogens, but provides further structural information by
identifying interconnected groups of indirectly spin-coupled
hydrogens. Useful pulse sequences for TOCSY measurement
are as follows: TPPI phase-sensitive mode, with water presa-
turation during relaxation delay, a spectral width 6 kHz in both
dimensions, a 2 s relaxation delay, an 80 ms mixing time, 1 K
data points in f2, and 512 increments in f1. Zero filling in f1 to
1 K real data points.
16. Instead of bins with constant width, it is often very helpful to
apply adaptive binning, i.e., the width of the bin is adjusted to
the position of peaks and this position varies from sample to
sample [19]. Further details about binning of NMR spectra can
be found in the review by Forshed [20].
17. The normalization of the data is a row operation (i.e., sample-
wise), while scaling is a column operation (variable-wise). Both
can be performed with Excel. However, it is more convenient
and faster to perform the processing steps prior to statistical
analysis using dedicated software that can be found in packages
within the R statistical environment (Metabolomic package,
MUMA, etc.) or MATLAB (The MathWorks Inc., Natick,
MA). AMIX (Bruker) also offers the option of normalization,
while SIMCA (Umea, Sweden) provides the option for several
scaling methods (autoscaling, Pareto, logarithmic transform,
etc.) to an already normalized data set.
18. Different types of multivariate data analysis can be applied, as
well as reviewed in two papers [21, 22]. In general, unsuper-
vised methods such as PCA and hierarchical cluster analysis
(HCA) are the first choice to obtain an overview of samples
 NMR Analysis of Fecal Samples 327

including their distribution and identification of outliers. The

next step is to apply supervised methods to exploit prior infor-
mation and assess the distributing metabolites for class or
group separation. Multivariate data analysis can be performed
using commercially available software such as SIMCA (Ume-
trics, Umea) and the PLS toolbox from MATLAB (The Math-
Works Inc., Natick, MA). A review on the different packages is
available [14].
19. When applying a PLS-DA model or PLS modeling, validation
must be performed. Without proper validation, there is a high
risk of over-fitting with these methods. The two most widely
used validation methods are cross-validation and test-set vali-
dation. Test-set validation involves having a separate data set to
check if the results of the statistical analysis are valid on new
data and how well they fit [21]. Permutation is another way of
validating the classification model [23].
20. Several free online databases can be used to assist in the identi-
fication of metabolites including the HMDB spectral database
(the human metabolome database, http://www.hmdb.ca),
BMRB peak server (Biological Magnetic Resonance Data
Bank, http://www.bmrb.wisc.edu/metabolomics/), and the
PRIMe server (the Platform for RIKEN Metabolomics,
http://prime.psc.riken.jp). AMIX (Bruker) and the Chenomx
NMR suite software (Canada) package also offer metabolite
databases or tools to identify metabolites. The pros and cons of
these databases are discussed in another review [24].
21. Some databases include spectra recorded at a different pH so
that it is essential to ensure that the queries match the experi-
mental conditions, in this case that they are recorded at a
neutral pH (7.0–7.4).

Acknowledgments

The authors thank Dr. E.G. Wilson for her comments and review of
the manuscript.

References
1. Beckonert O, Keun HC, Ebbels TMD, (2012) Host-gut microbiota metabolic inter-
Bundy J, Holmes E, Lindon JC, Nicholson JK actions. Science 336:1262–1267
(2007) Metabolic profiling, metabolomic and 3. Bjerrum JT, Wang Y, Hao F, Coskun M,
metabonomic procedures for NMR spectros- Ludwig C, G€ unther U, Nielsen OH (2015)
copy of urine, plasma, serum and tissue Metabonomics of human fecal extracts charac-
extracts. Nat Protoc 2:2692–2703 terize ulcerative colitis, Crohn’s disease and
2. Nicholson JK, Holmes E, Kinross J, healthy individuals. Metabolomics
Burcelin R, Gibson G, Jia W, Pettersson S 11:122–133
328 Hye Kyong Kim et al.

4. Le Gall G, Noor SO, Ridgway K, Scovell L, metabolomic data analysis packages. Prog
Janieson C, Johnson IT, Colquhoun IJ, Kems- Nucl Magn Reson Spectrosc 59:263–270
ley EK, Narbad A (2011) Metabolomics of 15. Astle W, De Iorio M, Richardson S,
fecal extracts detects altered metabolic activity Stephens D, Ebbels T (2012) A Bayesian
of gut microbiota in ulcerative colitis and irri- model of NMR spectra for the deconvolution
table bowel syndrome. J Proteome Res and quantification of metabolites in complex
10:4208–4218 biological mixtures. J Am Stat Assoc
5. Marchesi JR, Holmes E, Khan F, Kochhar S, 107:1259–1271
Scanlan P, Shanahan F, Wilson ID, Wang Y 16. Hao J, Liebeke M, Astle W, Maria De Iorio M,
(2007) Rapid and noninvasive metabonomic Bundy JG, Ebbels T (2014) Bayesian deconvo-
characterization of inflammatory bowel dis- lution and quantification of metabolites in
ease. J Proteome Res 6:546–551 complex 1D NMR spectra using BATMAN.
6. Kostidis S, Kokova D, Dementeva N, Saltykova Nat Protoc 9:1416–1427
KHK, Choi YH, Mayboroda OA (2017) 1H- 17. Zheng C, Zhang S, Ragg S, Raftery D, Vitek O
NMR analysis of feces: new possibilities in the (2011) Identification and quantification of
helminthes infections research. BMC Infect metabolites in 1H NMR spectra by Bayesian
Dis 17:275 model selection. Bioinformatics
7. Saccenti E, Hoefsloot HCJ, Smilde AK, Wes- 27:1637–1644
terhuis JA, Hendriks MMWB (2014) Reflec- 18. Findeisen M, Brand T, Berger S (2007) A 1H-
tions on univariate and multivariate analysis of NMR thermometer suitable for cryoprobes.
metabolomics data. Metabolomics Magn Reson Chem 45:175–178
10:361–374 19. De Meyer T, Sinnaeve D, Van Gasse B,
8. McKenzie JS, Donarski JA, Wilson JC, Charl- Tsiporkova E, Rietzschel ER, De Buyzere
ton AJ (2011) Analysis of complex mixtures ML, Gillebert TC, Bekaert S, Martins JC, Van
using high-resolution nuclear magnetic reso- Criekinge W (2008) NMR-based characteriza-
nance spectroscopy and chemometrics. Prog tion of metabolic alterations in hypertension
Nucl Magn Reson Spectrosc 59:336–359 using an adaptive intelligent binning algo-
9. Gratton J, Phetcharaburanin J, Mullish BH, rithm. Anal Chem 80:3783–3790
Williams HRT, Mark Thursz M, Nicholson 20. Forshed J, Torgrip RJO, Åberg KM,
JK, Holmes E, Marchesi JR, Li JV (2016) Karlberg B, Lindberg J, Jacobsson SP (2005)
Optimized sample handling strategy for meta- A comparison of methods for alignment of
bolic profiling of human feces. Anal Chem NMR peaks in the context of cluster analysis.
88:4661–4668 J Pharma Biomed Anal 38:824–832
10. Lamichhane S, Yde CC, Schmedes MS, Jensen 21. Liland KH (2011) Multivariate methods in
HM, Meier S, Bertram HC (2015) Strategy for metabolomics - from pre-processing to dimen-
nuclear magnetic resonance-based metabolo- sion reduction and statistical analysis. Trends
mics of human feces. Anal Chem Analyt Chem 30:827–841
87:59305937 22. Cevallos-Cevallos JM, Reyes-De-Corcuera JI,
11. Deda O, Gika HG, Wilson ID, Theodoridis Etxeberria E, Danyluk MD, Rodrick GE
GA (2015) An overview of fecal sample prepa- (2009) Metabolomic analysis in food science:
ration for global metabolic profiling. J Pharm a review. Trends Food Sci Technol 20:557–566
Biomed Anal 113:137–150 23. Smolinska A, Blanchet L, Buydens LMC, Wij-
12. Wu J, An Y, Yao J, Wang Y, Tang H (2010) An menga SS (2012) NMR and pattern recogni-
optimized sample preparation method for tion methods in metabolomics: from data
NMR-based faecal metabonomic analysis. Ana- acquisition to biomarker discovery: a review.
lyst 135:1023–1030 Anal Chim Acta 750:82–97
13. Wu P, Gottfried O (2005) Rapid pulse length 24. Lutz NW, Sweedler JV, Wevers RA (2013)
determination in high-resolution NMR. J Ch. 1. Exploring the human metabolome by
Magn Reson 176:115–119 NMR spectroscopy and mass spectrometry,
14. Izquierdo-Garcı́a JL, Villa P, Kyriazis A, del Wishart DS. In: Methodologies for metabolo-
Puerto-Nevado L, Pérez-Rial S, Rodriguez I, mics: experimental strategies and techniques.
Hernandez N, Ruiz-Cabello J (2011) Descrip- Cambridge University Press, London, pp 3–29
tive review of current NMR-based
 Chapter 25

Quantitative Analysis of Central Energy Metabolism

in Cell Culture Samples
Sarantos Kostidis

Abstract
Nuclear magnetic resonance (NMR) is one of the key analytical platforms used in the analysis of intracellular
and extracellular metabolites. Despite the technological advances that allow for the production of high-
quality data, the sampling procedures of cultured cells are less well standardized. Different cell lines and
culture media composition require adjustments of the protocols to result meaningful quantitative informa-
tion. Here we provide the workflow for obtaining quantitative metabolic data from adherent mammalian
cells using NMR spectroscopy. The robustness of NMR allows for the implementation of the here described
protocol to other cell types with only minor adjustments.

Key words NMR spectroscopy, Mammalian cells, Metabolomics, Sample preparation, Fingerprint,
Footprint

1 Introduction

In vitro cell-based metabolomics studies have found widespread use

in many areas of research, like toxicology [1], cancer metabolism
[2], regenerative medicine [3, 4], immune metabolism [5], and
many more. The main goal in such studies is to understand the
influence of metabolism on biological effects and mechanisms and,
ultimately, integrate this information onto metabolic maps
[6]. Such information is obtained by extracting quantitative meta-
bolic data from both the intracellular and extracellular compart-
ments. Several targeted metabolomics approaches have been
developed to accomplish this goal, including methods based on
mass spectrometry (MS) [7, 8] and nuclear magnetic resonance
(NMR) [9, 10] spectroscopy. MS-based techniques provide a
broader molecular window due to the superior sensitivity com-
pared to NMR, but they are often jeopardized by ionization sup-
pression, matrix effects, and linearity issues. On the other hand,
NMR-based analysis is capable to provide quantitative information
of the core metabolism (i.e., amino acids, glycolysis, and TCA cycle

Martin Giera (ed.), Clinical Metabolomics: Methods and Protocols, Methods in Molecular Biology, vol. 1730,
https://doi.org/10.1007/978-1-4939-7592-1_25, © Springer Science+Business Media, LLC 2018

329
330 Sarantos Kostidis

intermediates), spanning as much as six orders of magnitude

despite the method’s inherent lack of sensitivity. Moreover, NMR
is a nondestructive and robust platform, while quantification can be
carried out without the need for specific internal standards.
In order to acquire quantitative cellular data, either by MS or
NMR, the metabolites have to be extracted from the cells and the
culture medium and reconstituted in a proper solvent [11]. For
NMR-based analysis, cultured cells should grow to populations in
the order of a million or more before being harvested. The main
steps of the sample preparation process are the separation of the
cells from the medium and the quenching of metabolism, followed
by the extraction and recovery of intracellular and extracellular
components. Although several studies have been conducted, aim-
ing to optimize the conditions of these main steps [12–17], there is
not any standard operating procedure that can be applied to all cell
types. Prior to every study, several evaluation steps have to be
performed in order to establish the optimum conditions for cells
sampling.
In the present chapter, we provide a workflow for sampling and
analyzing adherent mammalian cells using NMR spectroscopy.
Emphasis is given to the key steps of effective quenching, metabolic
recovery by extraction, and quantification of metabolites in the
NMR data. The potential pitfalls that often accompany each of
these steps are highlighted and discussed in Subheading 4. Using
the described method, a total of about 60–70 extracellular and
intracellular metabolites can be quantified from adherent mamma-
lian cells.

2 Materials

2.1 Sampling 1. Centrifuge equipment capable to operate at temperatures of

of Culture Medium 4
C or below and 2 mL microcentrifuge tubes.
2. Cold LC-grade 100% methanol solution, stored overnight at
80
C.
3. Dry-ice.

2.2 Quenching 1. Solution of warm PBS at 37
C: 1.9 mM Na2HPO4, 8.1 mM

Intracellular NaH2PO4, 150 mM NaCl. For every 10 cm culture dish, about
Metabolism 5 mL of PBS is needed.
of Adherent Cells 2. Small cryostorage container with liquid nitrogen (LN2):
for every 10 cm culture dish, about 10 mL of LN2 are needed
(see Note 1).
3. Aspiration pump.
4. Disposable sterile plastic serological pipettes (10 mL) and
pipettors.
 Quantification of Cellular Metabolism 331

5. Dry-ice in a large box to accommodate the frozen culture petri

dishes post quenching.
6. 2 mL microcentrifuge tubes.

2.3 Extraction 1. Microcentrifuge tubes (2 mL) and centrifuge operating at

of Intracellular temperatures of 4
C or below.
Metabolites 2. Cell scrapers.
3. Cold extraction solvent: methanol/chloroform/water,
8.1:0.9:1 (v/v/v) (see Note 2), previously stored at 80
C.
About 1.5 mL is required per culture dish of 10 cm (see Note 3).
4. Nitrogen gas or other drying system.
5. Spectrophotometer for total protein quantitation.
6. Pierce BCA protein assay kit for total protein estimation.
7. Cell counter device: automatic or manual (if BCA kit is not an
option).

2.4 NMR 1. pH meter.

Spectroscopy 2. Phosphate buffer in D2O (K2HPO4/KH2PO4, pH 7.4): make
two solutions of 50 mL in deuterated water, the first consisting
of 1.5 M K2HPO4 and 2 mM NaN3 (1) and the second 1.5 M
KH2PO4 and 2 mM NaN3 (2). Add 2–1, dropwise, until pH is
7.4. Note the final volume and add 9 volumes of D2O (dilute
10 to 0.15 M concentration). Divide the final solution in two
parts and to the first part add sodium 3-trimethylsilyl (2,2,3,3-
d4) propionate (TSP) to reach a concentration of 0.4 mM
(buffer for culture medium), while in the second part add
TSP to 0.02–0.04 mM (cell extracts buffer) (see Note 4).
3. 96 Ritter well plates (if a robotic liquid handler is available).
4. 3 mm NMR tubes for Bruker SampleJet (Bruker Biospin, Ltd.)
in 96-tube racks (see Note 5).

3 Methods

3.1 Cell Culture 1. Aspirate 100–300 μL of culture medium (see Note 6).
Medium Sample 2. Mix immediately with two volumes of cold 100% methanol (see
Note 7) in a 2 mL microcentrifuge tube.
3. Place the tube on dry-ice and repeat steps 1–3 until all samples
are collected.
4. Vortex samples for 10 s.
5. Transfer samples to freezer (20
C or lower) and allow pro-
teins to precipitate for at least 20 min (see Note 8).
6. Centrifuge at 16,000  g for 20 min at 4
C.
332 Sarantos Kostidis

7. Collect supernatants and dry under gentle nitrogen gas stream

(see Note 9).
8. Repeat steps 1–7 for cell-free culture medium (3 samples).

3.2 Quenching 1. Take out the culture dishes from the incubator and in the fume
hood; immediately aspirate the culture medium using the
aspiration pump.
2. Add 5 mL of warm PBS and shake gently for 1 s.
3. Aspirate PBS and immediately add 10 mL LN2 (see Note 10).
4. Allow LN2 to evaporate (about 1 min) and transfer the dish
immediately to a box of dry-ice. It is critical that the frozen cells
are not allowed to thaw after quenching.
5. Repeat steps 1–4 until all samples are quenched. Keep all
samples frozen until extraction.

3.3 Extraction 1. Transfer all samples in a cold room (4
C) if available.

of Intracellular 2. Add 1.5 mL of the cold extraction solvent (see Note 11).
Metabolites
3. Detach cells from the culture dish surface using a cell scraper
(see Note 12).
4. Transfer the mixture of cell extract and debris to a 2 mL
microcentrifuge tube and place it on dry-ice until all samples
are extracted.
5. Centrifuge at 16,000  g for 10 min at 4
C.
6. Collect supernatants and dry under a gentle stream of
nitrogen gas.
7. Use the precipitated cell biomass that is remained after the
centrifugation to calculate each sample’s protein content. The
Pierce BCA protein assay kit is used for total protein quantifi-
cation [18]. Use the materials provided by the kit and follow
the accompanied instructions to calculate the protein content
using a spectrophotometer and the absorbance at 562 nm.
Alternatively, estimate the cell numbers of each cell line with a
cell counter (see Note 13).

3.4 NMR Sample 1. Reconstitute the dried extracts and culture medium samples
Preparation with 220–250 μL of the phosphate buffer in D2O (see Notes
14–16).
2. Vortex for 10 s.
3. Transfer the samples to 96 Ritter well plate with a pipette (see
Note 17).
4. Transfer 190 μL of each sample to 3 mm NMR tubes using a
liquid robotic sampler if available or a syringe.
 Quantification of Cellular Metabolism 333

5. From the leftovers of each sample, prepare pool samples, if

possible, one pool sample per biological group of samples (see
Note 18).
6. Place all samples on a cooling position in the SampleJet (see
Note 19).

3.5 NMR 1. Load a pure methanol sample and calibrate the temperature
Spectroscopy setting to get an actual sample temperature in the probe of
300 K (27
C), using a short single-pulse NMR experiment
[19]. The temperature is estimated by the distance Δδ (ppm)
between the two methanol peaks and using the formula T
(
C) ¼ 130.00  29.53Δδ  23.87Δδ2. In Topspin (Bruker
Biospin Ltd.), this step can be performed automatically with
the “calctemp” command (see Note 20).
2. Store the calibrated temperature setting and use it throughout
the remaining study.
3. Load one of the pool samples. Optimize shims and store shim
file (see Note 21).
4. Run the 1H 1D NOESY experiment (see Note 22), and set up
the number of scans based on the signal to noise ratio (SNR) of
less abundant peaks (e.g., aromatic protons signals, fumarate,
etc.) (see Note 23).
5. Calculate the receiver gain (RG) to be used for all experiments
in the study.
6. Store the 1D NOESY profiling experiment in the software’s
library (for Topspin: command: wpar
NAME_OF_EXPERIMENT).
7. Run a 2D J-resolved experiment (JRES) using the same RG as
the one used for 1D NOESY (see Note 24). Store the JRES
profiling parameters as before.
8. Start automatic measurements of all other samples using the
two saved NMR experiments. For each sample, the following
steps should occur automatically: loading of the sample into
the probe, wait for 5 min to adopt the 300 K temperature,
tuning and matching of the probe, field lock to D2O, load
stored shim file (from the pool sample), on axis (z-axis) shim-
ming, automatic 90
pulse calibration [20], and acquisition of
1D NOESY and JRES (see Note 25).
9. All collected NMR data in the form of free induction decays
(FIDs) are automatically Fourier transformed, baseline and
phase corrected, and referenced to TSP at 0.00 ppm. For the
1D NOESY, the FIDs are also zero filled to 128 k points (factor
of 2) and weighed using a line broadening factor of 0.3 Hz.
The baseline of all 1D spectra is evaluated and further corrected
334 Sarantos Kostidis

manually, using the Topspin (or other available software

routines) (see Note 26).
10. Once all samples have been measured, use the aliquots of the
stored pool samples to acquire 2D NMR experiments. For each
sample, the profiling experiments should be repeated, and in
addition, the 2D 1H-1H correlation spectroscopy (COSY) and
total correlation spectroscopy (TOCSY) and the 1H-13C het-
eronuclear single quantum correlation (HSQC) and hetero-
nuclear multiple quantum correlation (HMBC) experiments
are collected (see Note 27).

3.6 Metabolites 1. First, try to identify as many metabolites as possible using

Identification online databases and data from the literature. The most com-
and Quantification monly used and freely available metabolite databases are the
Human Metabolome Database (HMDB; http://hmdb.ca)
[21], the Biological Magnetic Resonance Data Bank (BMRB;
http://www.bmrb.wisc.edu), [22] and the Birmingham
Metabolite Library (BML; http://www.bml-nmr.org/)
[23]. Other well-known commercial databases are the Bbior-
efcode (Bruker Biospin Ltd.) and Chenomx library (Chenomx
Inc.).
2. Use the 2D NMR data of pool samples to verify all assignments
(see Note 28).
3. Import data to a NMR deconvolution software to quantify the
detected metabolites. The proprietary Chenomx NMR suite is
considered as the gold standard for this task. It uses an internal
database of more than 300 metabolites and an interphase
allowing for interactive fitting (see Note 29). Other options
are the also proprietary MNOVA (MestreLab, SL, Spain) and
the freely available BATMAN [24], which runs within the R
statistical environment (http://www.R-project.org/) (see Note
30). Detailed instructions and application notes are provided
by the developers of these software packages (see Note 31).
4. Following the here described protocol, a panel of more than
60 metabolites can be identified in mammalian cultured cell
lines. Use the information of selected peaks and their chemical
shifts provided in Table 1, as a basis for the assignment of the
intracellular and extracellular metabolites and as targets for
quantification.
5. Correct the relative or absolute concentrations of quantified
metabolites to the protein content of each sample
(or alternatively to the cell numbers).
6. The process is repeated for the extracellular metabolites. Com-
pare the results with the concentrations of the same metabolites
in cell-free medium samples to export uptake and release of
metabolites from and by the cells, respectively.
 Quantification of Cellular Metabolism 335

Table 1
Characteristic proton chemical shifts (at pH 7.4) of intracellular and extracellular metabolites of
cultured mammalian cells. The listed resonances can be used for quantification by deconvolution

Metabolite δ 1H (ppm) Metabolite δ 1H (ppm)

1-Methylnicotinamide 4.48, 8.18, 8.9, 8.97, Malate 2.34, 2.65, 4.29
9.28
α-Ketoglutarate 2.43, 2.99 Methionine 2.11, 2.12, 2.18, 2.63,
3.84
2-Oxoisocaproate 0.92, 2.60 Methylmalonate 1.23, 3.16
3-Methyl-2- 0.88, 1.08, 1.44, 1.68, Methylsuccinate 1.07, 2.11, 2.51, 2.61
oxovalerate 2.91
ATP 4.25, 4.39, 4.59, 6.13, N,N-Dimethylglycine 2.90
8.26, 8.52
Acetate 1.90 N-Acetylaspartate 2.00, 2.48, 2.67, 4.38
Alanine 1.48, 3.78 N-Acetylglutamine 1.91, 2.02, 2.31, 4.15
Arginine 1.64, 1.72, 1.91, 3.23, NADþ 4.22, 4.39, 4.53, 6.02,
3.76 6.08, 8.16, 8.18, 8.41,
8.82, 9.13, 9.32
Asparagine 2.84, 2.94, 3.98 Nicotinate adenine 4.40, 4.51, 6.03, 8.04,
dinucleotide 8.14, 8.42, 8.73, 8.99,
9.12
Aspartate 2.67, 2.80, 3.88 O-Phosphocholine 3.20, 3.58, 4.15
Betaine 3.25, 3.89 Ornithine 1.78, 1.92, 3.04, 3.77
Choline 3.19, 3.50, 4.05 Pantothenate 0.88, 0.92, 2.40, 3.97
Citrate 2.53, 2.64 Phenylalanine 3.12, 3.27, 3.98, 7.32,
7.36, 7.41
Creatine 3.02, 3.91 Proline 1.98, 2.05, 2.34, 3.33,
3.40, 4.12
Phosphocreatine 3.03, 3.93 Pyroglutamate 2.02, 2.39, 2.49, 4.16
Formate 8.44 Pyruvate 2.35
Fructose 3.55, 3.66, 3.69, 3.79, Serine 3.83, 3.93, 3.98
3.88, 3.98, 4.01, 4.10
Fumarate 6.50 Succinate 2.38
GTP 5.93, 8.13 Taurine 3.25, 3.41
Galactitol 3.68, 3.97 Threonine 1.31, 3.57, 4.24
Galactose 3.48, 3.63, 3.69, 3.72, Tryptophan 4.05, 7.19, 7.27, 7.31,
3.79, 3.84, 3.92, 3.98, 7.53, 7.72
4.07, 4.58, 5.26
(continued)
336 Sarantos Kostidis

Table 1
(continued)

Metabolite δ 1H (ppm) Metabolite δ 1H (ppm)

Glucose 3.23, 3.39, 3.45, 3.50, Tyrosine 3.05, 3.18, 3.92, 6.89,
3.71, 3.81, 3.88, 4.63, 7.18
5.22
Glutamate 2.04, 2.11, 2.34, 3.74 UDP-N- 2.06, 3.54, 4.27, 4.36,
acetylglucosamine 5.50, 5.96, 7.93
Glutamine 2.13, 2.44, 3.76 UDP-glucuronate 4.38, 5.62, 5.97, 7.93
Glutathione (reduced) 2.15, 2.54, 2.94, 3.76, UDP-glucose 4.19, 4.23, 4.27, 4.36,
4.55 5.59, 5.97, 7.94
Glycine 3.55 UDP-galactose 4.27, 4.35, 5.63, 5.97,
7.95
Histidine 3.12, 3.22, 3.98, 7.08, UMP 4.34, 4.41, 5.98, 8.09
7.87
Hypotaurine 2.63, 3.34 Valine 0.98, 1.03, 2.26, 3.59
Isoleucine 0.93, 1.00, 1.25, 1.46, Myoinositol 3.26, 3.52, 3.61, 4.05
1.97, 3.66
Lactate 1.31, 4.09 sn-Glycero-3- 3.22, 3.60, 3.64, 4.31
phosphocholine
Leucine 0.94, 0.95, 1.70, 3.73 β-alanine 2.54, 3.16
Lysine 1.46, 1.72, 1.89, 3.01, CoA 0.72, 0.84, 2.60
3.74
Cystine 3.18, 3.38, 4.08 2-Hydroxyisobutyrate 1.35
Hypoxanthine 8.18, 8.20 3-Hydroxybutyrate 1.19, 2.30, 2.40, 4.14
Pyridoxine 2.45, 7.64 Niacinamide 7.58, 8.24, 8.70, 8.93
Trans-4-Hydroxy-L- 2.14, 2.42, 3.35, 3.47, Galactose-1-phosphate 3.72, 3.76, 3.90, 3.99,
proline 4.33 5.48

4 Notes

1. LN2 temperature is below its boiling point of 195.56
C and

will cause cold burns in case of direct contact with the skin.
Always use protective gloves and glasses when handling LN2.
Pour LN2 into the culture dish using a Falcon tube attached to
tweezers to avoid direct contact with the tube.
2. Based on the fact that some residual water will remain in the
culture dish after quenching, it was estimated that the 90%
(v/v) methanol/chloroform 9:1 solution is actually close to
 Quantification of Cellular Metabolism 337

75% (v/v) methanol/chloroform 9:1 when added to the dish

[14]. Adjust volumes appropriately if different sizes of culture
dishes are used.
3. The selection of the extraction solvent is based on the resulting
metabolic recovery. The more polar the solvent system is, the
better the recovery of the polar metabolites will
be. Unfortunately, this comes at the expense of effective pro-
tein removal, which is desired for the highest possible quality of
the NMR spectra. Proteins induce several intense and broad
resonances in the proton spectrum, which hamper the accuracy
of metabolites quantification. Therefore, a compromise
between metabolic recovery and protein removal has to be
made by adjusting the polarity of the extraction solvent. The
here, proposed mixture was also proposed by Lorentz et al.
[14] and in general has worked well in our hands with several
cell lines (endothelial cells, cancer cells, and macrophages). We
recommend an evaluation of metabolic recovery using several
proposed extraction solvents (50% MeOH, 70% MeOH, 75%
MeOH/CHCl3, and 100% MeOH) before each study.
4. The intensity of the TSP signal in the NMR spectra should be
comparable to the other components of the sample. Since the
cell extracts are much less concentrated than the culture
medium samples, it is preferable to use different quantities of
internal standard among these two types of samples.
5. The 3 mm NMR tubes are selected because they require less
volume (190 μL) than the commonly used 5 mm tubes
(560 μL). This way, the dried extract is dissolved in less volume,
leading to more concentrated solution and increase NMR sen-
sitivity. However, this configuration works well for 5 mm
cryoprobes, but may suffer from increased thermal noise if a
5 mm room temperature probe is used. In that case, we recom-
mend a trial experiment with both 3 mm and 5 mm tubes with
the corresponding volumes of sample. Then, a direct compari-
son of the SNR between the two measurements will provide the
optimum setup for the study.
6. The volume of medium depends on the total culture medium,
i.e., the size of the culture dish. In general, the withdrawn
medium quantity should not induce any stress to the cells.
7. The extraction solvent methanol/chloroform/water, 8.1:0.9:1
(v/v/v) can also be used for this step. We encourage testing the
best system using blank culture medium with 10% fetal bovine
serum (FBS) prior to the actual study.
8. The samples can be stored for longer times if a 80
C freezer is
available. This can be practical since, typically, the quenching and
extraction steps of the intracellular metabolites should be per-
formed right after the medium collection. This way the medium
samples can be processed at later times (even several days later).
338 Sarantos Kostidis

9. Best is to perform the drying step at the day of analysis to avoid

unnecessary moisture in the samples.
10. The speed of the washing and quenching process is critical to
avoid the turnover of labile metabolites, for example, ATP,
NADH, and others. For optimum results, these steps should
be performed by two persons. In that case, our experience
shows that the cells can be quenched in about 5 s after the
medium is aspirated.
11. Keep the extraction solvent cold by placing it in a box with
dry-ice throughout the whole process.
12. An indication of the complete detachment of the cells from the
dish surface is when the scraping becomes smooth with no
resistance from anomalies of the surface.
13. In case that the Pierce BCA protein assay kit is not available and
cell counting is preferred, this cannot be performed on the
same samples used for extraction. Therefore, it is recom-
mended to culture two additional samples under identical
conditions and time and estimate the cell numbers in these
samples using either an automatic or a manual cell counter
method.
14. Although the 3 mm NMR tubes require just 190 μL of sample
volume, we recommend dissolving the dried extracts in more
volume so there is a sufficient quantity of leftovers to prepare
pool samples. The latter are very useful in two ways. First, to
prepare the NMR profiling parameters prior to automation
and, second, to be used for 2D NMR for assignment of
metabolites.
15. In case that 5 mm tubes are used, the dried extract should be
dissolved in 650 μL.
16. Use the buffer with 0.02–0.04 mM TSP for cell extracts and
the one with 0.4 mM TSP for culture medium samples.
17. If a robotic liquid handler is not available, then the samples are
transferred manually to the NMR tubes with a syringe.
18. Mix the collected leftovers so aliquots per biological group are
obtained (e.g., 3 samples of 60 μL), vortex 10 s, and store at
80
C until analysis. Directly use one of the aliquots for
setting up the NMR measurements.
19. If a SampleJet or other similar cooling setup is not available, do
not place all samples in an autosampler, as the sample integrity
might be affected after prolonged time at room temperature.
Instead, use only small batches of samples (e.g., 4–6) and store
the remaining ones in the fringe.
 Quantification of Cellular Metabolism 339

20. For accurate temperature calibration, the methanol sample

should be shimmed well. Process the FID using a line broad-
ening of 3 Hz before applying calctemp.
21. Use the peak of the internal standard to assess shimming
quality. When no apodization is applied (i.e., no line broaden-
ing), the width of TSP at half height should be less than
0.7 Hz. Use hwcal command (Topspin, Bruker) to automati-
cally calculate it.
22. The 1D NOESY experiment’s pulse sequence has the form -
RD—gz,1—90
—t—90
—tm—gz,2—90
—ACQ. RD is the
relaxation delay of 4 s, gz,1 and gz,2 are gradients with 1 ms
duration, t is a short delay of 3 μs between pulses, 90
is the 90

radiofrequency pulse (RF), tm is the mixing time of 10 ms, and
ACQ is the acquisition of the free induction decays (FIDs), set
up to 2.72 s. The residual water resonance is suppressed with a
continuous RF of 50 Hz (presaturation), applied during the
relaxation delay and the mixing time.
23. While a SNR of 3 is considered as the limit of detection (LOD),
the limit of quantification (LOQ) is dependent on how pre-
cisely a resonance can be quantified. This can be estimated by
calculating the coefficient of variation (CV %). We recommend
to use triplicates of blank culture media, for which the available
volume is not a restriction, and perform some 1D experiments
to determine the optimum SNR in order to get an acceptable
CV (often CV < 15–20% is sufficient).
24. The 2D JRES pulse sequence has the form -RD—90
—t1—
180
—t1—ACQ. RD is set to 2 s, and t1 increment between
pulses and ACQ is set to 12.8 ms. 12,288 data points are
collected in direct dimension (F2) with 2 transients over
40 increments in the indirect dimension (F1). The spectral
width is set to 16.66 ppm in F2 and 78 Hz in F1. The same
presaturation pulse as in the 1D NOESY is applied during the
relaxation delay.
25. The total experimental time can vary from 20 min to even 1 h
per sample for less concentrated samples.
26. At this point, it is critical to evaluate the quality of the spectra.
First, the TSP singlet is used to assess the shimming quality.
Poor shimming will reduce resolution and subsequently affect
the quality of quantification. Second, since the quantity of the
internal standard TSP is used as a reference for the quantifica-
tion of all other metabolites, it should be the same across all
samples (recall that the same quantity was added during sample
preparation). If inconsistent TSP intensities are observed, try
to correct the quantification results of those samples if possible,
or remove them from further analysis. With modern NMR
instruments, it is possible to generate artificial reference signals
340 Sarantos Kostidis

(e.g., ERETIC2 [25]) with which any TSP addition inconsis-

tencies can be corrected.
27. For the COSY experiment, use the parameters: the cosygpprqf
pulse program (Topspin, Bruker), 2048 points in the F2 and
256 increments in F1 with 16 scans per increment, 13.35 ppm
spectral width for both dimensions, and 2 s relaxation delay.
For the TOCSY experiment, use the dipsi2gpphpr pulse pro-
gram, which uses the DIPSI2 sequence for the homonuclear
Hartmann-Hahn transfer during the spinlock period of 60 ms.
Acquire 2048 points in the F2, with 32 scans and 256 incre-
ments in F1, 13 ppm spectral width for both dimensions, and
2 s relaxation delay. For the HSQC experiment, use the hsqce-
detgpprsisp2.2 pulse program. This pulse sequence includes
multiplicity selection for primary, secondary, and tertiary car-
bons. Collect 1024 data points in F2, with 128 scans and
180 increments. Set the spectra width to 13.01 ppm for 1H
(F2) and 165.65 ppm for 13C (F1) and the relaxation delay to
1.5 s. For the HMBC experiment, use the hmbcgplpndprqf
pulse program. It includes a low-pass J-coupling filter to
exclude one-bond correlations (use 1JCH ¼ 145 Hz) and set
the average for the long-range couplings at 7 Hz. Collect 2048
data points in F2, with 164 scans and 256 increments. Set the
spectral width to 13 ppm for 1H (F2) and 240 ppm for 13C
(F1), and use a relaxation delay to 1.5 s. Presaturation of
25–50 Hz can be applied in all 2D experiments.
28. We strongly recommend to verify all assignments made by the
available databases with 2D NMR data. Use either the 1H-1H
or the 1H-13C correlations to make sure the annotated meta-
bolites are correctly identified, and note their chemical shifts.
Use this information for quantification. In case of ambiguities,
perform spiking experiments with pure compounds rather than
reporting an ambiguous annotation. In order to better resolve
overlapped peaks, use also the 2D JRES spectra which are
collected for all samples.
29. When quantifying a compound with Chenomx, all of its pro-
tons are fitted. This method combined with the knowledge of
peak positions (from the 2D data), and tools like the sum
spectral line can have a beneficial influence in quantification
accuracy. However, some sources of variation have been
reported from user to user or for specific compounds [26]. In
order to avoid erroneous quantification results, we recommend
the following: study carefully and work with the application
notes provided on the website of the developer and practice
with samples of known composition until a good fitting meth-
odology is obtained. Ideal candidates for practice are the blank
culture medium, for which the composition is usually known
 Quantification of Cellular Metabolism 341

and their content is similar regarding the metabolites found in

the intracellular matrix.
30. In contrast to Chenomx, BATMAN only fits the resonances
which are defined by the user. As input files, it uses the infor-
mation of about 2500 peaks, freely available by HMDB. The
necessary inputs for each peak are its exact chemical shift, the
J-coupling, the number of protons that contribute to this peak
(e.g., 3 protons from a methyl group), and the compound
ID. The latter can also be unknown or a new compound (not
in the HMDB database) specified by the user. J-couplings can
be extracted from the JRES spectra. Based on our experience
with BATMAN, the fitting requires more computation time
and is less precise compared to Chenomx, especially in the cases
of overlapped peaks or complex multiplets. However, the fact
that it is freely available makes it an attractive alternative if
Chenomx is not available.
31. The extracted quantitative data is generally considered as rela-
tive concentrations of metabolites. In order to transform it to
absolute values (if needed), some corrections are necessary
(e.g., adjust for the relaxation times of each fitted peak)
[25]. On the other hand, Chenomx provides the option to
acquire data based on 1D NOESY experiments with specific
parameters and without gradients (see software’s available doc-
umentation and application notes). In that case, absolute con-
centrations are extracted based on the modeled compounds in
Chenomx library without any further corrections. It should be
noted, however, that the here proposed gradient-based
NOESY experiment provides better SNR and water suppres-
sion. Moreover, we obtain comparable results when we com-
pare the data from the two experiments. Based on this, the
selection of the profiling experiment is based on what software
will be used for quantification and how concentrated are the
samples of the study. The only critical point is that this selection
should be consistent throughout the study.

References
1. Liu W, Deng Y, Liu Y et al (2013) Stem cell 4. McNamara LE, Sjostrom T, Meek RMD et al
models for drug discovery and toxicology stud- (2012) Metabolomics: a valuable tool for stem
ies. J Biochem Mol Toxicol 27:17–27 cell monitoring in regenerative medicine. J R
2. Benjamin DI, Cravatt BF, Nomura DK (2012) Soc Interface 9:1713–1724
Global profiling strategies for mapping Dysre- 5. Kelly B, O’Neill LA (2015) Metabolic repro-
gulated metabolic pathways in cancer. Cell gramming in macrophages and dendritic cells
Metab 16:565–577 in innate immunity. Cell Res 25:771–784
3. Schlegel M, Köhler D, Körner A et al (2016) 6. Johnson CH, Ivanisevic J, Siuzdak G (2016)
The neuroimmune guidance cue netrin-1 con- Metabolomics: beyond biomarkers and
trols resolution programs and promotes liver towards mechanisms. Nat Rev Mol Cell Biol
regeneration. Hepatology 63:1689–1705 17:451–459
342 Sarantos Kostidis

7. Bennett BD, Yuan J, Kimball EH et al (2008) 16. Bi H, Krausz KW, Manna SK et al (2013) Opti-
Absolute quantitation of intracellular metabo- mization of harvesting, extraction, and analyti-
lite concentrations by an isotope ratio-based cal protocols for UPLC-ESI-MS-based
approach. Nat Protoc 3:1299–1311 metabolomic analysis of adherent mammalian
8. Paglia G, Hrafnsdóttir S, Magnúsdóttir M et al cancer cells. Anal Bioanal Chem
(2012) Monitoring metabolites consumption 405:5279–5289
and secretion in cultured cells using ultra- 17. Ser Z, Liu X, Tang NN et al (2015) Extraction
performance liquid chromatography quadru- parameters for metabolomics from cultured
pole–time of flight mass spectrometry cells. Anal Biochem 475:22–28
(UPLC–Q–ToF-MS). Anal Bioanal Chem 18. Smith PK, Krohn RI, Hermanson GT et al
402:1183–1198 (1985) Measurement of protein using bicinch-
9. Nagana Gowda GA, Abell L, Lee CF et al oninic acid. Anal Biochem 150:76–85
(2016) Simultaneous analysis of major coen- 19. Findeisen M, Brand T, Berger S (2007) A
zymes of cellular redox reactions and energy 1H-NMR thermometer suitable for cryop-
using ex vivo 1 H NMR spectroscopy. Anal robes. Magn Reson Chem 45:175–178
Chem 88:4817–4824 20. Wu PSC, Otting G (2005) Rapid pulse length
10. Goldoni L, Beringhelli T, Rocchia W et al determination in high-resolution NMR. J
(2016) A simple and accurate protocol for Magn Reson 176:115–119
absolute polar metabolite quantification in cell 21. Wishart DS, Jewison T, Guo AC et al (2013)
cultures using quantitative nuclear magnetic HMDB 3.0–the human Metabolome database
resonance. Anal Biochem 501:26–34 in 2013. Nucleic Acids Res 41:D801–D807
11. León Z, Garcı́a-Cañaveras JC, Donato MT et al 22. Ulrich EL, Akutsu H, Doreleijers JF et al
(2013) Mammalian cell metabolomics: experi- (2008) BioMagResBank. Nucleic Acids Res
mental design and sample preparation. Electro- 36:D402–D408
phoresis 34:2762–2775
23. Ludwig C, Easton JM, Lodi A et al (2012)
12. Dietmair S, Timmins NE, Gray PP et al (2010) Birmingham metabolite library: a publicly
Towards quantitative metabolomics of mam- accessible database of 1-D 1H and 2-D 1H
malian cells: development of a metabolite J-resolved NMR spectra of authentic metabo-
extraction protocol. Anal Biochem lite standards (BML-NMR). Metabolomics
404:155–164 8:8–18
13. Sellick CA, Hansen R, Stephens GM et al 24. Hao J, Liebeke M, Astle W et al (2014) Bayes-
(2011) Metabolite extraction from ian deconvolution and quantification of meta-
suspension-cultured mammalian cells for bolites in complex 1D NMR spectra using
global metabolite profiling. Nat Protoc BATMAN. Nat Protoc 9:1416–1427
6:1241–1249
25. Bharti SK, Roy R (2012) Quantitative 1H
14. Lorenz MA, Burant CF, Kennedy RT (2011) NMR spectroscopy. TrAC Trends Anal Chem
Reducing time and increasing sensitivity in 35:5–26
sample preparation for adherent mammalian
cell metabolomics. Anal Chem 83:3406–3414 26. Sokolenko S, Blondeel EJMM, Azlah N et al
(2014) Profiling convoluted single-dimension
15. Dettmer K, N€urnberger N, Kaspar H et al (2011) proton NMR spectra: a Plackett–Burman
Metabolite extraction from adherently growing approach for assessing quantification error
mammalian cells for metabolomics studies: opti- of metabolites in complex mixtures with appli-
mization of harvesting and extraction protocols. cation to cell culture. Anal Chem 86:
Anal Bioanal Chem 399:1127–1139 3330–3337
 Part VI

MALDI-Based Techniques and Mass Spectrometry

Imaging of Clinical Samples
 Chapter 26

Mass Spectrometry Imaging of Metabolites

Benjamin Balluff and Liam A. McDonnell

Abstract
Mass spectrometry imaging (MSI) is a technique which is gaining increasing interest in biomedical research
due to its capacity to visualize molecules in tissues. First applied to the field of clinical proteomics, its
potential for metabolite imaging in biomedical studies is now being recognized. Here we describe how to
set up experiments for mass spectrometry imaging of metabolites in clinical tissues and how to tackle most
of the obstacles in the subsequent analysis of the data.

Key words Mass spectrometry imaging, Tissue metabolomics, Biomarker discovery, Data analysis

1 Introduction

Mass spectrometry and NMR spectroscopy are the workhorse tech-

nologies in metabolomic studies [1]. Among mass spectrometry
techniques, mass spectrometry imaging (MSI) is unique in that it
allows visualizing metabolite distributions in biological samples
such as tissues. MSI involves the local extraction, desorption, and
ionization of the metabolites in a spatially correlated manner, for
example using matrix-assisted laser desorption/ionization
(MALDI) or desorption electrospray ionization (DESI) [2].
MALDI MSI can be used for imaging many different kinds of
compounds, ranging from proteins to small molecules including
metabolites or exogenous compounds. The latter application has
led to the widespread application of MALDI MSI in the pharma-
ceutical industry, for the determination of the abundances and
locations of drugs, their metabolized products, and interacting
molecules in the target and nontarget tissues [3].
Due to its histological specificity (the maximum resolution is
nowadays at a single cell level), MSI has also had a big impact for
biomarker discovery in complex solid tissues such as cancer
[2]. Although most biomarker studies have investigated proteins,
clinical MSI is slowly being expanded to the investigation of
metabolites [4].

Martin Giera (ed.), Clinical Metabolomics: Methods and Protocols, Methods in Molecular Biology, vol. 1730,
https://doi.org/10.1007/978-1-4939-7592-1_26, © Springer Science+Business Media, LLC 2018

345
346 Benjamin Balluff and Liam A. McDonnell

MSI of metabolites benefited from the introduction of new

sample preparation protocols, including new MALDI matrices
[5, 6], chemical stabilization of tissue [7], chemical derivatization
of target compounds [8], methods for the retrieval of metabolic
information from archived tissues [9], and the availability of high
mass resolution MSI instruments. The latter has proven important
to resolve specific metabolite ions from the sometimes complex
chemical background obtained from a tissue section [10].
A successful biomedical metabolomic study using MSI depends
on many factors: the sample selection, sample handling, sample
preparation, type of instrumentation—defining mass accuracy,
mass resolution power, and sensitivity—and last but not least also
the appropriate data processing and analysis procedures [11].
In this chapter we describe all the mentioned steps of how MSI
can be used for metabolite imaging and data analysis in clinical
tissues.

2 Materials

Prepare all solutions using ultrapure water and, where possible, use
analytical grade reagents. Prepare and store all reagents at room
temperature unless indicated otherwise. For clinical or preclinical
research, it is imperative that ethical approval is sought and
obtained prior to undertaking any of the steps described below.

2.1 Tissue Collection It is well established that the metabolome changes rapidly after
tissue sampling/animal sacrifice. Several methods have been
reported to maintain metabolic endogenous metabolite levels,
ranging from cold perfusion [12]/in situ freezing [7] to “freeze”
metabolic activity prior to organ excision, to heat
/microwave-
based denaturing of metabolic enzymes (see Note 1 for more
information, including published comparisons of the different
methods). Here for reasons of its broad applicability, we refer
to already excised tissue samples, specifically fresh frozen tissues
isolated via rapid tissue excision and freezing (preferably <1 min
postmortem time).

2.2 Tissue 1. Clean a stock of indium-tin-oxide (ITO)-coated glass slides

Sectioning (Bruker Daltonics, Bremen, Germany) by washing with deio-
nized water (3) and methanol (3); use a multimeter to
determine which side is conductive. Coat the ITO slides with
0.05% poly-L-lysine (poly-L-lysine coating used for greater
adherence of tissue sections, protocol used as reported in Aich-
ler et al. [13]). Once the poly-L-lysine coating is dry, mark the
conductive side with a Tippex pen for easy identification.
 Imaging of Metabolites 347

2. Prepare a stock of microscope slides for histological assessment

of the tissue sections.
3. Precool the ITO slides and slidebox by placing them within a
precooled (
20C) cryostat chamber.
4. Prepare stock solutions of hematoxylin and Eosin. For exam-
ple, using preprepared hematoxylin solution from Klinipath
BV, article code 4085.9002; Eosin Y 1% alcohol solution
from Klinipath BV, article code 640375.

2.3 Tissue 1. Prepare a stock solution of 2 mg/mL 9-aminoacridine in 70%

Preparation methanol. Measure 2 mg of 9-aminoacridine using a mass
balance that is accurate to 0.1 mg. Measure 0.7 mL methanol
and 0.3 mL deionized water. Combine the 0.7 mL methanol
with the 0.3 mL deionized water in a 1.5 mL Eppendorf tube,
and then dissolve the 2 mg 9-aminoacridine in the 7:3 metha-
nol/water solvent. Use a vortex mixer to ensure the matrix
solution is free of particulates.
2. Prepare a stock solution of 5 mg/mL 2,3-diphenyl-pyranylium
tetrafluoroborate (DPP-TFB) in 100% methanol. Measure 5 mg
of (DPP-TFB) using a mass balance that is accurate to 0.1 mg.
Measure 1 mL methanol and place in a 1.5 mL Eppendorf tube.
Then dissolve the 5 mg DPP-TFB in the methanol. Use a vortex
mixer to ensure the DPP-TFB solution is free of particulates.
3. Prepare a stock solution of 30 mg/mL 2,5-dihydroxybenzoic
acid (DHB) in 70% MeOH and 0.1% TFA. Measure 30 mg of
DHB using a mass balance that is accurate to 0.1 mg. Measure
0.7 mL methanol and 0.3 mL 0.1% TFA and place in a 1.5 mL
Eppendorf tube. Dissolve the 30 mg DHB in the methanol/
0.1% TFA solvent. Use a vortex mixer to ensure the DPP-TFB
solution is free of particulates.
4. Prepare a solution of 70% methanol.
5. Prepare a solution of 20 mg/mL 2,4,6-trihydroxyacetophenone
[THAP] in 50/50 ethanol/water.
6. Prepare 20 μM solutions of YGGFL, angiotensin II, [glu1]-
fibrinopeptide B (all available from Sigma-Aldrich) in 50/50
ethanol/water.

2.4 Data Acquisition 1. Prepare a MALDI-TOF/TOF instrument, such as RapiFlex

(Bruker Daltonics, Bremen, Germany) for positive ion and
negative ion MALDI, reflectron mode detection, m/z range
50–1000, using a laser spot size of 50 μm. Note 2 describes
how to determine laser spot size using a thin matrix layer
preparation. Ensure the mass spectrometer’s m/z scale is cali-
brated using a mixture of YGGFL, angiotensin II, and [glu1]-
fibrinopeptide B for positive mode calibration, or matrix clus-
ters of THAP for negative mode calibration.
348 Benjamin Balluff and Liam A. McDonnell

3 Methods

3.1 Sample 1. Precool a cryostat microtome (e.g., Leica CM1950) to the

Preparation desired cutting temperature, e.g.,
21  C. Mount the excised
tissue onto a precooled cryostat specimen holder. This may be
achieved using a small amount of water, gelatin solution, or
Tissue-Tek™ solution. On contact with the precooled speci-
men holder and frozen tissue block, the solution rapidly
freezes, to effectively fix the tissue specimen to the specimen
holder. Note: if using Tissue-Tek™, great care should be taken
to ensure only a small, localized droplet is used to “glue” the
tissue to the specimen holder. Tissue-Tek™ is known to cause
strong ionization bias effects, and so its presence should be
avoided on or near the tissue surfaces of interest.
2. Depending on the tissue specimen of interest, the tissue speci-
men may need to be “trimmed” to reach the desired location.
For example, if analyzing coronal tissue sections of a mouse
brain, then the desired region of the brain, e.g., between
0.10
and þ0.40 mm from bregma, needs to be reached before the
analytical tissue sections can be cut. Trimming may be per-
formed manually or automatically depending on cryostat
model. Brain location can be determined by expert comparison
of histological images (cut at 5 μm thickness, mounted onto
standard microscope slides and histologically stained) with a
tissue atlas, e.g., those contained in the Allen Brain Atlas. For
tumor specimens histopathological assessment during section-
ing is essential to ensure the analytical tissue sections contain
sufficient tumor material of the desired grade and cellularity
and free of excessive necrosis.
3. Cut tissue sections of 12 μm thick using the cryostat micro-
tome. Rapidly thaw-mount the tissue sections onto the marked
(conductive) side of the precooled ITO slide within the cryo-
stat chamber. Rapid thaw-mounting is achieved by localized
warming of the reverse side of the MALDI target using a finger
for maximum 3 s. Then place the slide þ tissue into the pre-
cooled slide holder. Store the slide-mounted tissues at
80  C
until use.
4. For the analysis of adenylate and TCA cycle metabolites, the
matrix 9-AA is used. First rapidly freeze-dry the slide-mounted
tissue sections and then place into an automated matrix depo-
sition system. Here we refer to the SunChrom SunCollect
system, but other spray-based systems may be used with
minor modification to the method. The 2 mg/mL 9-AA solu-
tion is sprayed onto the tissue section using two layers at 5 μL/
min followed by six layers at 10 μL/min, with complete drying
between layers (Y-offset ¼ 1.0 mm, Z (needle height) ¼ 25 mm,
 Imaging of Metabolites 349

speed X ¼ speed Y ¼ medium 1). After the matrix coating is

dry, use a Tippex pen to add fiducial markers at four corners
surrounding the tissue section (but not in contact with it).
5. For the analysis of amino metabolites and neurotransmitters,
the metabolites are first derivatized using DPP-TFB. First rap-
idly freeze-dry the slide-mounted tissue sections and then place
into an automated matrix deposition system. The 5 mg/mL
solution of DPP-TFB in 100% MeOH is then deposited onto
the tissue in 5 layers at a flow rate of 10 μL/min (Y-off-
set ¼ 1.0 mm, Z (needle height) ¼ 25 mm, speed X ¼ speed
Y ¼ medium 1). For complete derivatization, the brain sections
are kept overnight at room temperature and pressure. Follow-
ing complete derivatization the MALDI matrix DHB is depos-
ited onto the tissue section. The 30 mg/mL DHB solution in
70% MeOH and 0.1% TFA is sprayed onto the derivatized
tissue section using two layers at 5 μL/min followed by six
layers at 10 μL/min, with complete drying between layers.
After the matrix coating is dry, use a Tippex pen to add fiducial
markers at four corners surrounding the tissue section (but not
in contact with it).

3.2 Data Acquisition MALDI MSI experiments can be performed using many different
types of mass spectrometer, including MALDI-TOF/TOF,
MALDI-Q-TOF, MALDI-IMS-TOF, MALDI-FT-ICR, and
MALDI-Orbitrap systems. Here we refer to the MALDI-TOF/
TOF instruments from Bruker Daltonics, to date the most com-
mon systems used for MALDI MSI.
1. First acquire high-resolution optical images of the matrix-
coated tissue sections using a flatbed scanner, e.g., an optical
image recorded at 254 dpi corresponds to a pixel size of
100 μm. Mount the matrix-coated tissue section into the
microscope slide holder, being careful to ensure good electrical
contact between the stainless steel slide holder and the ITO
coating of the conductive surface (note: it may be necessary to
wipe the matrix coating from the edge of the slide to ensure a
good contact).
2. Start a new MSI experiment in FlexImaging. Import the optical
image of the matrix-coated tissue section into FlexImaging,
and align the instrument stage holder to the optical image by,
in-turn, selecting the fiducial markers in the imported optical
image and the system’s sample visualization camera. Once
aligned the MSI experiment may be defined. Select the tissue
section to be analyzed as the first measurement region, as well
as small region close to the tissue section to serve as a control
(useful for assessing potential analyte delocalization and the
ready identification of background ions). If more than one
350 Benjamin Balluff and Liam A. McDonnell

tissue section is present on the slide, they may be selected as

separate measurement regions. For each measurement region,
the spatial resolution then needs to be defined (e.g., 100 μm),
the number of laser shots (e.g., 300), and the FlexControl
method selected. If adenylate and TCA cycle intermediate are
analyzed, the FlexControl method corresponds to that for the
analysis of negative ions; if amino metabolites and neurotrans-
mitters are analyzed, the FlexControl method corresponds to
that for the analysis of positive ions.
3. Test the performance of the selected FlexControl method on
the tissue section and vary the laser power until high-quality
metabolite signals are obtained, past MALDI threshold but not
too high to avoid fragmentation of the metabolites and exces-
sive background noise. A data processing method may be
selected to preprocess each mass spectrum of the MALDI
MSI dataset; if such preprocessing is performed during data
acquisition, select a validated FlexAnalysis method for the
metabolite peaks of interest. Once the FlexImaging settings
have been finalized, data acquisition can be initiated.
4. After MALDI MSI data acquisition, any residual MALDI
matrix is removed by washing the slide-mounted tissue sections
in 70% ethanol (2  30 s dips). The tissue section is then
histologically stained, using hematoxylin and Eosin (tissue tis-
sue) or Nissl (mouse brain). Lou et al. recently reported a
method by which the histological staining can be optimized
for the thicker tissue sections analyzed by MALDI MSI [11].
5. A high-resolution histological image of the histologically
stained tissue section is then recorded using a digital slide
scanner (e.g., 3D Histech Pannoramic desk). This histological
image is then imported into FlexImaging for co-registration
with the MALDI MSI results, through registering the histo-
logical image (high-resolution optical image of the histology-
stained tissue section) to the optical image of the matrix-coated
tissue section, from which the experiment was defined. The
registration is performed by selecting the same fiducial markers
in both images.

3.3 Data Analysis There are many data analysis strategies, but the analysis of metabo-
lomic MSI datasets, which contain multiple samples (from different
patients and experiments), usually relies on a recurrent data proces-
sing routine, which is summarized in Fig. 1. To compare different
MSI datasets, it is necessary to compile them into one MSI dataset.
During the compilation, a minimum of data preprocessing steps is
considered important to enable and increase comparability between
metabolic MSI datasets. As we focus on Bruker instrumentation
and data, it is worth mentioning that most of the functionality to
merge and compare multiple MSI datasets is available through the
 Imaging of Metabolites 351

Fig. 1 Data analysis workflow for mass spectrometry imaging (MSI) data from multiple experiments (patients).
The pure data analysis workflow can be divided into three major steps, and which is normally followed by the
identification step. The three pure data analysis steps consist of (a) preparing the metadata, including the
annotation of the co-registered histological images, which is then passed to the data processing environment
together with the acquired MSI data. (b) data processing and compilation, in which a series of operations are
applied to increase the comparability between the samples (c) statistical analysis. Once the data has been
processed, summarized, and grouped, statistical analyses can be performed in two ways: supervised, which
requires a previous knowledge on the grouping of the data, and unsupervised, which requires no prior
information about the grouping of the data. The latter can be used to find new groups using clustering
techniques or to reduce the dimensionality of MSI datasets. Supervised analyses can be done to create
classifiers based on discriminatory metabolite patterns. This list of molecules is usually of interest for a
biochemical interpretation. However, MSI requires a separate step to assign names to that list of molecules (d)

MSI data analysis software SCiLS Lab. However, these functions

will be described in order to enable researchers to repeat those steps
in noncommercial data analysis environments such as Python,
Matlab (bioinformatics toolbox), or R (Cardinal MSI package).
1. As MSI provides spatially resolved molecular profiles, a careful
histological annotation using the co-registered microscopic
image for restriction to the relevant areas of the tissue can
increase specificity and sensitivity of the subsequent data analy-
sis procedure [14, 15]. The preciseness of annotation is natu-
rally limited by the pixel size of the MSI data. In FlexImaging,
for example, the annotations can be saved and exported as
XML files, which contain the names of spectra belonging to
each defined region of interest (ROI).
352 Benjamin Balluff and Liam A. McDonnell

2. This XML file can be used to read in the data into the data
analysis environment of choice corresponding to the selected
ROIs only, which reduces data and computational load.
3. During the import, some basic MS preprocessing operations
on a single spectrum level may be necessary to increase quality
and hence comparability between individual pixels and data-
sets. This includes smoothing and/or baseline subtraction on
the individual spectra, if not performed beforehand, for
instance, during data acquisition. For metabolomic data both
operations are less common and necessary as for proteomic
data but should be performed if data quality indicates that.
More details on the effect of these processing steps can be
found in [16].
4. For clinical MSI, the set of samples to compare should have
been acquired with the same instrument (FlexControl)
method. In case different methods with different m/z domains
have been used, a common m/z domain with a common mass
range and bin size has to be created. The common lower limit is
given by the maximum of all lower mass limits and the common
upper limit by the minimum of all upper mass limits. The bin
size should be no less than the original sampling rate, unless a
certain smoothing effect is desired by a down-sampling. How-
ever, in TOF/TOF-based metabolomic MSI, a linear bin size
of Δm/z 0.01 should be sufficient. Nonlinear binning is sup-
ported by SciLS Lab, which allows a more efficient way of
distributing the data points to describe a mass spectrum.
5. An important pillar for ensuring comparability between MSI
datasets is normalization of single pixels. Deininger et al. have
discussed several normalization techniques for MALDI MSI,
among them total-ion-count (TIC), root-mean-square (RMS),
and median normalization [17]. TIC is used when the baseline
is strongly influenced by chemical noise as in linear TOF. RMS,
in contrast, is determined by real peak intensities and therefore
popular for FT-ICR data. As metabolomic TOF/TOF data
exhibits less chemical baseline than linear TOF MSI, but
more than in FT-ICR, the application of the normalization
methods depends on the quality of the single spectra.
6. When comparing multiple MSI datasets from different experi-
ments, mass shifts might be observed (Fig. 2a, c) that can have
a strong impact on the final result outcome. This can be over-
come by spectral alignment which is less well described in MSI
literature. Alignment (or recalibration) of spectra is usually
done on the representative spectra of the different datasets,
since mass shifts within datasets are usually considered negligi-
ble. For this, a set of reference peaks has to be defined that
fulfill two criteria: they should be present in almost all of the
 Imaging of Metabolites 353

Fig. 2 Mass shifts and spectral alignment of mass spectrometry imaging (MSI) data. Due to the absence of
internal standards, MSI data can exhibit mass shifts within and between experiments. Within-experiment
mass shifts refer to the mass shifts occurring from pixel to pixel, which are usually very small and therefore
negligible. However, between-experiment mass shifts can be significant and therefore have a strong negative
effect on comparability between datasets. Panels (a, c) show examples of observed mass shifts (0.2–0.4 Da)
from 59 MSI experiments performed with the same method [4]. Alignment of the spectra can be performed by
a recalibration of the data using reference peaks. Matlab functions, such as msalign, require the definition of
two parameters. One that describes the peak width, which can be estimated by looking at the average FWHM
values of the detected peaks in the MSI experiment. Panels (e, f) show examples from the study of Lou et al.
where the peak width was estimated to be 0.05 Da. Panels (b, d) show the peaks of (a, c), respectively, after
successful alignment with msalign

spectra that are to be aligned, and they should cover the mass
range of interest. Useful reference peaks are matrix peaks or
molecules that are likely to be omnipresent such as AMP.
However, in case no a priori reference peaks are available, a
peak picking can be performed on the average spectrum of each
sample. The found peaks are clustered with a certain tolerance
which can be mass range-dependent, for instance, 1500,
800, and 600 ppm for the m/z blocks 0–200, 201–500, and
501–1000, respectively [4]. For alignment, we highly recom-
mend peak detection rates >90% and a minimum of five peaks
which are distributed across the whole mass range. Once a set
354 Benjamin Balluff and Liam A. McDonnell

of peaks has been identified, tools such as msalign in Matlab

can be used to align all spectra to the selected reference peaks.
Msalign is parameterized by default to the alignment of protein
spectra, which makes a tuning for metabolomic data necessary.
The most important parameters are MaxShiftValue and
WidthOfPulsesValue. The latter determines the peak width of
the reference peaks in the synthetic reference spectrum and
should be set to the average FWHM of the peaks observed
(Fig. 2e, f). MaxShiftValue depends on the observed peak shift
between measurements (Fig. 2a, c).
7. Once the data has been aligned (Fig. 2b, d), a global represen-
tative (e.g., average spectrum) can be created and a peak pick-
ing performed on that summarized spectrum. This enables a
direct comparison between datasets and groups.
8. Finally, a weak smoothing of the 2D data can be performed to
reduce the pixel-to-pixel variation in the MSI images, by, for
instance, a moving average filter of 3  3 pixels [16].
9. Statistical data analysis of metabolomic MSI can be done in
many ways but usually follows the same concepts as performed
in proteomic MSI, which has been discussed comprehensively
[18]. In principle, two approaches can be distinguished: super-
vised analysis, which aims for finding discriminating single
metabolites or patterns of those between known groups, and
unsupervised analysis, which aims for finding new groups based
on the detected metabolic profiles.

3.4 Metabolite Determination of the identity of differentiating metabolomic mass

Identification signals is usually of high interest to get an understanding of the
biological meaning of the metabolite(s) in the context of the proj-
ect or to enable the detection of the molecule(s) using other more
targeted techniques. The most common approach for MSI data is
depicted in Fig. 1.
1. Due to lack of comprehensive MALDI MS/MS databases and
search algorithms metabolites, a highly accurate mass is an
important indication of identity. FT-based mass spectrometers
provide the highest mass accuracy (<1 ppm) but offer lower
throughput. A common approach is therefore to perform
screening on high-throughput instruments, such as MALDI-
TOF/TOF systems, and determine the accurate mass of the
metabolites to interest in single subsequent experiments
[4, 10]. The accurate molecular weight can then be matched
with a small mass tolerance (1–3 ppm) to a public database,
such as HMDB (http://www.hmdb.ca/), METLIN (https://
metlin.scripps.edu), or lipid maps (http://www.lipidmaps.
org/). Furthermore, it can be checked if the observed frag-
mentation spectra are in line with the structure of the proposed
 Imaging of Metabolites 355

match [9, 19]. Ultimate confirmation can be obtained by

comparing the fragmentation pattern with one from a
corresponding standard compound [20].
2. Further biological interpretation can be offered by pathway anal-
ysis software, such as Reactome (http://www.reactome.org/) or
MetaboAnalyst (http://www.metaboanalyst.ca) [1]. The latter
also integrates a comprehensive set of statistical data analysis tools.

4 Notes

1. Several strategies have been reported for limiting postmortem

changes of the tissue metabolome:
Heat stabilization of ex vivo tissues (HS)—enzymes are inacti-
vated by rapidly heating the tissue to 95  C using high-power
heating blocks. It has been demonstrated to halt postmortem
degradation of adenine nucleotides that otherwise occurs in
ex vivo snap-frozen tissue [21].
In situ freezing (ISF) under anesthesia has been demonstrated
to be superior to ex vivo snap-frozen tissue [22] and heat-
stabilized ex vivo tissues for maintaining metabolite integrity
[7], including adenine nucleotides.
In situ focused microwave irradiation (FMW) uses focused
microwaves to rapidly heat (<2 s) the tissue and denature/
deactivate enzymes. It is the most rapid deactivation method
and has been reported that, while ISF and FMW provide similar
results for most metabolites, there are several metabolites that
are best analyzed using FMW because of their very rapid post-
mortem changes [23].
2. On-target laser spot size may be estimated by preparing a thin
layer of the matrix onto an ITO slide, by using a dry matrix
deposition method and a matrix solution containing a high
concentration of volatile solvent. A raster with a pixel-to-
pitch much larger than the laser spot diameter, and a large
number of laser shots (sufficient to ablate all matrix), results
in a laser burn pattern in the matrix coating. Through using
laser irradiation conditions identical to that used for MALDI
MSI, the on-target burn pattern provides an estimate of the
on-target laser spot size. Through the analysis of a raster of
burn spots, an average spot size can be determined.

Acknowledgments

This work has been made possible with the support of the Dutch
Province of Limburg. BB thanks the European Union (ERA-NET:
TRANSCAN 2), ITEA, and RVO (ITEA 151003/ITEA 14001)
for their financial support.
356 Benjamin Balluff and Liam A. McDonnell

References
1. Gika HG, Wilson ID, Theodoridis GA (2014) matrix-assisted laser desorption/ionization-
The role of mass spectrometry in nontargeted Fourier transform ion cyclotron resonance
metabolomics. In: Simó C, Cifuentes A, Gar- mass spectrometry? Anal Chem 88
cı́a-Cañas V (eds) Fundamentals of advanced (10):5281–5289. https://doi.org/10.1021/
Omics technologies: from genes to metabo- acs.analchem.6b00460
lites. Elsevier, Amsterdam 11. Lou S, Balluff B, Cleven AH et al (2016) An
2. Addie RD, Balluff B, Bovee JV et al (2015) experimental guideline for the analysis of his-
Current state and future challenges of mass tologically heterogeneous tumors by MALDI-
spectrometry imaging for clinical research. TOF mass spectrometry imaging. Biochim
Anal Chem 87(13):6426–6433. https://doi. Biophys Acta 1865:957–966. https://doi.
org/10.1021/acs.analchem.5b00416 org/10.1016/j.bbapap.2016.09.020
3. Nilsson A, Goodwin RJ, Shariatgorji M et al 12. Shariatgorji M, Nilsson A, Bonta M et al
(2015) Mass spectrometry imaging in drug (2016) Direct imaging of elemental distribu-
development. Anal Chem 87(3):1437–1455. tions in tissue sections by laser ablation mass
https://doi.org/10.1021/ac504734s spectrometry. Methods 104:86–92. https://
4. Lou S, Balluff B, Cleven AH et al (2017) Prog- doi.org/10.1016/j.ymeth.2016.05.021
nostic metabolite biomarkers for soft tissue 13. Aichler M, Elsner M, Ludyga N et al (2013)
sarcomas discovered by mass spectrometry Clinical response to chemotherapy in oesopha-
imaging. J Am Soc Mass Spectrom 28 geal adenocarcinoma patients is linked to
(2):376–383. https://doi.org/10.1007/ defects in mitochondria. J Pathol 230
s13361-016-1544-4 (4):410–419. https://doi.org/10.1002/path.
5. Zhou D, Guo S, Zhang M et al (2017) Mass 4199
spectrometry imaging of small molecules in 14. Rauser S, Deininger SO, Suckau D et al (2010)
biological tissues using graphene oxide as a Approaching MALDI molecular imaging for
matrix. Anal Chim Acta 962:52–59. https:// clinical proteomic research: current state and
doi.org/10.1016/j.aca.2017.01.043 fields of application. Expert Rev Proteomics 7
6. Wu Q, Chu JL, Rubakhin SS et al (2017) (6):927–941. https://doi.org/10.1586/epr.
Dopamine-modified TiO2 monolith-assisted 10.83
LDI MS imaging for simultaneous localization 15. Schwamborn K (2017) The importance of his-
of small metabolites and lipids in mouse brain tology and pathology in mass spectrometry
tissue with enhanced detection selectivity and imaging. Adv Cancer Res 134:1–26. https://
sensitivity. Chem Sci 8(5):3926–3938. doi.org/10.1016/bs.acr.2016.11.001
https://doi.org/10.1039/c7sc00937b 16. McDonnell LA, van Remoortere A, van Zeijl
7. Mulder IA, Esteve C, Wermer MJ et al (2016) RJ et al (2008) Mass spectrometry image cor-
Funnel-freezing versus heat-stabilization for relation: quantifying colocalization. J Prote-
the visualization of metabolites by mass spec- ome Res 7(8):3619–3627. https://doi.org/
trometry imaging in a mouse stroke model. 10.1021/pr800214d
Proteomics 16(11–12):1652–1659. https:// 17. Deininger SO, Cornett DS, Paape R et al
doi.org/10.1002/pmic.201500402 (2011) Normalization in MALDI-TOF imag-
8. Shariatgorji M, Nilsson A, Goodwin RJ et al ing datasets of proteins: practical considera-
(2014) Direct targeted quantitative molecular tions. Anal Bioanal Chem 401(1):167–181.
imaging of neurotransmitters in brain tissue https://doi.org/10.1007/s00216-011-4929-
sections. Neuron 84(4):697–707. https://doi. z
org/10.1016/j.neuron.2014.10.011 18. Jones EA, Deininger SO, Hogendoorn PC et al
9. Ly A, Buck A, Balluff B et al (2016) High- (2012) Imaging mass spectrometry statistical
mass-resolution MALDI mass spectrometry analysis. J Proteome 75(16):4962–4989.
imaging of metabolites from formalin-fixed https://doi.org/10.1016/j.jprot.2012.06.
paraffin-embedded tissue. Nat Protoc 11 014
(8):1428–1443. https://doi.org/10.1038/ 19. Dekker TJ, Jones EA, Corver WE et al (2015)
nprot.2016.081 Towards imaging metabolic pathways in tis-
10. Buck A, Balluff B, Voss A et al (2016) How sues. Anal Bioanal Chem 407(8):2167–2176.
suitable is matrix-assisted laser desorption/ion- https://doi.org/10.1007/s00216-014-8305-
ization-time-of-flight for metabolite imaging 7
from clinical formalin-fixed and paraffin- 20. Buck A, Ly A, Balluff B et al (2015) High-
embedded tissue samples in comparison to resolution MALDI-FT-ICR MS imaging for
 Imaging of Metabolites 357

the analysis of metabolites from formalin-fixed, 22. Hattori K, Kajimura M, Hishiki T et al (2010)
paraffin-embedded clinical tissue samples. J Paradoxical ATP elevation in ischemic penum-
Pathol 237(1):123–132. https://doi.org/10. bra revealed by quantitative imaging mass spec-
1002/path.4560 trometry. Antioxid Redox Signal 13
21. Blatherwick EQ, Svensson CI, Frenguelli BG (8):1157–1167. https://doi.org/10.1089/
et al (2013) Localisation of adenine nucleo- ars.2010.3290
tides in heat-stabilised mouse brains using ion 23. Sugiura Y, Honda K, Kajimura M et al (2014)
mobility enabled MALDI imaging. Int J Mass Visualization and quantification of cerebral
Spectrom 345:19–27. https://doi.org/10. metabolic fluxes of glucose in awake mice. Pro-
1016/j.ijms.2013.02.004 teomics 14(7–8):829–838. https://doi.org/
10.1002/pmic.201300047
 Part VII

Study Design, Data Analysis, and Bioinformatics

 Chapter 27

Quality-Assured Biobanking: The Leiden University Medical

Center Model
Rianne Haumann and Hein W. Verspaget

Abstract
Prospective or “de novo” biobanking is becoming increasingly popular. Biobanks are installed to provide
large collections of biological materials for future medical research. Quality assurance of biobank samples is
an important aspect of biobanking. Therefore, it is vital that all samples are collected and processed in a
similar manner according to standardized procedures to ensure high-quality samples and reduce variability
in the analytical process. We describe the processes of the centralized biobanking facility at the Leiden
University Medical Center (LUMC).

Key words Biobanking, Storage, Biological materials, Blood, Quality assurance, Quality control, Pre-
analytical variations, Standard operating procedures

1 Introduction

A biobank is a collection of biological materials coupled to asso-

ciated data consisting of donor information and sample processing
information [1]. The biobank ideally provides an infrastructure for
the collection, processing, and storage of biological materials
together with a supportive information technology (IT) system
for the registration and documentation of associated data
[2, 3]. Traditionally, biobanks consisted of tissue collections from
pathology departments that remained after the diagnostic process
and were accessible for researchers for secondary scientific use
under restricted conditions. Nowadays, many biobanks are more
specialized in “de novo” collection of various biological samples
and associated data obtained with an informed consent of the
participant. The establishment of a biobank is a relatively complex
process in which many aspects from sample collection to IT systems
have to be organized and coordinated [2].
Biobanking is of great importance for medical research in order
to guarantee the availability of patient material since these biological
materials are solely collected for research purposes [4]. “De novo”

Martin Giera (ed.), Clinical Metabolomics: Methods and Protocols, Methods in Molecular Biology, vol. 1730,
https://doi.org/10.1007/978-1-4939-7592-1_27, © Springer Science+Business Media, LLC 2018

361
362 Rianne Haumann and Hein W. Verspaget

biobanks collect biological materials prospectively for yet undefined

research purposes. Biological material that remains after diagnostic
process is not allowed to be directly used for research purposes unless
it is used anonymously and without associated clinical data or when
obtained with an informed consent. Centralizing the collection and
storage of biological material in biobanks for research purposes is
more efficient than setting up multiple small collections, and it also
reduces the patient burden. Biobanks have the sole purpose of
collecting biological materials for multiple research purposes, which
are undefined at the time of collection, using biological materials
more efficiently. Quality of samples is usually better controlled
because biological materials are processed in a similar manner. There-
fore, biobanking is becoming increasingly popular and more impor-
tant for biomedical research. In addition, it is quite difficult for
individual research groups or laboratories to establish large collec-
tions of biomaterials. Consequently, centralized biobank facilities are
advocated because they are better equipped for sample collection of
large cohorts.
The importance of large numbers of high-quality samples is
emphasized by reported difficulties with sample collection and
biomarker validation. Due to heterogeneity of diseases and in
order to achieve statistical power, large numbers of samples are
needed. The development of high-throughput analysis enables
simultaneous analysis of hundreds of samples, but the availability
of sufficient samples is often difficult without biobanks [3, 5,
6]. Poste [5] reported that over 150,000 papers documented
potentially interesting biomarkers; however, only less than 100 bio-
markers have eventually been validated for clinical use. In addition,
documentation of the standardization of sample collection is often
lacking, and researchers cannot obtain samples with uniform qual-
ity hampering high-quality research. Analysis of samples which in
retrospect were found to be of suboptimal or even bad quality is
costs ineffective and should be avoided. Freedman et al. [7] calcu-
lated that almost 50% of costs spent for biomedical research were
wasted due to irreproducible results in preclinical research related
to poor sample or analytical quality. These observations show an
enormous amount of research associated with high costs and little
high-quality output for improvement of diagnosis and treatment of
patients. Therefore, sample quality should be better monitored and
registered resulting in a better research output. In addition, jour-
nals nowadays also require documented sample handling and stor-
age information for the publication of research on biomarkers in
biological samples [8].
The process of sample analysis can be divided into three phases:
the pre-analytical, the analytical, and the post-analytical phase. The
pre-analytical phase comprises the sample collection, processing,
and storage of biological samples. The analytical phase consists
of the biochemical analysis of the samples followed by the
 Quality Assured Biobanking 363

post-analytical phase in which the data is analyzed. The

pre-analytical phase is particularly prone to errors with the highest
reported error rates ranging from 31% to 75%. During the analyti-
cal phase, the prevalence of errors is reported to be around 13–32%,
and during the post-analytical phase, an error rate of 9–31% is
thought to occur, both of which are considerably lower than the
pre-analytical phase [9, 10]. Sample errors in the pre-analytical
phase are, for example, hemolysis, precentrifugation delay, wrong
centrifuge settings, postcentrifugation processing delay, and wrong
temperature settings [9]. It is very important to reduce the
pre-analytical errors because it may lead to unnecessary patient
burden, higher costs, low-quality samples, and faulty interpretation
of analytical results. Green [11] estimated that a single
pre-analytical error costs approximately € 175. Knowledge of the
effect of a pre-analytical error on sample quality is therefore essen-
tial. Samples that contain pre-analytical errors might still be suitable
for a range of analytical techniques, depending on the analyte to be
assessed, and thus still valuable for research and biobanks.
Sample processing procedures should be described in standard
operating procedures (SOPs) to enable uniform processing of sam-
ples. Unfortunately, there is not a worldwide consensus on the best
method for sample processing that could be used by every labora-
tory or biobank. The National Cancer Institute (NCI) and the
International Society for Biological and Environmental Reposi-
tories (ISBER) are examples of organizations that propose best
practices for biobanking [12, 13]. One of the reasons of the publi-
cation of multiple best practices is the limited fundamental research
of the influence of pre-analytical variables on sample quality.
Research has shown that variations in pre-analytical processing
can have a pronounced effect on the results of the analysis of
biological samples [9, 10]. Thus, although challenging, it is imper-
ative to establish large collections of uniformly processed and
stored biological samples, and therefore it is important to docu-
ment the sample processing and storage conditions in the biobank-
ing process. Consequently, documentation of sample handling
should be coupled to the individual biological sample. Sample
track-and-trace data are important and can be registered in, for
example, a Biobank Information Management System (BIMS),
which will be explained in the following paragraph.
The Parelsnoer Institute (www.parelsnoer.org) is one of the
largest clinical biobank organizations in the Netherlands. The Par-
elsnoer Institute is a federative biobank of the eight university
medical centers in the Netherlands, which collect biological mate-
rials and associated data from patients of specific disease cohorts.
The Parelsnoer Institute provides an infrastructure for the collec-
tion of samples and data capture. At the moment, the Parelsnoer
collection encompasses 15 disease cohorts: abdominal aortic aneu-
rysm, congenital heart disease, cerebrovascular accident, diabetes,
364 Rianne Haumann and Hein W. Verspaget

hereditary colorectal cancer, inflammatory bowel disease, ischemic

heart disease, leukemia-myeloma-lymphoma, multiple endocrine
neoplasia, neurodegenerative diseases, esophageal and gastric can-
cer, pancreatic cancer and pancreatitis, renal failure, rheumatoid
arthritis, and Parkinson’s disease. For each cohort specific
biological samples are collected such as serum, plasma, DNA,
feces, urine, cerebrospinal fluid, pancreatic cyst fluid, viable cells,
fresh frozen tissues, and/or formalin-fixed paraffin-embedded
(FFPE) tissues. Clinical data and sample-specific data are stored in
a national disease-specific database. More detailed information
about the organization of the Parelsnoer Institute can be found in
the article of Manniën et al. [14].
The LUMC participates in the Parelsnoer biobank but also has
its own clinical biobanks organized in the Cura Rata biobank (www.
lumc.nl/org/cura-rata/) [15]. The Cura Rata biobank is a centra-
lized facility within the LUMC that provides an infrastructure for
the collection of biological samples and the storage of associated
data for more than 25 disease entities, similar to the Parelsnoer
Institute collections. The Cura Rata biobank and the Parelsnoer
biobank partially run parallel because samples are stored for both
biobanks and sample processing is identical. Furthermore, the Cura
Rata and the Parelsnoer biobanks use the same Biobank Informa-
tion Management System (BIMS), called SampleNavigator®, for
the registration of the processing data associated with the sample.
In this chapter we will describe the procedures for biobanking
within the LUMC [15, 16].

2 Software

2.1 HiX The healthcare information exchange system (HiX) (ChipSoft,

Amsterdam, the Netherlands) is a clinical data repository for the
registration of Electronic Health Records (EHRs) that is being
used by the LUMC. Unique is the integration of biobanks within
HiX at our institution. Registration of informed consents and order
management for biosample ordering are integrated in the EHRs
which enables physicians and research nurses to include patients
into the different biobanks. The process of sample ordering is not
part of the diagnostic ordering process but is a separate ordering
entity and is solely used for biobank purposes. HiX is used for the
registration of relevant clinical data that is associated with the
biobank sample. Information is transferred into a database, and
the personal data are encrypted in order to provide privacy protec-
tion for participants of the biobank. Figure 1 shows an overview of
the IT-directed logistics of the biobank.
 Quality Assured Biobanking 365

BIMS
Clinic Sample Navigator
Biobank Laboratory
Coding
√
Informed Consent Macro-tube
Physician/Nurse
Electronic
Health Record
√ Biobank Aliquong
(EHR) Order DNA isolaon
Clinical data Management

√ √
Micro-tubes DNA

UMC- Database Venipuncture

√
-80 °C +4 °C Freezers
Parelsnoer
Blood tube √ Facility
BIMS
ProMISe
Database

√ BIMS

Research
Evaluaon Distribuon
Proposal

Fig. 1 Schematic presentation of the information technology-directed logistics from the LUMC centralized
biobanking facility. BIMS Biobank Information Management System, √ electronic registration moment

2.2 SampleNavigator® software (Lelystad, the Netherlands) is

SampleNavigator® employed as the Biobank Information Management System
(BIMS) for the LUMC biobanks. SampleNavigator is designed to
capture relevant sample information such as track-and-trace, regis-
tration of standard operating procedures (SOP) deviations, and
user registration (Fig. 1). Track and trace of the biological samples
include registration of the time of collection, arrival time at the
biobank, processing time, storage location, and aliquot history.

3 Methods (Procedures)

3.1 Sample Order A physician informs a patient about the biobank according to the
rules and regulations and agreements of the LUMC. When a
patient decides to participate, the patient has to sign an informed
consent which is registered in HiX. After registration, a physician
orders samples specifically for the biobank through order manage-
ment in HiX.
366 Rianne Haumann and Hein W. Verspaget

3.2 Collection The biobank order from order management is directly coupled to
and Transport the BIMS with an HL7 message coupling. The central blood
collection facility (CBCF) receives the biobank order electronically,
and labels are printed for the collection tubes and containers just
before sample collection. Samples from the outpatient clinics and
wards are collected between 8.00 and 16.00, since the Biobank
Laboratory closes at 17.30. Samples that are collected outside
working hours will be processed the following morning with regis-
tration of SOP deviation(s) in BIMS. Sample processing has to be
completed preferably within 2 h after venipuncture.
The first time point for the track-and-trace registration of the
sample is printing the label for the sample collection tubes and
containers. The collection date and time are checked, and if the
collection time is incorrect, which can occur in the wards or when
samples are obtained from outlying hospital facilities, the time and
date will be adjusted on the printed labels. Biological samples are
collected in the SOP-defined labeled tubes and container. Blood
collection tubes are inverted on an automatic shake platform for the
exact number of inversions that are required for proper sample
mixing (see Note 1). A courier service transports the samples to
the Clinical Chemistry Laboratory where the biobank samples are
transferred to the Biobank Laboratory (see Note 2).

3.3 Processing The biological samples are registered and scanned (Handscanner
Opticon OPI-2201, Hoofddorp, the Netherlands) at the Biobank
Laboratory. This is a second time point in the BIMS track-and-trace
registration of the biological samples. The collection date and time
are checked and if needed adjusted and described in the BIMS. At
that moment patient information is still present on the collection
tubes and containers but is then encrypted with a sample code. New
labels (Salm and Kipp BY800537 BPT-21-461, Breukelen, The
Netherlands) are printed (Zebra printer GX420t, Lincolnshire,
Illinois, USA) with a unique sample code. The biological samples
are further processed under this specific sample code without any
visible personal data.
SOP deviations are registered in the BIMS. The first step is to
identify and indicate a SOP deviation by the biobank lab technician
and then to specify the deviation in the BIMS. Several predeter-
mined SOP deviations are documented: hemolysis, lipemic, icterus,
incorrect tube, inadequate filling of the citrate tube, incorrect
temperature before processing, prolonged precentrifugation time,
incorrect mixing or homogenizing, incorrect centrifuge setting,
storage problems, and others. In “others” the biobank lab techni-
cian can describe a SOP deviation that is not listed in the predeter-
mined SOP deviations.
Next, the biological samples are processed at the Biobank
Laboratory under the ISO 15189:2012. The biological samples
are aliquoted into pre-specified smaller volumes, as indicated in
 Quality Assured Biobanking 367

Table 1. New labels are printed for the cryotubes, which are a third
time point in the track-and-trace registration of the sample. SOP
deviations for the aliquots are also registered in the BIMS. After
registration the samples are processed according to the SOP.
Table 1 provides an overview of the processing steps for the various
biological samples. In short, the samples are centrifuged at a specific
force, time, and temperature. Next, the supernatant is transferred
into aliquots with a volume of 100–900 μL per aliquot (see Notes
3 and 4). Aliquots should be filled starting with the lowest micro-
sample number to the highest micro-sample number. The aliquot
tubes are closed and stored temporarily at the Biobank Laboratory
according to SOP protocol. Short-term storage temperature is
dependent on the biological material; DNA is temporarily stored
at 4  C, and the other biological samples are temporarily stored at
20  C, with the latter of a maximum duration of 4 h (see Note 5).
EDTA cell pellets from the plasma tubes are used for DNA
isolation. DNA isolation is registered in the BIMS, and new labels
are printed for the Autopure Qubes (AutoGen, Inc., Holliston,
MA, USA). Cell pellet is transferred to the Autopure Qubes, and
the EDTA tube is washed with PBS and the suspension added to
the Autopure Qubes. The Autopure Qubes must be filled with a
minimum of 6 mL and a maximum of 10 mL. The tube is closed
and transported to the Department of Clinical Genetics for robotic
DNA isolation. After isolation, DNA is transported back to the
Biobank Laboratory. At the Biobank Laboratory, new labels are
printed for the micro tubes containing DNA, which is registered
as a time point for the track-and-trace registration of the sample. At
the Department of Clinical Genetics, the quality parameters
260/280 absorbance ratio and DNA concentration are measured
which together with the volume of the aliquots is registered in the
BIMS. When the DNA stock concentration is >50 ng/μL, DNA is
marked as “yes DNA ok” in the BIMS, at 10–50 ng/μL the DNA is
marked as “no DNA not ok,” and at <10 ng/μL the DNA will not
be stored. DNA marked as “not ok” will result in a new venipunc-
ture at the next visit of the patient to again isolate high-quality
DNA. DNA is resuspended in Tris-EDTA buffer in the macro tube
and divided into two micro tubes. DNA is stored as suspension in
the fridge at 4  C.

3.4 Storage Aliquots are temporarily stored at 20  C for a maximum duration

of 4 h at the Biobank Laboratory except for DNA which is stored at
4  C. At noon or at the end of the day, the aliquots are transferred
to the long-term storage freezers. The sample boxes are transferred
to the long-term freezer facility on dry ice and insulated tempex
boxes to ensure limited temperature differences (see Note 6).
The refrigerator and freezer facility of the biobank are exclu-
sively used for the biobank. The same brand of freezers is used to
ensure uniform storage. The samples are predominantly stored at
 368

Table 1
Standard operating procedures at the LUMC biobank for processing of blood and urine

Biological
Max. processing Short-term Long-term
material Tube type Centrifugation Aliquots
time (h) storage ( C) storage ( C)
Volume Time Temperature Volume
Specified Additives Color (mL) RCF (min) ( C) Number (μL)
Blood Plasma NaHeparin Green 10 4 2350 10 20 5 500 20 80
EDTA Purple 4a, 10b 4 2350 10 20/4 2–3a, 5b 500 20 80
NaCitrate Blue 9 4 2350 10 20 5 500 20 80
PPP PPACK/ Blue 3 4 2350 10 20 5 100 20 80
Rianne Haumann and Hein W. Verspaget

citrate
Serum Serum Red 2c, 10d 4 2350 10 20 1–2c, 5d 500 20 80
e e f
Serum gel Red 3.5 , 4 2350 10 20 2–3 , 5 500 20 80
8.5f
DNA EDTA Red 4 2 500 4
RNA PAXgene 2.5 4 20 80
Urine Fresh 4 Yellow 9.5 4 2350 10 4 6 900 20 80
24 h 4 Yellow 9.5 4 2350 10 4 6 900 20 80
Pot 4 Yellow 9.5 4 2350 10 4 6 900 20 80
Tubes a to f are related to aliquot numbers a to f
 Quality Assured Biobanking 369

80  C. BIMS is used as a track-and-trace system for the registra-

tion of the drawer, sample box, and position of the samples in the
box. Multiple use of an aliquot is also registered in BIMS.
The freezers are cooled with wastewater from the hospital,
which saves hundreds of thousands of euros per year on
air-conditioning costs. The refrigerators and freezers are connected
to the internal alarm and temperature registration system of the
biobank facility. In case of an alarm, the infrastructural LUMC
emergency and breakdown service will contact the responsible
employee(s). In case of an emergency, employees of the centralized
biobank facility will carry out the emergency plan (see Note 7).

3.5 Sample Aliquots from the biobank can be requested by researchers after
Distribution they submit an approved protocol to the biobank. The scientific
biobank committee of the specific disease-related biobanks evalu-
ates the protocol and decides if permission is granted.
The data warehouse of the biobank consists of the integration
of EHRs and BIMS information (see Fig. 1). EHRs data are used
for the extraction of relevant clinical information related to the
specific biological sample. BIMS can be used for the extraction of
data associated with the biological sample such as processing time,
type of sample, and previous use of the sample. Researchers can
request, for example, samples of a specific gender or age. Research-
ers can, for example, request samples that have a processing time
with a maximum duration of 1 h and never been used before.
Biobanks using electronic bioresource data registration enable
researchers to request specific samples that are fit for their specific
studies and analysis.

4 Notes

1. Serum tubes should be left in an upright position at room

temperature for at least 30–60 min before processing the
tubes to allow proper clotting.
2. Urine is transferred on ice and should be processed before the
ice is melted.
3. Additional aliquots should be filled if more volume of the
biological material is left. The extra volume should not be
divided over the initial aliquots.
4. If there is less volume than required for an aliquot, one should
still fill the aliquot and register a SOP deviation.
5. PAXgene tubes for RNA analyses are directly frozen at 20  C
in an upright position. The patient identification information
should be removed for privacy reasons and replaced by a label
with a non-identifiable BIMS macro number on the tube.
370 Rianne Haumann and Hein W. Verspaget

6. Use cool elements or dry ice in insulated boxes for transport of

frozen material.
7. In case of emergency with storage of samples in the freezers, be
sure to have an emergency plan to ensure high quality of the
samples.

Acknowledgments

We would like to acknowledge Lilian Boonman for her assistance in

the management of the biobank organization.

References
1. Vaught J, Kelly A, Hewitt R (2009) A review of 9. Bonini P et al (2002) Errors in laboratory med-
international biobanks and networks: success icine. Clin Chem 48(5):691–698
factors and key benchmarks. Biopreserv Bio- 10. Carraro P, Plebani M (2007) Errors in a stat
bank 7(3):143–150 laboratory: types and frequencies 10 years later.
2. Vaught J (2016) Biobanking comes of age: the Clin Chem 53(7):1338–1342
transition to biospecimen science. Annu Rev 11. Green SF (2013) The cost of poor blood spec-
Pharmacol Toxicol 56:211–228 imen quality and errors in preanalytical pro-
3. Riegman PH et al (2008) Biobanking for bet- cesses. Clin Biochem 46(13):1175–1179
ter healthcare. Mol Oncol 2(3):213–222 12. Institute N.C. (2016) NCI best practices for
4. Womack C, Mager SR (2014) Human biospecimen resources
biological sample biobanking to support tissue 13. Lori D (2011) Best practices for repositories:
biomarkers in pharmaceutical research and collection, storage, retrieval and distribution of
development. Methods 70(1):3–11 biological materials for research. International
5. Poste G (2011) Bring on the biomarkers. Society for Biological and Environmental
Nature 469(7329):156–157 Repositories Vancouver, Canada
6. Button KS et al (2013) Power failure: why small 14. Manniën J et al (2017) The Parelsnoer Insti-
sample size undermines the reliability of neuro- tute: a national network of standardized clinical
science. Nat Rev Neurosci 14(5):365–376 biobanks in the Netherlands. Open J Bioresour
7. Freedman LP, Cockburn IM, Simcoe TS (2015) 4:3
The economics of reproducibility in preclinical 15. Available from: https://www.lumc.nl/org/
research. PLoS Biol 13(6):e1002165 cura-rata/. Cited 19 June 2017
8. Simeon-Dubach D, Burt AD, Hall PA (2012) 16. Simons J, Verspaget HW (2010) Traceerbaar
Quality really matters: the need to improve biobankbeheer in het LUMC. Laboratorium
specimen quality in biomedical research. J Magazine, 9
Pathol 228(4):431–433
 Chapter 28

Extracting Knowledge from MS Clinical Metabolomic Data:

Processing and Analysis Strategies
Julien Boccard and Serge Rudaz

Abstract
Assessing potential alterations of metabolic pathways using large-scale approaches today plays a central role
in clinical research. Because several thousands of mass features can be measured for each sample with
separation techniques hyphenated to mass spectrometry (MS) detection, adapted strategies should be
implemented to detect altered pathways and help to elucidate the mechanisms of pathologies. These
procedures include peak detection, sample alignment, normalization, statistical analysis, and metabolite
annotation. Interestingly, considerable advances have been made over the last years in terms of analytics,
bioinformatics, and chemometrics to help massive and complex metabolomic data to be more adequately
handled with automated processing and data analysis workflows. Recent developments and remaining
challenges related to MS signal processing, metabolite annotation, and biomarker discovery based on
statistical models are illustrated in this chapter considering their application to clinical research.

Key words Metabolomics, Mass spectrometry, Data processing, Data analysis

1 Introduction

Since its early days, metabolomics has been intensively used to

evaluate biochemical signatures and find biological readouts for
clinical diagnosis and prognosis [1]. According to the preponderant
roles of metabolites in many biological processes, metabolic phe-
notyping constitutes an essential element for providing mechanistic
insights into pathological states, detecting potential adverse effects
due to toxic exposure or evaluating beneficial effects of therapeutic
treatments [2]. Coupling a separation technique (chromatographic
and electrophoretic) with mass spectrometry (MS) today constitu-
tes a reference approach in metabolomics because of the large
number of signals that can be recorded at low concentration within
a single analytical run [3]. Consequently, raw data present an
intrinsic two-dimensional structure with a separation and a m/z
dimension. To take advantage from the wealth of biochemical

Martin Giera (ed.), Clinical Metabolomics: Methods and Protocols, Methods in Molecular Biology, vol. 1730,
https://doi.org/10.1007/978-1-4939-7592-1_28, © Springer Science+Business Media, LLC 2018

371
372 Julien Boccard and Serge Rudaz

information recorded, relevant signals need to be distinguished

from artifacts and noisy variables in these massive datasets.

2 Materials

2.1 Availability The different tools described in this chapter are available through
of the Algorithms diverse programing platforms such as R and Matlab or modules
integrated in global metabolomic processing workflow, including
XCMS, MZmine, apLCMS, xMSanalyzer, mzMatch, MetAlign,
MetSign, and OpenMS. The reader is invited to consult the
detailed review of metabolomic tools and resources proposed by
Misra et al. [4] or specific documentation for more details on the
usage of each method.

2.2 Biological The case study of Victor Yushchenko’s poisoning with an acute
Background dose of dioxin in 2004, when he was candidate for the presidency of
Ukraine, will be used as working example throughout this chapter.
Dioxin (2,3,7,8-tetrachlorodibenzo-p-dioxin, TCDD) is a
by-product of industrial processes involving chlorine, such as
waste incineration and chemical manufacturing. Human popula-
tions are mainly exposed to low levels of dioxin through food, and
the effects of chronic exposure to dioxin are unknown. However, in
case of acute intoxication, hepatitis, neuropathy, and skin damages
are just a few examples of pathologies that were observed.
Urine samples were collected at different time points after
intoxication and compared with urine of healthy volunteers. All
samples were analyzed by ultra high-pressure liquid chromatogra-
phy (UHPLC) coupled to quadrupole time-of-flight mass spec-
trometry (Acquity UPLC & Xevo G2 QTOF, Waters).
Multivariate analysis was carried out to distinguish the two groups
(Comet, Nonlinear Dynamics & SIMCA-P, Umetrics). Detailed
methodological aspects of the example (e.g., overall technical
modus operandi) can be found elsewhere [5].

3 Methods

3.1 Pre-Acquisition Among the multiple steps required for proper sample analysis,
Normalization ensuring that the total concentration of each sample is in the
same range constitutes a prerequisite for any comparison, even in
the case of relative quantitation based on peak intensity variations.
The sources of this variability include an intrinsic heterogeneity of
the samples, problems of stability, or inherent variations arising
from prior sample preparation. It is therefore crucial to remove or
at least reduce the overall discrepancies by adjusting the sample
volume or concentration prior to analysis.
 MS Data Processing and Analysis in Clinical Metabolomics 373

Because these differences are strongly dependent on the nature

of the sample, different pre-acquisition normalization strategies
should be specifically implemented:
1. Cell cultures and biopsies can be normalized according to the
total amount of biological material, evaluated as cell number or
total protein weight.
2. Fluctuations associated with biological fluids can be corrected
using either global indicators of the total solute concentration
(e.g., osmolality or specific gravity), or reference compounds
(e.g., creatinine for urine normalization).

3.2 MS Acquisition The choice of an experimental approach is often determined by the

availability of prior knowledge regarding a biological question. An
untargeted acquisition can be used to provide a global monitoring
of metabolites to derive data-driven hypotheses. Instrumental
developments recently improved the acquisition capacity offered
by MS. High-resolution MS, such as time-of-flight (TOF) and
Fourier transform technologies, today allows the untargeted and
simultaneous monitoring of thousands of analytes in full scan
mode, with the determination of elemental compositions from
accurate mass measurements.
1. Fully nontargeted metabolomic approaches are commonly
used for blind biomarker discovery, without any or low prior
knowledge about potentially altered metabolic pathways.
2. Alternatively, the abundances of groups of metabolites can be
specifically investigated within the full scan MS information
using a retrospective examination, namely, post-acquisition
data mining or post-targeted analysis (see Subheading 3.6).
Both strategies are complementary and can be simultaneously
implemented to provide different views of the metabolic informa-
tion. An example of UHPLC-HRMS data structure generated by the
untargeted acquisition of a urine sample from Victor Yushchenko is
provided in Fig. 1. Whatever the strategy’s choice, signal processing
and data mining constitute critical aspects that need to be carefully
considered to extract relevant biochemical knowledge in MS-based
clinical metabolomics. Accounting for both the characteristics of the
analytical setup and biochemical knowledge is mandatory for proper
peak detection, alignment, and metabolite annotation.

3.3 MS Signal Because irrelevant signals, including solvent contaminants and

Processing detector noise, may strongly deteriorate downstream data analysis,
a number of steps should be sequentially implemented to remove
analytical artifacts and emphasize valuable biochemical information
related to real analytes. It includes denoising, peak picking, decon-
volution, and alignment.
374 Julien Boccard and Serge Rudaz

Fig. 1 UHPLC-HRMS data with corresponding total ion chromatogram and total mass spectrum, recorded for a
urine sample collected from Victor Yushchenko after intoxication

3.3.1 Peak Detection Appropriate feature detection constitutes the starting point of MS
and Deconvolution data processing, and an additional challenge comes from the pres-
ence of multiple signals, such as isotopic peaks, fragment, or adduct
ions, produced by analytes in the sample.
1. Signal smoothing and denoising are often used in combination
with baseline correction as a first step of processing to increase
the data quality. It includes moving average windows, Savitzky-
Golay fitting, or wavelet transform to remove random noise,
and background subtraction, or Gaussian filters for chemical
denoising [6].
2. Peak picking is then achieved to extract relevant signals by
generating a series of mass features associated with specific
peak areas, m/z, and retention/migration time values. This is
done by detecting signals above a specified intensity or signal-
to-noise ratio threshold, and mathematical models (e.g.,
Gaussian) are used to assess peak consistency based on shape
constraints and integrate signals between peak boundaries [7].
3. Multiple signals can be generated for each analyte, including
isotopic peaks, fragment or adduct ions, n-mers, or degrada-
tion products. These derivative ions must be grouped using
deconvolution and deisotoping procedures. Peak annotation
associates signals sharing similar retention/migration times
and peak shapes based on specific m/z differences
corresponding to known chemical modifications.
Proper peak detection requires peak shape parameters defining
individual peak limits to be carefully adjusted according to the
 MS Data Processing and Analysis in Clinical Metabolomics 375

analytical conditions and the type of device (see Note 1). Other
approaches based on empirical peak modeling and isotopic patterns
are also available to evaluate the parameters from the data, e.g.,
from quality control samples.

3.3.2 Alignment Small differences are usually observed for both the separation and
the m/z dimensions for the same signal, when comparing several
samples. However, ensuring proper matching between occurrences
of the same feature across samples constitutes a prerequisite before
further analysis of the data. On the one hand, discrepancies in the
m/z dimension are usually small when using calibration ions during
MS data acquisition. On the other hand, shifts in the separation
dimension can originate from numerous factors including differ-
ences of temperature, mobile-phase pH, pump pressure, matrix
effects, sample carryover, and column degradation or clogging.
Appropriate alignment strategies are therefore used to correct or
at least limit variations between runs, while keeping chemical selec-
tivity intact [8].
Two types of strategies can be considered:
1. Warping methods account for the separation dimension as a
whole and have usually high computing time and resources
requirements. These approaches do not depend on prior peak
picking or deconvolution of the signals and lack the ability to
account for selectivity changes. Because warping methods max-
imize the overlap between samples by stretching or shrinking
the separation dimension, a reference template needs first to be
defined, and a sequential procedure is carried out to align
segments of chromatograms/electropherograms defined by a
width criterion. Classical warping methods include correlation
optimized warping (COW) and dynamic time warping (DTW).
Recent developments involve parametric and semi-parametric
strategies or piecewise cross-correlation warping
approaches [9].
2. Peak matching approaches grouping mass features across sam-
ples take advantage of prior deconvolution and peak picking
steps to condense the chemical information. Similar mass fea-
tures falling between the boundaries defined by a tolerance
threshold in the separation and the m/z dimension are usually
considered as the same analyte (see Note 2). A specific weight
can be associated with errors in each dimension according to
the characteristics of the analytical setup [7].
On a practical point of view:
1. Carefully choose the parameters for alignment according to the
characteristics of the data.
376 Julien Boccard and Serge Rudaz

2. Check the absence of detected peaks in some of the samples,

leading to missing values in the aligned data table (see Notes 3
and 4).

3.3.3 Post-Acquisition Post-acquisition normalization strategies can be required to

Normalization remove remaining unwanted systematic variations due to signal
intensity drift over time during the analytical sequence.
Compare sample-based strategies based on the evaluation of a
global correction factor including:
1. Total sum normalization (TSN, sum of all features).
2. MS total useful signal (MSTUS, sum of features common to all
samples).
3. Probabilistic quotient normalization (PQN, feature-dependent
correction).
Quality control samples (QCs) analyzed at regular intervals
should be used as landmarks to monitor the stability of the analyti-
cal process and guarantee proper data acquisition. By checking the
repeatability of the QCs and/or their response to dilution, irrele-
vant features such as noisy or saturated signals must be removed to
ensure data quality and relevance [10].
Investigate the potential drifts according to the processing
order of the analytical sequence using QC-based approaches.
These methods include an evaluation using local regression models
fitted for each feature detected in the QC samples. Use one of the
following by evaluating at least five important identified features:
1. Locally weighted scatterplot smoothing (LOESS).
2. Robust spline correction (QC-RSC).
3. Support vector regression correction (QC-SVRC).

3.3.4 Data Scaling The variability of abundance can strongly differ from one metabo-
lite to another with very different biological consequences, and the
most abundant compounds are not necessarily the most relevant.
Additionally, low concentrated metabolites are more prone to ana-
lytical errors than highly abundant compounds. Hence, at least two
scaling procedures need to be evaluated to make the features com-
parison effective:
1. Individual values should be standardized using a scaling factor,
e.g., standard deviation (SD) in the case of unit variance scaling
or the square root of the SD for Pareto scaling.
2. If the results are not satisfactory, variables transformations
(e.g., log function), are useful to standardize the signals and
correct heteroscedasticity [11].
 MS Data Processing and Analysis in Clinical Metabolomics 377

3.4 Multivariate Because a large number of signals are recorded for providing a
Statistics global picture of biological matrices, multivariate methods consti-
tute efficient solutions to highlight metabolic patterns [12]. Two
data analysis strategies can be sequentially carried out:
1. Exploratory approaches based on unsupervised learning
principles.
2. Predictive models involving supervised algorithms.

3.4.1 Unsupervised Unsupervised methods evaluating the dataset without accounting

Analysis for experimental groups or expected outcome offer reliable tools to
extract the major trends in the data (see Note 5). They provide an
efficient way to detect natural groupings among the distribution of
samples and should always be carried out as a first step.
For that purpose, principal component analysis (PCA) consti-
tutes the most widely used and convenient method for initial
exploratory data analysis. PCA summarizes the sources of variability
in the data by building a new subspace of low dimension [13]. Inter-
pretation is usually limited to the first PCs carrying the most salient
trends (see Note 6). Graphical outputs provide a reliable and con-
venient mean to detect groups of similar samples or signals:
1. The coordinates of the observations (scores) allow the distribu-
tion of the samples to be easily displayed and evaluated.
2. Relationships between variables can be highlighted by inspect-
ing their contributions (loadings) to the model.
These plots allow trends or unexpected patterns in the dataset
to be detected:
1. Check for any significant analytical drift by inspecting the QCs
projection.
2. Search for potential outliers, i.e., observation(s) outside the
Hotelling’s T2 ellipse (95% confidence interval) and consider
removing them from further analysis.

3.4.2 Supervised In contrast to the previous approaches, supervised methods make

Analysis use of prior knowledge linked to the experimental setup to drive the
analysis. Supervised models aim at predicting a response (depen-
dent variable Y ) from the measured metabolomic data (indepen-
dent variables X).
Partial least squares (PLS or projection to latent structures)
regression is one of the largely used methods in metabolomics.
The PLS algorithm builds a low-dimension subspace based on
linear combinations of the experimental variables. By these means,
the PLS model summarizes the X data in a synthetic manner with
the ability to predict the Y response [14]. Because the comparison
of experimental groups is often the goal of metabolomic studies,
PLS discriminant analysis (PLS-DA) is widely used to distinguish
378 Julien Boccard and Serge Rudaz

Score plot S-plot

70 VY samples 1
Control group
0.8
50
0.6

30 0.4

0.2

pp(corr)
10
to

0
-10
-0.2

-30 -0.4

-0.6
-50
-0.8

-70 -1
-60 -40 -20 0 20 40 60 -0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2

tp pp
Discriminant features:
Exogenous compounds: propofol, selvofurane, ibuprofen (drug therapy)
Endogenous compounds: mainly steroid metabolites and bile acids

Fig. 2 OPLS-DA model comparing the urine samples collected from Victor Yushchenko with age-matched
controls (3682 variables, Pareto scaling). (a) Score plot, (b) S-plot

known classes of observations. The orthogonal PLS algorithm

(O-PLS) [15] facilitating model interpretation constitutes a rou-
tinely applied method to analyze metabolomic datasets with effi-
cient diagnostic tools, e.g., the S-plot. Variable contributions to the
model can serve as an objective basis for detecting potential bio-
markers (see Note 7). An example of outputs from an OPLS-DA
model comparing urine samples from Victor Yushchenko and
healthy control volunteers matched for age is provided in Fig. 2.
This multivariate data analysis should be carefully interpreted,
particularly when the number of observation is limited (<20 per
group). The model validity and its capacity to handle new observa-
tions reliably need to be carefully evaluated. To this aim, the data
should be divided into a training set of observations, used to build
the model capturing the relation between metabolomic data and
the response variable(s), and a test set including different samples,
used to evaluate the reliability of the prediction.
Common procedures available to assess model validity with
respect to perturbations of the training and test sets include:
1. Cross-validation.
2. Permutation tests.
3. Bootstrap.
 MS Data Processing and Analysis in Clinical Metabolomics 379

In any case, the reliable estimation of model generalization

ability (validation) requires the prediction of another completely
new, independently generated, observations addressing the same
biological question (external test set).

3.5 Metabolite It is largely recognized that metabolite annotation today constitu-

Identification tes the major bottleneck of untargeted metabolomics. Because the
biochemical interpretation relies strongly on the quality of metab-
3.5.1 Annotation Levels
olite annotation, harmonization efforts including the metabolomic
and Databases
standard initiative (MSI) and the coordination of standards in
metabolomics (COSMOS) were made to propose standards for
reporting metabolite annotations [16, 17]. Therefore, four anno-
tation levels are considered:
l Level 1 annotation constitutes the highest annotation quality, as
it requires the comparison of two or more orthogonal properties
with reference material measured in the same experimental
conditions.
l Level 2 annotation involves compounds matching a molecular
formula and/or the physicochemical/spectral properties of
databases entries.
l Level 3 is based on a single molecular property and is thus
limited to chemical class annotation.
l Level 4 comprises unknown compounds that cannot be anno-
tated but distinguished from other analytes in the samples based
on their specific m/z and retention/migration time values.
Reliable metabolite annotation constitutes the cornerstone of
biochemical interpretation, and public databases including endog-
enous and exogenous metabolites were developed over the last
years, e.g., the Human Metabolite Database (HMDB), LipidMaps,
METLIN, the Golm Metabolome Database, and KEGG. A
detailed review of the different spectral databases was recently
proposed by Vinaixa et al. [18].

3.5.2 Annotation Using The starting point of the identification process is the definition of
MS and MS/MS Information an elemental formula, and the development of high-resolution MS
devices strongly facilitated this task. Modern instruments today
provide high mass accuracy and resolving power, leading to the
evaluation of reliable isotopic patterns, thus reducing the number
of possible molecular formulae. Additional information is then
needed to provide an unambiguous molecular structure:
1. A widely used approach is based on the acquisition of a frag-
mentation spectrum using MS/MS or MSn. The obtained data
are then compared with entries of experimental MS/MS spec-
tra databases using a similarity score based on matching peaks
or differences between spectra. Relative intensities can also be
380 Julien Boccard and Serge Rudaz

Fig. 3 In silico fragmentation leading to level 2 annotation of cholic acid glucuronide from an unknown
compound of molecular formula C30H48O11, with a m/z of 584.3188 and a retention time of 13.31 min

taken into consideration. This approach is straightforward and

reliable but remains limited by the number and quality of
MS/MS entries in public databases (see Note 8).
2. By applying fragmentation rules inferred from real reference
MS/MS data, a theoretical fragmentation spectrum can be com-
puted in silico for any molecular structure using dedicated algo-
rithms. Candidate compounds are first selected according to their
m/z and fragmented in silico. A score is then evaluated to rank the
candidates according to the similarity between the experimental
and predicted MS/MS spectra (see Notes 9 and 10).
Because they benefit from both strategies, hybrid approaches
combining systematic fragmentation and experimental MS/MS
spectra comparison constitute also a promising way to help metab-
olite identification [19]. The in silico fragmentation leading to level
2 annotation of cholic acid glucuronide from an unknown bio-
marker candidate detected in Victor Yushchenko’s urine samples
is provided in Fig. 3.

3.5.3 Annotation Using Metabolite annotation tools are based on exact mass, isotopic, and
Other Molecular Properties fragmentation patterns, but other physicochemical properties can
from the Separation be used to reduce the number of candidate compounds. When
possible, the separation dimension provides retention/migration
 MS Data Processing and Analysis in Clinical Metabolomics 381

times that provides complementary orthogonal information to

MS. This information can be measured or predicted (see Note 11):
1. Standard injection under the same analytical conditions allows
compounds sharing the same molecular formula and poten-
tially related fragmentation patterns to be distinguished. In
vitro synthesis can be carried out from precursors when sources
of putative metabolites are scarce or inexistent (e.g., phase II
metabolites) [20].
2. Retention/migration time prediction based on quantitative
structure-property relationship (QSPR) models offers an alter-
native to narrow down the number of candidates and improve
the reliability of metabolite annotation (see Note 12).
3. Emerging complementary properties such as the cross-
collisional section (CCS) obtained from ion mobility MS con-
stitute also valuable complementary criteria for reliable metab-
olite annotation [21].
Because unambiguous metabolite identification (level 1)
requires the injection of reference material (not always available in
exploratory studies), efficient solutions should be found to limit the
number of false-positive annotations. Combining experimental
information stored in public databases, in silico physicochemical
properties predictions, and combinatorial systematic searches pro-
vides relatively efficient metabolite annotation.

3.6 Retrospective Metabolomic-based approaches focusing on a subset of the meta-

Analysis bolome can be carried out by combining nontargeted data acquisi-
tion and automatic metabolite annotation. When particular ion
features are extracted from full scan MS datasets, specific metabolic
information can be obtained in a retrospective manner using post-
targeted approaches (see Note 13) [22]. This reexamination may
help to search for novel biomarkers within a chemical class or
specific biological pathway of interest. The data can therefore be
examined later “on demand” for other compound classes, given
additional biological information about possible metabolite
perturbations.

4 Notes

1. Peak detection is usually carried out by binning spectral data

into intervals that are processed separately, but defining an
optimal bin size often constitutes an issue. This parameter is
of crucial importance, because small bins will tend to split
peaks, while large bins will include unresolved signals.
382 Julien Boccard and Serge Rudaz

2. This type of approach remains however somewhat limited,

because it does not involve an explicit model for evaluating
systematic drifts.
3. Missing values can be due to a signal near the detection limit, a
distorted peak shape, or a true difference due to the factors of
the experimental setup. In such a situation, it is desirable to use
a gap-filling procedure to avoid adding too many zeros in the
final data table.
4. Analyzing skewed data including a large number of null values
may lead to irrelevant results and wrong conclusions. This is of
particular importance when downstream statistical methods
assume a normal distribution of the values.
5. Statistical workflows for analyzing metabolomic data should
always start with exploratory approaches to ensure data consis-
tency and check for potential outliers or unwanted sources of
systematic variability (e.g., analytical drift, batch effect, etc.).
6. Because biological phenomena are not necessarily orthogonal,
PCs are sometimes challenging to link with interpretable varia-
tions. In that case, other methods such as independent compo-
nent analysis (ICA) constitute relevant solutions. Alternatively,
PCA can serve as a pretreatment step before downstream anal-
ysis with clustering and classification methods.
7. The variable importance in the projection (VIP) provides an
efficient criterion to assess the predictive merit of each variable
in the whole PLS model.
8. The comparison of relative ratios between fragments is not
always straightforward when different MS devices are used. In
that context, multiple collision energies are usually included in
databases to provide a more exhaustive picture of fragmenta-
tion patterns.
9. Arbitrary fragmentation rules are not universal, and alternative
schemes can be applied for an extensive exploration of potential
candidates.
10. Fragmentation cascades can also be considered as a graph that
can be optimized for providing biochemically meaningful pat-
terns. These fragmentation trees help to rationalize the way
fragments are generated from precursor ions and can be com-
pared to experimental MS/MS spectra of unknown com-
pounds. Because similar fragmentation subtrees can be
detected, it may help compound identification by matching a
molecule sharing a close molecular structure, even if the true
compound is not present in the database.
11. As biochemical reactions give rise to a very large and diverse
range of compounds, it is not rare to observe structural, posi-
tional, and constitutional isomers. Therefore, retention/
 MS Data Processing and Analysis in Clinical Metabolomics 383

migration times are often considered as a mandatory informa-

tion in annotation workflows.
12. Predictions remain however limited to fixed analytical condi-
tions, and a specific database is often developed in each labora-
tory. These differences between platforms and experimental
conditions (stationary phase chemistry, mobile phase composi-
tion and temperature, running time, etc.) make standardiza-
tion difficult. Interestingly, this major limitation can be tackled
efficiently using knowledge about mechanisms of the separa-
tion process. For example, a prediction of retention times
related to specific gradient conditions is possible. In the case
of reversed-phase liquid chromatography (RPLC) analysis, the
linear solvent strength theory model (LSS) can be used.
13. For the sake of clarity, the classification could be summarized
according to the acquisition mode: (1) targeted metabolomics
(or metabolic profiling), i.e., the detection of a series of specific
metabolites, and (2) nontargeted metabolomics (metabolic
fingerprinting, global metabolomics), i.e., an analysis based
on the nontargeted monitoring of all signals in the samples.

References
1. Ellis DI, Dunn WB, Griffin JL, Allwood JW, (2012) Instrument and process independent
Goodacre R (2007) Metabolic fingerprinting binning and baseline correction methods for
as a diagnostic tool. Pharmacogenomics 8 liquid chromatography-high resolution-mass
(9):1243–1266. https://doi.org/10.2217/ spectrometry deconvolution. Anal Chim Acta
14622416.8.9.1243 740:12–19. https://doi.org/10.1016/j.aca.
2. Clayton TA, Lindon JC, Cloarec O, Antti H, 2012.06.014
Charuel C, Hanton G, Provost JP, Le Net JL, 7. Castillo S, Gopalacharyulu P, Yetukuri L, Oresic
Baker D, Walley RJ, Everett JR, Nicholson JK M (2011) Algorithms and tools for the prepro-
(2006) Pharmaco-metabonomic phenotyping cessing of LC-MS metabolomics data. Chemo-
and personalized drug treatment. Nature 440 metr Intell Lab 108(1):23–32. https://doi.
(7087):1073–1077. https://doi.org/10. org/10.1016/j.chemolab.2011.03.010
1038/nature04648 8. Lange E, Tautenhahn R, Neumann S, Gropl C
3. Boccard J, Veuthey JL, Rudaz S (2010) Knowl- (2008) Critical assessment of alignment proce-
edge discovery in metabolomics: an overview dures for LC-MS proteomics and metabolo-
of MS data handling. J Sep Sci 33(3):290–304. mics measurements. BMC Bioinformatics
https://doi.org/10.1002/jssc.200900609 9. https://doi.org/10.1186/1471-2105-9-
4. Misra BB, van der Hooft JJJ (2016) Updates in 375
metabolomics tools and resources: 9. Tomasi G, Savorani F, Engelsen SB (2011)
2014–2015. Electrophoresis 37(1):86–110. Icoshift: an effective tool for the alignment of
https://doi.org/10.1002/elps.201500417 chromatographic data. J Chromatogr A 1218
5. Jeanneret F, Boccard J, Badoud F, Sorg O, (43):7832–7840. https://doi.org/10.1016/j.
Tonoli D, Pelclova D, Vlckova S, Rutledge chroma.2011.08.086
DN, Samer CF, Hochstrasser D, Saurat JH, 10. Dunn WB, Broadhurst D, Begley P, Zelena E,
Rudaz S (2014) Human urinary biomarkers of Francis-McIntyre S, Anderson N, Brown M,
dioxin exposure: analysis by metabolomics and Knowles JD, Halsall A, Haselden JN, Nicholls
biologically driven data dimensionality reduc- AW, Wilson ID, Kell DB, Goodacre R, C
tion. Toxicol Lett 230(2):234–243. https:// HSMH (2011) Procedures for large-scale met-
doi.org/10.1016/j.toxlet.2013.10.031 abolic profiling of serum and plasma using gas
6. Krishnan S, Vogels JTWE, Coulier L, Bas RC, chromatography and liquid chromatography
Hendriks MWB, Hankemeier T, Thissen U coupled to mass spectrometry. Nat Protoc 6
384 Julien Boccard and Serge Rudaz

(7):1060–1083. https://doi.org/10.1038/ Hankemeier T, Luchinat C, Walther D, Stein-

nprot.2011.335 beck C (2015) COordination of Standards in
11. van den Berg RA, Hoefsloot HCJ, Westerhuis MetabOlomicS (COSMOS): facilitating
JA, Smilde AK, van der Werf MJ (2006) Cen- integrated metabolomics data access. Metabo-
tering, scaling, and transformations: improving lomics 11(6):1587–1597. https://doi.org/10.
the biological information content of metabo- 1007/s11306-015-0810-y
lomics data. BMC Genomics 7. https://doi. 18. Vinaixa M, Schymanski EL, Neumann S,
org/10.1186/1471-2164-7-142 Navarro M, Salek RM, Yanes O (2016) Mass
12. Boccard J, Rudaz S (2014) Harnessing the spectral databases for LC/MS- and GC/MS-
complexity of metabolomic data with chemo- based metabolomics: state of the field and
metrics. J Chemometr 28(1):1–9. https://doi. future prospects. Trends Anal Chem
org/10.1002/Cem.2567 78:23–35. https://doi.org/10.1016/j.trac.
13. Hotelling H (1933) Analysis of a complex of 2015.09.005
statistical variables into principal components. J 19. Lynn KS, Cheng ML, Chen YR, Hsu C,
Educ Psychol 24:417–441. https://doi.org/ Chen A, Lih TM, Chang HY, Huang CJ,
10.1037/h0071325 Shiao MS, Pan WH, Sung TY, Hsu WL
14. Wold S, Sjostrom M, Eriksson L (2001) (2015) Metabolite identification for mass
PLS-regression: a basic tool of chemometrics. spectrometry-based metabolomics using mul-
Chemometr Intell Lab 58(2):109–130. tiple types of correlated ion information. Anal
https://doi.org/10.1016/S0169-7439(01) Chem 87(4):2143–2151. https://doi.org/10.
00155-1 1021/ac503325c
15. Trygg J, Wold S (2002) Orthogonal projec- 20. Jeanneret F, Tonoli D, Hochstrasser D, Saurat
tions to latent structures (O-PLS). J Chemo- JH, Sorg O, Boccard J, Rudaz S (2016) Evalua-
metr 16(3):119–128. https://doi.org/10. tion and identification of dioxin exposure bio-
1002/Cem.695 markers in human urine by high-resolution
metabolomics, multivariate analysis and in vitro
16. Sumner LW, Amberg A, Barrett D, Beale MH, synthesis. Toxicol Lett 240(1):22–31. https://
Beger R, Daykin CA, TWM F, Fiehn O, doi.org/10.1016/j.toxlet.2015.10.004
Goodacre R, Griffin JL, Hankemeier T,
Hardy N, Harnly J, Higashi R, Kopka J, Lane 21. Paglia G, Williams JP, Menikarachchi L,
AN, Lindon JC, Marriott P, Nicholls AW, Reily Thompson JW, Tyldesley-Worster R,
MD, Thaden JJ, Viant MR (2007) Proposed Halldorsson S, Rolfsson O, Moseley A,
minimum reporting standards for chemical anal- Grant D, Langridge J, Palsson BO, Astarita G
ysis. Metabolomics 3(3):211–221. https://doi. (2014) Ion mobility derived collision cross sec-
org/10.1007/s11306-007-0082-2 tions to support metabolomics applications.
Anal Chem 86(8):3985–3993. https://doi.
17. Salek RM, Neumann S, Schober D, Hummel J, org/10.1021/ac500405x
Billiau K, Kopka J, Correa E, Reijmers T,
Rosato A, Tenori L, Turano P, Marin S, 22. Jeanneret F, Tonoli D, Rossier MF, Saugy M,
Deborde C, Jacob D, Rolin D, Dartigues B, Boccard J, Rudaz S (2016) Evaluation of ster-
Conesa P, Haug K, Rocca-Serra P, O’Hagan S, oidomics by liquid chromatography hyphenated
Hao J, van Vliet M, Sysi-Aho M, Ludwig C, to mass spectrometry as a powerful analytical
Bouwman J, Cascante M, Ebbels T, Griffin JL, strategy for measuring human steroid perturba-
Moing A, Nikolski M, Oresic M, Sansone SA, tions. J Chromatogr A 1430:97–112. https://
Viant MR, Goodacre R, Gunther UL, doi.org/10.1016/j.chroma.2015.07.008
Perspectives

The field of clinical metabolomics can be defined as the use of metabolomics-driven

techniques in a clinical context, studying disease-related changes of the metabolome,
which ultimately reflect possible alterations in genome, transcriptome, and proteome.
With the metabolome seen as “the closest link” to the phenotype, it is not surprising that
many researchers attribute a key role to metabolomics in the context of personalized
medicine. Nevertheless, metabolomics-driven clinical research has long struggled fulfilling
the high expectations and hopes which had been put into it. In the beginning, mainly
biomarker discovery-driven case-control studies dominated the field of clinical metabolo-
mics. While such research was mainly successful on the discovery of gene mutation-related
metabolic markers, giving rise to strong metabolic correlations [1], it did frequently fail in
the context of less pronounced metabolic phenotypes. However, a recent paradigm shift [2],
an advanced analytical technology, and a much more quantitative understanding of human
metabolism undoubtedly allow to paint a bright future for the field.
With respect to the analytical technologies and protocols applied in clinical metabolo-
mics, mass spectrometry and nuclear magnetic resonance spectroscopy stand central. Fur-
thermore, it has become recognized that a detailed molecular analysis is important for a
sound understanding of (patho-)biochemical mechanisms. Such is reflected in advanced
analytical methods, providing a more detailed mapping of chemical isomers, not solely
limited to diastereomers but also including geometrical isomers as well as enantiomers and
appreciating differential biological effects [3]. Additionally there has been a shift from
exploratory metabolic profiling toward large panel (quantitative) targeted metabolomics
approaches. The latter allowing for a more quantitative view of clinical data sets in correla-
tion with possible disease mechanisms. Analytically speaking, advanced mass spectrometric
techniques, increased quantitative capabilities and robustness, and the integration of differ-
ent analytical techniques play an important role. In this respect, ion mobility mass spec-
trometry has recently had a large impact particularly on clinical lipidomics, allowing for lipid
class and isomer separation [4, 5] and further more resulting in the first “plug-and-play”
commercially available lipidomics platform, designed for nonexperts in the field, quantifying
up to 1000 lipid species. It will be interesting to see if such approaches might enter the
clinical routine in the future. However, applicability, robustness, and ease of data interpreta-
tion are key features for driving clinical metabolomics forward. In addition it will be of
importance to facilitate data integration of several omics layers as well as pathway analysis [6]
as only in combination we will gain a better understanding of disease mechanisms and in
turn hopefully become capable of designing advanced treatments.

Leiden, The Netherlands Martin Giera

Martin Giera (ed.), Clinical Metabolomics: Methods and Protocols, Methods in Molecular Biology, vol. 1730,
https://doi.org/10.1007/978-1-4939-7592-1, © Springer Science+Business Media, LLC 2018

385
386 Perspectives

References

1. van Karnebeek CDM, Bonafe L, Wen X-Y et al (2016) NANS-mediated synthesis of sialic acid is required
for brain and skeletal development. Nat Genet 48(7):777–784. https://doi.org/10.1038/ng.3578.
http://www.nature.com/ng/journal/v48/n7/abs/ng.3578.html#supplementary-information
2. Johnson CH, Ivanisevic J, Siuzdak G (2016) Metabolomics: beyond biomarkers and towards mechan-
isms. Nat Rev Mol Cell Biol 17(7):451–459. https://doi.org/10.1038/nrm.2016.25. http://www.
nature.com/nrm/journal/v17/n7/abs/nrm.2016.25.html#supplementary-information
3. Serhan CN, Petasis NA (2011) Resolvins and protectins in inflammation resolution. Chem Rev 111
(10):5922–5943. https://doi.org/10.1021/cr100396c
4. Jónasdóttir HS, Papan C, Fabritz S et al (2015) Differential mobility separation of leukotrienes and
protectins. Anal Chem 87(10):5036–5040. https://doi.org/10.1021/acs.analchem.5b00786
5. Lintonen TPI, Baker PRS, Suoniemi M et al (2014) Differential mobility spectrometry-driven shotgun
lipidomics. Anal Chem 86(19):9662–9669. https://doi.org/10.1021/ac5021744
6. Ritchie MD, Holzinger ER, Li R a (2015) Methods of integrating data to uncover genotype-phenotype
interactions. Nat Rev Genet 16(2):85–97. https://doi.org/10.1038/nrg3868
 INDEX

A D
All ion fragmentation (AIF) .....................................46–57 Data analysis ....................... 32, 46, 69, 70, 97, 116, 137,
Amino acid .............................................9, 10, 12, 13, 32, 153, 155, 164, 170, 253, 262, 318, 321, 322,
33, 277, 305, 309, 310, 329 326, 346, 350, 351, 355, 373, 377, 378
ATP ....................................................................... 335, 338 Derivatization ..............................98, 138, 143, 153, 194,
Autoxidation................................... 88, 91, 267, 274, 284 205, 210, 212, 239, 241–243, 248, 254, 258,
263, 268, 270, 274, 280, 282, 286, 288, 291,
B 305, 346, 349
Bile acid ..................................... 103–105, 107–110, 176, Direct infusion .............................................137, 227–236
Disease ................................. v, 3, 5, 9, 10, 12–19, 29, 31,
177, 180, 184, 204
Biobanking ........................................................... 361–370 36–38, 59, 83, 93, 111, 124, 136, 175,
Bis(trimethylsilyl)trifluoroacetamide (BSTFA) ..........241, 213, 227, 257, 267, 283, 305, 317, 318,
362–364, 369, 385
243, 245, 270, 272, 274, 286, 291
Bligh and dyer ............................................. 228, 230, 231
E
Blood ......................10, 14, 16–18, 37, 83–91, 111–120,
168, 169, 213, 215, 218, 219, 221–223, 239, Electron capture negative ionization (ECNI) ............248,
242, 257, 259–261, 263, 283, 290, 318, 366, 368 258, 284, 289, 290
Butylated hydroxytoluene (BHT) ........... 85, 86, 90, 126, Electrospray ionization source (ESI) ........................7, 47,
128, 139, 141, 165, 168, 171, 193, 284, 285, 288 52, 73–75, 77–81, 85, 87, 95, 96, 114, 118, 120,
143, 144, 177, 194, 199, 201, 202, 209, 220,
C 222, 299, 306–310
Epoxides ............................................................... 123–132
Calibration............................................47, 88, 94, 95, 97,
101, 108, 119, 120, 132, 153, 176, 177, 184, Extraction .........................48, 59, 62, 64, 65, 68, 74, 75,
189, 194, 199, 202, 204–206, 208, 210, 211, 79, 84–87, 95, 99, 125, 126, 132, 137–139, 142,
244, 252–255, 259, 261–264, 269, 270, 272, 143, 153, 155–157, 159, 164, 165, 167–169,
171, 176, 193, 204, 210, 228, 230, 231,
277–282, 319, 321, 333, 339, 347, 375
Capillary electrophoresis (CE) ....................... v, 7, 13, 14, 234, 235, 239–244, 268, 269, 271, 274,
75, 77, 80, 81, 88, 90, 96, 189, 199, 210, 278, 279, 282, 286, 330–332, 337,
338, 345, 369
295–303, 309, 310
Capillary electrophoresis-mass spectrometry
F
(CE-MS) ........... v, 7, 13, 14, 295–299, 301, 302,
305–312 Fatty acid (FA) ............................7, 9, 13, 30, 31, 34, 36,
Cecum................................................................... 247–255 75, 77, 80, 83, 111–114, 118–120, 123, 124,
Chemical shift............................319, 321, 322, 326, 334, 135, 164, 171, 172, 175–178, 181–184, 186,
335, 340, 341 189, 190, 192, 194, 195, 198–202, 204–207,
Citrate ..........................................31, 138, 157, 224, 242, 209–211, 228, 247–255, 257, 259–261, 263,
277, 281, 335, 366 283, 317
Correlation spectroscopy (COSY) ............. 326, 334, 340 Feces........................................... 247–255, 317, 318, 321,
Cytochrome............................................................ 18, 123 322, 364
Folch ........................................................ 75, 86, 228, 288

Martin Giera (ed.), Clinical Metabolomics: Methods and Protocols, Methods in Molecular Biology, vol. 1730,
https://doi.org/10.1007/978-1-4939-7592-1, © Springer Science+Business Media, LLC 2018

387
 CLINICAL METABOLOMICS: METHODS AND PROTOCOLS
388 Index
Fourier transform ion cyclotron resonance-mass Lipoxins ....................................63, 64, 67, 178, 187, 195
spectrometry (FTICR-MS)............. 220, 221, 349 Liquid chromatography (LC).................... 29, 52, 74, 79,
Fructose ................................................................ 322, 335 85, 87, 105, 127, 132, 138, 156, 177,
180, 184, 204, 205, 209, 216, 222, 228,
G 229, 232, 235
Gas chromatography (GC) ........................ 7, 29, 84, 176, Liquid chromatography-mass spectrometry
189, 205, 210, 239–241, 243, 250, 253, 259, (LC-MS) ........................................ v, 7, 14, 46–57,
262, 268, 277–280, 282 59–65, 67, 69–71, 83–91, 94, 97, 98, 101,
103–105, 107–120, 123–132, 138–140,
Gas chromatography mass spectrometry
(GC-MS).................... v, 7, 13, 84, 176, 189, 194, 142–144, 153, 157, 158, 171, 175–177, 180,
239–245, 247–255, 257–264, 272, 278–281, 184, 193, 194, 213, 215, 218, 219, 221–223,
229, 230, 248, 253, 258, 262, 306
283–291
Glucose ................................. 10, 13, 15, 30, 34, 35, 268,
M
278, 322
Glutathione (GSH) .............................................. 284, 285 Macrophage.................................. 30–32, 34, 36, 38, 337
Malate ..................................................277, 281, 322, 335
H Maresin ............................................................................ 66
Heteronuclear multiple quantum correlation Mass spectrometry (MS)...........................v, 7, 29, 46, 61,
(HMBC) ................................................... 334, 340 74, 84, 94, 104, 112, 124, 137, 164, 175, 213,
Heteronuclear single quantum correlation 229, 239, 247, 257, 268, 277, 297, 305, 318,
330, 352, 371
(HSQC) ..........................321, 322, 326, 334, 340
Hydrolysis........................................ 83, 86, 91, 111, 124, Matrix ..........................................47, 57, 74, 91, 98, 101,
126, 128, 205, 210, 258, 263, 270, 287 110, 112, 116, 130, 132, 160, 176, 214, 224,
228, 239, 244, 251, 268, 295, 320, 329, 341,
I 345, 347–350, 353, 355, 375
Matrix-assisted laser desorption/ionization
Imaging...................................................... 9, 15, 345–355 (MALDI) ............................ 9, 345, 347–352, 355
Immunology....................................................... 15, 29, 38 Metabolite standard initiative (MSI) .......... 46, 345, 346,
Internal standard (IS) ..........................47, 48, 50, 62, 63, 349–355, 378
67, 69, 74, 75, 85, 88, 91, 94, 100, 104, 105, 108, Metabolome ................................ 3–8, 16, 33, 35, 36, 38,
110, 113, 115–118, 120, 126, 128, 130–132, 59, 213, 215, 218, 219, 221–223, 227, 318, 327,
138, 153, 157, 159, 160, 166–173, 177, 184, 334, 346, 355, 378, 381, 385
199, 205, 208, 210, 215, 217, 224, 230, 231, Metabolomics ............................... v, 3–19, 29–38, 46–57,
233–236, 241, 248, 258, 269, 271, 277, 279, 137, 175, 239, 295, 296, 305, 317, 318, 329,
282, 287–291, 306–308, 318–320, 325, 330, 345, 346, 352, 354, 371–373, 377,
337, 339, 353 378, 381, 383
Ion mobility...........................................75, 229, 381, 385 Methyl tert-butyl ether (MTBE) ........................... 95, 99,
Isofurans ...........................................................83, 89, 283 164, 165, 167–169, 171, 172, 228
Isoprostane .....................................................83, 283, 285
Isotope .............................. 15, 34, 81, 94, 104, 110, 116, N
118, 155, 159, 172, 213–224, 241,
264, 268, 282 Nuclear magnetic resonance (NMR) .................... v, 7, 13,
29, 34, 36–38, 248, 317–327, 329–334,
L 337–340, 345

Labeling .......................15, 213, 215, 218, 219, 221–223 O

Lipid mediator.....................59–65, 67, 69–71, 175–178,
181–184, 186, 189, 190, 192, 194, 195, Oxysterol .............................................................. 267–274
198–202, 204–207, 209–211
P
Lipidomics .......................... 32, 136, 137, 164–173, 228,
230, 232, 385 Pathway................................5, 6, 8, 9, 12–16, 19, 29–31,
Lipids ...................................... 7, 30, 47, 59, 74, 83, 103, 33, 35, 38, 59, 64, 69, 70, 136, 227, 228, 240,
123, 143, 164, 176, 227, 247, 257, 267, 373, 381, 385
271, 283, 351 Pentafluorobenzyl bromide (PFBBr)................. 248, 249,
Lipidyzer............................................................... 227, 230 252, 254, 258, 259, 261, 286, 288, 291
 CLINICAL METABOLOMICS: METHODS AND PROTOCOLS
Index 389
Plasma ............................... 13, 47, 48, 50, 52, 54, 56, 67, Short chain fatty acid (SCFA) .............. 36, 247–255, 317
75, 84, 86, 104, 112, 113, 115, 116, 136–139, Single ion monitoring (SIM) ............................. 248, 253,
141, 156, 157, 159, 168, 169, 175–178, 254, 258, 262, 263, 268, 272, 274, 289
181–184, 186, 189, 190, 192, 194, 195, Solid phase extraction (SPE) ............................62, 65, 85,
198–202, 204–207, 209–211, 214, 216–221, 87, 91, 125, 126, 128, 130, 132, 228,
224, 227, 230, 240–242, 252, 257, 259, 261, 268, 269, 272, 286
269, 271, 284, 287, 288, 317, 364, 367 Soluble epoxide hydrolase ............................................ 124
Protectins...................................................................66, 67 Sphingolipids .......................31, 135–139, 141–145, 153,
155–159, 164
Q Succinate.................................................15, 31, 277, 281,
Quenching.................................219, 224, 278, 279, 282, 322, 335
Supercritical fluid chromatography (SFC)........ 73, 74, 79
330–332, 336–338

R T

Resolvins .............................. 36, 164, 165, 170, 306, 378 Tissue ..............................5, 9, 10, 13, 15, 35, 37, 56, 64,
75, 83–91, 123, 124, 126, 128–132, 136, 156,
S 164, 165, 168, 171, 283, 290, 324, 345–351,
355, 361, 364
Serotonine ....................................................................... 17 Total correlation spectroscopy (TOCSY) .......... 321, 322,
Serum..........................9, 10, 13, 47, 48, 50, 56, 93–101, 326, 334, 340
104, 105, 113, 115, 116, 137–139, 141, 156, Tricarboxylic acid (TCA) cycle .............34, 329, 348, 350
157, 159, 216–221, 224, 227, 230, 240–242, Triglyceride...................................................111–120, 210
250, 252, 259, 261, 318, 337, 364, 368, 369