(Methods in Molecular Biology 1579) Charles A. Galea (Eds.) - Matrix Metalloproteases - Methods and Protocols-Humana Press (2017)

Methods in
Molecular Biology 1579
Charles A. Galea Editor
Matrix
Metalloproteases
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
Matrix Metalloproteases
Methods and Protocols
Edited by
Charles A. Galea
Drug Delivery, Disposition and Dynamics, Monash Institute of Pharmaceutical Sciences,
Monash University, Parkville, VIC, Australia
Editor
Charles A. Galea
Drug Delivery, Disposition and Dynamics
Monash Institute of Pharmaceutical Sciences
Monash University
Parkville, VIC, Australia
ISSN 1064-3745 ISSN 1940-6029 (electronic)

Methods in Molecular Biology
ISBN 978-1-4939-6861-9 ISBN 978-1-4939-6863-3 (eBook)
DOI 10.1007/978-1-4939-6863-3
Library of Congress Control Number: 2017930809
© Springer Science+Business Media LLC 2017

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is
concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction
on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation,
computer software, or by similar or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not
imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and
regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to
be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty,
express or implied, with respect to the material contained herein or for any errors or omissions that may have been made.
The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Printed on acid-free paper
This Humana Press imprint is published by 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
The matrix metalloprotease (MMP) field has witnessed enormous advances since Jerome
Gross and Charles Lapière first observed in 1962 an enzymatic activity (collagen degrada-
tion) associated with tadpole tail metamorphosis. Since the identification of this enzyme
(interstitial collagenase or MMP-1), more than 20 closely related and evolutionarily con-
served vertebrate MMPs have been discovered. These MMPs and their endogenous inhibi-
tors (TIMPS) are involved in a diverse range of functions including tissue remodeling,
immunity, inflammation, and angiogenesis. The first part of this book outlines recent
advances in the expression and purification of MMPs in various expression systems, high-
lighting the advantages and disadvantages of each. In Part II we highlight how various
biophysical methods such as X-ray crystallography, NMR spectroscopy, and small angle
X-ray scattering, in combination with computational analyses (Part III), can provide a
detailed understanding of the molecular mechanism of action of these enzymes. Part IV
details how experimental and bioinformatics approaches can be used to define the substrate
specificity of MMPs while Part V discusses methods for detecting MMP activity in vitro and
in vivo. In Part VI we present various methods for the development and characterization of
MMP-based inhibitors as potential therapeutics for the treatment of various diseases. The
final part presents an overview of the involvement of MMPs in various diseases and their
potential as diagnostic biomarkers.
Parkville, VIC, Australia Charles A. Galea
v
Contents
Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v
Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix
Part I Expression and Purification of Matrix Metalloproteases

1 Expression and Purification of Matrix Metalloproteinases
in Escherichia coli . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Krishna K. Singh, Ruchi Jain, Harini Ramanan, and Deepak K. Saini
2 Expression and Purification of a Matrix Metalloprotease
Transmembrane Domain in Escherichia coli . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Charles A. Galea
3 Heterologous Expression of the Astacin Protease Meprin β
in Pichia pastoris . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
Dagmar Schlenzig and Stephan Schilling
Part II Structural Characterization of Matrix Metalloproteases

4 Structural Studies of Matrix Metalloproteinase by X-Ray Diffraction . . . . . . . . . 49
Elena Decaneto, Wolfgang Lubitz, and Hideaki Ogata
5 Mapping Lipid Bilayer Recognition Sites of Metalloproteinases
and Other Prospective Peripheral Membrane Proteins . . . . . . . . . . . . . . . . . . . 61
Tara C. Marcink, Rama K. Koppisetti, Yan G. Fulcher,
and Steven R. Van Doren
6 Using Small Angle X-Ray Scattering (SAXS) to Characterize
the Solution Conformation and Flexibility of Matrix
Metalloproteinases (MMPs) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
Louise E. Butt, Robert A. Holland, Nikul S. Khunti, Debra L. Quinn,
and Andrew R. Pickford
Part III Computational Simulations of Matrix Metalloproteases

7 Molecular Dynamics Studies of Matrix Metalloproteases . . . . . . . . . . . . . . . . . . 111
Natalia Díaz and Dimas Suárez
Part IV Determining Matrix Metalloprotease

Substrate Specificity
8 Determining the Substrate Specificity of Matrix Metalloproteases
using Fluorogenic Peptide Substrates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
Maciej J. Stawikowski, Anna M. Knapinska, and Gregg B. Fields
9 Time-Resolved Analysis of Matrix Metalloproteinase Substrates
in Complex Samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
Pascal Schlage, Fabian E. Egli, and Ulrich auf dem Keller
vii
viii Contents
10 Identification of Protease Cleavage Sites by Charge-Based Enrichment

of Protein N-Termini . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
Zon W. Lai and Oliver Schilling
11 Mapping the Substrate Recognition Landscapes
of Metalloproteases Using Comprehensive Mutagenesis . . . . . . . . . . . . . . . . . . 209
Colin A. Kretz
Part V Detection of Matrix Metalloproteases

12 Detection of Matrix Metalloproteinases by Zymography . . . . . . . . . . . . . . . . . . 231
Rajeev B. Tajhya, Rutvik S. Patel, and Christine Beeton
13 Imaging Matrix Metalloproteases in Spontaneous Colon Tumors:
Validation by Correlation with Histopathology . . . . . . . . . . . . . . . . . . . . . . . . . 245
Harvey Hensley, Harry S. Cooper, Wen-Chi L. Chang,
and Margie L. Clapper
Part VI Matrix Metalloprotease Inhibitors

14 Virtual High-Throughput Screening for Matrix
Metalloproteinase Inhibitors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
Jun Yong Choi and Rita Fuerst
15 Computational Approaches to Matrix Metalloprotease Drug Design . . . . . . . . 273
Tanya Singh, B. Jayaram, and Olayiwola Adedotun Adekoya
16 A Simple Adaptable Blood-Brain Barrier Cell Model for Screening
Matrix Metalloproteinase Inhibitor Functionality . . . . . . . . . . . . . . . . . . . . . . . 287
Jennifer S. Myers, Joan Hare, and Qing-Xiang Amy Sang
Part VII Matrix Metalloproteases as Biomarkers

17 Matrix Metalloproteases as Biomarkers of Disease . . . . . . . . . . . . . . . . . . . . . . . 299
Fernando Luiz Affonso Fonseca, Beatriz da Costa Aguiar Alves,
Ligia Ajaime Azzalis, and Thaís Moura Gáscon Belardo
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313
Contributors
Olayiwola Adedotun Adekoya • Department of Pharmacy, University of Tromsø, Tromso,

Norway
Beatriz da Costa Aguiar Alves • Laboratório de Análises Clínicas—Anexo 3, Faculdade
de Medicina do ABC, Santo André, SP, Brazil
Ligia Ajaime Azzalis • Departamento de Ciências Biológicas, Instituto de Ciências
Químicas, Ambientais e Farmacêuticas, Universidade Federal de São Paulo, Diadema,
SP, Brazil
Christine Beeton • Department of Molecular Physiology and Biophysics, Baylor College
of Medicine, Houston, TX, USA
Thaís Moura Gáscon Belardo • Laboratório de Análises Clínicas—Anexo 3, Faculdade
de Medicina do ABC, Santo André, SP, Brazil
Louise E. Butt • Institute and Biomedical and Biomolecular Science (IBBS) and School
of Biological Sciences, University of Portsmouth, Portsmouth, UK
Wen-Chi L. Chang • Cancer Prevention and Control Program, Fox Chase Cancer Center,
Philadelphia, PA, USA
Jun Yong Choi • Department of Chemistry, The Scripps Research Institute, Jupiter, FL, USA
Margie L. Clapper • Cancer Prevention and Control Program, Fox Chase Cancer Center,
Philadelphia, PA, USA
Harry S. Cooper • Department of Pathology, Fox Chase Cancer Center, Philadelphia, PA,
USA; Cancer Prevention and Control Program, Fox Chase Cancer Center, Philadelphia,
PA, USA
Elena Decaneto • Max Planck Institute for Chemical Energy Conversion, Mülheim an
der Ruhr, Germany
Natalia Díaz • Dpto. Química Física y Analítica, Universidad de Oviedo, Oviedo,
Asturias, Spain
Steven R. Van Doren • Department of Biochemistry, University of Missouri, Columbia,
MO, USA
Fabian E. Egli • ETH Zurich, Department of Biology, Institute of Molecular Health
Sciences, Zurich, Switzerland
Gregg B. Fields • Department of Chemistry and Biochemistry, Florida Atlantic
University, Jupiter, FL, USA; Department of Chemistry, The Scripps Research Institute/
Scripps Florida, Jupiter, FL, USA; Departments of Chemistry and Biology, Torrey Pines
Institute for Molecular Studies, Port St. Lucie, FL, USA
Fernando Luiz Affonso Fonseca • Departamento de Ciências Biológicas, Instituto de
Ciências Químicas, Ambientais e Farmacêuticas, Universidade Federal de São Paulo,
Diadema, SP, Brazil; Laboratório de Análises Clínicas—Anexo 3, Faculdade de
Medicina do ABC, Santo André, SP, Brazil
Rita Fuerst • Department of Chemistry, The Scripps Research Institute, Jupiter, FL, USA
Yan G. Fulcher • Department of Biochemistry, University of Missouri, Columbia, MO, USA
Charles A. Galea • Drug Delivery, Disposition and Dynamics, Monash Institute
of Pharmaceutical Sciences, Monash University, Parkville, VIC, Australia
ix
x Contributors
Joan Hare • Institute of Molecular Biophysics, Florida State University, Tallahassee, FL, USA
Harvey Hensley • Biological Imaging Facility, Fox Chase Cancer Center, Philadelphia,
PA, USA
Robert A. Holland • Institute and Biomedical and Biomolecular Science (IBBS)
and School of Biological Sciences, University of Portsmouth, Portsmouth, UK
Ruchi Jain • Department of Molecular Reproduction, Development and Genetics, Indian
Institute of Science, Bangalore, India
B. Jayaram • Department of Chemistry, Indian Institute of Technology, HauzKhas,
New Delhi, India; Supercomputing Facility for Bioinformatics & Computational Biology,
Indian Institute of Technology, HauzKhas, New Delhi, India; Kusuma School
of Biological Sciences, Indian Institute of Technology, HauzKhas, New Delhi, India
Ulrich auf dem Keller • ETH Zurich, Department of Biology, Institute of Molecular
Health Sciences, Zurich, Switzerland
Nikul S. Khunti • Institute and Biomedical and Biomolecular Science (IBBS) and School
of Biological Sciences, University of Portsmouth, Portsmouth, UK; Diamond Light Source,
Diamond House, Harwell Science and Innovation Campus, Didcot, Oxfordshire, UK
Anna M. Knapinska • Department of Chemistry and Biochemistry, Florida Atlantic
University, Jupiter, FL, USA
Rama K. Koppisetti • Department of Biochemistry, University of Missouri, Columbia, MO,
USA; Department of Medical Microbiology and Immunology, Life Sciences Center,
University of Missouri, Columbia, MO, USA
Colin A. Kretz • Thrombosis and Atherosclerosis Research Institute and Department
of Medicine, McMaster University, Hamilton, ON, Canada
Zon W. Lai • Institute of Molecular Medicine and Cell Research, University of Freiburg,
Freiburg, Germany; Department of Genetics and Complex Diseases, Harvard T.H. Chan
School of Public Health, Boston, MA, USA
Wolfgang Lubitz • Max Planck Institute for Chemical Energy Conversion, Mülheim an
der Ruhr, Germany
Tara C. Marcink • Department of Biochemistry, University of Missouri, Columbia, MO,
USA
Jennifer S. Myers • Department of Chemistry and Biochemistry, Florida State University,
Tallahassee, FL, USA
Hideaki Ogata • Max Planck Institute for Chemical Energy Conversion, Mülheim an der
Ruhr, Germany
Rutvik S. Patel • Department of Molecular Physiology and Biophysics, Baylor College of
Medicine, Houston, TX, USA
Andrew R. Pickford • Institute of Biomedical and Biomolecular Science (IBBS) and
School of Biological Sciences, University of Portsmouth, Portsmouth, UK
Debra L. Quinn • Institute of Biomedical and Biomolecular Science (IBBS) and School of
Biological Sciences, University of Portsmouth, Portsmouth, UK
Harini Ramanan • Department of Molecular Reproduction, Development and Genetics,
Indian Institute of Science, Bangalore, India
Deepak K. Saini • Department of Molecular Reproduction, Development and Genetics,
Qing-Xiang Amy Sang • Department of Chemistry and Biochemistry, Florida State
University, Tallahassee, FL, USA; Institute of Molecular Biophysics, Florida State
University, Tallahassee, FL, USA
Contributors xi
Stephan Schilling • Department of Drug Design and Target Validation (IZI-IMWT),

Fraunhofer Institute for Cell Therapy and Immunology, Halle/Saale, Germany
Oliver Schilling • Institute of Molecular Medicine and Cell Research, University
of Freiburg, Freiburg, Germany; BIOSS Centre of Biological Signaling Studies, University
of Freiburg, Freiburg, Germany; German Cancer Consortium (DKTK) and German
Cancer Research Center (DKFZ), Heidelberg, Germany
Pascal Schlage • ETH Zurich, Department of Biology, Institute of Molecular Health
Sciences, Zurich, Switzerland
Dagmar Schlenzig • Department of Drug Design and Target Validation (IZI-IMWT),
Fraunhofer Institute for Cell Therapy and Immunology, Halle/Saale, Germany
Tanya Singh • Department of Chemistry, Indian Institute of Technology, HauzKhas,
New Delhi, India; Supercomputing Facility for Bioinformatics & Computational Biology,
Indian Institute of Technology, HauzKhas, New Delhi, India
Krishna K. Singh • Department of Molecular Reproduction, Development and Genetics,
Maciej J. Stawikowski • Department of Chemistry and Biochemistry, Florida Atlantic
University, Jupiter, FL, USA
Dimas Suárez • Dpto. Química Física y Analítica, Universidad de Oviedo, Oviedo, Spain
Rajeev B. Tajhya • Department of Molecular Physiology and Biophysics, Baylor College of
Medicine, Houston, TX, USA
Part I
Expression and Purification of Matrix Metalloproteases

Chapter 1
Expression and Purification of Matrix Metalloproteinases

in Escherichia coli
Krishna K. Singh, Ruchi Jain, Harini Ramanan, and Deepak K. Saini
Abstract
The MMP (matrix metalloproteinases) family of endopeptidases are involved in cleavage induced remodel-
ling of the extracellular matrix including collagen, fibrinogen, elastin, and gelatin. Owing to their proteo-
lytic activity which can cleave and degrade multiple intracellular substrates, the overexpression and
purification of these proteins tends to be toxic. Here we describe a novel “matrix assisted refolding” pro-
tocol to overcome the technical challenges associated with overexpression and purification of full-length
MMPs. The toxicity issue associated with MMP expression, is circumvented by expressing the recombi-
nant protein in Escherichia coli in an inactive insoluble form. The methodology used for obtaining full-
length MMP2 protein from these inclusion bodies, by its subsequent purification and refolding using
affinity chromatography, through a single-step matrix based refolding protocol is presented here. The
protocol described yields high concentrations of pure full-length and active MMP2 protein useful for
downstream applications.
Key words MMP2, Inclusion bodies, Affinity chromatography, Refolding, Zymography
1 Introduction
MMPs (matrix metalloproteinases) comprise a family of 23 pro-

teins which are metal ion dependent endopeptidases [1]. Owing to
their proteolytic activity towards matrix proteins such as collagen,
gelatin, and elastin, MMPs perform a number of physiological
roles in normal tissues including tissue remodelling, wound heal-
ing, and bone morphogenesis. However, the activity of this class of
proteins has been found to be highly upregulated under various
pathological conditions including cancer [2–5]. Therefore, the
expression and activity of MMP proteins can serve as a biomarker
for malignancies associated with cancer progression [5, 6]. MMPs
are also a promising target for cancer therapeutics, albeit with some
caveats on account of the physiologically relevant roles played by
MMPs in normal cellular physiology [7]. This arises due to a poor
understanding of the functional roles of these proteins and their
Charles A. Galea (ed.), Matrix Metalloproteases: Methods and Protocols, Methods in Molecular Biology, vol. 1579,
DOI 10.1007/978-1-4939-6863-3_1, © Springer Science+Business Media LLC 2017
3
4 Krishna K. Singh et al.
substrates in different biological conditions. It is hence imperative

to further study the MMP family of proteins in greater detail.
However, analysis of purified MMPs has been limited due to the
toxic nature of these proteins, which poses a technical challenge in
their overexpression, purification, and further characterization.
MMPs are secreted as zymogens in an inactive pro-MMP form,
that are proteolytically processed and activated by other proteases
(e.g., serine proteases) and can undergo an autocatalytic activation
process [8]. This proteolytic processing is critical for the activation
of pro-MMPs to yield the functional MMP protein. Under normal
physiological conditions, the activity of the full-length protein is
kept under regulatory check by members of another protein family,
known as tissue inhibitor of metalloproteases (TIMPs). Under
pathological conditions, such as cancer, the balance between
TIMPs and MMPs is altered and is shifted towards activation of
MMPs [9]. The activated MMPs lead to cleavage of a large variety
of extracellular substrates facilitating cytoskeletal remodelling nec-
essary for cancer metastasis and progression.
Traditional methods for MMP protein overexpression and
purification include expressing recombinant proteins in E. coli or
purification from human plasma [10] or from conditioned media
of MMP-expressing mammalian cells [11]. However, MMPs puri-
fied using human plasma or cell culture media are often found to
be contaminated with other associated cellular proteins including
TIMPs, fibronectins etc. [10]. These co-purified contaminants
interfere with the functional readouts used for MMP characteriza-
tion, thereby limiting the use of eukaryotic cells as a source for
MMPs. In this context, bacterial cells, specifically E. coli can serve
as a viable alternate host for expression and purification of MMP
proteins. However, due to the absence of inhibitory TIMPs, the
expression of MMP proteins in bacterial cells causes proteolysis
induced toxicity, thereby resulting in poor yield of the protein.
These limitations can be overcome if the toxic protein is overex-
pressed as inactive inclusion bodies (IBs). However, the expression
of proteins in IBs requires significant downstream processing,
where they have to be solubilized and refolded to restore their
bioactivity [12]. Refolding from IBs is generally tedious and
depends on temperature, pH, salt concentration, as well as on the
type and the concentration of the denaturants utilized. Conventional
methods of protein refolding based on dialysis of denatured pro-
teins against a large volume of refolding buffer or solvent-exchange
chromatography, in general results in poor yield of active protein
[13]. Keeping all this in mind, we developed a matrix assisted
refolding protocol for the purification and refolding of MMP2
[14]. This is an efficient, single step protocol, which results in sig-
nificantly higher yield of active protein from inclusion bodies
compared to any other protocol reported to date. The steps, salient
features, and advantages of the protocol are described below.
MMP2 Purification from E. coli 5
2 Materials
All the reagents utilized for protein work are from Sigma-Aldrich
(MO, USA), of molecular biology grade, free of DNase, RNase,
and proteases, unless otherwise mentioned.
Compositions of various media and buffers used in the proto-
cols described herein are listed below:
2.1 Plasmids and 1. MMP2 expression plasmid: cDNA for MMP2 was cloned into
E. coli Strains the pPROEx-HTc (Invitrogen, Carlsbad, CA) expression vec-
tor [14].
2. CFP-2RS expression plasmid: CFP-2RS CDS was cloned into
the pPROEx-HTa expression vector [14].
3. E. coli C43: F − ompT hsdSB (rB − mB − ) gal dcm (DE3) pLysS
(CmR ) [15].
4. E. coli BL21 ArcticExpressTM: E. coli B F− ompT hsdS(rB − mB
−
) dcm+ Tetr gal λ(DE3) endA Hte [cpn10 cpn60 Gentr] [16].
2.2 Protein 1. Terrific broth media (pH ~ 7.4): Dissolve 12 g tryptone, 24 g

Overexpression yeast extract, 9.4 g potassium phosphate dibasic, 2.2 g potas-
sium phosphate monobasic, and 4.0 ml glycerol in 1 l of deion-
ized water. Autoclave to sterilize before use.
2. Antibiotics: Prepare a stock solution of 100 mg/ml of ampicil-
lin or carbenicillin in water and filter-sterilize using a 0.22 μm
filter before use. Stock solutions are kept frozen at −20 °C.
3. 1 M IPTG (isopropyl β-D-1-thiogalactopyranoside): 238 mg/
ml stock solution in water. Dissolve 2.383 g of IPTG in a mini-
mum volume of water (~5 ml) until it completely dissolves and
then make up the volume to 10 ml. Filter-sterilize using a 0.22
μm filter before use. Sterile stock solutions are kept frozen at
−20 °C.
2.3 MMP2 1. 1 M sodium phosphate buffer, pH 7.4: Mix 77.4 ml of 1 M
Purification Na2HPO4 (monobasic) and 22.6 ml of 1 M NaH2PO4 (diba-
and Refolding sic) to obtain 100 ml of 1 M sodium phosphate buffer, pH 7.4.
2. 5 M NaCl: Dissolve 29.25 g of NaCl in 100 ml of water.
3. 2 M imidazole: Dissolve 2.72 g imidazole in 20 ml of water.
Store at 4 °C till use or prepare fresh.
4. 100 mM phenyl methyl sulfonylfluoride (PMSF): Dissolve
174 mg of PMSF in 10 ml of absolute ethanol and store at −20
°C in aliquots until use.
5. Native lysis buffer: 50 mM sodium phosphate buffer pH 7.4,
300 mM NaCl, 10 mM imidazole, 1.0 mM PMSF. (Tris–HCl
pH 8.0 can also be used instead of 50 mM sodium phosphate
buffer pH 7.4).
6. Inclusion bodies solubilization buffer: 20 mM Tris–HCl pH

8.0, 500 mM NaCl, 10% glycerol, and 8 M urea.
7. 100 mM benzamidine: Dissolve 120 mg of benzamidine hydro-
chloride in 10 ml of water. Store at 4 °C as aliquots till use.
8. 1 M dithiothreitol (DTT): Dissolve 154 mg of dithiothreitol
in water to a final volume of 1 ml. Store as aliquots at −20 °C
till use.
9. 0.5 M glutathione, oxidized (GSSG): Dissolve 3.06 g of GSSG
in water to a final volume of 10 ml. Store at −20 °C till use.
10. 0.5 M glutathione, reduced (GSH): Dissolve 1.53 g of GSH in
water to a final volume of 10 ml. Store at −20 °C till use.
11. Refolding buffer: 20 mM Tris–HCl pH 8.0, 500 mM NaCl,
10% glycerol, 20 mM imidazole, 0.5 mM oxidized glutathione
(GSSG), and 5 mM reduced glutathione (GSH).
12. Elution buffer: 20 mM Tris–HCl, pH 8.0, 500 mM NaCl,
10% glycerol and 250 mM imidazole.
13. Dialysis/storage buffer: 50 mM Tris–HCl, pH 8.0, 50% glyc-
erol, 50 mM NaCl and 1 mM DTT.
2.4 MMP2 Activity 1. 5× sample buffer: 250 mM Tris–HCl, pH 6.8, 5% SDS, 50%
Analysis glycerol, 0.5% bromophenol blue, and 6.25% β-mercaptoethanol.
2. 2× sample buffer (for zymography): 100 mM Tris–HCl, pH
6.8, 2% SDS, 20% glycerol, and 0.2% bromophenol blue.
3. 10× activation buffer: 500 mM Tris–HCl, pH 7.4, 50 mM
CaCl2, and 10 μM ZnCl2.
4. ARP100 (Cayman Chemical Co., USA): 5 mM ARP100 dis-
solved in DMSO. Aliquots stored at −20 °C till further use.
5. Staining solution: 0.5% CBB R250, 40% methanol, 10% acetic
acid, and 50% water.
6. Destaining solution: 40% methanol, 10% acetic acid, and 50%
water.
3 Methods
3.1 Cloning MMP2 1. PCR amplify cDNA template coding for MMP2, from total
Gene for Prokaryotic RNA extracted from HEK293 cells, using gene specific prim-
Expression ers (see Note 1).
2. The amplified MMP2 ORF is cloned into the EcoRI and XhoI
sites of the pPROEx-HT expression vector, which contains an
N-terminal 6× His tag and a trc promoter upstream of the
MMP2 ORF allowing for IPTG-dependent overexpression of
the fusion protein.
3. Clones were verified by DNA sequencing to ensure the in-
frame fusion of the purification tag with the MMP2 CDS.
3.2 Transformation 1. Chemically competent E. coli C43 (DE3) pLysS cells were
and Propagation transformed with the recombinant MMP2 expression plasmid
of the MMP DNA (see Notes 2 and 3).
Expression Strain 2. The transformed cells were plated onto Luria Agar (LA) plates
containing 100 μg/ml carbenicillin and incubate overnight at
37 °C. The colonies obtained were immediately used for pro-
tein overexpression (see Note 4).
3.3 Overexpression Given that inclusion body formation is favoured when the expres-
of Recombinant MMP2 sion of recombinant protein is high, conditions for MMP2 overex-
Protein pression had to be optimized to maximize expression and inclusion
in Inclusion Bodies body formation (see Note 5).
1. Inoculate a single colony of E. coli C43 cells into 10 ml of LB,
containing 100 μg/ml carbenicillin (see Note 4) and incubate
overnight at 37 °C.
2. Inoculate 1% of the overnight culture into 1 liter of terrific
broth (see Note 6) containing 35 μg/ml chloramphenicol and
100 μg/ml ampicillin and grow at 37 °C to an OD600 of
0.8–1.0.
3. Add IPTG to the culture to a final concentration of 1 mM and
incubate at 37 °C for a further 24 h at 180 rpm.
4. Pellet the induced culture by centrifuging at 6,000 × g for
10 min at room temperature.
5. Resuspend the pellet in 20 ml of native lysis buffer and sonicate
on ice (Branson Sonifier model S-450D with 1/8″ tapered
microtip) at 50% amplitude, six cycles of sonication with 10 s
of ON and OFF pulse (see Note 7).
6. Centrifuge the sonicated lysate at 20,000 × g for 30 min at 4
°C to separate the soluble and insoluble fractions.
7. The soluble supernatant and insoluble pellet containing the
inclusion bodies are analyzed by SDS-PAGE or western blotting
using anti-His antibody (Fig. 1). If the yield of MMP2 in inclu-
sion bodies is poor then it may be necessary to repeat steps 1–7
using a different concentration of IPTG (see Note 8).
3.4 Matrix Assisted 1. Resuspend the pellet containing the recombinant MMP2
Purification inclusion bodies in solubilization buffer (5 ml buffer for inclu-
and Refolding sion bodies obtained from 1 l of culture) and incubate at 37 °C
of Denatured MMP2 for 2 h at 180 rpm (see Note 9).
2. Centrifuged at 20,000 × g for 30 min at 15 °C to remove
insoluble debris.
3. Load the supernatant containing the denatured MMP2 pro-
tein onto a column containing Ni2+-NTA resin pre-equilibrated
with solubilization buffer (see Note 10). In general, 1 ml of
packed bead volume was used per litre of bacterial culture in a
column of 12–20 ml of total capacity.
Fig. 1 SDS-PAGE analysis for MMP2 protein expression. Bacterial cell lysates
were analyzed for MMP2 expression in soluble (supernatant) and insoluble (pel-
let) fraction. Lane M: Marker; lane 1, purified MMP2 protein; lane 2, soluble frac-
tion; lane 3, insoluble fraction containing inclusion bodies
4. After loading, incubate the column for 2 h at room tempera-

ture (see Note 11).
5. After the incubation, the flow through containing the unbound
proteins was collected and kept at room temperature until fur-
ther analysis.
6. Wash the column with 50 bed volumes (50 ml) of solubiliza-
tion buffer containing 20 mM imidazole under gravity flow
(see Note 12).
7. To refold the protein the column was washed with 10 ml of a
solution containing decreasing concentrations (from 8 to 0
M) of urea by mixing appropriate proportions of solubiliza-
tion buffer and refolding buffer (see Note 13). The washing
steps employed are described in Table 1.
8. Wash the column five times with 10 bed volumes of the refold-
ing buffer to ensure proper refolding of MMP protein (see
Notes 14 and 15).
9. Elute the refolded MMP2 protein in 5 bed volumes of elution
buffer, in steps of 1 ml each on ice (see Note 16).
Table 1
Washing steps used for on-column refolding of the denatured MMP2 protein
Washing Final urea concentration Volume of solubilization buffer Volume of refolding buffer
step (M) (ml) (ml)
1 6 7.50 2.5
2 5 6.25 3.75
3 4 5.0 5.0
4 3 3.75 6.25
5 2 2.50 7.50
6 1 1.25 8.75
10. Fractions containing purified and refolded MMP2 protein are

analyzed by microplate based Bradford’s assay.
11. Pool peak elution fractions containing the refolded MMP2
protein (see Note 17) and dialyze against dialysis buffer at
4 °C for 4 h using a 12 kDa cutoff dialysis tube. Aliquot the
dialyzed protein and store at −20 °C.
12. Confirm protein purity by SDS-PAGE. Mix 12 μl of purified
protein with 3 μl of 5× loading dye. Heat the mixture at 95 °C
for 10 min and resolve on a 12% SDS-PAGE gel (Fig. 2).
13. Determine total protein yield for the purified protein using
the Bradford protein assay (typically approx. 2–4 mg/l of
culture).
3.5 Activity Analysis The MMP2 protein activity was determined using a conventional
of Purified and Refolded “gelatin zymography” methodology as well as using an advanced
MMP2 Protein “form invariant substrate cleavage assay” as described previously [14].
3.5.1 Gelatin The biological activity of the purified MMP2 protein was assessed
Zymography using gelatin based zymography as reported in various studies [17,
18], which is described below:
1. Add 2 μg of purified and refolded MMP2 protein to an equal
volume of 2× sample buffer (see Note 18).
2. Resolve the sample on a 12% SDS-PAGE gel co-polymerized
with 0.1% gelatin at 100 V for 2 h.
3. Carefully remove the gel and wash with activation buffer con-
taining 2.5% Triton X-100 for 1 h at room temperature (see
Note 19).
4. Briefly rinse the gel with deionized water and incubate in acti-
vation buffer (without Triton X-100) overnight at 37 °C.
Fig. 2 SDS-PAGE analysis of refolded MMP2 protein. MMP2 protein purified and
refolded from IBs using affinity chromatography were analyzed on 12% SDS-
PAGE and stained with Coomassie brilliant blue
5. Stain the gel by submerging it in staining solution for 30 min

at room temperature followed by destaining for 1–2 h at room
temperature using the destaining solution.
6. Wash the gel with deionized water and scan using a image doc-
umentation system. The presence of light or unstained bands
as a result of proteolytic cleavage of gelatin substrate on a dark
background reveals the activity of the purified protein (see
Note 20) (Fig. 3).
3.5.2 Form Invariant The total activity of purified MMP2 protein can be further tested
Substrate Cleavage Assay using an advanced substrate cleavage assay (see Note 21) [14].
1. In brief, a synthetic substrate site for MMP2 proteins com-
posed of CFP fluorescent protein fused with MMP2 substrate
recognition sequence IPVS↓LRSG (CFP-2RS), was overex-
pressed, purified and used for activity assessment as previously
described [14].
2. Incubate 2 μg of purified CFP-2RS protein with 500 ng of
purified MMP2 in the presence of 1× activation buffer with or
without the MMP2 specific inhibitor, ARP-100 [19] for 16 h
at 37 °C.
Fig. 3 Gelatin zymography for analysis of refolded MMP2 protein activity.

Zymography was used to analyze the bioactivity of purified full-length MMP2
protein. The two bands corresponding to pro-MMP2 and MMP2 are as marked by
arrows
3. Add 5× loading dye and heat denature at 95 °C for 10 min.

4. Resolve the samples on a 15% SDS-PAGE gel followed by
staining for 30 min and destaining.
5. Rinse the destained gel with deionized water and image the gel
using a gel image acquisition system (Fig. 4). In this assay,
action of MMP2 protein on the CFP-2RS substrate leads to
reduction in its size by approx. 3 kDa which can be recorded
after resolving the proteins on SDS-PAGE. The refolded pro-
tein should show the presence of clear digested protein bands,
which are abrogated when ARP100 is present in the reaction.
4 Notes
1. For amplification of MMP2 CDS, cDNA prepared using RNA

isolated from HEK 293 cells was used. The RNA was reverse
transcribed to cDNA in a single step RT-PCR protocol using
oligo-dT primers and Superscript cDNA synthesis kit (Life
Technologies, USA). The use of oligo-dT primer was preferred
Fig. 4 Form invariant substrate cleavage assay. Activity of purified and refolded
MMP2 was monitored using recombinant CFP-2RS protein containing a synthetic
MMP2 substrate recognition sequence. SDS-PAGE analysis of CFP-2RS sub-
strate protein incubated with purified and refolded MMP2 in the absence and
presence of the MMP2 inhibitor, ARP100. Lane 1, CFP-2RS only; lane 2, CFP-2RS
+ MMP2; lane 3: CFP-2RS + MMP2 + 1 μM ARP 100; and lane 4, CFP-2RS +
MMP2 + 10 μM ARP 100
over the random hexamer to ensure the amplification of full-

length MMP2 CDS. One hundred nanogram of cDNA was
used for PCR amplification of MMP2 CDS using gene specific
primers and Pfu polymerase enzyme. The PCR conditions
used had extension step done at 72 °C for 4 min, to ensure
amplification larger amplicons corresponding to full-
length
MMP2. Details have been described previously [14].
2. The use of the E. coli C43 (DE3) pLysS strain is critical for
efficient expression of toxic proteins like MMPs. It is a deriva-
tive of E. coli B strain containing the DE3 lysogen, a λ pro-
phage carrying the T7 RNA polymerase gene and lacIq [15].
The pLysS plasmid encodes for T7 phage lysozyme, an inhibi-
tor of T7 RNA polymerase (T7RNAP) that reduces the activ-
ity of T7 RNAP and confers resistance to cell death associated
with toxic protein overexpression. The strain also contains
additional mutations, which confer greater tolerance to the
presence of toxic proteins. Alternatively, C43 (DE3) pLysS
can be replaced by the BL21 ArcticExpress E. coli strain (see
Note 3).
3. E. coli BL21 ArcticExpress (Agilent Inc., USA) is the preferred
strain for expression of proteins in soluble form. The strain is
genetically engineered to co-express the Cpn10 and Cpn60
chaperonins from O. antarctica which assists improved pro-

tein refolding at low temperature [16] .
4. Usage of carbenicillin was preferred over ampicillin because of
improved stability of carbenicillin against the action of β-lactamase.
This ensures that only cells containing the recombinant plasmid
grow during overnight growth, thereby providing good primary
cultures for protein overexpression in the next step.
5. The colonies obtained were immediately used for protein
overexpression. The transformed plates stored beyond 4 days
were not utilized for protein expression as the yield of MMP2
protein was consistently lower or absent from them.
6. Terrific broth (TB), nutrient rich media with glycerol as an
extra carbon source is preferred as it confers robust growth of
bacteria. Phosphate salts present in TB prevent changes in pH
during bacterial growth which is critical for obtaining the high
cell densities required for maximum expression and yield of
the MMP2 protein.
7. The sonication protocol should be optimized for time and the
power of the sonicator used for cell lysis. The process should
be carefully optimized to maximize removal of contaminating
soluble proteins, which may interfere with downstream
processing. Additionally, sonication should be done on ice to
avoid overheating, that can cause protein denaturation and
thereby contamination of the inclusion bodies.
8. If MMP2 overexpression localizes the protein in the soluble
fraction, its toxicity leads to extremely poor yield, in such case
fresh induction is performed with higher concentration (2 mM)
of IPTG to ensure higher expression of the recombinant
protein, which facilitates its localization in inclusion bodies.
9. The following points can be considered if issues are faced with
solubilization and recovery of MMP2 from the inclusion
bodies:
(a) Vortexing the pellet vigorously, in general, ensures better
solubilization of inclusion bodies.
(b) 6 M Guanidine hydrochloride (Gm.HCl) can be used as a
denaturant in place of 8 M urea, however samples contain-
ing Gm.HCl cannot be analyzed on SDS-PAGE unlike
urea containing fractions.
(c) Do not centrifuge the solubilized protein at temperature
less than 15 °C, as it can lead to the precipitation of the
denaturant (urea/Gm.HCl), thereby reducing the recov-
ery of soluble protein.
10. If Gm.HCl is used for solubilization of inclusion bodies instead
of urea in the solubilization buffer, then the same buffer
should be used for Ni2+-NTA resin equilibration.
11. To facilitate efficient binding of the protein on the Ni+2-NTA

resin, resuspend the loaded protein and resin every 15 min by
gentle mixing.
12. In our hands, the flow rate does not have significant bearing
on the outcome of the folding, but faster flow rates always
provide better yields.
13. In the refolding steps, the urea concentration is reduced con-
comitantly with an increase in the concentration of GSH/
GSSH (contributed by the refolding buffer). This facilitates
appropriate disulfide bond formation, which is critical for
appropriate folding of the protein [20].
14. The final refolding step is performed at 4 °C to prevent pro-
tein degradation.
15. GSH and GSSG are added fresh in powder form to the refold-
ing buffer. The gradual increase in the concentration of GSH
during the refolding steps can be effectively visualized by the
reduction of Ni+2 on the column, because the colour of the
resin changes from blue to pink/white.
16. At this step since bound and refolded protein are eluted from
the Ni+2-NTA resin, the color of the reduced Ni+2 is restored
and the resin becomes blue again. The restoration of the color
is a good indicator of elution of the protein.
17. For rapid analysis of fractions containing high amounts of
eluted MMP2 protein, we utilized a rapid Bradford’s assay.
Here, we use a 96-well microplate where 100 μl of Broadford’s
reagent is added to ten wells and kept ready. In one well
marked as a control, 5 μl of elution buffer is added to deter-
mine the baseline colour. In the remaining wells, as the frac-
tions are collected, 5 μl each of the eluate is added to rapidly
identify fractions containing protein. Peak fractions (generally
fractions 2–6) are then pooled for subsequent dialysis.
18. For SDS-PAGE based activity assays such as zymography, pro-
tein samples are processed in non-reducing conditions, without
heat denaturation. For this, 2× sample buffer devoid of reducing
agent is used. Also heating of samples before loading on SDS gel
is avoided to prevent loss of MMP protein activity.
19. The complete exchange of SDS in the gel by Triton X-100 in
the activation buffer is critical to restore the activity of MMP
proteins. To ensure this the gel is pre-washed with dH2O
before treatment with activation buffer.
20. During the activation step, not only is the activity of denatured
MMP2 protein restored but the inactive pro-MMP2 protein
also undergoes activation. This leads to detection of both the
pro- as well as the active form of MMP2 in the zymogram,
which is evident by the presence of two bands, one of high

molecular weight (pro form, corresponding to the purified
protein) and a smaller molecular weight band (corresponding
to the processed form of MMP2, which is highly active).
21. The substrate cleavage assay utilizes a chimeric reporter protein
(CFP-2RS) whose size reduces by approx. 3 kDa after cleavage
by MMP2. The assay records activity from both proMMP2 as
well as from processed and active MMP2 by a simple SDS-
PAGE analysis, which cannot be recorded from conventional
zymography. The design of the assay and construction of the
chimeric reporter has been described previously [14]. The
reporter protein consists of a fusion of ORF coding for CFP
fluorescent protein, PCR amplified from pECFP vector
(Clontech, USA) with a synthetic nucleotide sequence
(5’GGATCCGGCGGAAGCATCCCCGTCAG
CCTCCGTAGCGGCGGAAGCGTCGAC 3′) coding for a
16-amino acid long peptide, SGSGGSIPVSLRSGGS contain-
ing the MMP2 recognition sequence IPVS↓LRSG [21]. The
CFP-2RS (CFP+MMP2 recognition sequence encoding pep-
tide) fusion was cloned in EcoRI and SalI sites of pProEx-HTa
bacterial expression vector. The chimeric protein was overex-
pressed in E. coli Origami™ strain and was purified from the
soluble protein fraction using Ni+2-NTA chromatography as a
33 kDa protein.
Acknowledgments
Financial assistance to DKS from Department of Science and

Technology (EMR/2014/000997); Department of Biotechnology
and Indian Institute of Science partnership program, and DST-
FIST for equipment support and a Research Fellowship to RJ from
University Grants Commission is acknowledged.
References
1. Massova I, Kotra LP, Fridman R, Mobashery S cancer progression and their pharmacological
(1998) Matrix metalloproteinases: structures, targeting. FEBS J 278(1):16–27
evolution, and diversification. FASEB 4. Schmalfeldt B, Prechtel D, Härting K, Späthe
J 12(12):1075–1095 K, Rutke S, Konik E, Fridman R, Berger U,
2. Baker AH, Edwards DR, Murphy G (2002) Schmitt M, Kuhn W, Lengyel E (2001)
Metalloproteinase inhibitors: biological actions Increased expression of matrix metalloprotein-
and therapeutic opportunities. J Cell Sci ases (MMP)-2, MMP-9, and the urokinase-
115(Pt 19):3719–3727 type plasminogen activator is associated with
3. Gialeli C, Theocharis AD, Karamanos NK progression from benign to advanced ovarian
(2011) Roles of matrix metalloproteinases in cancer. Clin Cancer Res 7(8):2396–2404
5. Vihinen P, Kähäri V-M (2002) Matrix metal- and activity assessment using a ‘form invariant’
loproteinases in cancer: prognostic markers and assay for matrix metalloproteinase 2 (MMP2).
therapeutic targets. Int J Cancer Mol Biotechnol 56(12):1121–1132
99(2):157–166 15. Miroux B, Walker JE (1996) Over-production
6. Roy R, Yang J, Moses MA (2009) Matrix of proteins in Escherichia coli: mutant hosts
metalloproteinases as novel biomarkers and that allow synthesis of some membrane pro-
potential therapeutic targets in human cancer. teins and globular proteins at high levels. J Mol
J Clin Oncol 27(31):5287–5297 Biol 260(3):289–298
7. Hadler-Olsen E, Winberg JO, Uhlin-Hansen L 16. Ferrer M, Chernikova TN, Yakimov MM,
(2013) Matrix metalloproteinases in cancer: Golyshin PN, Timmis KN (2003) Chaperonins
their value as diagnostic and prognostic mark- govern growth of Escherichia coli at low tem-
ers and therapeutic targets. Tumour Biol peratures. Nat Biotechnol 21(11):1266–1267
34(4):2041 17. Hawkes SP, Li H, Taniguchi GT (2010)
8. Ra HJ, Parks WC (2007) Control of matrix Zymography and reverse zymography for
metalloproteinase catalytic activity. Matrix Biol detecting MMPs and TIMPs. Methods Mol
26(8):587–596 Biol 622:257–269
9. Nagase H, Visse R, Murphy G (2006) Structure 18. Hu X, Beeton C (2010) Detection of func-
and function of matrix metalloproteinases and tional matrix metalloproteinases by zymogra-
TIMPs. Cardiovasc Res 69(3):562–573 phy. J Vis Exp 45:2445
10. Steffensen B, Chen Z, Pal S, Mikhailova M, Su 19. Rossello A, Nuti E, Orlandini E, Carelli P,
J, Wang Y, Xu X (2011) Fragmentation of Rapposelli S, Macchia M, Minutolo F, Carbonaro
fibronectin by inherent autolytic and matrix L, Albini A, Benelli R, Cercignani G, Murphy G,
metalloproteinase activities. Matrix Biol Balsamo A (2004) New N-arylsulfonyl-N-
30(1):34–42 alkoxyaminoacetohydroxamic acids as selective
11. Wang X, Yi J, Lei J, Pei D (1999) Expression, inhibitors of gelatinase A (MMP-2). Bioorg Med
purification and charaterization of recombi- Chem 12(9):2441–2450
nant mouse MT5-MMP protein products. 20. Saini DK, Pant N, Das TK, Tyagi JS (2002)
FEBS Lett 462(3):261–266 Cloning, overexpression, purification, and
12. Carrió MM, Villaverde A (2002) Construction matrix-assisted refolding of DevS (Rv 3132c)
and deconstruction of bacterial inclusion bod- histidine protein kinase of Mycobacterium
ies. J Biotechnol 96(1):3–12 tuberculosis. Protein Expr Purif
13. Tsumoto K, Ejima D, Kumagai I, Arakawa T 25(1):203–208
(2003) Practical considerations in refolding 21. Turk BE, Huang LL, Piro ET, Cantley LC
proteins from inclusion bodies. Protein Expr (2001) Determination of protease cleavage
Purif 28(1):1–8 site motifs using mixture-based oriented
14. Singh KK, Jain R, Ramanan H, Saini DK peptide libraries. Nat Biotechnol
(2014) Matrix-assisted refolding, purification 19(7):661–667
Chapter 2
Expression and Purification of a Matrix Metalloprotease

Transmembrane Domain in Escherichia coli
Charles A. Galea
Abstract
Membrane tethered matrix metalloproteases are bound to the plasma membrane by a
glycosylphosphatidylinositol-anchor or a transmembrane domain. To date, most studies of membrane-bound
matrix metalloprotease have focused on the globular catalytic and protein–protein interaction domains of
these enzymes. However, the transmembrane domains have been poorly studied even though they are known
to mediate intracellular signaling via interaction with various cellular proteins. The expression and purifica-
tion of the transmembrane domain of these proteins can be challenging due to their hydrophobic nature. In
this chapter we describe the purification of a transmembrane domain for a membrane-bound matrix metal-
loprotease expressed in E. coli and its initial characterization by NMR spectroscopy.
Key words Membrane-anchored MMP, Matrix metalloproteases, NMR spectroscopy, Transmembrane

domain, Protein expression, Isotopically labeled
1 Introduction
Matrix metalloproteases comprise a large family of zinc-dependent

endopeptidases that function at neutral pH to degrade extracellu-
lar matrix proteins, cleave cell surface receptors, release apoptotic
ligands, and activate chemokines and cytokines [1–3]. Matrix
metalloproteases (MMPs) participate in a wide range of cellular
processes including tissue remodeling, cell proliferation, cell migra-
tion, differentiation, angiogenesis, apoptosis, and the immune
response [4–9]. Aside from these normal physiological roles,
MMPs have been implicated in a large number of pathological dis-
orders such as arthritis, Alzheimer’s disease, atherosclerosis, vascu-
lar disease, central nervous system disease, liver cirrhosis, and
various cancers [10, 11].
MMP activity is regulated at multiple levels, including biosyn-
thesis (transcription/translation) [12], zymogen activation [10,
12], compartmentalization [10, 13, 14], and inactivation [14–16].
Several MMPs are active during development and normal
17
18 Charles A. Galea
physiology and may play a role in homeostasis [7, 9, 17–20], while

others play a role in tissue injury and infection [4, 7]. In chroni-
cally inflamed tissues and most cancers MMPs contribute to vari-
ous pathological processes, including tissue degradation, tumor
progression and invasion [21–27].
MMPs are expressed and transported through the cell as inac-
tive zymogens and secreted or anchored to the plasma membrane,
confining their activity to the extracellular environment or the cell
surface [8]. Evidence suggests that secreted MMPs bind to specific
cell-surface receptors, membrane-anchored proteins and cell-asso-
ciated extracellular matrix (ECM) molecules, and function pericel-
lularly at focused locations [14]. MMPs also function in the cell
nucleus (MMP2, 3, 9, 13 and MT1-MMP), cytoplasm (MMP1, 2,
26, 23), and various organelles, where their localization is facili-
tated by interactions with other proteins, proteoglycan core pro-
teins and/or their glycosaminoglycan chains, and other molecules
[1, 10, 13, 28–31].
Unlike the majority of MMPs that function as soluble prote-
ases, several are anchored to the cell membrane [32–34]. The
membrane tethered MMPs include the membrane-type MMPs
(MT-MMPs) and MMP23, a membrane anchored MMP contain-
ing a domain architecture that differs significantly from that of the
MT-MMPs [35]. There are six MT-MMPs denoted MT1-MMP
(MMP14), MT2-MMP (MMP15), MT3-MMP (MMP16), MT4-
MMP (MMP17), MT5-MMP (MMP24), and MT6-MMP
(MMP25). MT4-MMP and MT6-MMP are tethered to the plasma
membrane via a glycosylphosphatidylinositol (GPI)-anchor while
the remaining MT-MMPs are anchored via a transmembrane (TM)
domain [32]. The MT-MMPs contain a common domain architec-
ture consisting of an N-terminal signal peptide followed by a pro-
domain, a catalytic domain, a hinge (linker-1) region, a
hemopexin-like (Hpx) domain, and a stalk region (linker-2).
Linker-2 is followed by a TM domain or GPI-anchor and a
C-terminal cytoplasmic tail (Fig. 1) [32]. In contrast, the domain
architecture of MMP23 comprises an N-terminal signal sequence
followed by a cytoplasmic tail, a membrane-anchoring TM domain,
a linker region, a pro-domain, a catalytic domain, a small cysteine-
rich toxin-like domain (TxD), and a C-terminal immunoglobulin-
like cell adhesion molecule (IgCAM) domain (Fig. 1) [35].
To date, studies of membrane-anchored MMPS have mostly
focused on the characterization of the catalytic and globular pro-
tein–protein interaction domains. However, evidence indicates
that the single helical TM domain and unstructured cytoplasmic
domain of these enzymes can play a key role in the function of
membrane-anchored MMPs. For example, the formation of MT1-
MMP homo-dimers mediated by both the Hpx [36] and TM
domains [37] is required for proMMP2 activation on the cell sur-
face [36] and degradation of collagen [38].
Purification of MMP Transmembrane Domain 19
Fig. 1 Structural domains of the membrane-anchored MMPs. MT1-MMP, MT2-MMP, MT3-MM, MT4-MMP,
MT5-MMP, and MT6-MMP share a common domain structure consisting of an N-terminal signal peptide (SP)
that targets the protease to the cell surface, a pro-domain (Pro), a catalytic domain, a linker domain (L1), a
hemopexin (Hpx) domain involved in protein–protein interactions, and a second linker region (L2). MT1-MMP,
MT2-MMP, MT3-MMP, and MT5-MMP contain a C-terminal single helical transmembrane (TM) domain fol-
lowed by a cytoplasmic (CP) domain while MT4-MMP and MT6-MMP possess a C-terminal GPI anchor. MMP23
differs in its domain structure, possessing an N-terminal signal peptide followed by a single type-II helical
transmembrane domain, a pro-domain, a catalytic domain, a cysteine-rich toxin-like (TxD) domain, and a
protein–protein interaction IgCAM domain. The pro-domain (Pro) in most membrane-anchored MMPs (except
MMP23) contains an unpaired cysteine sulfhydryl group that interacts with the active site zinc atom and main-
tains the enzyme in a latent inactive form (the cysteine switch). The MT-MMPs are transported in their inactive
latent form to the cell surface, where they are cleaved at the recognition motif (RXKR) within the pro-domain
by a furin-like protease, resulting in activation of the enzyme
Previous studies have highlighted the importance of the cyto-

plasmic domain of membrane-anchored MMPs in mediating vari-
ous intracellular interactions. For example, the 20 amino acid
residue cytoplasmic tail of MT1-MMP has been reported to inter-
act with a number of proteins and is also the target of posttransla-
tional modifications. Interaction of the MT1-MMP cytoplasmic
domain (CPD) with FIH-1 (Factor inhibiting HIF-1) triggers a
cascade of protein–protein interactions resulting in activation of
HIF-1α thereby regulating ATP production in macrophages dur-
ing normoxia [39]. While interaction of phosphorylated Tyr573 in
the MT1-MMP CPD with p130Cas regulates Rac1 signaling in
osteoclast formation [40, 41]. The MT1-MMP CPD has also been
20 Charles A. Galea
shown to interact with p27RF-Rho (LAMTOR1) which regulates

RhoA activation [42]. Other studies have shown that the MT1-
MMP CPD is required for its localization to focal adhesions via
interactions with the focal adhesion kinase (FAK)-p130Cas [43]
and an eight amino acid residue loop structure, MT-Loop, located
in the CPD is required to degrade underlying matrices in an effi-
cient manner [44]. MT1-MMP has a half-life of less than 30 min
on the cell surface and is endocytosed in a clathrin- and caveolae-
dependent manner. The clathrin-dependent endocytosis of MT1-
MMP is attributed to the interaction of residues LLY573 with
adapter protein 2, a component of clathrin-coated pits [45].
Endocytosed MT1-MMP is also recycled to the cell surface and a
sequence (DKV582) within the CPD is required for recycling [46].
Other studies have shown that the last three residues (EWV645) in
the CPD of MT5-MMP play a similar role in its recycling to the
cell surface [47]. These residues act as a PDZ binding motif and
bind to Mint-3 which contains two PDZ domains. While our
recent studies have shown that the pro-domain of MMP23, con-
taining a TM and cytoplasmic domain, interacts with the voltage-
gated potassium ion channel Kv1.3 and regulates the trafficking of
this ion channel to the cell surface [35, 48].
Considering the important role played by the TM and cyto-
plasmic domains of membrane-anchored MMPs, very few studies
have been undertaken to define their structure and dynamics. In
this chapter we focus on the expression and purification of a trun-
cated protein containing the transmembrane and cytoplasmic
domains of a membrane-anchored MMP (MMP23) for the charac-
terization of these domains by NMR spectroscopy.
2 Materials
2.1 Expression 1. E. coli BL21(DE3) or E. coli BL21(DE3)pLysS chemically

of Isotopically Labeled competent cells (Novagen).
Peptides in E. coli 2. pET32a expression vector (Novagen) containing cDNA for
in M9 Minimal Media your peptide of interest covering the TM and cytoplasmic
domains (TM/CPD) of your membrane-anchored MMP and
an N-terminal 3C protease recognition site codon optimized
for expression in E. coli (see Note 1).
3. Luria-Bertani (LB) broth: 10 g tryptone, 10 g NaCl, 5 g yeast
extract per liter of deionized water. Adjust the pH to 7.0 with
0.5 N NaOH and autoclave. Add ampicillin to a final concen-
tration of 100 μg/ml immediately prior to use.
4. Isopropyl β-d-1-thiogalactopyranoside (IPTG).
5. 50% (w/v) glucose: 125 g of d-glucose per 250 ml of deion-
ized water. Sterilize by passing through a 0.2 μm filter and
store at room temperature.
6. 13C-labeled d-glucose (99% 13C): This can be obtained from a

number of suppliers including Cambridge Isotope Labs and
Sigma-Aldrich.
7. 1 M MgSO4: 24.7 g of MgSO4∙7H2O per 100 ml of deionized
water. Sterilize by autoclaving and store at room
temperature.
8. 15N–labeled ammonium chloride (NH4Cl, 99% 15N): This can
be obtained from a number of suppliers including Cambridge
Isotope Labs and Sigma-Aldrich.
9. Vitamin solution (5 mg/ml): 0.5 g thiamine hydrochloride
and 0.5 g nicotinic acid per 100 ml of deionized water. Filter-
sterilize and store at 4 °C.
10. Trace element solution (1,000×): 0.60 g of CaCl2∙2H2O, 0.60
g of FeSO4∙7H2O, 0.12 g of MnCl2∙4H2O, 0.08 g of
CoCl2∙6H2O, 0.07 g of ZnSO4∙7H2O, 0.03 g of CuCl2∙2H2O,
2 mg of boric acid, 25 mg of (NH4)6Mo7O24∙4H2O, and 0.50
g of EDTA per 100 ml of deionized water [49] (see Note 2).
11. M9 salts (5×): 2.5 g NaCl, 15 g anhydrous KH2PO4, 34 g
anhydrous Na2HPO4 per liter of deionized water. Autoclave to
sterilize (see Note 3).
12. Antibiotic for plasmid selection (1,000×): 100 mg/ml ampi-
cillin in deionized water (see Note 4).
13. M9 minimal medium: 780 ml of deionized water, 200 ml of
5 × M9 salts, 6 ml of vitamins, 12 ml of d-glucose (50% w/v),
1 ml of antibiotic (1,000×), 1 ml of trace elements (1000×),
3 ml of 1 M MgSO4 and 1 g of 15NH4Cl (see Note 5).
14. Filter-sterilize M9 minimal medium and add ampicillin to a
final concentration of 100 μg/ml immediately prior to use.
2.2 Expression This method was originally described by Studier [50].

of Isotopically Labeled 1. 20 × N: 14.2 g of Na2SO4, 136 g of KH2PO4, and 142 g of
Peptides in E. coli Na2HPO4 per liter of deionized water. Add each in sequence
in Autoinduction and stir until dissolved. Sterilize by autoclaving and store at
Media room temperature.
2. 50 × 5052: Dissolve 250 g of glycerol in 730 ml of deionized
water then add 25 g of d-glucose and 100 g of α-lactose. Add
each component in sequence and stir until dissolved. Lactose
can take several hours to dissolve at room temperature. Sterilize
by autoclaving and store at room temperature (see Note 6).
3. NG medium: 47 ml of sterile water, 300 μl of vitamins, 50 μl
of 1 M MgSO4, 50 μl of trace elements (1,000×), 500 μl of d-
glucose (50% w/v), 2.5 ml of 20 × N, 50 μl of antibiotic, and
125 mg of 15NH4Cl (see Note 7).
22 Charles A. Galea
4. N-5052 medium: 922 ml of sterile water, 6 ml of vitamins,

1 ml of MgSO4, 1 ml of trace elements (1,000×), 20 ml of 50 ×
5052, 50 ml of 20 × N, 1 ml of antibiotic (1,000×), and 2.5 g
of 15NH4Cl.
5. Additional components for preparing the autoinduction
medium include 1 M MgSO4, 15NH4Cl, vitamins, trace ele-
ment solution (1,000×), and antibiotics (1,000×) as described
in Subheading 2.1.
2.3 Immobilized 1. BugBuster protein extraction reagent (Merck Millipore).

Metal Ion Affinity 2. cOmplete, EDTA-free protease inhibitor cocktail tablets
Chromatography (Sigma-Aldrich) (see Note 8).
(IMAC) Purification
3. Human Rhinovirus (HRV) 3C protease (Novagen) (see Note 9).
of Expressed Peptide
4. 3-((3-cholamidopropyl) dimethylammonio)-1-
propanesulfonate (CHAPS) (Critical micelle concentration of
8–10 mM) (see Note 10).
5. Lysis buffer: BugBuster containing 10 mM CHAPS, 1.0 mg/
ml lysozyme, 5 μg/ml DNaseI, and 1 × cOmplete EDTA-free
protease inhibitor cocktail.
6. Ni2+ affinity column: 5 ml HiTRAP chelating HP column charged
with nickel (GE Healthcare Life Sciences) (see Note 11).
7. NTA charging solution: 250 mM NiSO4 in deionized water.
8. NTA buffer A: 20 mM Tris–HCl buffer, pH 8.0 containing
150 mM NaCl, 10 mM CHAPS and 5 mM imidazole.
9. NTA buffer B: 20 mM Tris–HCl buffer, pH 8.0 containing
10. NTA buffer C: 20 mM Tris–HCl buffer, pH 8.0 containing
2.4 HPLC Purification 1. 0.2 μm PTFE syringe filters.

of Peptides 2. C8 reversed-phase semi-preparative HPLC column: 100 Å,
C8, 5 μm (Phenomenex).
3. Trifluoroacetic acid (TFA).
4. Acetonitrile (HPLC grade).
5. Solvent A: 0.1% TFA (v/v). Add 1 ml of TFA to 1 L of deion-
ized water.
6. Solvent B: acetonitrile–water (80:20) + 0.085% TFA. To
700 ml of acetonitrile add 300 ml of deionized water and
0.85 ml TFA.
7. Sample buffer: 20 mM sodium citrate, pH 5.0, containing 20
mM TCEP and 10 mM CHAPS.
8. Liquid nitrogen.
2.5 NMR Sample 1. 5 mm Shigemi tube (Shigemi Inc.).

Preparation 2. Deuterated dodecylphosphocholine (d38-DPC) (Sigma-Aldrich).
3. Tris(2-carboxyethyl)phosphine hydrochloride (TCEP).
4. Sodium azide (NaN3).
5. Dueterium oxide (D2O).
6. NMR buffer: 20 mM sodium citrate pH 5.0, 100 mM deuter-
ated DPC, 20 mM TCEP, 0.02% NaN3 in 90% H2O/10% D2O.
7. 1,4-Dioxane (NMR reference standard) (Sigma-Aldrich) (see
Note 12).
3 Methods
The truncated membrane-anchored MMP containing the TM and

CPD domains of the protein (the TM/CPD protein) can be over-
expressed as a recombinant protein in E. coli BL21(DE3) follow-
ing induction by the addition of IPTG. The production of
uniformly isotopically labeled protein is achieved using minimal
media containing 15N–labeled NH4Cl and 13C–labeled glucose as
the sole nitrogen and carbon sources, respectively.
3.1 Protein
Overexpression 1. Inoculate 2 ml of LB medium containing the appropriate anti-
of Isotopically Labeled biotic with a single colony of freshly transformed E. coli
Protein in M9 BL21(DE3)pLysS cells harboring the pET32a expression vec-
Minimal Media tor containing the TM/CPD cDNA (see Notes 13 and 14).
2. Grow overnight at 37 °C with shaking at approximately
200 rpm.
3. Inoculate 2 ml of M9 minimal medium with 20 μl of overnight
culture. Grow at 37 °C for 6–8 h until the culture is visibly
turbid (see Note 15).
4. Inoculate 48 ml of M9 minimal medium with the entire 2 ml
of culture and grow overnight at 37 °C (see Note 16).
5. Inoculate 500 ml of M9 minimal medium in a 2 L Erlenmeyer
flask (see Note 17) with 25 ml of overnight starter culture.
Grow at 37 °C until the optical density at 600 nm (OD600) is
approximately 0.5–1.0 (about 2–3 h although it may take lon-
ger to reach this OD600).
6. Add IPTG to a final concentration of 1 mM to induce the
expression of your target protein.
7. Grow at 37 °C for 3–5 h or 20 °C overnight (see Note 18).
8. Centrifuge at 5,000 × g for 20 min at 4 °C to pellet the cells.
9. Decant the supernatant and store the pellet at −20 °C.
24 Charles A. Galea
3.2 Protein Autoinduction typically produces significantly higher cell masses

Overexpression and yields of expressed protein compared to conventional induc-
of Isotopically Labeled tion protocols using IPTG (Fig. 2) [50].
Peptides 1. Inoculate 2 ml of LB medium containing the appropriate anti-
in Autoinduction biotic with a single colony from a fresh plate of E. coli
Medium (See Notes 19 BL21(DE3) cells harboring an expression plasmid containing
and 20) your peptide (or protein) of interest.
2. Grow overnight at 37 °C with constant shaking at approxi-
mately 200 rpm.
3. Inoculate 2 ml of NG minimal medium with 20 μl of overnight
culture.
4. Grow at 37 °C for 6–8 h until the culture is visibly turbid.
5. Inoculate 48 ml of NG minimal medium in a sterile 250 ml
Erlenmeyer flask with the entire 2 ml of culture and grow over-
night at 37 °C (see Note 21).
6. Inoculate 500 ml of N-5052 autoinduction medium in a sterile
2 L Erlenmeyer flask (see Note 17) with 25 ml of overnight
starter culture and grow for 24–48 h at 30 °C with constant
shaking at approximately 200 rpm (see Note 22).
7. Centrifuge for 20 min at 5,500 × g and 4 °C to pellet the cells.
8. Decant the supernatant and store the pellet at −20 °C.
3.3 Purification 1. Weigh the frozen pellet and add 10 ml of lysis buffer per 1 g of
of Isotopically Labeled pellet.
His6-Tagged TM/CPD 2. Place on a rotary mixer and gently resuspend the pellet over
Peptide 30–60 min.
3. Lyse the cells by using either sonication (see Note 23) or a
French press.
4. Centrifuge the lysate at 16,000 × g for 20 min at 4 °C.
Fig. 2 Expression of 15N-labeled His6-tagged Trx-TM/CPD. Protein expression was induced by (a) the addition
of 1 mM IPTG and the culture was grown for 3 h at 37 °C or (b) autoinduction at 28 °C over 48 h
5. Collect the supernatant and load onto a 5 ml nickel affinity

column equilibrated with NTA buffer A (see Note 24). Collect
the flow-through and take a 50 μl sample for analysis by
SDS-PAGE.
6. Extensively wash the column with 50 ml of NTA buffer A fol-
lowed by 50 ml of NTA buffer B (see Note 25). Collect the flow-
through and take a 50 μl sample for analysis by SDS-PAGE.
7. Elute the His6-tagged peptide using a linear gradient of
0–100% NTA buffer C over 30 min at a flow rate of 2 ml/min
and collecting 5 ml fractions. Take a 50 μl sample of each frac-
tion for analysis by SDS-PAGE.
8. Analyse samples taken during the purification of the TM/
CPD protein by SDS-PAGE (Fig. 3).
9. Pool fractions containing the protein and store at 4 °C.
10. Add 3C protease to the pooled fractions and incubate with
constant stirring at 4 °C to cleave the His6-tagged protein (see
Note 26).
3.4 Reverse-Phase 1. Add an equal volume of solvent A and then filter the cleaved
HPLC Purification peptide through a 0.22 μm membrane.
of TM/CPD Peptide 2. Load onto a semi-preparative C8 reverse-phase HPLC column
at a flow-rate of 1 ml/min.
3. The column is developed at 0% solvent B for 5 min followed by
a gradient of 0–40% solvent B over 10 min and 40–80% solvent
B over 30 min. Monitor the absorbance at 214 nm and collect
eluting peaks.
4. Freeze the peak fractions in liquid nitrogen and lyophilize.
Fig. 3 Ni2+-affinity (NTA) purification of the His6-tagged TRX-TM/CPD protein. A

and B refer to NTA buffers A and B, respectively
26 Charles A. Galea
5. Dissolve the dried powder for each fraction in 0.5 ml of sam-

ple buffer and analyse by SDS-PAGE.
6. Pool peak fractions containing the TM/CPD peptide, add an
equivalent volume of solvent A and then load onto the same
semi-preparative C8 reverse-phase HPLC column.
7. Develop the column with a gradient of 0–50% solvent B over
5 min followed by 50–70% solvent B over 40 min. Monitor
the absorbance at 214 nm and collect eluting peaks.
8. Freeze the peak fractions in liquid nitrogen, lyophilize, and
store dried powder at −20 °C (see Note 27).
9. Dissolve a small amount of dried powder for each peak frac-
tion in sample buffer and analyse by SDS-PAGE (Fig. 4a) (see
Note 28).
10. Determine the purity and mass of the purified TM/CPD pep-
tide by analytical C8 reverse-phase HPLC and MALDI-TOF
mass spectrometry.
3.5 NMR Sample 1. Dissolve the dried TM/CPD peptide in 550 μl of NMR buffer
Preparation (see Notes 29 and 30).
2. Vortex briefly for about 30 s to promote the incorporation of
the TM/CPD peptide into the DPC micelles.
3. Take a 50 μl sample to determine the concentration of the
protein and for analysis by circular dichroism (CD) spectros-
copy (Fig. 4b) (see Note 31).
4. Transfer the remaining 500 μl of material to a 5 mm NMR tube
(see Note 32). Store at 4 °C prior to NMR data acquisition.
Fig. 4 The purified TM/CPD protein. (a) C8 reverse-phase HPLC chromatogram for the purification of the TM/
CPD protein. The blue line indicates the percentage of solvent B in the mobile phase. Inset: SDS-PAGE analysis
of HPLC peaks 1 and 2 showing the presence of the TM/CPD protein in both fractions. Native PAGE gel analysis
shows peak 2 contains the dimeric form of the TM/CPD protein (data not shown). (b) CD spectrum recorded at
25 and 40 °C for the TM/CPD protein prepared in the NMR buffer which contains DPC micelles. The spectrum
indicates that part of the protein adopts a helical structure (minima at 208 and 222 nm)
3.6 NMR Screening 1. Record a 1D 1H spectrum and a 2D 1H-15 N-TROSY spectrum

of the TM/CPD protein at different temperatures to deter-
mine the optimal temperature to acquire NMR spectra (see
Notes 33 and 34).
2. Process the NMR data using TopSpin (Bruker), NMRView
(One Moon Scientific Inc.) [51], NMRPipe/NMRDraw [52],
or another suitable software package.
3. Examine the spectra (i.e., spectral dispersion, line widths, peak
intensities and number of resonance peaks compared to num-
ber expected based on the amino acid sequence) to determine
whether they are of sufficient quality for structure determina-
tion (Fig. 5).
4 Notes
1. The pET32a expression vector contains an ampicillin resistance

gene and N-terminal His6-tag and thioredoxin fusion tag. There
are a wide range of commercially available expression vectors
that contain other selectable markers for expression in the pres-
ence of an antibiotic (e.g., kanamycin, chloramphenicol), prote-
ase recognition sites (e.g., tobacco etch virus (TEV) protease,
thrombin, enterokinase, and factor Xa), and one or more N- or
C-terminal fusion tags. Fusion tags can facilitate detection on
Western blots (e.g., FLAG, S-tag, c-myc, HSV, glutathione
S-transferase (GST), His6), solubility (e.g., thioredoxin (Trx),
SUMO, maltose binding protein (MBP), NusA), and purifica-
tion (e.g., His6, GST, MBP) of the target protein.
2. Trace metals are required for maximal growth in minimal
media. Each salt should be added in the order outlined in item
9 in Subheading 2.1 after the previous salt has fully dissolved
into the solution. Filter-sterilize and store in the dark at room
temperature.
3. M9 salts can be stored at room temperature for up to 6 months
after autoclaving.
4. The majority of T7-based expression vectors use either ampi-
cillin or kanamycin (35 mg/ml in deionized water) for plas-
mid selection.
5. 15NH4CL and U13C d-glucose can be substituted as the nitro-
gen and carbon source instead of unlabeled NH4Cl and d-glu-
cose for the expression of uniformly 13C/15N-labeled peptides
(or proteins).
6. Uniformly labeled 13C d-glucose can be added instead of
d-glucose for the expression of 13C-labeled peptides.
7. 13C/15N-labeled peptides (or proteins) can be produced if

13
C-glycerol is used as the carbon source instead of unlabeled
glycerol.
28 Charles A. Galea
Fig. 5 NMR spectra for the TM/CPD protein in DPC micelles. 2D 1H/15N TROSY spectra were acquired at 20, 30,
40, and 50 °C on a Varian Inova 600 MHz NMR spectrometer equipped with a cryogenically cooled triple-
resonance probe. Optimal spectral resolution and peak intensities was obtained at 40 °C
8. The cOmplete EDTA-free protease inhibitor cocktail com-

prises tablets that contain inhibitors for serine and cysteine
proteases. The cOmplete EDTA-free cocktail is recommended
since EDTA interferes with the IMAC (immobilized metal
affinity chromatography) purification of His6-tagged fusion
proteins.
9. HRV 3C protease is a recombinant cysteine protease used to
remove fusion tags from proteins containing the HRV 3C
protease cleavage site (Leu-Glu-Val-Phe-Gln*Gly-Pro). This
protease exhibits a high degree of specificity and has optimal
activity at 4 °C.
10. CHAPS is a zwitterionic detergent that is used for the solubi-
lization of membrane proteins over a wide range of pH (2–12).
It has a high critical micelle concentration (CMC) and for this
reason can be easily removed by dialysis or reverse-phase
HPLC.
11. The HiTrap chelating HP column is prepacked with chelating

sepharose high performance agarose. The medium is charged
with nickel for Immobilized Metal Ion Affinity Chromatography
(IMAC) and subsequent purification of polyhistidine tagged
proteins.
12. Chemical shifts are measured relative to the reference com-
pound 1,4-dioxane (3.69 ppm). Other reference compounds
that are commonly used include: 2,2-dimethyl-2-silapentane-
5-
sulfonate (DSS, 0.00 ppm) and tetramethylsilane (TMS,
0.00 ppm). D2O is included as a deuterium lock to correct for
drift in the NMR spectrometers magnetic field with time.
NaN3 (sodium azide) is an antimicrobial that is included to
prevent microbial growth.
13. It is often convenient to prepare a 50% glycerol stock of E. coli
cells transformed with the expression vector containing your
gene of interest and store frozen at −80 °C. To do this inocu-
late 5 ml of LB broth containing an appropriate antibiotic
with a single colony from an LB agar plate and grow overnight
at 37 °C. Take 0.5 ml of overnight culture and add 0.5 ml of
sterile glycerol. Freeze on liquid nitrogen and then store at
−80 °C. Fresh cultures can be prepared by scraping a small
amount of material from the surface of the frozen glycerol
stock and using this to inoculate the LB broth.
14. E. coli BL21(DE3)pLysS contains the pLysS plasmid which
expresses T7 lysozyme that binds to the T7 promoter and
represses expression of the target protein (or peptide) in the
absence of IPTG. Inhibition of basal expression of proteins
(or peptides) prior to induction with IPTG is especially impor-
tant if the expressed protein (or peptide) is toxic to the cells.
15. This step allows the cells to adapt to M9 minimal medium
before inoculating into a large volume of minimal media.
16. This provides a starter culture for inoculating a much larger
volume of M9 minimal medium the following day. The v olume
of this initial starter culture should be approximately 5% of the
final larger volume.
17. Good aeration is particularly important during induction. The
use of baffled flasks and culture volumes no greater than 25%
of the flask volume will help to improve yields.
18. Prior to isotopic labeling, optimal protein expression condi-
tions (i.e., temperature and concentration of IPTG) should be
determined using small-scale cultures. Parameters such as
temperature, IPTG concentration, OD600 at induction, expres-
sion strain, and induction time can critically influence the yield
and solubility of the expressed protein or peptide.
19. The autoinduction medium contains glucose, glycerol, and
lactose as carbon sources. Glucose is the initial primary carbon
and energy source promoting culture growth to high cell den-
30 Charles A. Galea
sities while inhibiting protein expression. Glucose inhibits

recombinant protein expression by preventing induction of
operons responsible for metabolizing lactose to allolactose,
the actual inducer. Once the glucose has been consumed lac-
tose is imported into the host cells and induces the expression
of the recombinant protein. This allows the cell culture to
reach very high OD600 values of ~20 prior to the expression of
the target protein. Glycerol is provided in the autoinducing
media as a good carbon and energy source that does not pre-
vent glucose depletion during growth, glucose inhibition of
protein expression, or the uptake of lactose upon glucose
depletion.
20. For autoinduction, do not use an E. coli strain harboring a
pLysS plasmid if your expression vector contains a T7lac pro-
moter. The presence of the lac repressor in combination with
T7 lysozyme expressed by the pLysS plasmid (which binds to
and inactivates the T7 promoter) leads to significantly lower
levels of protein expression in the autoinduction medium.
21. This starter culture is used to inoculate 1 L of medium the
next day. Increase the size of the starter culture proportionally
(i.e., 50 ml of starter culture per 1 L of minimal medium) if
you plan to inoculate a larger batch of medium.
22. Grow the cells to saturation when using autoinduction

medium. The cultures will grow to a high OD600 before induc-
tion starts. It may take at least 24 h or more for the cultures to
reach saturation when grown at lower temperatures.
23. Sonication should be performed on ice with five rounds of
short bursts of 1.0 min at 2.0 min intervals. It is important to
avoid frothing or overheating the sample as these may cause
protein denaturation.
24. Load the lysate onto the column at a flow-rate of about 2–3
ml/min. The clear lysate can be kept at room temperature at
this stage as it contains protease inhibitors and becomes cloudy
at lower temperatures due to the presence of the CHAPS
detergent.
25. The same flow-rate should be used as for loading the lysate
onto the column. Monitor the absorbance at 280 nm and keep
washing the column until OD280 no longer decreases. It should
be noted that imidazole exhibits a significant absorbance at
280 nm.
26. Perform a small-scale pilot trial in order to determine ratio of
protease to purified protein to use and the time of incubation.
We prefer to use 3C protease since it is a highly specific prote-
ase and has an optimal activity to 4 °C. However, expression
vectors containing a variety of protease recognition sites (e.g.,
TEV, thrombin, enterokinase, and Factor Xa) are also available
but it is important that you check to ensure that the protein

that you wish to express does not also contain an identical
protease cleavage site.
27. Most proteins and peptides are very stable when store for long
periods as a lyophilized powder.
28. Place the sealed container containing the lyophilize powder on
the bench for 10–20 min to equilibrate to room temperature
before opening to avoid the powder absorbing moisture from
the atmosphere.
29. The final concentration of the TM/CPD peptide should be at
least 1 mM. Concentrations lower than 1 mM are acceptable
but will result in weaker NMR signals and an increase in the
amount of time required to acquire NMR spectra. If possible
concentrate the protein to about 300 μl using a centrifugal
concentrator such as an Amicon Ultra-0.5 Centrifugal Unit
(Merck Millipore) with an appropriate molecular weight cut-
off. The high concentration of DPC included in the NMR
buffer ensures that only one molecule of peptide will be incor-
porated into each DPC micelle.
30. The buffer and pH used for the NMR buffer will depend on
the pI of the TM/CDP peptide, as proteins tend to aggregate
close to their pI value where they have neutral charge. If suf-
ficient protein is available screen a range of buffer conditions.
31. Determine the concentration of the TM/CPD protein by

measuring the absorbance at 280 nm and using the equation
A = εbc (where A is the absorbance, ε is the molar extinction
coefficient, b is the path length, and c is the concentration).
32. If the sample volume is less than 500 μl transfer to a 5 mm
Shigemi NMR tube. These NMR tubes can take a minimum
volume of 300 μl.
33. We usually record spectra at 10, 20, 30, 40, and 50 °C

(depending on the stability of the protein). Single helical
transmembrane proteins often exhibit higher quality spectra
(improved spectral resolution and peak intensity) at higher
temperatures. However, transient structure within the extra-
membrane regions (e.g., the cytoplasmic domain) may only be
observed at lower temperatures.
34. 2D–TROSY spectra exhibit improved spectral resolution com-
pared to 2D-HSQC (heteronuclear single quantum coher-
ence) spectra due to the large size of the TM/CPD
protein-DPC micelle complex (typically ≥ 30 kDa). Typical
2D datasets are usually recorded with 1024 × 256 complex
points with 8 to 32 scans per increment (depending on the
sample concentration) on a 600 MHz NMR spectrometer.
Data acquisition times are typically 30–60 min.
32 Charles A. Galea
References
1. Butler GS, Overall CM (2009) Updated bio- inhibitors of metalloproteinases (TIMPs):

logical roles for matrix metalloproteinases and Positive and negative regulators in tumor cell
new "intracellular" substrates revealed by adhesion. Semin Cancer Biol 20(3):161–168
degradomics. Biochemistry 16. Brew K, Nagase H (2010) The tissue inhibitors
48(46):10830–10845 of metalloproteinases (TIMPs): an ancient
2. Morrison CJ, Butler GS, Rodríguez D, Overall family with structural and functional diversity.
CM (2009) Matrix metalloproteinase pro- Biochim Biophys Acta 1803(1):55–71
teomics: substrates, targets, and therapy. Curr 17. Gustafsson T (2011) Vascular remodelling in
Opin Cell Biol 21(5):645–653 human skeletal muscle. Biochem Soc Trans
3. Rodríguez D, Morrison CJ, Overall CM 39(6):1628–1632
(2010) Matrix metalloproteinases: what do 18. Kraiem Z, Korem S (2000) Matrix metallopro-
they not do? New substrates and biological teinases and the thyroid. Thyroid
roles identified by murine models and pro- 10(12):1061–1069
teomics. Biochim Biophys Acta 19. Ortega N, Behonick D, Stickens D, Werb Z
1803(1):39–54 (2003) How proteases regulate bone morpho-
4. Gill SE, Parks WC (2008) Metalloproteinases genesis. Ann N Y Acad Sci 995:109–116
and their inhibitors: regulators of wound heal- 20. Parks WC, Shapiro SD (2001) Matrix metallo-
ing. Int J Biochem Cell Biol proteinases in lung biology. Respir Res
40(6–7):1334–1347 2(1):10–19
5. Mott JD, Werb Z (2004) Regulation of matrix 21. Apte SS, Parks WC (2015) Metalloproteinases:
biology by matrix metalloproteinases. Curr a parade of functions in matrix biology and an
Opin Cell Biol 16(5):558–564 outlook for the future. Matrix Biol 44-46:1–6
6. Nagase H, Visse R, Murphy G (2006) Structure 22. Rohani MG, Parks WC (2015) Matrix remod-
and function of matrix metalloproteinases and eling by MMPs during wound repair. Matrix
TIMPs. Cardiovasc Res 69(3):562–573 Biol 44-46:113–121
7. Page-McCaw A, Ewald AJ, Werb Z (2007) 23. Hu Y-FF, Chen Y-JJ, Lin Y-JJ, Chen S-AA
Matrix metalloproteinases and the regulation (2015) Inflammation and the pathogenesis of
of tissue remodelling. Nat Rev Mol Cell Biol atrial fibrillation. Nat Rev Cardiol
8(3):221–233 12(4):230–243
8. Sternlicht MD, Werb Z (2001) How matrix 24. Nissinen L, Kähäri V-MM (2014) Matrix
metalloproteinases regulate cell behavior. Annu metalloproteinases in inflammation. Biochim
Rev Cell Dev Biol 17:463–516 Biophys Acta 1840(8):2571–2580
9. Vu TH, Werb Z (2000) Matrix metalloprotein- 25. Houghton AM (2015) Matrix metalloprotein-
ases: effectors of development and normal ases in destructive lung disease. Matrix Biol
physiology. Genes Dev 14(17):2123–2133 44-46:167–174
10. Hadler-Olsen E, Fadnes B, Sylte I, Uhlin- 26. Duarte S, Baber J, Fujii T, Coito AJ (2015)
Hansen L, Winberg J-O (2011) Regulation of Matrix metalloproteinases in liver injury, repair
matrix metalloproteinase activity in health and and fibrosis. Matrix Biol 44-46:147–156
disease. FEBS J 278(1):28–45
27. Mukherjee A, Swarnakar S (2015) Implication
11. Sbardella D, Fasciglione GF, Gioia M, Ciaccio of matrix metalloproteinases in regulating neu-
C, Tundo GR, Marini S, Coletta M (2012) ronal disorder. Mol Biol Rep 42(1):1–11
Human matrix metalloproteinases: an ubiqui-
tarian class of enzymes involved in several path- 28. Ip YC, Cheung ST, Fan ST (2007) Atypical
ological processes. Mol Aspects Med localization of membrane type 1-matrix metal-
33(2):119–208 loproteinase in the nucleus is associated with
aggressive features of hepatocellular carcinoma.
12. Piperi C, Papavassiliou AG (2012) Molecular Mol Carcinog 46(3):225–230
mechanisms regulating matrix metalloprotein-
ases. Curr Top Med Chem 12(10):1095–1112 29. Kwan JA, Schulze CJ, Wang W, Leon H,
Sariahmetoglu M, Sung M, Sawicka J, Sims
13. Mannello F, Medda V (2012) Nuclear localiza- DE, Sawicki G, Schulz R (2004) Matrix metal-
tion of matrix metalloproteinases. Prog loproteinase- 2 (MMP-2) is present in the
Histochem Cytochem 47(1):27–58 nucleus of cardiac myocytes and is capable of
14. Murphy G, Nagase H (2011) Localizing matrix cleaving poly (ADP-ribose) polymerase (PARP)
metalloproteinase activities in the pericellular in vitro. FASEB J 18(6):690–692
environment. FEBS J 278(1):2–15 30. Limb GA, Matter K, Murphy G, Cambrey AD,
15. Bourboulia D, Stetler-Stevenson WG (2010) Bishop PN, Morris GE, Khaw PT (2005)
Matrix metalloproteinases (MMPs) and tissue Matrix metalloproteinase-1 associates with
intracellular organelles and confers resistance 42. Hoshino D, Tomari T, Nagano M, Koshikawa
to lamin A/C degradation during apoptosis. N (2009) A novel protein associated with
Am J Pathol 166(5):1555–1563 membrane-type 1 matrix metalloproteinase
31. Si-Tayeb K, Monvoisin A, Mazzocco C, binds p27kip1 and regulates RhoA activation,
Lepreux S, Decossas M, Cubel G, Taras D, actin remodeling, and matrigel invasion. J Biol
Blanc J-FF, Robinson DR, Rosenbaum Chem 284(40):27315–27326
J (2006) Matrix metalloproteinase 3 is present 43. Wang Y, McNiven MA (2012) Invasive matrix
in the cell nucleus and is involved in apoptosis. degradation at focal adhesions occurs via prote-
Am J Pathol 169(4):1390–1401 ase recruitment by a FAK-p130Cas complex.
32. Itoh Y (2015) Membrane-type matrix metal- J Cell Biol 196(3):375–385
loproteinases: Their functions and regulations. 44. Woskowicz AM, Weaver SA, Shitomi Y, Ito N
Matrix Biol 44-46:207–223 (2013) MT-LOOP-dependent localization of
33. Zucker S, Pei D, Cao J, Lopez-Otin C (2003) membrane type I matrix metalloproteinase
Membrane type-matrix metalloproteinases (MT1-MMP) to the cell adhesion complexes
(MT-MMP). Curr Top Dev Biol 54:1–74 promotes cancer cell invasion. J Biol Chem
34. Hernandez-Barrantes S, Bernardo M, Toth M, 288(49):35126–35137
Fridman R (2002) Regulation of membrane 45. Uekita T, Itoh Y, Yana I, Ohno H, Seiki M
type-matrix metalloproteinases. Semin Cancer (2002) Cytoplasmic tail-dependent internal-
Biol 12(2):131–138 ization of membrane-type 1 matrix metallopro-
35. Galea C, Nguyen H, Chandy K, Smith B, teinase is important for its invasion-promoting
Norton R (2014) Domain structure and func- activity. J Cell Biol 155(7):1345–1356
tion of matrix metalloprotease 23 (MMP23): 46. Wang X, Ma D, Keski-Oja J, Pei D (2004)
role in potassium channel trafficking. Cell Mol Co-recycling of MT1-MMP and MT3-MMP
Life Sci 71(7):1191–1210 through the trans-Golgi network identification
36. Itoh Y, Takamura A, Ito N, Maru Y, Sato H of DKV582 as a recycling signal. J Biol Chem
(2001) Homophilic complex formation of 279(10):9331–9336
MT1-MMP facilitates proMMP-2 activation 47. Wang P, Wang X, Pei D (2004) Mint-3 regu-
on the cell surface and promotes tumor cell lates the retrieval of the internalized membrane-
invasion. EMBO J 20(17):4782–4793 type matrix metalloproteinase, MT5-MMP, to
37. Itoh Y, Ito N, Nagase H, Seiki M (2008) The the plasma membrane by binding to its car-
second dimer interface of MT1-MMP, the boxyl end motif EWV. J Biol Chem
transmembrane domain, is essential for 279(19):20461–20470
ProMMP-2 activation on the cell surface. 48. Nguyen H, Galea C, Schmunk G, Smith B,
J Biol Chem 283(19):13053–13062 Edwards R, Norton R, Chandy K (2013)
38. Itoh Y, Ito N, Nagase H, Evans RD (2006) Intracellular trafficking of the K(V)1.3 potas-
Cell surface collagenolysis requires homodi- sium channel is regulated by the prodomain of
merization of the membrane-bound collage- a matrix metalloprotease. J Biol Chem
nase MT1-MMP. Mol Biol Cell 288(9):6451–6464
17(12):5390–5399 49. Cai M, Huang Y, Sakaguchi K, Clore MG,
39. Sakamoto T, Seiki M (2010) A membrane pro- Gronenborn AM, Craigie R (1998) An effi-
tease regulates energy production in macro- cient and cost-effective isotope labeling proto-
phages by activating hypoxia-inducible factor-1 col for proteins expressed in shape Escherichia
via a non-proteolytic mechanism. J Biol Chem coli. J Biomol NMR 11(1):97–102
285(39):29951–29964 50. Studier FW (2005) Protein production by
40. Gonzalo P, Guadamillas MC, Hernández- auto-induction in high-density shaking cul-
Riquer MV (2010) MT1-MMP is required for tures. Protein Expr Purif 41(1):207–234
myeloid cell fusion via regulation of Rac1 sig- 51. Johnson BA (2004) Using NMRView to visu-
naling. Dev Cell 18(1):77–89 alize and analyze the NMR spectra of macro-
41. Gingras D, Michaud M, Tomasso DG, Béliveau molecules. Methods Mol Biol 278:313–352
E (2008) Sphingosine-1-phosphate induces 52. Delaglio F, Grzesiek S, Vuister GW, Zhu G,
the association of membrane-type 1 matrix Pfeifer J, Bax A (1995) NMRPipe: a multidi-
metalloproteinase with p130Cas in endothelial mensional spectral processing system based on
cells. FEBS Lett 582(3):399–404 UNIX pipes. J Biomol NMR 6(3):277–293
Chapter 3
Heterologous Expression of the Astacin Protease Meprin

β in Pichia pastoris
Dagmar Schlenzig and Stephan Schilling
Abstract
Meprins are zinc-dependent proteases of the metzincin superfamily of metalloproteases. The enzymes are
extracellular multi-domain proteins which are stabilized by disulfide bridges, dimerization, and glycosyl-
ation. Due to their complex structure, recombinant expression was first established in mammalian and
insect cells. However, these methods have several disadvantages such as high costs and the low yields. For
this reason, yeast is often considered a preferable expression system. Here, we describe the manipulation
and secretory expression of human meprin β in the methylotrophic yeast P. pastoris. We show that the posi-
tion of the affinity tag strongly influences the yield of expression, favoring fusion of the affinity tag at the
C-terminus.
Key words Metzincin, Endoprotease, Meprin, yeast, Pichia pastoris
1 Introduction
Meprins (meprin α and β) are endoproteases of the astacin family.

Together with matrix metalloproteases (MMPs), astacins constitute
the superfamily of metzincins. Meprin α and meprin β are two
evolutionary related enzymes only found in vertebrata. Both are
formed as zymogenes containing a transmembrane sequence (type
1 transmembrane proteins) and a distinct multi-domain structure.
The signal sequence is followed by an inhibitory pro-sequence and
an astacin-like catalytic domain bearing the typical zinc binding
motif HExxHxxGFxHExxRxDRD [1]. C-Terminal to the catalytic
domain are a MAM (meprin/A5 protein/protein tyrosine phos-
phatase μ) domain and a TRAF (tumor necrosis factor receptor-
associated factor) domain which are thought to play a role in
protein–protein interactions, an EGF (epidermal growth factor)-
like domain, the transmembrane domain and a cytosolic region.
Additionally, meprin α contains a so-called “inserted” domain
N-terminal to the EGF-like domain. Proteolytic processing occurs
35
36 Dagmar Schlenzig and Stephan Schilling
at this site during passage through the secretory p athway, prior to

secretion of soluble meprin α into the extracellular matrix. Meprin
α and β are disulfide-linked homodimers which in the case of
meprin α tends to form large oligomers complexes up to 6 MDa in
size [2]. Both enzymes are heavily glycosylated and the protein–
protein interaction domains as well as glycosylation are necessary
for correct folding, stability, and localization.
In men and mouse, meprins are highly expressed in epithelial
cells of the kidney and the intestine, and they have been localized
in intestinal leukocytes, skin, and certain cancer cells [2]. Both
enzymes show a unique substrate specificity with a preference for
acidic amino acids in the P1′-position [3]. A number of in vitro
substrates of meprins have been identified including extracellular
matrix proteins (procollagen I + III, occludin, E-cadherin) and
biologically active peptides (interleukin 1β, interleukin 18, trans-
forming growth factor-α, vascular endothelial growth factor-A and
fibroblast growth factor 19) [4–6]. Due to the nature of these sub-
strates, there is compelling evidence linking meprins with collagen
assembly, inflammation, intestinal immune response, and neurode-
generation [7].
Unlike MMPs, meprins are not inhibited by tissue inhibitors of
metalloproteinases (TIMPs). However, since the catalytic mecha-
nism and active site architecture of MMPs, ADAMs and meprins
are similar many small molecule inhibitors exhibit limited selectiv-
ity between these enzymes. Considering the compelling evidence
for a role of meprins in pathophysiological processes like fibrosis,
nephropathy, and inflammatory bowel disease [7, 8] the lack of
effective and selective small molecule inhibitors is an obstacle. To
meet the requirements of large amounts of protein for the develop-
ment of selective, potent, and drug-like inhibitors, we report here
the utilization of yeast for the cost-effective production of recom-
binant meprin β [9].
2 Materials
If not otherwise stated, all chemicals and solvents used are of gradi-
ent grade. All materials are sterilized prior to use. Procedures are
carried out under semi-sterile conditions.
2.1 Plasmid 1. Yeast expression vector pPICZα (Thermo Fischer Scientific,

Linearization Waltham, MA USA).
and Transformation 2. PmeI and 10× CutSmart buffer (NEB BioLabs, Ipswich, USA).
of Pichia pastoris
3. Isopropyl alcohol.
X-33 Cells
4. Electro competent X33 cells (Life Technologies GmbH,
Darmstadt, Germany).
5. 0.2 cm electroporation cuvettes.
Expression and Purification of Merpin β in Pichia pastoris 37
6. YPDS medium: 10 g yeast extract, 20 g peptone, 182 g sorbitol,

900 ml mono-distilled water, supplemented with 100 ml 20%
Dextrose (filter sterilized) after autoclaving.
7. YPDS-agar: add 20 g agar to the YPDS medium before
autoclaving.
8. YPDS agar plates containing increasing concentrations of
Zeocin: 100, 200, and 500 μg/ml.
2.2 Test Expression 1. BMGY medium: 10 g yeast extract, 20 g peptone, 700 ml
mono-distilled water supplemented with 100 ml 1 M K2SO4
pH 6.0, 100 ml YNB (34 mg yeast nitrogen base w/o amino
acids, 100 mg (NH4)2SO4, and 1 l of water), 2 ml 0.02% bio-
tin, and 100 ml 10% glycerol (filter sterilized) after
autoclaving.
2. BMMY medium: the same as BMGY but replace the glycerol
with 100 ml of 5% methanol.
3. Methanol (99% HPLC grade).
2.3 Activity Assay. 1. Assay buffer 1: 50 mM Tris–HCl, pH 7.5, 20 mM CaCl2.
2. Assay buffer 2: 40 mM Tris, pH 8.0.
3. Trypsin: 400 μg/ml trypsin in assay buffer 2 (see Note 1).
4. Substrate: 50 μM solution of Abz-YVADAP(Dnp)G-OH (or
Mca-YVADAPK(Dnp)-OH) in assay buffer 2 (see Note 2).
2.4 Purification 1. Equilibration buffer 3: 3 M NaCl, 300 mM Tris, pH 7.6.

2. Equilibration buffer 4: 300 mM NaCl, 30 mM Tris, pH 7.6.
3. Wash buffer 5: 20 mM imidazole, 300 mM NaCl, 30 mM Tris,
pH 7.6.
4. Elution buffer 6: 200 mM imidazole, 100 mM NaCl, 30 mM
Tris, pH 7.6.
5. 1 mM CaCl2 in water.
6. Trypsin: 2 mg/ml trypsin in assay buffer 2.
7. Trypsin inhibition: 1 M 4-(2-aminoethyl)benzenesulfonylfluo-
ride (AEBSF) in water.
8. Equilibration buffer 7: 30 mM Tris, 1.5 M (NH4)2SO4.
9. Elution buffer 8: 30 mM Tris, 100 mM NaCl.
10. Equilibration buffer 9: 40 mM Tris, 100 mM NaCl, pH 7.6.
3 Methods
The open reading frame of meprin α or β are integrated into

pPICZαC using the ClaI and NotI restriction sites. The start
codon and signal sequences are replaced by the plasmid-encoded
Scheme 1 Schematic representation of constructs used for expression of human meprins α and β in
P. pastoris. The native N-terminal signal sequence was replaced by the vector-encoded α-factor pre pro-
sequence of Saccharomyces to enable efficient secretion of the protein. The C-terminal transmembrane
and cytosolic regions were omitted. A His6-tag was introduced either N-terminally between the α-factor
and pro-sequence or C-terminally to accelerate the purification process. Domain structure of meprins
according to UniProt Knowledgebase: S signal sequence, Pro pro-sequence, MAM meprin, 5A protein,
tyrosine phosphatase μ domain, MATH meprin and TRAF homology domain, EGF epidermal growth factor-
like domain, T transmembrane region, I insertion in meprin α which contains a cleavage site. Modified
from [9] with permission of Elsevier
α-factor prepro-sequence of S. cerevisiae. The transmembrane and

cytosolic regions are removed to yield secreted proteins. The
resulting sequences correspond to meprin α (22–600) and meprin
β (23–652). A His6-tag is fused to the N- or C-terminus by means
of appropriate primers in order to facilitate the purification
(see Scheme 1). For termination of translation, a stop codon is
introduced at the end of the sequence. All cloning procedures
are performed according to standard methods described in the
technical guidelines obtained from New England Biolabs. The
ligation efficiency could be enhanced by a dephosphorylation step
utilizing shrimp alkaline phosphatase after restriction digestion of
the vector.
3.1 Plasmid 1. Incubate expression vector (2 μg) with 20 U of PmeI and 5 μl
Linearization 10× CutSmart buffer in a total volume of 50 μl at 37 °C for
and Transformation 1 h. Check for complete linearization using agarose gel elec-
of P. pastoris trophoresis (0.8% gel).
2. Add 500 μl of isopropanol to the reaction for precipitation of

DNA and incubate for at least 2 h at −20 °C. Centrifuge the
mixture at 4 °C and 14400 × g for 30 min, decant the super-
natant, air-dry the pellet for about 2 h and dissolve it in 30 μl
of water. Estimate the DNA concentration.
3. Thaw 50 μl electro competent X33 cells on ice and add 5 μl of
linearized plasmid. Place the electroporation cuvette on ice
and transfer the cells. Place the dried cuvettes in the electro-
porator and apply a pulse of 2 kV (see Note 3).
4. Immediately add 1 ml of ice-cold YPDS medium to the cells
and incubate for 2 h at 30 °C while gently shaking from time
to time.
5. Spread 300 μl of cells per YPDS-agar plate with increasing
Zeocin concentrations (100, 200, 400, and 600 μg/ml) and
incubate for 3 days at 30 °C (see Note 4).
6. Select in total 70–100 colonies from the plates and transfer
them onto labeled master plates containing 100 μg/ml Zeocin
(see Note 5). Incubate at 30 °C for 2 days.
3.2 Test Expression 1. Prepare 70–100 cultivation tubes containing 2 ml of

BMGY. Inoculate each with a single colony from the master
plates and let them grow in a shaking incubator (200 rpm) at
30 °C for 24 h (see Notes 6 and 7).
2. Centrifuge the tubes at 2,000 × g for 5 min and decant the
supernatant. Add 2 ml of BMMY to the cells, resuspend the
cells and incubate with shaking (200 rpm or higher) at 30 °C
for 24 h.
3. Maintain the methanol concentration at about 0.5% (v/v) for
the next 2 days by daily addition of 10 μl of 100% methanol.
4. Centrifuge the samples at 2,000 × g for 10 min and collect the
supernatant. Keep it on ice until performing the activity assay.
3.3 Activity Assay 1. Pipet 50 μl of media supernatant to a 96-well plate, add 100 μl
of assay buffer 1, 50 μl trypsin and incubate for 30 min at
37 °C.
2. Transfer 50 μl of each well to a black clear bottom 96-well plate,
add 150 μl of assay buffer 2 and incubate for 10 min at 30 °C. Start
kinetic fluorescence measurement immediately after addition of
50 μl of substrate solution and measure time-dependent conver-
sion of substrate at excitation and emission wavelengths of 340
and 420 nm, respectively (Fig. 1, see Note 8).
3. Select samples with the highest RFU/min and save the corre-
sponding colonies as glycerol stocks (Figs. 1 and 2, see Note 9).
Fig. 1 (a) Meprin β activity determined in the medium after small-scale expression of C-terminally or
N-terminally His6-tagged meprin β in P. pastoris. Meprin β activity in the media of at least 50 clones after
an activation step with trypsin depicted as logarithmic scatter plot. On average, the activity for NHis-meprin
β is more than 20-fold lower compared with CHis-meprin β. (b) The difference in expression efficiency is
also reflected by Western blot analysis. A total amount of 100 μg protein from the fermentation medium was
applied to each lane. Meprin β was detected using a polyclonal antibody. (c) No enzymatic activity could be
detected after test expression of CHis-meprin α in P. pastoris. On the protein level, a signal of reduced size
was detected with a polyclonal antibody. ServaGelTM TG PrimeTM 4–20%, Modified from [9] with
permission of Elsevier
Fig. 2 Activity measurement of three selected colonies expressing CHis-meprin β or NHis-meprin β (Fig. 1),
respectively. Prior to measurement, samples were activated by incubation with trypsin. Assays were started by
addition of substrate. Samples of CHis-meprin β were diluted 1:10
3.4 Scale 1. Expression was scaled up by fermentation in a bioreactor, basi-

Up of Expression cally as described in the Pichia Fermentation Process Guidelines
(See Note 10) (Thermo Fischer Scientific, Waltham, MA USA). Briefly, fer-
mentation was carried out in a 5 l reactor. The cells were grown
in fermentation basal salt medium (pH 5.5, T = 30 °C) supple-
mented with trace salts and glycerol. The process was started
by inoculation with 200 ml pre-culture with an OD600 of about
2–3. A 24 h initial batch phase followed by a 6 h glycerol feed
batch phase results in an accumulation of cell mass (OD600 of
about 400–450). Application of the three-step methanol feed-
ing protocol outlined in the Process Guideline for 45–50 h is
accompanied by a steady increase of meprin activity [9]
(see Note 11).
2. Separate the cells from the supernatant by centrifugation at
6,000 × g for 15 min at 4 °C. Store the supernatant at
−80 °C. Long-term storage for months does not negatively
affect the yield of meprin.
3.5 Purification 1. Thaw the supernatant and dilute it with 1:10 of the final total
(Fig. 3, See Note 12) volume with equilibration buffer 3. Centrifuge at 30,000 × g
for 20 min at 4 °C and apply the supernatant to a Streamline
chelating expanded bed column (25 × 280 mm in non-
expanded state; GE Healthcare Life Science, Piscataway, USA,
see Note 13) previously equilibrated with equilibration buffer
4 in reverse flow at a flow rate of 12 ml/min. After a washing
step with wash buffer 5 the enzyme is eluted in forward flow
with elution buffer 6. The eluted protein is stored at 4 °C
overnight.
2. The eluted enzyme is supplemented with CaCl2 (final concen-
tration of 10 mM). One milligram of trypsin is added per
15 mg of protein and incubated with gentle shaking at room
temperature for 15 min. Afterwards, add AEBSF (final concen-
tration of 5 mM) for inhibition of trypsin.
3. The volume of the sample is doubled by the addition of 3 M
(NH4)2SO4 in the cold (see Note 14). Syringe filter through a
0.46 μM filter membrane to remove turbidity. Load the sample
onto a pre-equilibrated Butyl Sepharose column (25 × 100 mm,
GE Healthcare Life Science, Piscataway, USA, equilibrated
with buffer 7) at a flow rate of 3 ml/min. Elute with an ammo-
nium sulfate gradient (1.5–0 M ammonium sulfate acheived by
mixing buffers 7 and 8) for 10 column volumes at the same
flow rate.
4. Eluted protein is concentrated by ultrafiltration (MWCO
30,000 Da) and applied onto a Superdex 200 column
(26 × 850 mm, material GE Healthcare Life Science, Piscataway,
USA) equilibrated with buffer 9 (see Notes 15 and 16). Eluted
Fig. 3 Purification of CHis-meprin β expressed in P. pastoris. (a) Chromatogram of purification of the fermenta-
tion broth by immobilized metal chelating chromatography (IMAC). (b) Chromatogram of purification of the
eluate from IMAC using hydrophobic interaction chromatography. (c) Final purification performing size exclu-
sion chromatography. Bars indicate meprin containing fractions
protein can be concentrated again, aliquoted and stored at

−80 °C (see Note 17). Analyze the eluted protein by SDS-
PAGE to determine its purity (Fig. 4, see Note 18).
4 Notes
1. Aliquots of 20 mg/ml trypsin stock solution can be prepared

in 0.1 M HCl for long term storage at −20 °C. Prepare fresh
trypsin working solution from these aliquots prior to use by
the addition of assay buffer 1.
Fig. 4 SDS-PAGE analysis illustrating the purification of CHis-meprin β from the

fermentation supernatant of P. pastoris. The combined fractions after each step
have been applied. Lanes: M molecular mass standards, FB fermentation broth
(50 μg), 1 after IMAC (5 μg), 2 HIC following zymogen activation by trypsin (5 μg),
3 after final SEC (5 μg). Asterisk trypsin, ServaGel™ TG Prime™ 4–20%.
Reproduced from [9] with permission of Elsevier
2. Aliquots of 1 mM substrate stock solution can be prepared in

DMSO for long term storage at −20 °C. Prepare fresh
substrate working solution from these aliquots prior to use by
the addition of assay buffer 2.
3. A pulse time of about 5 ms is optimal for electroporation.
4. Increasing the Zeocin concentration for selection of transfor-
mants to more than 500 μg/ml leads to fewer colonies which
did not show higher expression rates.
5. For longer storage keep master plates at 4 °C. After 1 month,
transfer the colonies on new plates if further use is intended.
6. Addition of Zeocin to test expression medium is not essential.
7. Include untransformed X33 cells in the test expression to per-
form negative control assays.
8. Subtract the slope obtained for a negative control test expres-
sion from that of the samples.
9. Further evaluation of selected colonies before scale-up is
advised. The protein concentration in the supernatant after
test expression is very low. Since the target protein can hardly
be detected on Coomassie-stained SDS gels a Western blot
should be used to verify the molecular mass and integrity of
the expressed protein (i.e., presence of His-tag protein).
10. Due to the low amount of expressed protein, process-controlled

high density fermentation is appropriate for fast expansion of
cell mass and production of larger amounts of enzyme.
11. Scaling up expression to 2 l of media in shaking flasks is inef-
fective due to low cell mass accumulation.
12. Purification steps are performed at room temperature as
meprin β is very stable. However, samples were kept on ice
prior to application onto columns and after elution.
13. It is possible to make use of other standard IMAC columns.
The expanded bed material has the advantage of application of
crude material at very high flow rates.
14. Add the 3 M ammonium sulfate dropwise and slowly with gen-
tle stirring.
15. Equilibrate the column with 4–5 column volumes (CV) equili-
bration buffer. If the column is stored in buffer containing
ethanol, start with 2 CV of water prior to buffer.
16. Apply not more than 4.5 ml of sample (1% of total CV) to the
gel filtration column for efficient separation.
17. The protein is highly glycosylated. About 15% of the protein is
glycosylated with N-linked carbohydrates. Therefore, the pro-
tein can be concentrated to 30 mg/ml without precipitation.
Due to the high stability of human meprin β, freezing aliquots
at −80 °C does not lead to decreased enzyme activity.
18. Human meprin β has been successfully expressed in P. pastoris.
The location of the His6-tag dramatically influenced the overall
yield. Thirty milligram of functional enzyme purified to homo-
geneity could be produced from 2 l of process-controlled high
density fermentation of C-terminally tagged protein. For
N-terminally His6-taged enzyme the yield was tenfold lower.
Due to protein degradation, human meprin α could not be
isolated, despite a high sequence homology of about 45%. The
utilization of protease A deficient SMD5896 cells and the vari-
ation of culture conditions like pH, temperature, and supple-
mentation with casein hydrolysate did not show any positive
effect on expression of meprin α.
References
1. Dumermuth E, Sterchi EE, Jiang WP et al side specificity for the astacin metalloprotease
(1991) The astacin family of metalloendopepti- family reflected by physiological substrates. Mol
dases. J Biol Chem 266(32):21381–21385 Cell Proteomics 10(9):M111.009233
2. Sterchi EE, Stocker W, Bond JS (2008) 4. Kronenberg D, Bruns BC, Moali C et al (2010)
Meprins, membrane-bound and secreted asta- Processing of procollagen III by meprins: new
cin metalloproteinases. Mol Aspects Med players in extracellular matrix assembly? J Invest
29(5):309–328 Dermatol 130(12):2727–2735
3. Becker-Pauly C, Barre O, Schilling O et al 5. Bao J, Yura RE, Matters GL et al (2013) Meprin
(2011) Proteomic analyses reveal an acidic prime A impairs epithelial barrier function, enhances
monocyte migration, and cleaves the tight junc- cancer and fibrosis. Biochem J 450(2):
tion protein occludin. Am J Physiol Renal 253–264
Physiol 305(5):F714–F726 8. Biasin V, Marsh LM, Egemnazarov B et al
6. Huguenin M, Muller EJ, Trachsel-Rosmann S (2014) Meprin β, a novel mediator of vascular
et al (2008) The metalloprotease meprinbeta remodelling underlying pulmonary hyperten-
processes E-cadherin and weakens intercellular sion. J Pathol 233(1):7–17
adhesion. PLoS ONE 3(5):e2153 9. Schlenzig D, Wermann M, Ramsbeck D et al
7. Broder C, Becker-Pauly C (2013) The metal- (2015) Expression, purification and initial char-
loproteases meprin α and meprin β: unique acterization of human meprin β from Pichia
enzymes in inflammation, neurodegeneration, pastoris. Protein Expr Purif 116:75–81
Part II
Structural Characterization of Matrix Metalloproteases

Chapter 4
Structural Studies of Matrix Metalloproteinase by X-Ray

Diffraction
Elena Decaneto, Wolfgang Lubitz, and Hideaki Ogata
Abstract
Matrix Metalloproteinases (MMPs) are a family of proteolytic enzymes whose endopeptidase activity is
dependent on the presence of specific metal ions. MT1-MMP (or MMP-14), which has been implicated
in tumor progression and cellular invasion, contains a membrane-spanning region located C-terminal to a
hemopexin-like domain and an N-terminal catalytic domain. We recombinantly expressed the catalytic
domain of human MT1-MMP in E. coli and purified it from inclusion bodies using a refolding protocol
that yielded significant quantities of active protein. Crystals of MT1-MMP were obtained using the vapour
diffusion method. Here, we describe the protocols used for crystallization and the data analysis together
with the resulting diffraction pattern.
Key words Metalloproteinase, MMP-14, MT1-MMP, Crystallization, Diffraction
1 Introduction
Matrix metalloproteinases (MMPs) are a family of proteolytic

enzymes whose endopeptidase activity is dependent on the pres-
ence of specific metal ions [1]. These enzymes are conserved across
the animal kingdom and are mostly involved in various physiologi-
cal processes such as embryonic development, morphogenesis, tis-
sue remodeling and wound healing. Abnormalities in the
mechanisms of regulation that control the multiple functions of
MMPs may lead to severe inflammatory and degenerative condi-
tions such as arthritis, cancer, and cardiovascular diseases [2, 3].
MMPs share a common domain structure characterized by (1) an
N-terminal pro-peptide which inhibits protein activity prior to
zymogenic activation, (2) a catalytic domain containing two zinc
binding sites and at least two calcium binding sites, and (3) an
hemopexin-like domain connected to the catalytic domain through
a flexible “hinge region.” While MT1-MMP (or MMP-14), an
important MMP involved in tumor progression and cellular inva-
sion, contains an additional membrane-spanning domain located
49
50 Elena Decaneto et al.
C-terminal to the hemopexin-like domain [4, 5]. Of the 176 X-ray

structures of human MMP catalytic domains, only MMP-1 [6],
MMP-3 [7], MMP-8, and MMP-12 [8] were crystallized in the
active form. In addition, no catalytic domain structure for a
membrane- associated MMP has been solved except for MT1-
MMP and MT3-MMP complexed with natural and synthetic
inhibitors, respectively [8–10]. Detailed structural knowledge of
the catalytic domain of MMPs in all stages of enzymatic activity is
indispensable for understanding the catalytic process and thus for
controlling enzymatic activity, improving the design criteria for
specific inhibitors, and enhancing the sensitivity of ligand screen-
ing [7, 11].
Herein, we describe the recombinant expression of the cata-
lytic domain of human MT1-MMP in E. coli and its purification
from inclusion bodies utilizing a refolding protocol that yields sig-
nificant quantities of active protein. The purified protein was used
to obtain crystals of MT1-MMP [12]. We also outline protocols
used for the crystallization of the protein and analysis of data to
generate the diffraction pattern.
2 Materials
2.1 Chemicals 1. pET-3 expression system (Novagen; Madison, WI).

and Solutions 2. Human MT1-MMP cDNA.
for Purification
3. E. coli strains BL21 (DE3).
4. SOC (Super optimal broth) sterile media: 20 g tryptone, 5 g
yeast extract, 2 mL of 5 M NaCl, 2.5 mL of 1 M MgCl2, 10
mL of 1 M MgSO4, and 20 mL of 1 M glucose for 1 L.
5. 2YT agar plates: 2.5 g NaCl, 8 g tryptone, 5 g yeast extract,
and 7.5 g agar for 0.5 L.
6. LB (Lysogeny broth) sterile media: 10 g NaCl, 10 g tryptone,
and 5 g yeast extract for 1 L.
7. Potter-Elvehjem homogenizer (Heidolph; Elektro GmbH,
Kelheim, Germany).
8. Sonifier cell disrupter (Branson, Danbury, CT).
9. 0.9% NaCl.
10. Wash buffer I: 30 mM Tris–HCl pH 7.4, 30 mM NaCl, 1 mM
EDTA, 2% (v/v) Tween 20, and 5 mM β-mercaptoethanol.
11. Wash buffer II: 30 mM Tris–HCl pH 7.4, 500 mM NaCl, 1
mM EDTA, and 5 mM β-mercaptoethanol.
12. Wash buffer III: 30 mM Tris–HCl pH 7.4, 30 mM NaCl, 1
mM EDTA, and 5 mM β-mercaptoethanol.
13. Refolding buffer I: 50 mM Tris–HCl pH 8.5 and 6 M urea.
X-ray Diffraction Studies of Matrix Metalloproteases 51
14. Refolding buffer II: 50 mM Tris–HCl pH 8.5, 100 mM NaCl,

10 mM CaCl2, 10 μM Zn(OAc)2, and 2 M urea.
15. Refolding buffer III: 50 μM Tris–HCl pH 7.5, 100 mM NaCl,
10 mM CaCl2, 10 μM Zn(OAc)2, and 1 M urea.
16. Refolding buffer IV: 50 mM Tris–HCl pH 7.5, 100 mM

NaCl, 5 mM CaCl2 and 1 μM Zn(OAc)2.
17. Refolding buffer V: 50 μM Tris–HCl pH 7.5, 100 mM NaCl,
and 5 mM CaCl2.
18. 10 kDa Amicon Ultra-0.5 centrifugal filters (EMD Millipore,
Darmstadt, Germany).
19. SDS-PAGE (sodium dodecyl sulfate–polyacrylamide gel elec-
trophoresis) equipment.
20. Running buffer: 25 mM Tris–HCl pH 6.8, 1 mM Na2EDTA,
10% (v/v) glycerol, 5% (v/v) β-mercaptoethanol, and 0.025%
(w/v) bromophenol blue.
21. Staining solution: 0.1% (w/v) Coomassie Brilliant Blue

G-250 in 10% (v/v) acetic acid and 40% (v/v) methanol.
22. Destaining solution: 10% (v/v) acetic acid and 20% (v/v)

methanol.
23. Protein Test Mixture 6 (Serva Electrophoresis, Heidelberg,
Germany).
24. Fluorogenic substrate
Mca-Lys-Pro-Leu-Gly~Leu-Lys(Dnp)-Ala-Arg-NH2.
2.2 Chemicals 1. Cryoprotectant buffer (0.3 M ammonium nitrate, 25% (w/v)

and Instrumentations PEG 3350 (pH 6.2), and 20% glycerol).
for Crystallization 2. Thermal Seal RT film, 2mil (Sigma-Aldrich).
3. Clear tape, HD clear (ShurTech Brands, LLC).
4. Microscope, Leica M125 (Leica Microsystems).
5. 96-well “Intelli-Plate,” LBR, Low profile, three wells (Dunn
Labortechnik GmbH, Germany).
6. Liquid dispensing robot, PHOENIX (Art Robbins Instruments,
CA, USA).
7. Crystallization buffer kits (PEG ION/PEG ION2) (Hampton
research, CA, USA).
8. Compressor-cooled incubator, IPP500 (Memmert, Schwabach,
Germany).
9. Izit Crystal Dye (Hampton research, CA, USA).
2.3 Instrumentation 1. Magnetic cryovials and cryocaps (Molecular Dimensions, UK).

for X-Ray Data 2. Foam dewar (Molecular Dimensions, UK).
Collection
3. Vial clamp, curved (Hampton Research, USA).
4. Crystalwand Magnetic (Hampton Research, USA).

5. Cryo tong, 18 mm (Hampton Research, USA).
6. Cryoloop 10 μm (0.1–1.0 mm) (Hampton Research, USA).
7. Cryocane (Molecular Dimensions, UK).
8. Dryshipper, CX100B-11 M (Taylor Wharton).
2.4 Computer The computer programs for data analysis can be found on the
Programs Internet websites.
1. Linux operation system (Ubuntu12.04).
2. XDS (X-ray Detector Software), XSCALE and XDSCONV
(http://xds.mpimf-heidelberg.mpg.de/).
3. CCP4i (http://www.ccp4.ac.uk/).
4. PHENIX (http://www.phenix-online.org/).
5. COOT (http://www2.mrc-lmb.cam.ac.uk/personal/pems-
ley/coot/).
6. PYMOL (https://www.pymol.org/).
3 Methods
3.1 Purification 1. Transform E. coli BL21 (DE3) chemically competent cells with
the pET-3a plasmid containing the catalytic domain of human
3.1.1 Induction MT1-MMP together with the hinge-linker to the hemopexin-
and Harvest like domain (residues 112–292) [5]. Transfer 1 μL of DNA to
100 μL of E. coli strain BL21 (DE3) competent cells in a sterile
tube and kept on ice for 30 min.
2. Incubate the cells at 40 °C for 90 s and then place on ice for
2 min.
3. Add 60 μL of cells to 1 mL of SOC medium in a 15 mL sterile
tube (see Note 1).
4. Incubate the cells at 37 °C for 1 h at 180 rpm.
5. Streak cells on 2YT agar plates containing ampicillin (100 μg/
mL) and incubate overnight at 37 °C.
6. Inoculate a single colony into LB media containing 100 μg/
mL of ampicillin and grow overnight at 37 °C (see Note 2).
7. Inoculate 1 L of LB media containing ampicillin with an ali-
quot (50–100 mL) of the overnight culture and grow at 37 °C
to an OD600 of 0.8.
8. Induce cells with IPTG (1 mM) for a further 5 h at 37 °C.
9. Following induction, harvest the cells by centrifugation
(6,000 × g, 4 °C, 15 min) and suspend the cell pellet in 0.9%
NaCl (see Note 3).
3.1.2 Refolding 1. Homogenize the cells using a Potter–Elvehjem homogenizer.

and Purification 2. Sonicate the cells on ice for five cycles of 1 min with a sonifier
with a 1 s pulse duration at 70% output power (see Note 4).
3. Centrifuge (6,000 × g, 4 °C, 15 min) to harvest the inclusion
bodies in the pellet.
4. Homogenize the inclusion bodies and wash them twice with
Wash buffer I.
5. Then, wash the inclusion bodies again twice with Wash
buffer II.
6. Finally wash them twice with Wash buffer III.
7. Homogenize the inclusion bodies in Refolding buffer I and
keep at 4 °C for 2 h.
8. Ultracentrifuge the inclusion bodies (100,000 × g, 4 °C, 30
min).
9. Dilute the supernatant in the same buffer to a protein concen-
tration of less than 1 mg/mL (see Note 5).
10. Dialysis (12–14 kDa molecular weight cutoff membrane)

against a ten-fold volume of Refolding buffer II for 12 h at
room temperature.
11. Change the solution to Refolding buffer III for 24 h.
12. Dialyze the protein against Refolding buffer IV for 24 h.
13. Concentrate the solution to approximately 10 mg/mL using
an Amicon Ultra-0.5 centrifugal filter device (10 kDa molecu-
lar weight cutoff) and then exchange the solution for zinc-free
Refolding buffer V.
14. Incubate the protein solution for 24 h at room temperature
until the C-terminal hinge region is fully digested (Fig. 1).
The band at ~25 kDa in Fig. 1 corresponded to the full-length
construct while the lower band at 19–20 kDa correlated with
the size of the domain without the C-terminal hinge region.
After approximately 1 day, only the lower band corresponding
to the MT1-MMP catalytic domain was observed and this was
stable over the next 3 days. The freshly purified MT1-MMP
catalytic domain lacking the hinge region was used for crystal-
lization and structure determination (see Note 6).
15. Measure the activity of the mature MT1-MMP using a fluoro-
genic assay (λex = 340 nm, λem = 400 nm). We typically use an
assay which involves monitoring the change in fluorescence
upon cleavage of the peptide Mca-Lys-Pro-Leu-Gly~Leu-
Lys(Dnp)-Ala-Arg-NH2 (see Note 7).
3.2 Crystallization Screening of crystallization conditions was carried out by the sit-
ting vapour-diffusion method with a liquid dispensing robot,
PHOENIX (Art Robbins Instruments, CA, USA).
Fig. 1 Autoproteolytic activity of MT1-MMP. Day 1: recombinantly expressed

human MT1-MMP catalytic domain together with the hinge region following
dialysis in refolding buffer. Day 2, 3, and 4: the sample shown in lane 1 was kept
at 20 °C for 1, 2, and 3 days, respectively prior to analysis. Lane M contains the
molecular-mass marker (labeled in kDa). Samples were run on a 12% SDS–
PAGE gel
1. Precool the crystallization screening buffer kits (PEG ION

and PEG ION2) at 4 °C.
2. Spin the concentrated MT1-MMP samples (10 mg/mL) at
10,000 rpm (~9,000 g) for 5 min at 4 °C.
3. Wash the nano-needle and the 96-syringes of the dispensing
robot PHOENIX with distilled water.
4. Aspirate each 100 μL reservoir solution of the 96 different
buffers (PEG ION and PEG ION2) with the 96-syringes.
Aspirate the 50 μL protein solution with the nano-needle.
5. Dispense 80 μL of the 96 different buffers into the reservoir
well of the “Intelli-plate” with a dispensing robot, PHOENIX
(see Note 8).
6. Dispense 0.3 μL of the protein solution into the middle of the
three wells of the “Intelli-plate” (see Note 9).
7. Dispense 0.3 μL of the screening buffer into the middle of the
three wells of the “Intelli-plate” (see Note 9).
8. Seal the plate with “ThermalSeal” and store at 4 °C.
9. Recover the remaining crystallization buffer in the 96-syringes
and the remaining protein solution in the nano-needle.
10. Wash the 96-syringes and the nano-needle with distilled water.
11. Check the crystallization plates with a microscope (×50–×100
magnitude). Crystals of MT1-MMP were obtained in a solu-
tion containing 0.2 M ammonium nitrate and 20% (w/v) PEG

(polyethylene glycol) 3350 (pH 6.2) (Fig. 2a).
12. To stain the protein crystals add Izit Crystal Dye to one of the
crystallization wells and incubate for a few hours (Fig. 2b).
3.3 Crystal Handling 1. Prepare the cryoprotectant buffer (see Note 10).
for the Cryo 2. Open the seal of the crystallization plate.
Experiments
3. Pick the crystals with a cryo-loop under the microscope.
4. Soak the crystals in the cryoprotectant buffer and wait for sev-
eral minutes (1─5 min).
5. Pick a single crystal and freeze in liquid nitrogen (see Note
11).
6. Put the cryocap on the cryo-loop vial in liquid nitrogen and
position it on the cryo-cane (see Note 12).
7. Store the cryo-cane in the precooled dryshipper for future
analysis (see Note 13).
3.4 X-Ray Diffraction Diffraction data sets were collected on beamline BL14.2 at
Data Collection BESSYII (Hermholtz-Zentrum Berlin, Germany).
1. The CCD detector is a Rayonix MX-225 (Rayonix, USA). Set
the detector-to-crystal distance to 300 mm.
2. Set the cryo nitrogen gas-stream temperature at 100 K.
3. Mount the crystal with the cryo-loop on the goniometer. Then
align the crystal to the center of the X-ray beam position.
4. Set the X-ray wavelength of 0.91841 Å (maximum X-ray beam
intensity).
5. Collect a diffraction image and check the highest resolution
diffraction spot. If the resolution is higher, set the detector-to-
crystal distance shorter.
6. Repeat step 5 until the maximum resolution of the diffraction
spots is covered.
Fig. 2 MT1-MMP crystals. (a) The original MT1-MMP crystals and (b) stained with the blue colored Izit dye
7. Collect a complete data set. The data collection conditions of

the MT1-MMP crystal are: X-ray exposure time of 3 s, oscilla-
tion angle of 1.0°, and total number of diffraction images of
100. The detector-to-crystal distance was 250 mm.
8. Copy the image files to a USB hard drive.
3.5 Data Analysis The computer programs described below were used for data analy-
sis on a Linux operating system (Ubuntu12.04).
1. The diffraction images are indexed and integrated using the
program XDS [13]. The input file template to use is supplied
by the beamline facility. The input parameters are as follows;
JOB = XYCORR INIT COLSPOT IDXREF DEFPIX XPLAN
INTEGRATE CORRECT, DETECTOR_DISTANCE =
250.0, OSCILLATION_RANGE = 1.0, X-RAY_
WAVELENGTH = 0.91841, NAME_TEMPLATE_OF_
DATA_FRAMES = mmp_1_???.img, DATA_RANGE = 1100
(XDS will substitute the appropriate image number for the
question marks), SPOT_RANGE = 1 11, SPOT_RANGE =
42 52, SPOT_RANGE = 73 78, SPACE_GROUP_
NUMBERS = 96, UNIT_CELL_PARAMETERS = 63 63
123 90 90 90, INCLUDE_RESOLUTION_RANGE = 50.0
2.24, FRIEDEL’S_LAW = TRUE (see Note 14).
2. Scaling is carried out using XSCALE. In the input file, select
the integrated data (XDS_ASCII.HKL) as the input file. The
output file is named as MT1-MMP.xscale. Set the parameters,
INCLUDE_RESOLUTION_RANGE = 50 2.24, OUTPUT_
FILE = MT1-MMP.xscale.
3. The scaled file (MT1-MMP.xscale) is converted to the MTZ
file format (CCP4) by XDSCONV. Set the parameters:
GENERATE_FRACTION_OF_TEST_REFLECTIONS =
0.05, FRIEDEL’S_LAW = TRUE.
(a) For molecular replacement, an initial model, the MT1-
MMP/TIMP-2 complex (PDB entry 1BQQ) [10], was cho-
sen from the PDB database (http://www.rcsb.org/).
4. Before molecular replacement, remove the coordinates for
TIMP-2 from the PDB file with a text editor and save this
(1BQQ-MT1-MMP.pdb) as the search model.
5. In the program CCP4i [14], create a project (MT1-MMP)
using the “Change Project” tab.
6. Select “Run Molrep –auto MR” [15] in the “Molecular
Replacement” tab.
7. Select the reflection file MT1-MMP.mtz from XDSCONV as
the input file. Set the parameters; Model in “1BQQ-MT1-
MMP.pdb”, Coords out “MT1-MMP-MR.pdb”, Use data to
maximum resolution “2.24” in the Experimental Data tab,
Search for “1” monomers in the asymmetric unit in the Search
Parameters, then run the program.
8. In the output file from MOLREP, check the molecular replace-

ment solution.
9. In the refinement program PHENIX [16], create the new proj-
ect “MT1-MMP” and select “refinement” in the menu tab.
10. For the first refinement, select the files (MT1-MMP.mtz, MT1-
MMP-MR.pdb, ZN.cif and CA.cif) as the input data (see Note
15). Select the parameter, “Rigid-body” in the Refinement
strategy. Run the refinement setting the number of cycles to 3
and the output file name as MT1-MMP_refine_1.pdb.
11. For the second refinement, select the files MT1-MMP.mtz,
MT1-MMP_refine_1.pdb, ZN.cif and CA.cif as the input
data. Set the parameters; “xyz coordinates”, “Real-space”,
“Individual B-factors”, “Occupancies” in the Refinement
strategy. Set the other parameters: “Update waters” and
“Automatically correct N/Q/H error”. Run the refinement
with three cycles.
12. For the model building, run COOT [17]. Open the coordi-
nate file and the reflection file, MT1-MMP_refine_2.pdb and
MT1-MMP_refine_2.mtz (including the phases from the
PHENIX output), respectively. Set the contour levels (e.g., for
the 2Fo-Fc map: 1 rmsd, for the Fo-Fc map: 3 rmsd).
13. Check the electron density maps and the refined model (Fig.
3). Modify the model to fit into the electron density. Save the
coordinates as MT1-MMP_refine_2-coot1.pdb.
14. Repeat steps 11–13 until the R-factor is converged.
15. Check the refinement statistics using POLYGON and the ste-
reochemical geometry (Ramachandran plot) using
MOLPROBITY in the output tab of the refinement in
PHENIX.
4 Notes
1. The nutrient-rich microbial broth SOC instead of LB is rec-

ommended for transformation because of its refined glucose
and salts balance which maximizes the transformation
efficiency.
2. Care should be taken during the selection of a single colony
from the plate to avoid cross contamination from the other
colonies.
3. This step was repeated three times. Usually about 3–4 g of
pellet was obtained per liter of media. Note that after resus-
pension of the pellet, the cells can either be directly extracted
or frozen at −80 °C for up to a month before extraction with-
out a significant degradation of the protein.
4. Alternatively, cells were also lysed using a high pressure
Emulsiflex C5 homogenizer (Avestin, Mannheim, Germany).
Fig. 3 MT1-MMP catalytic domain model generated using the program

COOT. Snapshot showing the stick model for the MT1-MMP catalytic domain and
the corresponding electron density map (2Fo-Fc map in blue mesh). The coordi-
nation of a calcium ion to various protein residues and water molecules is also
illustrated
The cells were passed twice through a French press at 10,000–

15,000 psi to ensure complete cell lysis.
5. Dilution is important to reduce steric hindrance by providing
space for refolding to occur, thus allowing for more efficient
protein refolding and decreased protein precipitation. The
protein concentration was determined by measuring the
absorbance at 280 nm using the calculated molar extinction
coefficient of 35,410 M−1 cm−1 at 280 nm in water [18].
6. Following digestion of the hinge region, the protein can be
frozen in liquid nitrogen after addition of glycerol (10% final
concentration) and stored at −80 °C until use. Repeated freez-
ing and thawing cycles should be avoided.
7. Mca is (7-methoxycoumarin-4-yl)-acetyl and Dnp is N-3-
(2,4-dinitrophenyl)-l-2,3-diaminopropionyl. This is a FRET-
based activity assay for MMPs in which peptide hydrolysis can
be monitored by fluorescence spectroscopy [19, 20].
8. If a dispensing robot is not available, dispense each 80 μL ali-
quot manually.
9. The crystallization drop size can be increased up to 1.0 μL in the
96-well Intelli-plate. If a dispensing robot is not available, dis-
pense each 0.5 μL aliquot manually. A 24-well “Cryschem Plate”
(Hampton research) can be used to obtain larger crystals.
10. Optimize the concentration of the cryoprotectant. It is better

to use a slightly higher concentration of salts and
precipitants.
11. Wear gloves and goggles for protection when you use liquid
nitrogen.
12. The cryo cap should be precooled in liquid nitrogen and the
frozen crystal should be kept at liquid nitrogen temperature.
13. The dryshipper should be precooled with liquid nitrogen

before use (at least 1 day before).
14. A graphical user interface, XDSAPP [21], can be used. It auto-
mates the data processing (XDS, SCALE, and XDSCONV)
and generates the suitable reflection files (https://www.
helmholtz-berlin.de/forschung/oe/em/soft-matter/forsc-
hung/bessy-mx/xdsapp/index_en.html).
15. The cif files can be found under the monomer library directory
of PHENIX (e.g., /usr/local/phenix/chem_data/mon_
lib/z/ZN.cif).
Acknowledgment
We thank Dr. Moran Grossman and Prof. Irit Sagi for providing
the plasmid containing human MT1-MMP and Yvonne
Brandenburger and Ingeborg Heise for technical assistance. This
work was supported by the Cluster of Excellence RESOLV (grant
No. EXC 1069) funded by the Deutsche Forschungsgemeinschaft
and the Max Planck Society.
References
1. Tallant C, Marrero A, Gomis-Ruth FX (2010) metalloprotease active site. Nat Struct Mol Biol
Matrix metalloproteinases: Fold and function 18:1102–1108
of their catalytic domains. Biochim Biophys 6. Bertini I et al (2012) The catalytic domain of
Acta Molecular Cell Research 1803:20–28 MMP-1 studied through tagged lanthanides.
2. Vandenbroucke RE, Libert C (2014) Is there Febs Lett 586:557–567
new hope for therapeutic matrix metallopro- 7. Pavlovsky AG et al (1999) X-ray structure of
teinase inhibition? Nat Rev Drug Discov human stromelysin catalytic domain com-
13:904–927 plexed with nonpeptide inhibitors: implications
3. Gialeli C, Theocharis AD, Karamanos NK for inhibitor selectivity. Protein Science
(2011) Roles of matrix metalloproteinases in 8:1455–1462
cancer progression and their pharmacological 8. Bertini I et al (2006) Snapshots of the reaction
targeting. FEBS J 278:16–27 mechanism of matrix metalloproteinases.
4. Decaneto E et al (2015) Pressure and tempera- Angew Chem Int Ed 45:7952–7955
ture effects on the activity and structure of the 9. Grossman M et al (2010) The intrinsic protein
catalytic domain of human MT1-MMP. Biophys flexibility of endogenous protease inhibitor
J 109:2371–2381 TIMP-1 controls its binding interface and
5. Grossman M et al (2011) Correlated structural affects its function. Biochemistry
kinetics and retarded solvent dynamics at the 49:6184–6192
10. Fernandez-Catalan C (1998) Crystal structure 16. Adams PD, Afonine PV, Bunkoczi G et al
of the complex formed by the membrane type (2010) PHENIX: a comprehensive Python-
1-matrix metalloproteinase with the tissue based system for macromolecular structure
inhibitor of metalloproteinases-2, the soluble solution. Acta Cryst Sec D 66:213–221
progelatinase A receptor. EMBO 17. Emsley P, Lohkamp B, Scott WG et al (2010)
J 18:5238–5248 Features and development of coot. Acta Cryst
11. Bertini I et al (2003) X-ray structures of binary Sec D 66:486–501
and ternary enzyme-product-inhibitor com- 18. Wilkins MR et al (1999) Protein identification
plexes of matrix metalloproteinases. Angew and analysis tools in the ExPASy server.
Chem Int Ed 42:2673–2676 Methods Mol Biol 112:531–552
12. Ogata H et al (2014) Crystalization and pre- 19. Decaneto E et al (2015) A caged substrate pep-
liminary X-ray crystallographic analysis of tide for matrix metalloproteinases. Photochem
the catalytic domain of membrane type 1 Photobiol Sci 14:300–307
matrix metalloproteinase. Acta Cryst F 20. Neumann U et al (2004) Characterization of
70:232–235 M c a - L y s - P r o - L e u - G l y - L e u - D p a - A l a -
13. Kabsch W (2010) XDS. Acta Cryst Arg-NH2, a fluorogenic substrate with
D66:125–132 increased specificity constants for collagenases
14. Winn MD et al (2011) Overview of the CCP4 and tumor necrosis factor converting enzyme.
sute and current developments. Acta Cryst Sec Analytical Biochemistry 328:166–173
D 67:235–242 21. Krug M, Weiss MS, Heinemann U, Mueller U
15. Vagin AA, Teplyakov A (2010) An approach to (2012) XDSAPP: a graphical user interface for
multi-copy search in molecular replacement. the convenient processing of diffraction data
Acta Cryst D 56:1622–1624 using XDS. J Appl Cryst 45:568–572
Chapter 5
Mapping Lipid Bilayer Recognition Sites

of Metalloproteinases and Other Prospective Peripheral
Membrane Proteins
Tara C. Marcink*, Rama K. Koppisetti*, Yan G. Fulcher*,
and Steven R. Van Doren*
Abstract
Peripheral binding of proteins to lipid bilayers is critical not only in intracellular signaling but also in metal-
loproteinase shedding of signaling proteins from cell surfaces. Assessment of how proteins recognize fluid
bilayers peripherally using crystallography or structure-based predictions has been important but incom-
plete. Assay of dynamic protein–bilayer interactions in solution has become feasible and reliable using
paramagnetic NMR and site-directed fluor labeling. Details of preparations and assay protocols for these
spectroscopic measurements of bilayer proximity or contact, respectively, are described.
Key words Protein–lipid interactions, Peripheral membrane protein, Metalloproteinase, NMR,

Paramagnetic relaxation enhancement, Environment-sensitive fluorophore, Structure-based
prediction
1 Introduction
Protein-mediated membrane-proximal events are critical in cell

biology. The best characterized peripheral membrane interactions
involve intracellular signaling proteins, and include many studies of
proteins that bind phosphoinositides or head groups of phospho-
lipids [1]. Fewer studies have examined proteins that bind periph-
erally to the cell surface and the extracellular leaflet of plasma
membranes. The matrix metalloproteinases (MMPs) have impor-
tant proteolytic targets on the cell surface such as receptors, growth
factors, growth factor binding proteins, and cell adhesion mole-
cules [2] while the ectodomains of some of these proteins may be
shed from the cell surface by MMPs [2, 3]. Related ADAM
*
These authors contributed equally to this work.
61
62 Tara C. Marcink et al.
proteases (a disintegrin and metalloproteinase) are prominent as

sheddases [3–5] while soluble MMP-2, -7, -8, -9, and -12 have
been observed in compartments at or near cell surfaces [6–12] and
may also be involved in cleavage of these cell surface proteins.
Paramagnetic NMR and fluorescence studies have revealed the
dual modes of peripheral binding of MMP-12 and -7 to model
membranes and cellular membranes [13, 14]. Comparison of four
methods for mapping the interfaces of these proteins with mem-
branes (and interfaces of other proteins under study), suggests the
following relative reliability and priority of the mapping methods:
paramagnetic NMR (Subheading 3.6) > site-directed fluor labeling
(Subheading 3.8) > bilayer-induced shifts or broadening of NMR
peaks (Subheading 3.6) ≥ predictions from structural coordinates
(Subheading 3.1). The approaches developed for studying bilayer
interactions for soluble MMP-7 and -12 [13, 14] should be appli-
cable for other soluble MMPs and potential peripheral membrane-
binding proteins. The methodology we propose for studying
peripheral membrane interactions involves the following steps.
1. Anticipate potential sites for peripheral binding to membranes

using recently reported predictive methods that use high-
resolution structural coordinates for the protein of interest.
2. Isotopically label the protein and assign NMR spectral peaks.
3. Prepare spin-labeled membrane bilayer mimics.
4. Localize and quantify paramagnetic NMR line broadening
resulting from proximity to the spin-labeled lipid.
5. Use site-directed fluor labeling (SDFL) to verify the NMR-
mapped binding sites for recognition of (A) liposomes and (B)
cells, as well as (C) effects of lipid composition.
The cornerstone of this strategy for accurately determining

membrane binding sites and orientation is to measure strongly
distance-dependent NMR line broadening emanating from mobile
electron spin-labeled phospholipids placed in the membrane [13–
16]. Our experience in mapping protein–protein and other pro-
tein–macromolecular interfaces is that paramagnetic NMR line
broadening is superior in accuracy and interpretability to more
conventional mapping of shifted NMR peaks onto the structure of
the protein [17–23]. Paramagnetic relaxation enhancements
(PREs) provide insightful assessment of dynamic, transient and
light occupation of binding modes [24–28], including with bilayer-
like partners [13, 14]. SDFL can be used to confirm modes of
binding determined by NMR methods [13, 14]. Alternatively,
other investigators have utilized site-directed mutagenesis to con-
firm these NMR-derived phospholipid binding interfaces [29].
Two types of disk-like mimics of lipid bilayers are promising
for NMR studies in solution: (a) a class of small bicelles and (b)
Mapping Peripheral Membrane Binding by Matrix Metalloproteases 63
nanodiscs. Bilayered micelles or bicelles form disks containing an

interior long-chain phospholipid bilayer circumscribed by an annu-
lus of short-chain phospholipids [30]. Bicelles are well-suited to
solution NMR studies of membrane-associated proteins [31–34].
The bicelles are discoidal, tumble rapidly and isotropically, and
have a low critical micelle concentration (CMC) [33].
Nanodiscs are disks of lipid bilayers encircled by two belts of
molecular scaffolding proteins, which define their size and planar-
ity [35]. There are several reasons why nanodiscs are well suited for
biophysical studies of membrane proteins. These include: (1) a
diameter near 10 nm that is large enough to accommodate two
interacting proteins, (2) homogeneity, (3) long-term stability, and
(4) high fluidity at the high concentrations needed for NMR
experiments [36]. A marked disadvantage for solution NMR of
proteins embedded in nanodiscs is their high molecular weight,
resulting in slow tumbling, rapid decay of NMR signals, and
spectra with broad peaks [37–40]. This led to efforts to decrease
their size by engineering shorter molecular scaffolding proteins
which has in turn resulted in enhanced NMR spectra [40]. The
peripheral binding of proteins to nanodiscs is transient or dynamic
so their NMR spectra are usually less affected by slow tumbling
associated with embedded membrane proteins. The solubility limit
of nanodiscs is ~500 μM which constrains mixtures and titrations
with proteins binding them peripherally to concentrations that are
comparatively low for NMR studies. To summarize, nanodiscs pro-
vide a stabilizing environment for membrane proteins but this
comes at the cost of weaker NMR signal intensity due to the slow
tumbling of the protein/nanodisc complex.
2 Materials
2.1 Protein 1. Beckman ultracentrifuge, and an ultracentrifuge rotor such as

Expression a Ti 70.
and Purification 2. Bath sonicator (Laboratory Supplies Co., Hicksville, NY).
3. Sorvall or other centrifuge.
4. Cell homogenizer such as a motor-driven Potter-Elvehjem tis-
sue grinder.
5. Bacterial cell disruptor such as a French press.
6. Low to medium-pressure chromatography system.
7. Column for gel permeation chromatography such as Superdex
200 10/300 GL.
8. Ni-NTA column.
9. Trace minerals [41] (5,000× stock).
10. MEM vitamin solution (100× stock; HyClone).
11. Minimal growth medium: 50 mM Na2HPO4, 50 mM KH2PO4,

5 mM Na2SO4, 2 mM MgSO4, 50 mM NH4HCl, 0.5% (w/v)
glucose, and 0.2× trace metals (see Note 1).
12. Isopropyl β-d-1-thiogalactopyranoside (IPTG,
Sigma-Aldrich).
13. Lysis buffer: 20 mM Tris–HCl, 10 mM EDTA, pH 7.5.
14. 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC)
(Avanti Polar Lipids).
15. 1,2-diheptanoyl-sn-glycero-3-phosphocholine (D7PC)
(Avanti Polar Lipids).
16. Dipalmitoylphosphatidylcholine (DPPC) variants with doxyl
spin-label at 5, 10, or 14 position in one acyl chain (i.e., 1-pal
mitoyl-2-stearoyl-(X-doxyl)-sn-glycero-3-phosphocholine,
where X is 5, 10, or 14) (Sigma-Aldrich). Store lipids at 20 °C
in a tightly sealed container.
17. Chloroform (ACS-grade) (Sigma-Aldrich).
18. Sodium cholate (Sigma-Aldrich).
19. Bio-Beads SM2 (BioRad).
20. NMR buffer for MMPs: 20 mM imidazole (pH 6.6), 10 mM
CaCl2, 20 μM ZnCl2, 0.02% NaN3, and 5% 2H2O.
21. Molecular scaffolding protein MSP1D1 (Sigma-Aldrich). See
ref. 42 and Subheading 3.5, step 1.
22. MSP buffer: 20 mM Tris–HCl (pH 7.4), 0.1 M NaCl, 0.5
mM EDTA, 0.01% NaN3.
23. TEV protease (Sigma-Aldrich).
24. Tris buffer: 20 mM Tris–HCl (pH 7.2).
25. TNC buffer: 20 mM Tris–HCl (pH 7.4), 150 mM NaCl, 10
mM CaCl2.
2.2 Fluorescence 1. Freshly prepared 10 mM N,N′-dimethyl-N-(iodoacetyl)-N′-

Spectroscopy (7-nitrobenz-2-oxa-1,3-diazol-4-yl) ethylenediamine
(IANBD) (ThermoFisher).
2. Dimethylsulfoxide (DMSO; Sigma-Aldrich).
3. G-25 desalting resin (Sigma-Aldrich).
4. Dulbecco’s modified Eagle’s medium (DMEM) with 10% fetal
bovine serum, supplemented with l-glutamine and nonessen-
tial amino acids (NEAA).
5. Cell lines (ATCC).
6. Biotek Synergy MX plate reader (see Note 2).
7. Zeiss LSM 510 Meta confocal microscope for imaging live
cells [13, 14].
2.3 NMR 1. NMR spectrometer operating at a 1H frequency of 600 MHz

Spectroscopy or higher (see Note 3).
2. 5 mm NMR tubes (Norell) (see Note 4).
3 Methods
3.1 Navigating Two excellent servers are available to predict the atomic structural
Servers for Predicting coordinates of proteins that peripherally bind to membranes:
Protein Binding Membrane Optimal Docking Area (MODA) [43] and Positioning
Peripherally of Proteins in Membranes (PPM) [44]. MODA is designed for
to Membranes predicting membrane binding sites of peripheral membrane pro-
teins [43]. PPM is designed for predicting the positioning of both
peripheral and integral membrane proteins with respect to bilayers
[44]. PPM was preceded by the well-developed OPM database of
predictions of positions and orientations of proteins in and on
membranes [44–47]. The predictions may be used to determine
whether it is worthwhile undertaking experimental testing of
membrane interactions. For example, a comparison of a PPM pre-
diction for the catalytic domain of MMP-12 with results of experi-
mental docking studies using paramagnetic NMR [13] is shown in
Fig. 1. The PPM prediction failed to predict the α-interface (Fig.
1a). However, it did predict one of the three experimentally deter-
mined loops (the II–III loop) of the β-interface [13]. This predic-
tion bears resemblance to the measured orientation, but approaches
the bilayer at a different angle (Fig. 1b). While MODA predicted
residues for the other two experimentally defined loops (III–IV
and IV–V loops) within the β-interface [13].
3.1.1 MODA and PPM 1. First obtain the PDB file containing the structural coordinates
Predictions for the protein of interest from the Protein Data Bank (www.
rcsb.org).
2. MODA predictions (http://molsoft.com/~eugene/moda/
modamain.cgi).
(a) Enter the PDB accession code or upload the PDB coordi-
nate file. If the PDB file contains multiple chains, indicate
which chain should be used for the prediction, e.g., A, B,
C. Select “Predict” and wait for the results (see Note 5).
(b)
After the server identifies potential sites of membrane
binding, it outputs a tabulated list of MODA scores for
each of the amino acid residues in the sequence. The plain
MODA score is scaled down by curvIndex to the more
conservative curvMODA score.
(c) If the curvMODA score for a residue is above 40 [48] and
it has one or more neighbors (in sequence or space) with a
high curvMODA score, that patch of residues is likely to
Fig. 1 Comparison of the PPM predicted and experimentally determined mode of bilayer binding by MMP-12.
The protein backbone shown is used both to illustrate the experimental dockings and the predicted docking.
The lipid chains plotted are from the experimental docking. The blue grid indicates the predicted location of
bilayer head groups where the bilayer contacts the II–III loop of MMP-12. (a) The experimental structural model
bound to the membrane via the α-interface (PDB: 2MLR, with lipid chains) is superposed with predicted inter-
faced marked by the blue grid. Molecules comprising the bilayer are shown and phosphorous atoms of the
DMPC head groups are illustrated as orange spheres. The protein chain is rainbow colored from blue at the
N-terminus to red at the C-terminus. (b) The experimental structural model bound to the membrane via the
β-interface (PDB: 2MLS, with lipid chains) which includes not only the II–III loop but also the III–IV and IV–V
loops, resulting in a different tilt angle than predicted by PPM (blue grid)
interact with membrane bilayers. Residues with curv-

MODA score between 20 and 40 that cluster with other
high-scoring residues possibly also interact with lipid
bilayers.
(d) Download the spreadsheet readable text file in .CSV for-
mat. One should either manually mark high scoring resi-
dues on the protein structure with a molecular graphics
software such as Pymol. Alternately, residues predicted to
interact with membranes can be semi-automatically
denoted by opening the ICB file in the program MolSoft.
3. PPM predictions (http://opm.phar.umich.edu/server.php).
(e) Upload the PDB file of interest.
(f) Specify the topology of the protein (N-terminus being in
or out of the membrane) and whether you wish to include
nonstandard residues or atoms in the prediction. Submit
and wait for the results to appear on a fresh webpage.
(g) The calculated predictions include the depth that the pro-
tein penetrates into the membrane, the water-to-mem-
brane transfer free energy (ΔGtrans in kcal/mol), and the
tilt angle (°) of the protein in the membrane.
(h) To display the predicted model of the protein–membrane
interaction select the “jmol” link under the heading
“Image of the protein in membrane”. This model is best
viewed using a web browser that supports Java applets,

such as Firefox.
(i) Download the PDB file generated for the predicted model
to examine the hypothetical protein–membrane interface
in a molecular graphics software package like PyMol or
Chimera. The projected location of the membrane bilayer
in the predicted protein–membrane model is defined by a
grid of dummy atoms.
3.2 Preparation 1. Prepare at least 100 ml of unlabeled PG medium [41] and at

of E. coli Inclusion least 250 ml of PG medium using 99% D2O in lieu of H2O, as
Bodies Harboring well as 15NH4Cl (Sigma-Isotec or Cambridge Isotope
Recombinant Laboratories) as the sole nitrogen source (see Note 6).
Isotopically Labeled 2. Transform competent E. coli BL21(DE3) Gold cells with the
MMP (or Other expression plasmid and plates 20–50 μl of the mixture onto a
Eukaryotic Protein) PG medium agar plate containing an appropriate antibiotic.
for NMR Studies 3. After incubating at 37 °C overnight, pick a few colonies from
the plate and add to 1 ml of PG medium containing the anti-
biotic and incubate for 8 h at 37 °C on a shaker operating at
250 rpm.
4. Transfer the entire 1 ml of culture into 50 ml of PG medium
containing antibiotics in a 250 ml flask. Incubate overnight at
37 °C in the shaker at 250 rpm.
5. Subculture 1–2% of the overnight culture into 250 ml of
labeled PG medium containing antibiotic in a 2-l shake flask.
6. When A600 reaches 0.4–0.8, add 0.5 mM IPTG to induce pro-
tein expression and incubated overnight at 37 °C in a shaking
incubator at 250 rpm.
7. Collect the bacterial cell pellet by centrifugation and store at
−20 °C until needed.
8. Resuspend the cell pellet with lysis buffer and homogenize the
suspension with a motor-driven Potter-Elvehjem tissue
grinder.
9. Rupture the cell suspension, preferably by two slow passes
through a French pressure cell, until the milky suspension
darkens due to breakage of the cells.
10. Centrifuge at 20,000 × g for 40 min at 4 °C. Discard the
supernatant as the MMP construct should be in the pellet in
inclusion bodies. This pellet may optionally be washed by
resuspension in lysis buffer containing 0.01% Triton X-100
and centrifuging again.
11. The inclusion bodies are dissolved in concentrated urea or
guanidinium∙HCl, centrifuged to remove insoluble debris,
refolded, chromatographically purified, and concentrated in a
centrifugal filter unit (Millipore). Optimal procedures for the
purification and refolding of the protein are specific to the con-

struct and have been described for MMP-1 [49, 50], MMP-3
[51], MMP-7 [52], MMP-12 [21, 53, 54], MMP-13 [55],
and MMP-14 [56, 57].
3.3 Assignment 1. Initially check to determine whether the assignments for the
of NMR Spectral Peaks backbone amide NMR peaks are not available in the literature
or in the BioMagResBank database (http://www.bmrb.wisc.
edu/).
2. If de novo NMR peak assignments are required, prepare the
15
N/13C labeled protein at a concentration of at least 150 μM
using the procedure outlined in Subheading 3.2 (see Notes 7
and 8).
3. Acquire standard NMR spectra for spectral peak assignments
(including HNCA, HN(CO)CA, CBCA(CO)NH and
HNCACB triple resonance spectra) (see Note 9).
4. Semi-automated assignments for a majority of the backbone
resonance peaks can be readily determined by non-specialists
using the program PINE, a downloadable, integrated software
environment which comes with a set of explanatory tutorials
[58]. The software relies heavily upon NMR spectral analysis
software NMRFAM-Sparky for visualizing the spectra [59]
and PONDEROSA for automation of peak assignments [60,
61].
5. Assignments of methyl peaks for hydrophobic side chains are
optional, but are desirable in order to provide additional dis-
tance constraints between protein residues and atoms within
the lipid bilayer. This typically involves selective “ILV” labeling
of isoleucine, leucine, and valine residues in a perdeuterated
medium supplemented with α-ketoacids containing 1H/13C–
labeled methyl group(s) and optional deuteration [62] (see
Note 10).
6. Methyl peaks can be assigned for smaller proteins using
HMCM[CG]CBCA (HMCACB) [63] and CC(CO)NH or
HNCACB triple resonance spectra, supplemented with
NOESY-derived NOEs to the backbone [64] and methyl-
amides [56]. This strategy works well for single domain pro-
teins. To extend assignments of methyl peaks to multiple
domain protein constructs such as MMPs, the assignments of
methyl peaks from the single domains can simply be combined
with NOEs from methyl groups in the larger construct [64,
65]. Software designed to aid in the determination of methyl
peak assignments that utilizes additional experimental data
such as structural coordinates, methyl-methyl NOEs, and
PREs is also available [66].
3.4 Preparation Bilayered micelles or bicelles that are small, discoidal, and isotropic
of Small Bicelles in tumbling are preferred for solution NMR [30, 32, 33, 67].
for NMR Studies These are composed of 2–3 equivalents of detergent-like short
chain lipids per equivalent of long chain phospholipid, which are
designated q = 0.5 to q = 0.33 (ratio of long to short chains),
respectively. Short chain lipids are amendable to forming curved
surfaces, in this case, the rims of the bicelle disks [30] (Fig. 2b).
Long chain phospholipids, such as DMPC, form the planar bilay-
ers in the interior of the disks (Fig. 2b). Dihexanoylphosphatidylcholine
(D6PC) with its 6-carbon acyl chains has dominated biophysical
studies [30] but suffers the limitation of not aggregating into
micelles or bicelles until reaching a critical micelle concentration
(CMC) as high 14 mM in monomers [33]. D7PC with its 7-carbon
acyl chains has the much lower and more favorable CMC of 1.2
mM allowing D7PC-DMPC bicelles (q ≤ 0.5) to form at much
lower lipid concentrations with only 1.2 mM monomers in solu-
tion [33]. Preparation of bicelles incorporating the latter and using
our recommended formulation is outlined below.
1. Weigh 20 mg of the powder form of D7PC into a clean 5 ml
glass tube and dissolve in 1.5 ml of chloroform.
2. Blow a gentle stream of argon or nitrogen gas into the glass
tube for 30–45 min to evaporate the chloroform. Disturbance
Fig. 2 Phosphatidylcholine molecules with 7- and 14-carbon acyl chains (a) form the annulus and center,
respectively, of disk-like bicelles (b)
of the liquid by the nitrogen (or argon) stream should be slight

and barely visible, with no splashing of liquid. Upon drying, a
solid lipid film will form at the bottom of the glass tube. Dried
lipids should be white in appearance. If the film appears glassy,
it means that some residual chloroform remains in the lipid
film.
3. Cover the glass tube with a piece of parafilm and pierce this
several times. Place the covered tube in a jar containing desic-
cant or under vacuum for drying overnight.
4. To the dried D7PC, add 35 μl of NMR buffer. Cover with
parafilm and hydrate at 42 °C for 1 h. Frequently and gently
centrifuge using either a hand-cranked centrifuge or a
microfuge on its slowest setting in order to collect the micelles
from the sides of the tube.
5. Dissolve 20 mg of DMPC in 2 ml of Tris buffer. Hydrate at 42
°C for 2 h. Vortex occasionally to ensure an even suspension.
6. Place the glass tube in a bath sonicator. Adjust the glass tube
so that the water barely covers the liposome dispersion.
7. Sonicate for 10 min (see Note 11). The solution will turn from
milky white to translucent (see Note 12). Incubate at 42 °C
for 1 h.
8. Ultracentrifuge the DMPC liposomes in a Ti 70 rotor for 1 h
at ~64,000 × g and 20 °C. Decant the supernatant which con-
tains the small unilamellar vesicles (SUVs), freeze and store at
−80 °C for later use in fluorescence assays. The nearly white
flocculent material at the bottom of the centrifuge tube con-
tains large unilamellar vesicles (LUVs) and medium-sized
unilamellar vesicles (MUVs) which will be used for preparing
bicelles in the next step.
9. Resuspend the DMPC liposomes (LUVs and MUVs) with the
solution containing the D7PC micelles. Incubate at 42 °C
until an even suspension is achieved. Be careful to minimize
frothing.
10. Plunge the suspension into liquid nitrogen and freeze. Thaw
at 42 °C.
11. Transfer to a 500 μl microfuge tube and bring the volume up
to 100 μl with NMR buffer. Freeze in liquid nitrogen again
and thaw at 42 °C.
12. Transfer 25 μl aliquots to a fresh tube and freeze in liquid
nitrogen. Store at −80 °C until needed.
3.5 Preparation This is a synopsis of the previously reported protocol for preparing
of Nanodiscs for NMR nanodiscs [42, 68, 69]. This protocol uses the MSP1D1 variant of
or Fluorescence the molecular scaffolding protein to encircle a DMPC phospho-
Assays lipid disc. Other lipids and MSP variants may also be assembled
into nanodiscs.
1. MSP1D1 [42] may be purchased from Sigma-Aldrich or the

His-tagged protein can be expressed from a plasmid (Addgene)
and purified by affinity chromatography on Ni-NTA agarose in
a buffer containing sodium cholate [68].
2. Pool fractions containing the His-tagged MSP1D1 and dialysis
against MSP buffer.
3. Treat with TEV protease to release the His-tag. Remove the
peptide containing the His tag by capturing it on the Ni-NTA
column. Measure the concentration of the cleaved protein
using a molar extinction coefficient (ε280) of 18,200 M−1 cm−1
[69].
4. Dissolve sodium cholate in MSP1D1 or NMR buffer to make
a solution that is 20–40 mM in sodium cholate monomers (see
Note 13). Add this solution to 40 mg of dry DMPC at a molar
ratio of two detergent molecules per phospholipid molecule.
Vortex and heat in a water bath at 60 °C for 1 h. Then sonicate
in a bath sonicator until clear.
5. Add the MSP1D1 to the cholate-solublized DMPC solution
to achieve a final ratio of 1:80, while keeping the concentration
of sodium cholate above 20 mM. Then mix (see Note 14).
6. Incubate with 0.5–0.8 g of BioBeads SM2 per ml of sample for
2–3 h to remove the detergent micelles.
7. Wash the BioBeads with an excess of MSP or NMR buffer in
order to release more nanodiscs.
8. If turbidity or precipitation is present, pellet the particles by
briefly spinning in a microcentrifuge. Inject the clarified solution
of nanodiscs onto a gel filtration column such as Superdex 200
10/300 GL and run with a flow rate of ~0.5 ml/min. Collect
1 ml fractions. The elution profile for a preparation of nanodiscs
composed of DMPC and MSP1D1 superimposed with that for
several molecular weight standards is shown in Fig. 3.
3.6 NMR to Measure The main idea here is to estimate distances from protein amide and
Proximity to Mimics methyl groups to doxyl spin labeled DPPC, with around one or
of Membrane Bilayers two such DPPC added per leaflet of the bilayer mimetic. PRE
(paramagnetic relaxation enhancement) measurements from the
unpaired electron of the doxyl group to the amide or methyl pro-
ton of the protein depends strongly (α r−6) on the distance r
between them and less strongly on the time constant τc of their
rotational diffusion. The PREs are measured from the increased
paramagnetic relaxation rate enhancement Γ2 resulting from the
addition of an unpaired electron.
1. Prepare a 15N-2H labeled protein sample according to
Subheading 3.1.1 and concentrate to between 200 and 600
μM in NMR buffer in a volume of at least 400 μl, typically
using a centrifugal membrane concentrator (Millipore or
Vivaspin).
Fig. 3 Gel permeation chromatography of the assembled nanodiscs comprised of

DMPC and MSP1D1. The elution profiles for several hydrodynamic standards are
superimposed onto the chromatogram for the nanodiscs. The nanodiscs migrated
with an apparent molecular weight that was less than that of γ-globulin (158
kDa) on a Superdex 200 10/300 GL column. The major peaks comprised uni-
formly assembled nanodiscs while a significantly smaller peak corresponding to
incomplete assemblies was also observed
2. Titrate the protein sample with the bicelles, nanodiscs, or lipo-

somes. Record 1H-15 N TROSY NMR spectra of the protein
amide spectral region at each lipid concentration in the titra-
tion. If protein methyl groups are specifically labeled, also col-
lect 13C HMQC spectra of the methyl spectral region at each
concentration. Titrate the membrane mimic to at least half of
the molar equivalents of the discoidal bilayer assembly per
equivalent of protein, in order to provide at least one discoidal
leaflet as a potential binding site for every protein molecule
(expecting instances of two protein molecules per disc).
3. Continue the titration to a ratio of 1:1 and beyond, in order
to maximize the number of assemblies harboring one protein
bound per disc-like bilayer (see Notes 15 and 16).
4. Estimate the rotational correlation time τc to confirm bilayer-
protein association and calibrate distance estimates. 15N NMR
relaxation indicative of τc is measured, preferably at each con-
centration point in the titration of Subheading 3.4, step 1
(see Note 17) Acquire a 1D relaxation series with at least five

increments of constant-time relaxation using an appropriate
pulse sequence for measuring 15N NMR transverse dipole–
dipole/CSA cross-correlated relaxation [70]. Set the longest
relaxation period to attenuate peak heights to around 35–50%
of the initial intensity.
5. Integrate the 1D spectral envelope at each relaxation time,
while omitting the central region from 7.6 to 8.4 ppm which
is enriched in signal from unstructured loops. An exponential
relaxation rate constant is fitted to these integrals. This con-
stant is known as the transverse cross-correlated relaxation
rate, or ηxy.
6. Use ηxy to estimate τc. The relationship between ηxy and τc is
given by the following equations of ref. 71:
hxy = pd N (4 J (0) + 3 J (wN ))(3 cos 2 q - 1) (1)

where p is the dipole–dipole coupling between the 1H and 15N of

the amide group.
p = m0g H g Nh / (16p 2 2rHN

3
) (2)
δN is the chemical shift anisotropy of the 15N nucleus defined as
d N = g N B0 Dd N / (3 2) (3)
where γH are γN are the respective 1H and 15N gyromagnetic ratios,

h is Planck’s constant, rHN is the distance between amide 1H and
15
N nuclei, ∆δN is the difference between the two main compo-
nents of the 15N chemical shift tensor, and J(ω) is the spectral
density function that depends on frequency ω.
J (w) = 0.4t C / [1 + (wt C )2 ] (4)
These equations, the constants, and the 15N resonance frequency

of an 800 MHz spectrometer define the relationship of ηxy and
τc as:
hxy = 9.075e8 ´ 0.4 ´ t c ´ (4 + 3 / (1 + t c ´ 5.0948e8)2 ) (5)

Using Eq. (5), ηxy is plotted versus τc and is almost linear. At the
measured ηxy rate, the τc value corresponding to this point on
the line is read from this plot. This estimate of τc represents the
hydrodynamics. It can be monitored throughout the titration.
An increase in τc during the titration indicates the binding of
protein to the membrane and slower tumbling of the protein
due to its association with the membrane.
7. Measure the 1H NMR transverse (R2) relaxation rate constants

of the diamagnetic state of the lipid–protein mixture. (These
exponential rate constants will be used as the reference values
to which the distance-dependent Γ2 values increase relaxation.)
Use an NMR pulse sequence modified to include a PROJECT-
CPMG train that suppresses proton-proton J-couplings [72].
For amide 1H relaxation, either a 15N TROSY sequence with a
prepended PROJECT-CPMG train or an 15N HSQC with the
PROJECT-CPMG train appended are recommended [13, 14,
56]. For methyl 1H relaxation, use a 13C HMQC incorporat-
ing PROJECT-CPMG [13, 56]. Set the 1H transmitter fre-
quency offset to the water resonance peak near 4.7 ppm.
Determine and set the 1H 90° pulse length at high power.
Attenuate the 1H power level for the PROJECT-CPMG train
by 3 dB and set its constituent 90° pulse widths 2 -fold lon-
ger than the high power pulse width. Acquire trial 1D spectra
as a series of 4 ms steps for the length of the CPMG train in
order to monitor the degree of exponential reduction in over-
all peak height. Use this spectral series to determine the CPMG
periods (multiples of 4 ms) to use for acquiring a series of 2D
spectra. For relaxation of amide peaks, try CPMG periods
from 0 to 32 ms. For methyl relaxation, try CPMG periods
out to 40 or 48 ms. Choose a maximum CPMG period in
which the spectrum has decayed to 35–50% of the intensity
with a CPMG period of 0 ms.
8. Copy the optimized parameters into a number of 2D experi-
ments with increasing the value of the CPMG period by a
multiple of 4 ms for each successive experiment. Increase the
number of transients acquired to achieve a S/N ≥ 20 for the
spectrum with the longest CPMG period. Measure the 1H
NMR transverse relaxation with a series of five or six relaxation
delays [13, 14, 56].
9. To prepare the paramagnetic lipid DPPC (modified with the
doxyl spin label at the 5, 10, or 14 position) dissolve 1 mg of
doxyl-substituted DPPC in 250 μl of methanol to make a
stock solution of 4.6 mM. Store at 4 °C.
10. Estimate the volume of this solution needed to incorporate an
average of one (possibly two) doxyl-DPPC molecules per leaf-
let of the bicelles or nanodiscs. Transfer this volume of solution
to a fresh tube and evaporate using a gentle stream of nitrogen
or argon gas. Add the sample of labeled protein with bicelles
or nanodiscs to the dried doxyl-DPPC and incubate for
10–15 min to incorporate the spin label in the bilayers (see
Note 18). The incorporation of the doxyl-DPPC introduces
its unpaired electron to the membrane bilayer. This broadens
the 1H NMR peaks of nearby hydrogen atoms. The breadth of
the NMR peak is measured in the time domain as relaxation
rate constant R2, which decays exponentially with the length

of the CPMG time period used in the NMR pulse sequence.
The increase in 1H R2 relaxation due to the presence of the
unpaired electron is Γ2 and results in the distance-dependent
PRE. Γ2 values may be measureable on protons on the protein
as far as 25 Å from the doxyl group.
11. Repeat the series of CPMG experiments for the paramagnetic
doxyl-DPPC incorporated membrane. Use the same buffer,
temperature and membrane mimic to protein ratio as used for
the diamagnetic sample (see Notes 19 and 20). Fit an exponen-
tial decay function and rate constant of decay to each resolved
and confidently assigned peak. The exponential decay rate in
the paramagnetic sample minus the exponential decay rate of
the diamagnetic sample is equal to Γ2, the rate constant of the
paramagnetic relaxation for a given NMR resonance peak.
12. These experiments should be repeated using other DPPC ana-
logues, where the doxyl label is incorporated at different posi-
tions along the acyl chain, in order to sample varying depths
and magnitudes of PREs (Fig. 4). This provides additional
information about the depth (or lack of) penetration of the
protein into the membrane. These experiments provide addi-
tional distance constraints that can be used for structural mod-
eling of the peripheral membrane binding.
13. Estimate the distance r from the mobile doxyl spin label to
each amide or methyl resonance peak using the expression:
r = 4K t c / G 2 (6)
Fig. 4 Options for the location of doxyl substitution on phosphatidylcholine (PC)

inserted into the membrane bilayer. Doxyl substitution at the 5-position is the
closest choice in the acyl chain to the head group. It introduces more numerous
PREs to the protein interacting peripherally. Doxyl substitution at the 14-position
deep in the bilayer results in fewer PREs in the protein, but which are more likely
to be central to the interface. The blue ribbon depicts MMP-12. Interfacial resi-
dues are plotted by ball-and-stick
Here K = 1.23 × 10−44 m6 s−2 and τc is the rotational correla-

tion time. These estimates can then be used as distance constraints
in rigid body structural calculations using a docking program, such
as HADDOCK [13, 29], to define areas of interaction between the
protein and the membrane mimic, and in subsequent restrained
molecular dynamics calculations. Computational details of the
docking calculations are complex and beyond this chapter’s scope,
but have been introduced previously [13, 14].
3.7 Preparation Fluorescence assays are recommended to confirm modes of bilayer

of Liposomes binding determined by the paramagnetic NMR measurements
for Fluorescence described in Subheading 3.6 [13, 14]. These fluorescence assays can
Assays be undertaken using classic liposomes. Small unilamellar vesicles
(SUVs) are recommended over larger vesicles in order to decrease
light scattering that adds unwanted background to the detected flu-
orescence signal. Overall, the protocol outlined below involves pre-
paring a stock suspension of SUVs of DMPC that is 10 mM in lipid
monomers. SUVs may be also prepared from other lipids.
1. Dissolve 3.4 mg of DMPC (MW 677.94) in 500 μl of assay

buffer (see Note 21).
2. Hydrate the lipids by placing the solution in a 42 °C water
bath for 2 h. Vortex every 30 min in order to suspend the
lipids.
3. Sonicate the suspension until it is clear. This should take
10–15 min. If it does not become clear or if the concentration
of lipid is low, use freeze–thaw cycles in a plastic tube with
liquid nitrogen as described for preparing bicelles in Subheading
3.4. That is, incubate the plastic tube for 3 min in liquid nitro-
gen followed by 6 min in a 42 °C water bath. Remove MUVs
by spinning for 1–2 min in a microcentrifuge at 9300 × g. The
SUVs can be stored in the 42 °C water bath. It is best to use
them within one to 2 days of preparation, but they may be use-
able for up to a week. SUVs stored at 4 °C will form MUVs or
aggregates (see Notes 22 and 23).
3.8 Site-Directed A fluorophore that exhibits enhanced fluorescence upon insertion

Fluor Labeling (SDFL) into lipid bilayers has proven useful to experimentally determine
to Interrogate Binding the site and orientation of interaction of peripheral proteins with
to Liposomes membrane bilayers of both liposomes and cells [13, 14, 73, 74]
(Fig. 5). We recommend using site-directed fluor labeling (SDFL)
3.8.1 Preparation to confirm peripheral membrane binding sites determined by para-
of Fluorescently Labeled magnetic NMR. Alternately, SDFL can be utilized as the primary
Protein method for mapping the membrane–protein binding interface
when it is not feasible to use paramagnetic NMR. The general
approach for using SDFL to define the interaction of peripheral
binding protein with membrane bilayers is outlined below:
1. To confirm membrane-binding sites identified by paramag-

netic NMR, a number of single cysteine residue protein
mutants need to be prepared containing cysteine residues
located at positions corresponding to the interfacial binding
site and one or more distant locations to serve as negative
control(s) (Fig. 5) (see Notes 24 and 25). The substituted cys-
teine residues should be placed at a surface-accessible site, typi-
cally in a loop, inside or outside the proposed interface (see
Note 26).
2. Reduce the single reactive Cys residue of the mutated pro-
tein (typically ≥1 μM) with at least a tenfold molar excess
of thiol reductant, e.g., 1 mM β-mercaptoethanol (BME)
(see Note 27).
3. Dialyze against TNC buffer to remove excess reductant (see
Note 28).
4. To a solution containing 1 μM protein add a tenfold molar
excess of IANBD (see Note 29 and Fig. 6).
5. Let the conjugation reaction proceed for 2 h at room tempera-
ture or overnight at 4 °C. Perform this in an anaerobic cham-
ber under argon.
6. Add excess BME or glutathione to consume excess thiol-
reactive IANBD.
Fig. 5 The site-directed fluor labeling (SDFL) approach for defining peripheral membrane-binding sites for a
protein. When the IANBD (red) conjugated to the protein (blue) inserts in a hydrophobic compartment such as
the bilayer, its fluorescence emission increases
7. Concentrate the reaction mixture before loading onto a G-25

desalting column equilibrated in 10 column volumes of Tris
buffer containing between 5 and 10 mM CaCl2.
8. Collect the eluted conjugated protein and determine the extent
of the conjugation reaction as follows. Determine the protein
concentration using a protein assay such as a Bradford assay or
absorbance at 280 nm. Determine the concentration of
IANBD groups by measuring the maximum absorbance
between 472 and 480 nm and using an extinction coefficient
of 23,700 M−1 cm−1 [75]. The ratio of the concentration of
protein to that of the fluorphore (IANBD) equals the apparent
percentage labeling, which is typically ~85–105%. The excess
over 100% could suggest incomplete removal of free IANBD.
3.8.2 Fluorescent 1. In a 96-well plate prepare a series of 200 μl solutions (in qua-
Binding Assay druplicate) containing 10–100 nM fluorophore-conjugated
protein. Measure fluorescence at 541 nm using an excitation
wavelength of 478 nm to establish sensitivity and fluorescence
linearity (see Note 30).
2. Each 200 μl solution should contain 250–300 μM liposome
monomers, typically prepared by diluting 5–6 μl from a 10
mM stock solution. Measure fluorescence intensity every
10–30 min for 2–3 h in order to allow the protein time to
equilibrate (i.e., insert into the liposomes), before photo-
bleaching takes a toll on emission from the IANBD. Normalize
the fluorescence intensity in the presence of the liposomes
(FSUV) by dividing it by the fluorescence in the absence of
membranes (F0), (i.e., F0/FSUV). Bilayer insertion of the
NBD moiety is accompanied by an increase in fluorescence
intensity with FSUV/F0 ≥ 1.5 [13, 73] (see Note 31).
3.8.3 Confocal Imaging 1. The IANBD-tagged protein may also be used to investigate
of Binding to Live Cells interactions with cell membranes and their compartments.
We have used human HeLa cells or Raw264.7 murine macro-
phages grown in suspension [13, 14], (Macrophages secrete
Fig. 6 Conjugation of the iodoacetamide-containing, environment-sensitive NBD fluorochrome (IANBD) to a

single surface-exposed, reactive cysteine thiol
several MMPs). Like the assays with liposomes, the degree

of association with cell membranes is quantified using the
plate reader measurements of the fluorescence emission with
cells present (Fcells) divided by the emission in their absence
(i.e., Fcells/F0). Use of three to five million cells per well has
been effective in our experiments [13].
2. The mammalian cell line is seeded on sterile MatTek glass bot-
tom culture dishes and grown overnight at 37 °C in a 5% CO2
incubator until reaching about 60–80% confluency.
3. The cells are washed three times with DMEM without phenol
red and incubated with media containing a fluorescent staining
reagents specific for the membrane compartment of interest
and nuclei for 10 min at 37 °C in a 5% CO2 incubator.
4. The cells are then washed three times with 20 mM Tris, pH
7.2, 5 mM CaCl2, 0.1 mM ZnCl.
5. The culture dish is then placed on the stage of a confocal
microscope (in our case a Leica TCP SP8 MP equipped with
405 nm and tunable white light lasers). The excitation/emis-
sion bandpass wavelengths used to detect IANBD, Hoechst
33,342, and Alexa Fluor 594 WGA are set to 472/485–525,
405/415–470, and 594/610–680 nm, respectively.
6. Confocal fluorescence images are taken immediately (control
images). IANBD-labeled protein is then added to the dish and
images are collected every 90 s for 15–30 min.
4 Notes
1. This minimal growth medium for high-level expression in E.

coli is Studier’s non-inducing PG medium (P-0.5G) [41], with
the addition of vitamins and use of an alternative choice of
ammonium salts.
2. The authors used a Biotek Synergy MX plate reader, an instru-
ment with the advantage of monochromator selection of exci-
tation and detections wavelengths 478 and 541 nm,
respectively. With a number of other plate readers, an
appropriate set of filters are used instead on the excitation and
emission channels.
3. A cryogenic probe is highly recommended for the sensitivity
needed given the dilution and line broadening accompanying
association with a slowly tumbling bilayer disc, and the quan-
titative nature of measurements of paramagnetic NMR relax-
ation. The authors have used a Bruker Avance III 800 MHz
NMR spectrometer with TCI cryoprobe for proteins of 20
kDa and greater, and a Bruker Avance HD 600 MHz system

with TCI/F cryoprobe for proteins less than 20 kDa.
4. Five millimeter NMR tubes are recommended for sufficient
sample volume to accommodate the dilution of the protein
sample upon mixing with a discoidal membrane mimic.
5. The option to incorporate electrostatics appears to be inconse-
quential. The option of clustering spreads the scores across
more residues, without filling in the omissions of binding
modes that we have observed anecdotally; we do not recom-
mend clustering.
6. To prepare double-labeled protein for assignment of NMR
chemical shift peaks also substitute 0.3% (w/v) 13C6-glucose as
the sole carbon source. In some circumstances (e.g., severe
spectral overlap) it is possible to selectively label specific amino
acids using α-keto acids or other precursors [62].
7. Some assignment projects may be especially challenging due to
line broadening caused by the slow tumbling of the membrane-
associated protein, low protein concentration, or overlapped
peaks. In these cases, combinatorial or selective labeling strate-
gies for partial assignment of specific amino acid residues or
protein segments can be worthwhile [76–78]. If insect cells are
required for high-level expression and proper folding, media
for labelling in insect cells can now be obtained affordably by
supplementing with 15N/13C/2H–labeled yeast extracts [79].
8. Proteins exceeding 20 kDa in mass should be fractionally deu-
terated to at least 60% (preferably 99%) by inclusion of D2O in
the growth medium [79] in order to achieve narrower line-
widths and increased sensitivity [80].
9. For proteins greater than 20 kDa, TROSY versions of these
and other triple resonance spectra are recommended [81, 82].
For proteins less than 20 kDa with sharper line widths, rapid
acquisition BEST versions of triple resonance spectra are rec-
ommended [83, 84].
10. Selective labeling of the methyl groups of alanine, methionine,
and threonine is also feasible [62, 85].
11. While using the bath sonicator, monitor the temperature of
the water to ensure that it does not rise above about
40 °C. Turning the sonicator off for a while will allow it to
cool toward room temperature.
12. A translucent, blue tinted solution is a clear sign of the forma-
tion of liposomes. If the suspension has not clarified, then
liposomes have not yet formed adequately. Continue sonicat-
ing for a further 5–10 min.
13. This forms detergent micelles since the CMC of sodium cho-
late is 3 mM [86].
14. The optimal molar ratio of MSP1D1 to DMPC is 1:80.
15. The 1:1 assemblies should tumble faster and have sharper

NMR peaks than 2:1 assemblies.
16. Note residues with NMR peaks broadened or shifted by addi-
tion of the membrane mimics. Such spectral perturbations
suggest that the protein is binding to the membrane mimic
and the general location of binding or conformational pertur-
bations linked to binding.
17. This is performed before doping the bilayers with doxyl-

substituted DPPC, which is paramagnetic.
18. We have had success with D7PC/DMPC (q = 0.5) bicelles at
300 μM discoidal aggregates, corresponding to 100 mM in
total lipid monomers. These conditions require addition of
the doxyl-DPPC to 600 μM for an average of one per leaflet.
19. The R2 1H NMR relaxation rate constants fitted to the series
should be insensitive to experimental choices of length of sig-
nal averaging, spectral window, and exact concentrations.
20. The R2 values do depend on the hydrodynamics, which are
sensitive to temperature and degree of molecular association,
which should be kept uniform for making comparisons. Using
similar protein and lipid concentrations for the diamagnetic
(control) and paramagnetic experiments ensures similar
degrees of molecular association are obtained in both
experiments.
21. Minimize exposure to air while weighing the dry lipids. Weigh
out a 10% excess of DMPC in order to obtain the desired con-
centration of SUVs since ~10% of the DMPC will form MUVs
which will subsequently be removed by microcentrifugation.
22. Lipids dissolved in chloroform can be used to prepare the
SUVs, provided an additional day is allowed to evaporate the
chloroform and dry the lipids overnight under vacuum.
23. The aggregates could be sonicated to recover SUVs, but it is
not recommended.
24. If one or more such cysteine already exists in the protein at an
experimentally undesirable location, it should first be removed
by site-directed mutagenesis.
25. If mapping by NMR (preferably paramagnetic NMR) is not
available, the interfaces predicted by MODA and PPM should
be regarded as the hypotheses to test, realizing that these algo-
rithms have each missed one of the dual membrane-binding
sites that we located experimentally [13, 14] (Fig. 1a).
26. Since basic and hydrophobic residues are frequently found
within peripheral membrane interfaces, care should be taken
not to substitute Arg, Lys, Phe, Tyr, Leu, and Ile residues
within this region as this could disrupt binding at the protein–
membrane interface.
27. We recommend performing these steps in an anaerobic or

semi-anaerobic environment in order to prevent oxidation of
thiols to disulfides. All buffers should be deoxygenated, e.g.,
by bubbling through argon and reactions should be carried
out under an inert atmosphere of argon.
28. Constructs containing an MMP catalytic domain require 5–10
mM CaCl2 to stabilize the catalytic domain.
29. Use a fresh 10 mM stock solution of IANBD.
30. Linearity of fluorescence deteriorates as the sum of the absor-
bance at the excitation and emission wavelengths exceeds 0.08
[87].
31. Examination of other lipids that are found in cell membranes
is also recommended in order to explore biological specificity.
These may include anionic lipids like phosphatidylserine [88],
signaling lipids potentially relevant to your target protein (e.g.,
phosphoinositides) [1], unsaturated lipids, sterols, sphingho-
lipids, glycolipids, etc.).
32. Our studies have been utilyzing the Image-It™ kit (Life
Technologies) containing the Alexa Fluor 594 conjugate of
wheat germ agglutinin (WGA) at 5 μg/ml to dye the glycosyl-
ation red on the outer face of the plasma membranes and 2 μM
of Hoechst 33342 to label chromatin and nuclei blue.
Acknowledgment
This work was supported primarily by NIH grant GM057289, but

also by NIH grant CA098799.
References
1. Moravcevic K, Oxley CL, Lemmon MA (2012) metalloproteinase expressed on the surface of
Conditional peripheral membrane proteins: invasive tumour cells. Nature 370:61–65
facing up to limited specificity. Structure 7. Strongin AY, Collier I, Bannikov G, Marmer
20:15–27 BL, Grant GA, Goldberg GI (1995) Mechanism
2. Sternlicht MD, Werb Z (2001) How matrix of cell surface activation of 72-kDa type IV col-
metalloproteinases regulate cell behavior. Annu lagenase. Isolation of the activated form of the
Rev Cell Dev Biol 17:463–516 membrane metalloprotease. J Biol Chem
3. Clark P (2014) Protease-mediated ectodomain 270:5331–5338
shedding. Thorax 69:682–684 8. Yu WH, Woessner JF Jr, McNeish JD,
4. Rose-John S (2013) ADAM17, shedding, Stamenkovic I (2002) CD44 anchors the
TACE as therapeutic targets. Pharmacol Res assembly of matrilysin/MMP-7 with heparin-
71:19–22 binding epidermal growth factor precursor and
5. Edwards DR, Handsley MM, Pennington CJ ErbB4 and regulates female reproductive organ
(2008) The ADAM metalloproteinases. Mol remodeling. Genes Dev 16:307–323
Aspects Med 29:258–289 9. Berton A, Selvais C, Lemoine P, Henriet P,
6. Sato H, Takino T, Okada Y, Cao J, Shinagawa Courtoy PJ, Marbaix E, Emonard H (2007)
A, Yamamoto E, Seiki M (1994) A matrix Binding of matrilysin-1 to human epithelial
cells promotes its activity. Cell Mol Life Sci (2006) Hsc70 contacts helix III of the
64:610–620 J domain from polyomavirus T antigens:
10. Owen CA, Hu Z, Lopez-Otin C, Shapiro SD addressing a dilemma in the chaperone hypoth-
(2004) Membrane-bound matrix metallopro- esis of how they release E2F from pRb.
teinase-8 on activated polymorphonuclear cells Biochemistry 45:6917–6929
is a potent, tissue inhibitor of metalloproteinase- 21. Palmier MO, Fulcher YG, Bhaskaran R, Duong
resistant collagenase and serpinase. J Immunol VQ, Fields GB, Van Doren SR (2010) NMR
172:7791–7803 and bioinformatics discovery of exosites that
11. Owen CA, Hu Z, Barrick B, Shapiro SD tune metalloelastase specificity for solubilized
(2003) Inducible expression of tissue inhibitor elastin and collagen triple helices. J Biol Chem
of metalloproteinases–resistant matrix metallo- 285:30918–30930
proteinase-9 on the cell surface of neutrophils. 22. Gobl C, Madl T, Simon B, Sattler M (2014)
Am J Respir Cell Mol Biol 29:283–294 NMR approaches for structural analysis of
12. Cobos-Correa A, Trojanek JB, Diemer S, Mall multidomain proteins and complexes in solu-
MA, Schultz C (2009) Membrane-bound tion. Prog Nucl Magn Reson Spectrosc
FRET probe visualizes MMP12 activity in pul- 80:26–63
monary inflammation. Nat Chem Biol 23. Hennig J, Warner LR, Simon B, Geerlof A,
5:628–630 Mackereth CD, Sattler M (2015) Structural
13. Koppisetti RK, Fulcher YG, Jurkevich A, Prior analysis of protein–RNA complexes in solution
SH, Xu J, Lenoir M, Overduin M, Van Doren using NMR paramagnetic relaxation enhance-
SR (2014) Ambidextrous binding of cell and ments. In: Sarah AW, Frédéric HTA (eds)
membrane bilayers by soluble matrix metallo- Methods in Enzymology. Academic, New York,
proteinase-12. Nat Commun 5:5552 NY, pp 333–362
14. Prior SH, Fulcher YG, Koppisetti RK, Jurkevich 24. Prior, S. H., Byrne, T. S., Tokmina-Roszyk, D.,
A, Van Doren SR (2015) Charge-triggered Fields, G. B., and Van Doren, S. R. (2016)
membrane insertion of matrix metalloprotein- Path to collagenolysis: collagen V triple-helix
ase-7, supporter of innate immunity and model bound productively and in encounters
tumors. Structure 23:2099–2110 by matrix metalloproteinase-12. J Biol Chem
15. Kutateladze TG, Capelluto DG, Ferguson CG, 291:7888
Cheever ML, Kutateladze AG, Prestwich GD, 25. Iwahara J, Clore GM (2006) Detecting transient
Overduin M (2004) Multivalent mechanism of intermediates in macromolecular binding by
membrane insertion by the FYVE domain. paramagnetic NMR. Nature 440:1227–1230
J Biol Chem 279:3050–3057 26. Fawzi NL, Doucleff M, Suh JY, Clore GM
16. Hilty C, Wider G, Fernandez C, Wuthrich K (2010) Mechanistic details of a protein-protein
(2004) Membrane protein-lipid interactions in association pathway revealed by paramagnetic
mixed micelles studied by NMR spectroscopy relaxation enhancement titration
with the use of paramagnetic reagents. measurements. Proc Natl Acad Sci U S A

Chembiochem 5:467–473 107:1379–1384
17. Arumugam S, Van Doren SR (2003) Global 27. Schilder J, Ubbink M (2013) Formation of
orientation of bound MMP-3 and N-TIMP-1 in transient protein complexes. Curr Opin Struct
solution via residual dipolar couplings. Biol 23:911–918
Biochemistry 42:7950–7958 28. Bashir Q, Volkov AN, Ullmann GM, Ubbink
18. Arumugam S, Hemme CL, Yoshida N, Suzuki M (2010) Visualization of the encounter
K, Nagase H, Berjanskii M, Wu B, Van Doren ensemble of the transient electron transfer
SR (1998) TIMP-1 contact sites and perturba- complex of cytochrome c and cytochrome c
tions of stromelysin 1 mapped by NMR and a peroxidase. J Am Chem Soc 132:241–247
paramagnetic surface probe. Biochemistry 29. Dancea F, Kami K, Overduin M (2008) Lipid
37:9650–9657 interaction networks of peripheral membrane
19. Takeda M, Terasawa H, Sakakura M, proteins revealed by data-driven micelle dock-
Yamaguchi Y, Kajiwara M, Kawashima H, ing. Biophys J 94:515–524
Miyasaka M, Shimada I (2003) Hyaluronan 30. Glover KJ, Whiles JA, Wu G, Yu N, Deems R,
recognition mode of CD44 revealed by cross- Struppe JO, Stark RE, Komives EA, Vold RR
saturation and chemical shift perturbation (2001) Structural evaluation of phospholipid
experiments. J Biol Chem 278:43550–43555 bicelles for solution-state studies of membrane-
20. Garimella R, Liu X, Qiao W, Liang X, associated biomolecules. Biophys
Zuiderweg ER, Riley MI, Van Doren SR J 81:2163–2171
31. Morrison EA, DeKoster GT, Dutta S, PPM web server: resources for positioning of
Vafabakhsh R, Clarkson MW, Bahl A, Kern D, proteins in membranes. Nucleic Acids Res
Ha T, Henzler-Wildman KA (2012) 40:D370–D376
Antiparallel EmrE exports drugs by exchang- 45. Lomize MA, Lomize AL, Pogozheva ID,
ing between asymmetric structures. Nature Mosberg HI (2006) OPM: orientations of pro-
481:45–50 teins in membranes database. Bioinformatics
32. Liu Y, Kahn RA, Prestegard JH (2010) 22:623–625
Dynamic structure of membrane-anchored 46. Lomize AL, Pogozheva ID, Lomize MA,
Arf*GTP. Nat Struct Mol Biol 17:876–881 Mosberg HI (2007) The role of hydrophobic
33. Lu Z, Van Horn WD, Chen J, Mathew S, Zent interactions in positioning of peripheral pro-
R, Sanders CR (2012) Bicelles at low concen- teins in membranes. BMC Struct Biol 7:44
trations. Mol Pharm 9:752–761 47. Lomize AL, Pogozheva ID, Lomize MA,
34. Song Y, Mittendorf KF, Lu Z, Sanders CR Mosberg HI (2006) Positioning of proteins in
(2014) Impact of bilayer lipid composition on membranes: a computational approach. Protein
the structure and topology of the transmem- Sci 15:1318–1333
brane amyloid precursor C99 protein. J Am 48. Kufareva, I., and Overduin, M. (2015) http://
Chem Soc 136:4093–4096 w w w. s l i d e s h a r e . n e t / o v e r d u i n /
35. Schuler MA, Denisov IG, Sligar SG (2013) moda-slideshare.
Nanodiscs as a new tool to examine lipid- 49. Spurlino JC, Smallwood AM, Carlton DD,
protein interactions. Methods Mol Biol Banks TM, Vavra KJ, Johnson JS, Cook ER,
974:415–433 Falvo J, Wahl RC, Pulvino TA et al (1994)
36. Denisov IG, Sligar SG (2016) Nanodiscs for 1.56 A structure of mature truncated human
structural and functional studies of membrane fibroblast collagenase. Proteins 19:98–109
proteins. Nat Struct Mol Biol 23:481–486 50. Bertini I, Fragai M, Luchinat C, Melikian M,
37. Raschle T, Hiller S, Yu TY, Rice AJ, Walz T, Mylonas E, Sarti N, Svergun DI (2009)
Wagner G (2009) Structural and functional Interdomain flexibility in full-length matrix
characterization of the integral membrane pro- metalloproteinase-1 (MMP-1). J Biol Chem
tein VDAC-1 in lipid bilayer nanodiscs. J Am 284:12821–12828
Chem Soc 131:17777–17779 51. Ye Q-Z, Johnson LJ, Baragi V (1992) Gene
38. Gluck JM, Wittlich M, Feuerstein S, Hoffmann synthesis and expression in E. coli for PUMP, a
S, Willbold D, Koenig BW (2009) Integral human matrix metalloproteinase. Biochem
membrane proteins in nanodiscs can be studied Biophys Res Commun 186:143–149
by solution NMR spectroscopy. J Am Chem 52. Fulcher YG, Sanganna Gari RR, Frey NC,
Soc 131:12060–12061 Zhang F, Linhardt RJ, King GM, Van Doren
39. Yu TY, Raschle T, Hiller S, Wagner G (2012) SR (2014) Heparinoids activate a protease,
Solution NMR spectroscopic characterization secreted by mucosa and tumors, via tethering
of human VDAC-2 in detergent micelles and supplemented by allostery. ACS Chem Biol
lipid bilayer nanodiscs. Biochim Biophys Acta 9:957–966
1818:1562–1569 53. Zheng X, Ou L, Tong X, Zhu J, Wu H (2007)
40. Hagn F, Etzkorn M, Raschle T, Wagner G Over-expression and refolding of isotopically
(2013) Optimized phospholipid bilayer nano- labeled recombinant catalytic domain of human
discs facilitate high-resolution structure deter- macrophage elastase (MMP-12) for NMR
mination of membrane proteins. J Am Chem studies. Protein Expr Purif 56:160–166
Soc 135:1919–1925 54. Bertini I, Calderone V, Fragai M, Jaiswal R,
41. Studier FW (2005) Protein production by Luchinat C, Melikian M, Mylonas E, Svergun
auto-induction in high-density shaking cul- DI (2008) Evidence of reciprocal reorientation
tures. Protein Expr Purif 41:207–234 of the catalytic and hemopexin-like domains of
42. Denisov IG, Grinkova YV, Lazarides AA, Sligar full-length MMP-12. J Am Chem Soc
SG (2004) Directed self-assembly of monodis- 130:7011–7021
perse phospholipid bilayer Nanodiscs with con- 55. Lovejoy B, Welch AR, Carr S, Luong C, Broka
trolled size. J Am Chem Soc 126:3477–3487 C, Hendricks RT, Campbell JA, Walker KA,
43. Kufareva I, Lenoir M, Dancea F, Sridhar P, Martin R, Van Wart H, Browner MF (1999)
Raush E, Bissig C, Gruenberg J, Abagyan R, Crystal structures of MMP-1 and -13 reveal
Overduin M (2014) Discovery of novel mem- the structural basis for selectivity of collagenase
brane binding structures and functions. inhibitors. Nat Struct Biol 6:217–221
Biochem Cell Biol 92:555–563 56. Zhao Y, Marcink TC, Sanganna Gari RR,
44. Lomize MA, Pogozheva ID, Joo H, Mosberg Marsh BP, King GM, Stawikowska R, Fields
HI, Lomize AL (2012) OPM database and GB, Van Doren SR (2015) Transient collagen
triple helix binding to a key metalloproteinase 69. Bayburt, T. H., Dennisov, I. G., Grinkova,
in invasion and development. Structure Y. V., and Sligar, S. G. Nanodisc technology:
23:257–269 protocols for preparation of nanodiscs.
57. Udi Y, Fragai M, Grossman M, Mitternacht S, University of Illinois, Urbana-Champaign, IL.
Arad-Yellin R, Calderone V, Melikian M, http://sligarlab.life.uiuc.edu/nanodisc/pro-
Toccafondi M, Berezovsky IN, Luchinat C, tocols.html
Sagi I (2013) Unraveling hidden regulatory 70. Liu Y, Prestegard JH (2008) Direct measure-
sites in structurally homologous metalloprote- ment of dipole-dipole/CSA cross-correlated
ases. J Mol Biol 425:2330–2346 relaxation by a constant-time experiment.
58. Lee W, Cornilescu G, Dashti H, Eghbalnia J Magn Reson 193:23–31
HR, Tonelli M, Westler WM, Butcher SE, 71. Lee D, Hilty C, Wider G, Wuthrich K (2006)
Henzler-Wildman KA, Markley JL (2016) Effective rotational correlation times of pro-
Integrative NMR for biomolecular research. teins from NMR relaxation interference.
J Biomol NMR 64:307–332 J Magn Reson 178:72–76
59. Lee W, Tonelli M, Markley JL (2015)
72. Aguilar JA, Nilsson M, Bodenhausen G,
NMRFAM-SPARKY: enhanced software for Morris GA (2012) Spin echo NMR spectra
biomolecular NMR spectroscopy. without J modulation. Chem Commun
Bioinformatics 31:1325–1327 48:811–813
60. Lee W, Kim JH, Westler WM, Markley JL 73. Schulz TA, Choi MG, Raychaudhuri S, Mears
(2011) PONDEROSA, an automated JA, Ghirlando R, Hinshaw JE, Prinz WA
3D-NOESY peak picking program, enables (2009) Lipid-regulated sterol transfer between
automated protein structure determination. closely apposed membranes by oxysterol-
Bioinformatics 27:1727–1728 binding protein homologues. J Cell Biol
61.
Lee W, Stark JL, Markley JL (2014) 187:889–903
PONDEROSA-C/S: client-server based 74. Kim YE, Chen J, Chan JR, Langen R (2010)
software package for automated protein 3D Engineering a polarity-sensitive biosensor for
structure determination. J Biomol NMR time-lapse imaging of apoptotic processes and
60:73–75 degeneration. Nat Methods 7:67–73
62. Ruschak AM, Kay LE (2010) Methyl groups as 75. Song Y, Hustedt EJ, Brandon S, Sanders CR
probes of supra-molecular structure, dynamics (2013) Competition between homodimeriza-
and function. J Biomol NMR 46:75–87 tion and cholesterol binding to the C99
63. Tugarinov V, Kay LE (2003) Ile, Leu, and Val domain of the amyloid precursor protein.
methyl assignments of the 723-residue malate Biochemistry 52:5051–5064
synthase G using a new labeling strategy and 76. Löhr F, Tumulka F, Bock C, Abele R, Dötsch
novel NMR methods. J Am Chem Soc V (2015) An extended combinatorial 15N,
125:13868–13878 13
Cα, and 13C′ labeling approach to protein
64. Sprangers R, Kay LE (2007) Quantitative backbone resonance assignment. J Biomol
dynamics and binding studies of the 20S pro- NMR 62:263–279
teasome by NMR. Nature 445:618–622 77. Ikeya T, Takeda M, Yoshida H, Terauchi T, Jee
65. Sinha K, Jen-Jacobson L, Rule GS (2013) JG, Kainosho M, Guntert P (2009) Automated
Divide and conquer is always best: sensitivity of NMR structure determination of stereo-array
methyl correlation experiments. J Biomol isotope labeled ubiquitin from minimal sets of
NMR 56:331–335 spectra using the SAIL-FLYA system. J Biomol
66. Chao FA, Kim J, Xia Y, Milligan M, Rowe N, NMR 44:261–272
Veglia G (2014) FLAMEnGO 2.0: an enhanced 78. Hefke F, Bagaria A, Reckel S, Ullrich SJ,
fuzzy logic algorithm for structure-based Dotsch V, Glaubitz C, Guntert P (2011)
assignment of methyl group resonances. Optimization of amino acid type-specific 13C
J Magn Reson 245:17–23 and 15N labeling for the backbone assignment
67. Poget SF, Cahill SM, Girvin ME (2007) Isotropic of membrane proteins by solution- and solid-
bicelles stabilize the functional form of a small state NMR with the UPLABEL algorithm.
multidrug-resistance pump for NMR structural J Biomol NMR 49:75–84
studies. J Am Chem Soc 129:2432–2433 79. Opitz C, Isogai S, Grzesiek S (2015) An eco-
68. Bayburt TH, Grinkova YV, Sligar SG (2002) nomic approach to efficient isotope labeling in
Self-assembly of discoidal phospholipid bilayer insect cells using homemade 15N-, 13C- and
nanoparticles with membrane scaffold pro-
2
H-labeled yeast extracts. J Biomol NMR
teins. Nano Lett 2:853–856 62:373–385
80. Sattler M, Fesik SW (1996) Use of deuterium NMR experiments to a few minutes. J Am
labeling in NMR: overcoming a sizeable prob- Chem Soc 128:9042–9043
lem. Structure 4:1245–1249 85. Velyvis A, Ruschak AM, Kay LE (2012) An
81. Tugarinov V, Muhandiram R, Ayed A, Kay LE economical method for production of (2)H,
(2002) Four-dimensional NMR spectroscopy (13)CH3-threonine for solution NMR stud-
of a 723-residue protein: chemical shift assign- ies of large protein complexes: application to
ments and secondary structure of malate syn- the 670 kDa proteasome. PLoS One
thase g. J Am Chem Soc 124:10025–10035 7:e43725
82. Revington M, Zuiderweg ER (2004) TROSY- 86. Gennis RB (1989) Biomembranes: molecular
driven NMR backbone assignments of the structure and function. Springer, New York,
381-residue nucleotide-binding domain of the NY
Thermus Thermophilus DnaK molecular chap- 87. Palmier MO, Van Doren SR (2007) Rapid
erone. J Biomol NMR 30:113–114 determination of enzyme kinetics from fluores-
83. Lescop E, Schanda P, Brutscher B (2007) A set cence: overcoming the inner filter effect. Anal
of BEST triple-resonance experiments for Biochem 371:43–51
time-optimized protein resonance assignment. 88. Li L, Shi X, Guo X, Li H, Xu C (2014) Ionic
J Magn Reson 187:163–169 protein-lipid interaction at the plasma mem-
84. Schanda P, Van Melckebeke H, Brutscher B brane: what can the charge do? Trends Biochem
(2006) Speeding up three-dimensional protein Sci 39:130–140
Chapter 6
Using Small Angle X-Ray Scattering (SAXS) to Characterize

the Solution Conformation and Flexibility of Matrix
Metalloproteinases (MMPs)
Louise E. Butt, Robert A. Holland, Nikul S. Khunti, Debra L. Quinn,
and Andrew R. Pickford
Abstract
Small angle X-ray scattering (SAXS) provides information about the conformation and flexibility of pro-
teins in solution, and hence provides complementary structural information to that obtained from X-ray
crystallography and nuclear magnetic resonance spectroscopy. In this chapter, we describe the methods for
the preparation of matrix metalloproteinase (MMP) samples for SAXS analyses, and for the acquisition,
processing and interpretation of the SAXS data.
Key words Protein dynamics, Protein flexibility, Multi-domain, Oligomer, Structure–function

relationship
1 Introduction
Matrix metalloproteinases (MMPs) are a family of zinc-dependent

proteases that perform vital roles in essential processes such as
embryogenesis, development, and wound healing [1]. Impaired
regulation of MMP activity contributes to disease states such as
aneurism, arthritis, cancer, and fibrosis [1]. Although primarily
known for their degradation of extracellular matrix (ECM) com-
ponents, MMPs are also involved in signaling through their libera-
tion of ECM-bound growth factors and cytokines [2]. A more
thorough understanding of the roles of MMPs in homeostasis and
disease requires detailed knowledge of the relationship between
their structure and function.
The consensus minimal architecture for an active MMP com-
prises a catalytic (CAT) domain—the location of the hydrolytic
apparatus—connected to a hemopexin-like (HPX) domain by a
polypeptide linker [3]. This holds true for all except MMP-7,
which lacks a HPX domain. In many MMPs, the minimal
87
88 Louise E. Butt et al.
AT-
C linker-
HPX arrangement is supplemented by additional
domains for substrate binding (e.g., the fibronectin type 2 domains
in MMP-2 and -9), or cell surface localization, as seen in the
membrane-type (MT) MMPs. The linker connecting the CAT and
HPX domains varies in length from 14 residues in the prototypic
collagenase MMP-1, to 68 residues in the gelatinase MMP-9. In
their zymogenic proMMP forms, a small pro-peptide (PRO)
domain docks into the active site of the CAT domain sterically
blocking substrate access, and chemically blocking hydrolytic activ-
ity by chelation of the reactive Zn2+ ion with a Cys residue (the
so-called “cysteine switch”).
1.1 Structural This domain architecture of MMPs makes them amenable to

Analyses of Intact molecular dissection through recombinant expression of isolated
MMPs domains for subsequent structural analysis by macromolecular
X-ray crystallography (MX) and nuclear magnetic resonance
(NMR) spectroscopy. To date, MX has been the predominant
technique in such investigations, providing us with high-resolution
structural information, not only of isolated MMP domains, but
also of their complexes with small molecule inhibitors, TIMPs, and
other MMPs (reviewed in [3]).
The first structure of an intact MMP, that of porcine MMP-1,
was solved in 1995 [4]. It revealed a defined relative location and
orientation of the HPX domain with respect to the CAT domain,
despite the two being connected by what had been presumed to be
a flexible linker. The belief that the domains interacted in a pre-
ferred manner was reinforced with the later MX structures of
human MMP-1 [5], proMMP-1 [6] and proMMP-2 [7]. The
recent MX structure of MMP-13 bound to peptide fragments of its
own PRO domain revealed a similar arrangement, albeit with a 30°
rotation of the HPX domain [8]. In each case, a consensus inter-
face (nicknamed the “ball-and-socket” joint [9]) exists between
the CAT and HPX domains; a cluster of three residues (the “ball”)
from the first blade of the HPX domain docks into a depression
(the “socket”) in the CAT domain surface comprised of residues
from the Met-turn, the S1’ wall forming segment and the specific-
ity loop. The subsequent crystal structure of MMP-1 in complex
with a collagen-like triple helical peptide suggested that this joint
remains intact when the enzyme binds its substrate [10].
1.2 Flexibility Although these MX studies have revealed intricate intramolecular

in MMPs and intermolecular interactions with exquisite precision, we must
keep in mind that the MMPs (or fragments thereof) are not in their
native, solution state. Instead, the macromolecules are constrained
within the crystal by protein–protein contacts with their neighbors,
an enforced order that may restrict their dynamic behavior.
Furthermore, the very process of crystallization requires the pro-
tein of interest to become more ordered, a more entropically
SAXS Studies of Matrix Metalloproteases 89
challenging prospect for an inherently flexible macromolecule. In a

solution of such a protein, there will be multiple conformations,
and some of these will be more amenable to crystallization than
others under a given set of conditions. Hence, those crystals that
do form may be trapping a “snapshot” structure that is not fully
representative of the ensemble population in the solution state.
It is well established that proteins are inherently flexible, that
they exhibit conformational changes of varying magnitudes and on
a variety of timescales, and that these molecular motions are inex-
tricably linked with their function [11–13]. Therefore, in order to
fully understand the structure–function relationship in (pro)
MMPs, we must establish how their structure changes with time.
In MX analyses, flexibility is often inferred from high B-factors or
a lack of electron density for some stretches of the polypeptide. For
example, the protease-sensitive “bait region” in the PRO domain
of many MMP zymogens is believed to be flexible because its elec-
tron density is absent in the crystal structures of proMMP-1 [6],
proMMP-3ΔHPX [14] and proMMP-9ΔHPX [15]. But does the pres-
ence of defined electron density in a crystal structure imply that the
protein is rigid in solution?
In a landmark paper, Bertini and coworkers both solved the
crystal structure of MMP-12, and characterized its solution con-
formation using small angle X-ray scattering (SAXS) and NMR
[16]. The medium resolution MX study revealed an unambiguous
and specific relative domain orientation quite different to the “ball-
and-socket” joints seen in proMMP-1, MMP-1, and proMMP-2
(see Subheading 1.1, above); instead, in MMP-12, the fourth blade
of the HPX domain contacts a point on the opposite side of the
CAT domain to the above “socket”. However, the authors showed
that this MX structure is not fully representative of the enzyme in
solution, as their data from both NMR and SAXS were inconsis-
tent with that single static conformation. Furthermore, no single
structure from a pool of possible MMP-12 conformations (with
differing CAT and HPX domain locations/orientations generated
by randomizing the linker connecting them) could, by itself,
explain the SAXS data. Instead, using an ensemble optimization
modeling (EOM) approach [17], the authors were able to select
from that pool a family of conformers that, on average, did satisfy
the SAXS data. Whilst approximately one-half of the conformers in
solution were of similar size to the MX structure, the other half
were significantly more extended [16].
These results on MMP-12, which reveal a far more dynamic
solution conformation than the “snapshot” crystal structures would
suggest, led to a reappraisal of the structural plasticity of MMP-1
using SAXS and NMR [18]. The combined study revealed that
MMP-1 is far more dynamic in solution than the previous crystal
structures suggest, with the population ensemble including
extended conformers where the non-covalent interface between the
CAT and HPX domains is disrupted [18]. This transient dislocation

of the “ball-and-socket” interface between the two domains may
reveal a collagen binding exosite on the HPX domain that is other-
wise concealed [9]. This structural flexibility in MMP-1 is impor-
tant for function, as shown by a subsequent NMR study which
revealed that the enzyme initially binds to a collagen-like triple heli-
cal peptide with its CAT-HPX interface dislocated, then condenses
to the compact structure prior to collagenolysis [19, 20].
In contrast, separate SAXS analyses have shown that the crystal
structure of proMMP-1 [6] is a good representation of the solu-
tion conformation [9]; the non-covalent PRO–HPX interface seen
in the zymogen crystal structure appears to maintain a tightly com-
pact domain arrangement in solution. This highlights our inability
to predict the overall flexibility of multi-domain proteins purely
from MX analyses alone, and emphasizes the importance of per-
forming complementary studies on MMPs (and proMMPs) in
solution by methods such as SAXS and/or NMR.
1.3 Small Angle In MX, diffraction of X-rays from well-ordered atoms in oriented
X-Ray Scattering molecules within the crystal lattice gives rise to discrete reflections
(SAXS) of varying intensity on the two-dimensional (2D) detector; trans-
formation of this diffraction pattern gives rise to the electron den-
sity map, which can precisely locate atomic positions within the
protein. However, in small-angle X-ray scattering (SAXS), the ran-
domized orientation of the isotropically tumbling molecules in the
sample gives rise to a circularly symmetric pattern of elastically scat-
tered X-rays [21, 22]. As with MX, a 2D detector is utilized, but
the rotational symmetry means the SAXS data is, in effect, a one-
dimensional profile, varying in intensity with angular deviation
(2θ) out from the direction of the incident collimated X-ray beam
(of wavelength λ). The resultant scattering profile, which is
described as intensity I(q) versus q, where q = (4π sin θ)/λ (see
Note 1), can be interpreted in terms of the protein conformation
but the rotational averaging in solution means that both the infor-
mation content and resolution (typically about 10 Å) are much
lower than that obtained from MX [21, 22]. Nevertheless, SAXS
provides an effective way not only of validating those protein struc-
tures solved by MX, but also of characterizing the mixed popula-
tions of the varying conformations that exist in solution.
1.4 Acquisition Protein solution SAXS is an inherently insensitive technique requir-

of SAXS Data ing a synchrotron X-ray radiation source and high concentrations
of purified protein (typically 1-5 mg/mL) for satisfactory results
[21–24]. Needless to say, protein solubility can be a limiting factor,
and attempting SAXS at the limit of a protein’s solubility is prob-
lematic since even minimal aggregation can cause disproportion-
ately large distortions to the scattering profile. To alleviate this
problem, many modern SAXS beamlines (e.g., B21 at Diamond
Light Source (DLS), BM29 at the European Synchrotron Radiation

Facility (ESRF), and P12 at the Deutsches Elektronen-Synchrotron
(DESY)) offer in-line size-exclusion chromatography (SEC) to
allow removal of aggregated material immediately before the sam-
ple enters the X-ray beam. Typically, protein SAXS data is acquired
over the range of q values from approximately 0.01–0.5 Å−1, which
corresponds to an interatomic spacing (d) range of 628–13 Å
(since d = 2π/q). Even with a well-behaved, highly concentrated,
monodisperse (i.e., non-aggregated) protein, most of the scatter-
ing arises from the buffer, and this contribution must be subtracted
prior to interpretation of the protein signal. If more than one con-
former is present in the sample, the scattering profile will be an
average of their individual contributions, weighted by their relative
abundance.
The high protein concentrations required for SAXS analyses
render samples containing uninhibited CAT domains susceptible
to self-hydrolysis (autolysis), most readily in the protease-
susceptible CAT-HPX linker. Hence, studies on intact, mature
MMPs (i.e., lacking the PRO domain) have utilized hydrolytically
impaired enzymes in which the active site Glu residue is mutated to
Ala (i.e., mutant E219A in both MMP-1 [9, 18, 20] and MMP-12
[16]). The equivalent mutation in MMP-7 reduces hydrolytic
activity by approximately 99.9% [25]. In some studies, further
autolytic inhibition is achieved through the addition of a
hydroxamic acid-based inhibitor, for example N-isobutyl-N-[4-
methoxyphenyl-sulfonyl] glycyl hydroxamic acid (NNGH), and by
substitution of the active site Zn2+ ion by Cd2+ by exhaustive dialy-
sis [26]. However, metal ion chelating agents, such as ethylenedi-
aminetetraacetic acid (EDTA) or 1,10-phenanthroline should not
be used, as this will strip Zn2+ and/or Ca2+ ions from the enzyme
and thus disrupt its structure.
1.5 Interpretation Scattering at the lower end of the q range gives information about
of the Scattering the size of the protein in solution, that is the apparent molecular
Profile weight (MW) and the approximate radius of gyration (Rg) [21–
24]. The MW reveals the oligomeric state of the sample, and the
Rg provides a quantitative measure of the protein’s compactness.
These parameters can be used to reveal any monomer-dimer
transitions, for example as exhibited by the HPX domain of MMP-
14 in solution [27]. Scattering at higher q values gives information
about the molecular shape, albeit at low resolution compared to
MX. A Kratky plot (q2I(q) vs. q), a mathematical manipulation of
the scattering profile, can distinguish between folded, unfolded
and flexible proteins.
An indirect Fourier transform of the reciprocal-space scatter-
ing profile (I(q) vs. q) over a finite q range yields the real-space
pairwise distance function (P(r) vs. r), which is in effect a histo-
gram of the possible interatomic distances (r) in the protein
[21–24]. From the P(r), we can determine the maximum dimen-

sion of the protein (Dmax) in solution, and a more accurate mea-
surement of the Rg (often called the real-space Rg). A P(r)
distribution can be calculated from a particular atomic resolution
structure (such as that derived from MX), but the reverse is not
possible; the P(r) distribution contains insufficient detail to be
transformed into a unique atomic resolution structure. Previously
solved MX/NMR structures can be validated by back-calculation
of either reciprocal-space scattering data or real-space P(r) from
the atomic coordinates, and then comparing with the experimental
SAXS data.
Methods for generating three-dimensional (3D) models from
the SAXS data fall into two broad categories, namely volumetric
modeling and atomistic modeling [21–24]. In volumetric (also
known as ab initio) modeling, a molecular envelope is generated
from the P(r) distribution using dummy atoms (or “beads”) of fixed
volume with no assumptions made about the protein structure.
Subsequently, a previously determined MX/NMR structure may be
superimposed with this envelope by rigid body minimization. In the
atomistic approach, all-atom models are generated that satisfy the
SAXS data, and that use previously determined atomic resolution
structures of the protein or fragments thereof (typically, individual
domains). When studying flexible proteins, the scattering cannot be
explained by a single atomistic structure, so an ensemble optimiza-
tion modeling (EOM) approach [17] is commonplace.
1.6 Disclaimer This chapter describes generalized methods for assessing the confor-
mational state of (pro)MMPs in solution using SAXS. The strategies
employed are adapted from those published previously [9, 16, 18,
20, 27]. However, they remain “first-shot” approaches, and may
require some optimization (e.g., sample buffer conditions) in order
to obtain the highest quality data. As with other structural methods,
the quality of results will depend on the MMP (or proMMP) under
study, and the purity of the sample. In recent years, the amount of
available SAXS experiment time has grown with the expansion of
dedicated synchrotron beamlines, and data acquisition is becoming
more automated and routine. However, we strongly advise anyone
interested in performing SAXS analyses to consult their local beam-
line scientist and identify any potential pitfalls prior to planning
experiments and production of their MMP samples.
2 Materials
2.1 Preparation 1. N-isobutyl-N-[4methoxyphenylsulfonyl] glycyl hydroxamic

of Protein Samples acid (NNGH).
and Buffers 2. Acetohydroxamic acid (AHA).
3. Cadmium (II) chloride (toxic).

4. Bovine serum albumin (BSA).
5. HiLoad 26/60 Superdex 75 pg (GE Healthcare).
6. ÄKTAprime plus chromatography system (GE Healthcare).
7. Standard SAXS buffer: 50 mM Tris, pH 7.4, 150 mM NaCl,
10 mM CaCl2, 0.02% (w/v) NaN3.
8. Cd2+ SAXS buffer: 20 mM Tris, pH 7.2, 10 mM CaCl2, 0.3 M
NaCl, 0.2 M AHA, 0.3 mM CdCl2.
9. Glycerol.
10. Ascorbic acid (vitamin C).
11. NanoDrop 2000 spectrophotometer (Thermo Scientific).
12. Refrigerated bench-top centrifuge.
13. Refrigerated microcentrifuge.
14. Vivacon-500 10,000 Da MW cutoff (MWCO) ultrafiltration
device (Sartorius).
15. Vivacon-500 100,000 Da MW cutoff (MWCO) ultrafiltration
device (Sartorius) (see Note 2).
16. Dialysis membrane (10,000 Da MWCO).
2.2 Data Acquisition 1. BioSAXS™ sample handling robot (Arinax).

Using the BioSAXS 2. BioSAXS™-compatible 96-well microplates (see Note 3).
Robot
2.3 Data Acquisition 1. Agilent 1200 high-pressure liquid chromatography (HPLC)

Using Inline SEC-SAXS system.
2. Shodex KW402.5 4F (4.6 mm internal diameter × 300 mm
length) HPLC SEC column (Showa Denko).
2.4 Processing 1. Data reduction software (beamline-dependent).

of SAXS Data 2. ATSAS software package [28, 29], version 2.7.1 or later (www.
embl-hamburg.de/biosaxs) (see Note 4).
3. ScÅtter software, version 2.3 or later (www.bioisis.net/scatter).
4. Access to the EOM online server (www.embl-hamburg.de/
biosaxs/atsas-online/eom.php).
3 Methods
3.1 Procedures To date, all SAXS analyses on MMPs have been performed on
for Preparation recombinant proteins, expressed either from E. coli (proMMP-1
of Protein Samples [9], MMP-1 [9, 18, 20], MMP-12 [16] and the HPX domain of
and Buffers MMP-14 [27]) or Sf9 insect cells (for proMMP-9 [30]).
Recombinant expression more readily provides the substantial
(milligram) amounts of protein required for SAXS analysis, and

offers the potential of studying mutant enzymes. Use of higher
eukaryotic host systems is required for native-like glycosylation on
some MMPs, e.g., proMMP-9 [30]. Methodology for recombi-
nant expression is beyond the scope of this chapter, but the inter-
ested reader is referred to suggested expert texts [31–33].
MMPs are generally amenable to SAXS analysis at near-neutral
pH and physiological salt concentrations. Calcium is added to
ensure occupancy of binding sites in the CAT and HPX domains.
Addition of zinc is not considered necessary as it is retained by the
folded enzymes with very high affinity (see Note 5). Extended
exposure to X-rays damages proteins inducing aggregation and
thus ruining the data. Addition of 3–5% (v/v) glycerol or 1–2 mM
ascorbic acid to the buffer may reduce such damage (see Note 6).
If the (pro)MMP sample or the BSA MW standard contains
aggregated protein then the interpretation of the scattering data
will be extremely challenging. Therefore, we recommend using the
following SEC scheme as the final step in sample purification prior
to SAXS analysis. We use the same scheme regardless of whether
the sample will subsequently be delivered into the beam by the
BioSAXS™ robot (Subheading 3.2) or inline SEC (Subheading
3.3) (see Note 7). In addition, we always purify the BSA to be used
as a MW standard in the same way as the (pro)MMP of interest.
1. Pre-rinse a 10,000 Da MWCO ultracentrifugation device with
the buffer to be used for preparative SEC. The buffer composi-
tion will depend on the (pro)MMP under study.
2. Concentrate the protein using the rinsed device to a concen-
tration of at least 5 mg/mL. Use multiple short spins (less than
3 min duration) to avoid excessively high protein concentra-
tion building up at the membrane surface (see Note 8).
3. Purify the protein by low-pressure SEC using the HiLoad
26/60 Superdex 75 pg column on an ÄKTAprime plus chro-
matography system (or equivalent) collecting 5 mL fractions
across the elution range of void volume (v0) to total volume
(vt) (see Note 9).
4. Assess the purity of the fractions by sodium dodecyl sulfate
(SDS) polyacrylamide electrophoresis (PAGE), and pool the
desired fractions containing the (pro)MMP of interest.
5. If required to prevent autolysis, add equimolar NNGH and/or
an excess of AHA to the SEC-purified (pro)MMP sample and
incubate for 1 h at 4 °C to equilibrate.
3.2 Procedure 1. Dialyse the sample overnight at 4 °C against the SAXS buffer.
for Data Acquisition The precise buffer used will depend on the (pro)MMP under
Using the BioSAXS study. Do not discard the dialysate (see Note 10).
Robot 2. Clarify the dialysate using a 0.2 μm filter and keep as the SAXS
reference buffer.
3. Measure the protein concentration of the dialysed sample by

spectrophotometry using the Nanodrop 2000. If the protein
concentration is below 5 mg/mL (or the solubility limit, if that
is lower), concentrate the sample by ultrafiltration as in
Subheading 3.1, step 1 above (see Note 11).
4. Centrifuge the sample in the refrigerated bench-top microfuge
for 5 min at 15,000 × g and 4 °C. Retain the supernatant.
5. Check that the sample is monodisperse. A variety of methods
can be used for this, for example native PAGE, multi-angle
laser light scattering or dynamic light scattering (see Note 12).
Each of these techniques is beyond the scope of this chapter.
6. Prepare suitable dilutions of the concentrated stock sample
using the clarified SAXS reference buffer. Typical protein con-
centrations for SAXS are 1.0, 2.0, and 5.0 mg/mL. Precise
measurements of the protein concentration are needed (see
Note 13) for accurate determination of MW from the forward
scattering I(0) (see Subheading 3.4.4).
7. Centrifuge the sample again, as in step 4.
8. Load 30 μL of each concentration of the sample and an excess
of buffer reference into appropriate wells of the BioSAXS™-
compatible microplate. The samples should be analysed in the
order buffer-sample-buffer, so the volume of buffer used will
be twice the total volume of sample.
9. Program the BioSAXS™ robot to deliver the sample into the
X-ray beam once the shutter is open, collecting multiple
tandem 10-s exposures (frames) of raw scattering data with the
buffer/sample under flow (see Note 14).
3.3 Procedure In SEC-SAXS, as the sample (or BSA standard) passes through the
for Data Acquisition SEC column it becomes more dilute resulting in a lower scattering
Using SEC-SAXS signal. Hence, the initial concentration required is slightly higher
than that for the BioSAXS™ robot.
1. Measure the protein concentration of the dialysed sample by
spectrophotometry using the Nanodrop 2000. If the concen-
tration is below 7 mg/mL (or the solubility limit, if that is
lower), concentrate the sample by ultrafiltration as in
Subheading 3.1, step 1 above.
2. Prior to the first analysis, pass at least five column volumes of
SAXS buffer through the SEC column to ensure that it is thor-
oughly equilibrated (see Note 9).
3. Program the HPLC to inject 40-50 μL of (pro)MMP sample
onto the SEC column and into the X-ray beam once the shut-
ter is open. Start monitoring the A280 with the sample running
onto the column at a flowrate of 0.16 mL/min, and begin
acquiring tandem 3-s exposures (frames) of raw scattering
data.
4. Use selected frames recorded prior to elution of the column void

volume (v0) as the reference frames for buffer subtraction.
5. Stop recording the A280 and scattering data once a complete
column volume (vt) of buffer has passed through the column.
6. With the recommended setup, each SEC-SAXS run will take
approximately 30 min. Prepare the next protein sample for
analysis during this time.
3.4 Procedures Parameters for data normalization and reduction will be beamline
for Processing dependent, and some (e.g., beam center position) may need calibrat-
of SAXS Data ing. Data reduction and normalization is usually performed at the
synchrotron beamline using bespoke software (i.e., neither ScÅtter
3.4.1 Data Normalization nor the ATSAS package); in many cases it will be an automated step
and Reduction with no user input necessary. If unsure, consult your local beamline
scientist for assistance and advice with the following steps.
1. Apply a mask to all frames of interest (both buffer and sample) to

eliminate those pixels behind the beamstop (and any supporting
structures), and any pixels from inactive areas of the detector.
2. Radially average the datasets through an appropriate sector
(typically between 60° and 120°) to reduce the 2D SAXS
images to one-dimensional dataset (intensity vs. radial distance
from beam center).
3. Normalize this dataset using the camera length (the distance
from sample to detector) and the intensity of the transmitted
beam (measured at the beamstop diode) to obtain the scatter-
ing curve (I(q) vs. q).
3.4.2 Buffer Subtraction Data reduction provides us with a one-dimensional profile of the
and Data Quality scattering from both protein and buffer [21–24]. For both the
(pro)MMP sample and the BSA standard, we obtain the protein’s
scattering profile by subtracting the contribution from the buffer
(Fig. 1). Buffer subtraction and inspection of the data quality can
be performed in either PRIMUS [34] or ScÅtter.
1. Check for consistency of the multiple buffer frames (either the

pre- and post-sample buffers if using the BioSAXS™ robot
setup, or the pre-void volume buffer if using SEC-SAXS) by
overlaying their scattering profiles.
2. Average those buffer frames showing consistent scattering for
use in buffer subtraction.
3. Subtract this buffer scattering profile from each of the sample
frames in order to visualize the scattering from just the protein
of interest.
4. If SEC-SAXS was not used, every buffer-subtracted sample
frame must be checked for signs of aggregation (see Note 15).
In each case, examine the low-q2 (<~0.005) region of the
Guinier plot of ln(I(q)) vs. q2. A nonlinear relationship between

ln (I(q)) and q2 with upward deviation at low q2 (i.e., a positive
second derivative) indicates the presence of aggregation
which may have been X-ray induced.
5. Do not continue analysing any frames for which aggregation is
suspected, since scattering from aggregates strongly influences
the entire dataset.
6. Examine the data for signs of inter-particle repulsion; this is
indicated by a downward deviation from linearity in the Guinier
plot at low q2 (i.e., a negative second derivative). Inter-particle
repulsion may be alleviated by using more dilute samples or
increasing the ionic strength of the buffer.
7. For a given sample, combine all good quality frames by averag-
ing the scattering (I(q)) data over the entire q-range. If using
SEC-SAXS, average those frames which correspond to an
absorbance (A280) greater than or equal to half the maximum
peak intensity (see Note 16). If the BioSAXS™ robot setup was
used, average those frames for a given sample concentration
that show no obvious signs of aggregation.
8. Where multiple sample concentrations have been analysed
using the BioSAXS™ robot, a single scattering profile should
be generated by merging the data sets (see Note 17). Using a
concentration- weighted average, combine the data at low
angle (e.g., q < 0.15 Å−1) from the lowest protein concentra-
tion (e.g., 1.0 mg/mL), and the data at higher angle (e.g., q >
0.10 Å−1) from the highest protein concentration (e.g., 5.0
mg/mL). Visually check for consistency between the two data-
sets in the overlapping region (i.e., 0.10 Å−1 < q < 0.15 Å−1)
before merging; a systematic error in this range is indicative of
inaccurate protein concentration measurement.
Fig. 1 Buffer subtraction and Guinier analysis. SAXS data were acquired on proMMP-1 at DLS beamline B21.
(a) Overlay of buffer-only (grey) and proMMP-1 (black) scattering profiles; (b) Buffer-subtracted scattering
profile for proMMP-1; (c) Guinier analysis of proMMP-1 (above) with residuals (below), showing the reciprocal-
space approximation to Rg and extrapolation of I(0). Graphs were produced using ScÅtter
3.4.3 Determination In the absence of aggregation, the forward scattering, I(0), is one of
of Forward Scattering, I(0), the most accurate parameters that can be derived from SAXS data
and Reciprocal-Space Rg [21–24]. However, I(0) is not measured directly (the beamstop pre-
by Guinier Analysis vents us from doing so), but instead must be obtained by backward
extrapolation of the scattering profile to zero q. Guinier analysis (i.e.,
inspecting ln(I(q)) vs. q2 at low q2) is a convenient way of obtaining
I(0). The same process yields the reciprocal-space Rg, an excellent
approximation of the real-space Rg (see Subheading 3.4.5). For globu-
lar proteins, Guinier approximation of Rg is only robust when utiliz-
ing the range of q-values where q.Rg ≤ 1.3 [21–24]. Interactive
Guinier analysis can be performed in either PRIMUS [34] or ScÅtter.
1. Initiate an interactive Guinier analysis on the buffer-subtracted

scattering profile (either the sample or BSA standard).
2. Truncate the low-q2 datapoints where interference from the
beamstop is suspected (see Note 18).
3. Truncate the high-q2 datapoints until q.Rg ≤ 1.3.
4. Note the q2-range used, the resultant reciprocal-space Rg and
the reverse-extrapolated I(0) value which can used for deter-
mination of the apparent MW (see Subheading 3.3, step 4).
3.4.4 Determination The forward scattering, I(0), has been shown to scale linearly with
of Apparent MW MW if one assumes a constant protein density [21–24]. It may be
obtained from the Guinier region (see Subheading 3.4.3) or the
P(r) distribution (see Subheading 3.4.5). If using the ATSAS suite,
use PRIMUS [34] for Guinier analysis, and GNOM [35] for P(r)
generation. ScÅtter can perform both of these functions.
1. Determine the I(0) value for each concentration of BSA. The

I(0) value should scale linearly with BSA concentration.
Choose one pair of BSA concentration and resultant I(0) for
the calculation of MW in step 3 below.
2. Determine the I(0) value for each concentration of the (pro)
MMP sample.
3. For each sample concentration, calculate the apparent MW
from the concentration and forward scattering using the fol-
lowing equation:
I (0)( pro )MMP / [(pro)MMP ]
MW( pro )MMP = MWBSA
I (0)BSA / [BSA]
The MW of BSA is 66,463 Da. All concentrations are in mg/mL.
4. If the apparent MW increases with increasing concentration, it
indicates either aggregation (although this should be apparent
from the nonlinearity of the Guinier plot—see Subheading
3.4.2) or changes in the multimerization state of the protein.
A similar increase in Rg value will also usually be observed if

this is the case.
3.4.5 Generating the P(r) The real-space pairwise distance function, P(r), is generated by
Distribution indirect Fourier transformation of the reciprocal-space scattering
profile over a finite q range [21–24]. From the P(r), we can deduce
the Dmax (the r where P(r) returns to zero) and a more accurate
(real-space) measurement of the Rg than that obtained from
Guinier approximation (see Subheading 3.4.3). In addition, the
P(r) profile gives an indication of the global shape of the protein in
solution, and can be used for volumetric modeling (see Subheading
3.4.8). Dmax is actually a “soft” parameter that must be iteratively
optimized by the user in the original GNOM [35] software, or by
a recursive algorithm in the more recent AUTOGNOM program
[28], and in ScÅtter. The following instructions apply to GNOM
and ScÅtter:
1. Load your buffer-subtracted, merged scattering data (see

Subheading 3.4.2) into the program.
2. Specify the desired data range, excluding data at low q that may
be distorted by the beamstop, and data at high q that are exces-
sively noisy.
3. Begin the iterative P(r) analysis, starting with a Dmax of approx-
imately three times the reciprocal-space Rg that was obtained
from Guinier analysis (see Subheading 3.4.3).
4. Optimize the P(r) distribution by changing the Dmax value.
The P(r) distribution should decay smoothly to the x-axis
towards Dmax (Fig. 2). If Dmax is set too low, there will be a
sharp drop in P(r) as r approaches Dmax. If the function drops
below zero, then Dmax has been set too high.
5. Note down the values of the real-space Rg and the forward-
scattering, I(0). Discrepancies between these values and those
calculated from Guinier analysis (Subheading 3.4.3) typically
arise from small amounts of aggregation that have a greater
impact on the low-q data.
6. Assess the profile of the P(r) distribution which informs the
shape of the protein in solution; a single, near-symmetrical
peak is indicative of a single globular domain, whilst multi-
domain proteins give rise to multiple peaks or a shoulder on
the main peak.
3.4.6 Back-Calculation Known high-resolution atomic structures can be compared to

of Scattering Data experimental intensity data using CRYSOL [36], a program which
from Atomic Coordinates generates theoretical solution scattering profiles that can be subse-
quently fitted to experimental scattering data. Importantly, these
simulated profiles are created using spherical harmonic reconstruc-
tions that take into account the hydration shell. Fitting is achieved
by minimization of the χ2 discrepancy between the back-calculated

and experimental scattering profiles through varying the following
parameters: (1) average displaced solvent volume per atomic group,
(2) contrast of the hydration shell, and (3) relative background.
Familiarity with command-line arguments is strongly advised, but
the program may also be run in interactive mode allowing the user
to easily define option parameters. The default values for these
parameters are usually sufficient but can be adjusted accordingly by
the user, if deemed appropriate (see the CRYSOL manual for fur-
ther details).
1. Enter the required Protein Data Bank (PDB) file as the initial
input file. CRYSOL will read the atomic coordinates and evalu-
ate scattering profiles for all models and chains present, unless
otherwise specified.
2. If comparison with experimental data is required, specify the
buffer-subtracted experimental scattering data (see Subheading
3.4.2), and select the appropriate angular units (Å−1 or nm−1)
for this data.
3. After execution, multiple output files are generated (see Note
19). The file with extension “.log” contains the experimental
and theoretical Rg values. The fit to the experimental scatter-
ing data and the χ2 value are located in the file with extension
“.fit”.
4. Inspect the χ2 values; low values indicate good agreement
between the known atomic structure and the experimental
SAXS data. However, for highly flexible proteins, one expects
variations throughout the sampled q range. Examine the data
for signs of any deviation; upward deviation of the experimen-
tal data at low q range (< 0.2 Å−1) is suggestive of molecular
elongation, whilst deviation within the medium q range (> 0.2
Å−1) may indicate a difference in relative domain orientation/
location between the two structures.
Fig. 2 Optimizing Dmax in the P(r) distribution. SAXS data were acquired on proMMP-1ΔHPX at DLS beamline B21.
Pairwise distance distributions are shown for (a) Dmax = 60 Å; (b) Dmax = 80 Å, the correct value; and (c) Dmax =
100 Å. In (a), the Dmax is too low, giving rise to a truncated P(r) (vertical arrow). In (c), the Dmax is too high, making
the P(r) descend below the x-axis (vertical arrow). Data displayed using GNOM
3.4.7 Volumetric For generating three-dimensional ab initio reconstructions from 1D

Modeling scattering data, we recommend using the program DAMMIF [37]
which uses the GNOM output file (see Subheading 3.4.5) as input to
produce low-resolution envelopes of dummy atoms (see Note 20).
Thereafter, DAMAVER [38] is used to align, average, and filter
these envelopes according to the frequency of dummy atom occu-
pancy. The process in brief involves comparison of all reconstructed
envelopes to find the most probable one, outliers (defined as models
with two standard deviations from the mean) are removed, and the
remaining models are aligned. A probability map is created from the
average aligned models, then filtered and refined to generate a final
averaged ‘core’ molecular envelope. If a high-resolution atomic
structure is available, SUPCOMB [39] can be used to superimpose
this structure onto the envelope by minimizing the normalized spa-
tial discrepancy (NSD) between the envelope and crystal structure.
Users are referred to the DAMMIF, DAMAVER and SUPCOMB
manuals for detailed information regarding command-line argu-
ments and output files (see Note 21).
1. In DAMMIF, enter the required GNOM file (see Subheading

3.4.5); DAMMIF will read the previously computed distance
distribution function from this file.
2. Select an appropriate annealing configuration, either “slow”
(which we recommend), “fast” or “interactive” (see Note 22).
3. Enforce the appropriate symmetry, typically P1 (see Note 23).
4. Define appropriate anisometry: prolate, oblate, or unknown.
This may be inferred from previous experiments.
5. Once submitted, DAMMIF will output a runtime log (see
Note 24) which includes step-by-step χ2 values which indicate
the level of agreement between the experimental scattering
profile and one simulated from the reconstructed molecular
envelope.
6. Generate multiple DAMMIF reconstructions (see Note 25)
prior to submitting the modeled envelopes (contained in the
−1.pdb output files) to DAMAVER [38] for processing in
batch mode (see Note 26).
7. Check the DAMAVER damsel.log file for the NSD values. A
mean NSD value of less than 0.9 indicates a unique model.
8. Where a high-resolution MX/NMR structure is available,
compare it to the reconstructed envelope (the output damfilt.
pdb file from DAMAVER) using SUPCOMB. A single .pdb
output file (appended with “r,” unless otherwise specified) is
generated containing the successful superposition. Note the
NSD value, in which a low value (< 1.0) indicates an accept-
able fit. Higher NSD values may indicate that the (pro)MMP
has disordered regions or conformational freedom.
3.4.8 Atomistic Atomistic modeling involves the generation of all-atom conform-

and Ensemble Optimization ers by assembling predetermined (by MX/NMR) domain struc-
Modeling tures in a particular arrangement (i.e., position and orientation)
that satisfies the experimental SAXS data. For (pro)MMPs, where
interdomain flexibility may be present, a particularly powerful
approach is to generate a population of conformers which together
satisfy the SAXS data using the EOM [17] program within the
ATSAS suite. EOM online is a web portal-based access to an EOM
server at EMBL, but has some restrictions on parameter selection
(see Note 27). The following instructions apply to EOM online:
1. Identify the domains in your (pro)MMP of interest for which

atomic resolution structures have been solved previously (see
Note 28). Using a text editor, generate a separate PDB file for
each chosen domain containing the atomic coordinate entries
(lines starting with the “ATOM” keyword) (see Note 29).
2. In EOM online, provide a file containing the (pro)MMP amino
acid sequence in one-letter code.
3. Select the overall symmetry for the protein, typically P1 (see
Note 23).
4. Upload the PDB files for the specified domains, and specify
whether each of these domains should be fixed in space or free
to move (see Note 30).
5. Specify the chain type, either “random coil” or “native-like.”
This influences the rotamer distribution of the built polypep-
tides—“random coil” gives a torsional distribution typical of
chemically denatured proteins, whilst “native-like” gives a
(slightly more restrictive) distribution consistent with intrinsi-
cally disordered proteins.
6. Provide one or more scattering curve files (I(q) vs. q), together
with the q units (either nm−1 or Å−1) used in those files.
7. When run, the program will generate a pool of 10,000 models by
randomizing the linker(s) between the ordered domains, and
then use a genetic algorithm to select a subset ensemble which, as
a minimum population, gives the best fit to the experimental data.
8. Download the results files. These contain the Rg, Dmax, and vol-
ume distributions for the initial pool and selected ensemble
(Fig. 3b), the ensemble-averaged theoretical scattering curve
(Fig. 3c), and PDB coordinate files for representative conformers.
4 Notes
1. Those new to SAXS should note that there are a variety of

abbreviations used in the literature for the scattering angle: q
(as used here), Q, s, and S.
2. High MW filtration device for removal of protein aggregates.
Fig. 3 Ensemble optimization modeling of proMMP-1. (a) Preparing the crystal structure for an EOM calcula-
tion. The PDB file was divided into four separate ordered “domains”: the first helix of the PRO domain, the
remainder of the structured PRO domain, the CAT domain and the HPX domain. All except the HPX domain are
fixed in space during the EOM calculation. The program generates Cα atomic positions for the unstructured
regions that connect these, i.e., the N-terminal tail, “bait” region, and PRO-CAT linker (white spheres in the
diagram). Although precisely positioned in the crystal structure, the CAT-HPX linker (black spheres) was
allowed to flex during the EOM calculation to accommodate varying positions/orientations of the HPX domain;
(b) Results of the EOM calculation showing the Rg distribution for the starting pool of proMMP-1 conformers
(dashed line) and the ensemble of selected conformers (solid line). Unlike the mature enzyme, proMMP-1
remains compact in solution [9]; (c) Superposition of the theoretical scattering from the ensemble of selected
conformers (black line) over the experimental data (grey crosses)
3. We recommend Greiner Bio-One 96-well microplates.

4. The ATSAS software package incorporates numerous pro-
grams for SAXS data processing and analysis, such as PRIMUS,
GNOM, CRYSOL, DAMMIF, SUBCOMP, and EOM
5. Adding excessive Zn2+ to the samples may promote protein
cross-linking and aggregation.
6. If using SEC-SAXS, addition of glycerol increases the column
back-pressure, so 3% (v/v) glycerol is the recommended maxi-
mum concentration. Although commonly used to protect
against radiation damage in other protein SAXS studies [40],
the reducing agents DTT and tris(2-carboxyethyl)phosphine
(TCEP) are not recommended (even when studying MMP
fragments that do not contain disulfide bonds) as each can
chelate Zn2+ ions.
7. Although SEC purification of the sample may seem redundant
if the SAXS analysis is to use inline SEC as the delivery method,
the purification step serves to remove aggregates that will oth-
erwise serve as nucleation points for further aggregation (and
hence sample loss) during any concentrating step.
8. Between spins, ensure to mix the retentate by gentle pipetting.
Using a NanoDrop 2000 spectrophotometer, measure the
A280 of 1 μL aliquots of both the retentate and filtrate at
r egular intervals (for example, every five spins) to ensure that

the protein is being retained by the membrane and is increas-
ing in concentration.
9. Refer to the SEC column manufacturer for the column’s total
volume (vt) and void volume (v0). Alternatively, the v0 may be
measured using dextran blue, a high-MW polysaccharide to
which dye molecules are covalently bound. If MW calibration
of the column is desired, we recommend using a mixture of
aprotinin (6.5 kDa), ribonuclease A (13.7 kDa), carbonic
anhydrase (29 kDa), ovalbumin (43 kDa), and conalbumin
(75 kDa).
10. It is imperative that the chemical composition of the buffer
exactly matches that of the sample. Even with the most care-
ful pipetting, separately prepared buffers are usually insuffi-
ciently matched leading to erroneous buffer subtraction. In
our experience with the BioSAXS™ robot, the best results
are obtained when using the dialysate from the final dialysis
step. However, it is also possible to use the filtrate from cen-
trifugal ultrafiltration.
11. During protein concentration, low levels of aggregates can
serve as nucleation sites for further aggregation, so it is advis-
able to remove these by passage through a high-MWCO (e.g.,
100,000 Da) ultrafiltration device beforehand. If the sample
does not aggregate on standing, pre-prepared high concentra-
tion stocks can be taken to the beamline. Otherwise, we rec-
ommend concentrating the sample on-site at the beamline
immediately prior to SAXS analysis.
12. Although a convenient technique, observing a single band on
an SDS-PAGE gel of the sample is not indicative of sample
monodispersity.
13. For accurate apparent MW determination, it is imperative that
the (pro)MMP concentrations are themselves accurate. We
recommend using nanoliter-scale spectrophotometry. Unless
they are particularly well standardized for the protein of inter-
est, generic protein concentration assays (e.g., Bradford and
BCA) are usually not sufficiently precise.
14. Having the sample under flow reduces X-ray induced damage
(and resultant aggregation) as the irradiated protein is con-
stantly moving out of the beam and being replaced.
15. In theory, it is possible to observe aggregation in SEC-SAXS,
but only if the flow rate is too low (and hence X-ray damage is
occurring), or if the protein being analysed is eluting at the
void volume (i.e., it is aggregated even before it enters the
X-ray beam). In principle, it is possible to reduce protein dam-
age by attenuating the beam and hence lowering the X-ray flux
(consult your local beamline scientist). However, the reduced
number of photons passing through the sample will also

reduce the data quality by diminishing the signal–noise ratio.
16. Addition of further data below this cutoff will mostly add
noise to the scattering profile.
17. I(q) decays roughly exponentially as a function of q, so the
data at wider angles has a much poorer signal–noise ratio.
Hence, for good quality data at high q, a high protein concen-
tration is required. However, this renders the sample more
susceptible to X-ray induced aggregation, which has a greater
effect on the low q data.
18. If present, beamstop interference will be apparent in both the
buffer and sample scattering profiles (see Fig. 1a) as deviations
at very low q (<0.01 A−1).
19. The following files will be saved as outputs after CRYSOL has
run: .log (ASCII file—CRYSOL log), .sav (binary file—CRY-
SOL information), .flm (ASCII file—multipole coefficients),
.int (ASCII file—scattering intensities), .fit (ASCII file—fit to
experimental data), and .alm (binary file—net partial
amplitudes).
20. DAMMIF [37] is an improvement over the earlier software
DAMMIN [38], maintaining the key features but offering
algorithm improvements which substantially increase the
speed of model reconstruction.
21. Interactive configuration modules are available in DAMMIF,
DAMAVER, and SUBCOMP, but note that some parameter
changes will incur significant penalties in CPU usage.
22. In DAMMIF, the “slow” mode will generate a more accurate
reconstruction of the envelope than in “fast” mode by increas-
ing both the number of harmonics and iterations per step dur-
ing the annealing process; this may be beneficial for elongated
molecules. The “interactive” mode allows customization of
the annealing procedures as required.
23. This will usually be “P1 (no symmetry)”, but for (pro)MMP
fragments that dimerize via the HPX domain (e.g., MMP-14
[27]), “P2 (dimer)” may be more appropriate.
24. The following files are saved as outputs after DAMMIF has
run: a .log file containing the screen output; a −0.pdb file
containing the search volume; a −1.pdb containing the mod-
eled envelope; a .fit file of the simulated scattering profile
compared to the smoothed experimental data; a .fir file of the
simulated scattering profile compared to the experimental
data; and a .in file, which can be used to rerun DAMMIF with-
out changing parameters.
25. Ten independent reconstructions are recommended. However,
it may be necessary to increase the number of DAMMIF
reconstructions to ensure a valid averaged “core” molecular

envelope is generated.
26. In DAMAVER, the modeled envelope PDB input files are
processed by a suite of functions, namely damsel, damsup,
damaver, damfilt, and damstart. These may be run individu-
ally, but it is recommended to run them all automatically in
batch mode.
27. EOM online will suit most (pro)MMP studies. However, if
greater versatility is required (e.g., changing the size of the
conformer pool, or simultaneous fitting to multiple SAXS
datasets), we suggest using the downloadable EOM execut-
able. For further details, see the EOM manual (http://www.
embl-hamburg.de/biosaxs/manuals/eom.html).
28. If high-resolution (MX/NMR) structures are not available for
the (pro)MMP of interest then approximate domain folds can
be produced by homology modeling using the online SWISS-
MODEL (swissmodel.expasy.org) [41].
29. The current version of the EOM program (2.0) is unable to
handle PDB coordinate files that have a noncontiguous amino
acid sequence. This is problematic for any domains containing
unstructured regions, e.g., the PRO domain “bait” region in
crystal structure of proMMP-1 [6], as their MX PDB coordi-
nate files will have missing atomic entries. Our solution is to
split these PDB files into two separate sub-domains.
30. If a (pro)MMP domain has been split into two (or more) sub-
domains for EOM (see Note 29), then they must be fixed in
space, otherwise they will dissociate during the calculation.
Acknowledgments
We gratefully acknowledge the financial support of IBBS and the

BBSRC.
References
1. Murphy G, Nagase H (2008) Progress in of full-length porcine synovial collagenase

matrix metalloproteinase research. Mol Aspects reveals a C-terminal domain containing a
Med 29(5):290–308 calcium-linked, four-bladed beta-propeller.
2. Mott JD, Werb Z (2004) Regulation of matrix Structure 3:541–549
biology by matrix metalloproteinases. Curr 5. Iyer S, Visse R, Nagase H, Acharya KR (2006)
Opin Cell Biol 16(5):558–564 Crystal structure of an active form of human
3. Nagase H, Visse R, Murphy G (2006) Structure MMP-1. J Mol Biol 362(1):78–88
and function of matrix metalloproteinases and 6. Jozic D, Bourenkov G, Lim NH, Visse R,
TIMPs. Cardiovasc Res 69(3):562–573 Nagase H, Bode W, Maskos K (2005) X-ray
4. Li J, Brick P, O'Hare MC, Skarzynski T, Lloyd structure of human proMMP-1—new insights
LF, Curry VA, Clark IM, Bigg HF, Hazleman into procollagenase activation and collagen
BL, Cawston TE, Blow BM (1995) Structure binding. J Biol Chem 280(10):9578–9585
7. Morgunova E, Tuuttila A, Bergmann U, 18. Bertini I, Fragai M, Luchinat C, Melikian M,

Isupov M, Lindqvist Y, Schneider G, Mylonas E, Sarti N, Svergun DI (2009)
Tryggvason K (1999) Structure of human pro- Interdomain flexibility in full-length matrix
matrix metalloproteinase-2: activation mecha- metalloproteinase-1 (MMP-1). J Biol Chem
nism revealed. Science 284(5420):1667–1670 284(19):12821–12828
8. Stura EA, Visse R, Cuniasse P, Dive V, Nagase 19. Bertini I, Fragai M, Luchinat C, Melikian M,
H (2013) Crystal structure of full-length Toccafondi M, Lauer JL, Fields GB (2012)
human collagenase 3 (MMP-13) with peptides Structural basis for matrix metalloproteinase
in the active site defines exosites in the catalytic 1-catalyzed collagenolysis. J Am Chem Soc
domain. FASEB J 27(11):4395–4405 134(4):2100–2110
9. Arnold LH, Butt LE, Prior SH, Read CM, 20. Cerofolini L, Fields GB, Fragai M, Geraldes
Fields GB, Pickford AR (2011) The interface CF, Luchinat C, Parigi G, Ravera E, Svergun
between catalytic and hemopexin domains in DI, Teixeira JM (2013) Examination of matrix
matrix metalloproteinase-1 conceals a collagen metalloproteinase-1 in solution: a preference
binding exosite. J Biol Chem for the pre-collagenolysis state. J Biol Chem
286(52):45073–45082 288(42):30659–30671
10. Manka SW, Carafoli F, Visse R, Bihan D, 21. Mertens HD, Svergun DI (2010) Structural
Raynal N, Farndale RW, Murphy G, Enghild characterization of proteins and complexes
JJ, Hohenester E, Nagase H (2012) Structural using small-angle X-ray solution scattering.
insights into triple-helical collagen cleavage by J Struct Biol 172(1):128–141
matrix metalloproteinase 1. Proc Natl Acad Sci 22. Rambo RP, Tainer JA (2015) Modeling mac-
U S A 109(31):12461–12466 romolecular motions by X-ray-scattering-
11. Pastor N, Amero C (2015) Information flow constrained molecular dynamics. Biophys
and protein dynamics: the interplay between J 108(10):2421–2423
nuclear magnetic resonance spectroscopy and 23. Svergun DI, Koch MH (2002) Advances in
molecular dynamics simulations. Front Plant structure analysis using small-angle scattering
Sci 6:306 in solution. Curr Opin Struct Biol
12. Orozco M (2014) A theoretical view of protein 12(5):654–660
dynamics. Chem Soc Rev 43(14):5051–5066 24. Trewhella J, Hendrickson WA, Kleywegt GJ,
13. Schwartz SD (2013) Protein dynamics and the Sali A, Sato M, Schwede T, Svergun DI, Tainer
enzymatic reaction coordinate. Top Curr JA, Westbrook J, Berman HM (2013) Report
Chem 337:189–208 of the wwPDB small-angle scattering task
14. Becker JW, Marcy AI, Rokosz LL, Axel MG, force: data requirements for biomolecular
Burbaum JJ, Fitzgerald PM, Cameron PM, modeling and the PDB. Structure
Esser CK, Hagmann WK, Hermes JD et al 21(6):875–881
(1995) Stromelysin-1: three-dimensional 25. Cha J, Auld DS (1997) Site-directed mutagen-
structure of the inhibited catalytic domain and esis of the active site glutamate in human
of the C-truncated proenzyme. Protein Sci matrilysin: investigation of its role in catalysis.
4(10):1966–1976 Biochemistry 36(50):16019–16024
15. Elkins PA, Ho YS, Smith WW, Janson CA, 26. Bertini I, Fragai M, Lee YM, Luchinat C, Terni B
D'Alessio KJ, McQueney MS, Cummings MD, (2004) Paramagnetic metal ions in ligand screen-
Romanic AM (2002) Structure of the ing: the Co(II) matrix metalloproteinase 12.
C-terminally truncated human ProMMP9, a Angew Chem Int Ed Engl 43(17):2254–2256
gelatin-binding matrix metalloproteinase. Acta 27. Tochowicz A, Goettig P, Evans R, Visse R,
Crystallogr D Biol Crystallogr 58(Pt Shitomi Y, Palmisano R, Ito N, Richter K,
7):1182–1192 Maskos K, Franke D, Svergun D, Nagase H,
16. Bertini I, Calderone V, Fragai M, Jaiswal R, Bode W, Itoh Y (2011) The dimer interface of
Luchinat C, Melikian M, Mylonas E, Svergun the membrane type 1 matrix metalloproteinase
DI (2008) Evidence of reciprocal reorientation hemopexin domain: crystal structure and bio-
of the catalytic and hemopexin-like domains of logical functions. J Biol Chem
full-length MMP-12. J Am Chem Soc 286(9):7587–7600
130(22):7011–7021 28. Petoukhov MV, Konarev PV, Kikhney AG,
17. Bernado P, Mylonas E, Petoukhov MV, Svergun DI (2007) ATSAS 2.1 – towards auto-
Blackledge M, Svergun DI (2007) Structural mated and web-supported small-angle scatter-
characterization of flexible proteins using ing data analysis. J Appl Cryst 40:S223–S228
small-angle X-ray scattering. J Am Chem Soc 29. Petoukhov MV, Franke D, Shkumatov AV, Tria
129(17):5656–5664 G, Kikhney AG, Gajda M, Gorba C, Mertens
HD, Konarev PV, Svergun DI (2012) New 36. Svergun D, Barberato C, Koch MHJ (1995)
developments in the program package for CRYSOL – a program to evaluate x-ray solu-
small-angle scattering data analysis. J Appl tion scattering of biological macromolecules
Cryst 45(Pt 2):342–350 from atomic coordinates. J Appl Cryst
30. Rosenblum G, Van den Steen PE, Cohen SR, 28:768–773
Grossmann JG, Frenkel J, Sertchook R, Slack 37. Franke D, Svergun DI (2009) DAMMIF, a
N, Strange RW, Opdenakker G, Sagi I (2007) program for rapid ab-initio shape determina-
Insights into the structure and domain flexibil- tion in small-angle scattering. J Appl Cryst
ity of full-length pro-matrix metalloproteinase- 42(2):342–346
9/gelatinase B. Structure 15(10):1227–1236 38. Volkov VV, Svergun DI (2003) Uniqueness of
31. Nie J, Pei D (2010) Expression and purifica- ab initio shape determination in small-angle scat-
tion of membrane-type MMPs. Methods Mol tering. J Appl Crystallogr 36(3 Part 1):860–864
Biol 622:99–110 39. Kozin MB, Svergun DI (2001) Automated
32. Rodgers UR, Clark IM (2010) Expression of matching of high- and low-resolution struc-
recombinant MMP-28 in mammalian cells. tural models. J Appl Cryst 34(1):33–41
Methods Mol Biol 622:55–65 40. Jeffries CM, Graewert MA, Svergun DI,
33. Windsor LJ, Steele DL (2010) Expression of Blanchet CE (2015) Limiting radiation dam-
recombinant matrix metalloproteinases in age for high-brilliance biological solution scat-
Escherichia coli. Methods Mol Biol 622:67–81 tering: practical experience at the EMBL P12
34. Konarev PV, Volkov VV, Sokolova AV, Koch beamline PETRAIII. J Synchrotron Radiat
MHJ, Svergun DI (2003) PRIMUS: a Windows 22(Pt 2):273–279
PC-based system for small-angle scattering data 41. Biasini M, Bienert S, Waterhouse A, Arnold K,
analysis. J Appl Cryst 36:1277–1282 Studer G, Schmidt T, Kiefer F, Cassarino TG,
35. Svergun DI (1992) Determination of the regu- Bertoni M, Bordoli L, Schwede T (2014) SWISS-
larization parameter in indirect-transform MODEL: modelling protein tertiary and quater-
methods using perceptual criteria. J Appl Cryst nary structure using evolutionary information.
25:495–503 Nucleic Acids Res 42(W1):W252–W258
Part III
Computational Simulations of Matrix Metalloproteases

Chapter 7
Molecular Dynamics Studies of Matrix Metalloproteases

Natalia Díaz and Dimas Suárez
Abstract
Matrix metalloproteases are multidomain enzymes with a remarkable proteolytic activity located in the
extracellular environment. Their catalytic activity and structural properties have been intensively studied
during the last few decades using both experimental and theoretical approaches, but many open questions
still remain. Extensive molecular dynamics simulations enable the sampling of the configurational space of
a molecular system, thus contributing to the characterization of the structure, dynamics, and ligand bind-
ing properties of a particular MMP. Based on previous computational experience, we provide in this chap-
ter technical and methodological guidelines that may be useful to and stimulate other researchers to
perform molecular dynamics simulations to help address unresolved questions concerning the molecular
mode of action of MMPs.
Key words Free energy calculations, Ligand binding, Molecular dynamics, Molecular modeling,
Multidomain enzymes, Peptide hydrolysis
1 Introduction
Matrix metalloproteinases (MMPs), as their name implies, display

hydrolytic activity against extracellular matrix structural elements
like collagen. However, the zinc-dependent MMPs also cleave bio-
active molecules thus acting as pivotal effectors in many physiolog-
ical processes [1]. Moreover, MMPs contribute to the initiation
and progression of prevalent pathologies like cancer and, accord-
ingly, numerous laboratories have pursued the design of MMP
inhibitors. Unfortunately, compounds developed over the last 20
years have failed in clinical trials probably due to their broad-
spectrum activity (i.e., they inhibit several MMPs and other
metallo-enzymes). Nowadays, MMPs are still considered drugga-
ble targets but it is now recognized that more specific inhibitors
need to be developed [2]. Nevertheless, the design of specific
MMP inhibitors, which could certainly benefit from computational
aided drug design approaches, remains a challenge due to the
inherent flexibility of these structurally similar macromolecules.
111
112 Natalia Díaz and Dimas Suárez
MMPs are multi-domain proteins that exhibit specific struc-

tural features which must be taken into account when planning
molecular simulations [3]. NMR and X-ray crystal structures have
shown that the ~165-residue catalytic domain displays a similar
fold for all MMPs and consists of an upper N-terminal subdomain
with a five-stranded β-sheet, two α-helices and several loops, and a
lower C-terminal subdomain formed by one α-helix and a large
Ω-loop. The narrow active site cleft separates both subdomains and
binds the catalytic zinc ion (Zn1) involved in the hydrolysis reac-
tion. A second zinc (Zn2) and several calcium ions are also observed
in all the catalytic domains and these play a structural role in the
folding of several loops. Peptide substrates bind within the active
site via polar and hydrophobic contacts with the so-called S3–S3′
binding subsites. The S1′ hydrophobic pocket has been suggested
to be particularly important for increasing selectivity during inhibi-
tor design, since it is partially delimited by the Ω-loop and displays
some variability in size and composition across the MMPs.
However, induced-fit effects have been observed at this site and it
is likely that the dynamics of the S1′ pocket should be included in
future drug design efforts [4]. Structures available for the mature
catalytic domain also differ in the length and the position of the
N-terminal coil which, in turn, affects the proteolytic activity of
MMPs [5]. In the physiologically relevant form (i.e., superactive
form) of the enzyme, an N-terminal tyrosine or phenylalanine
remains packed against a concave hydrophobic surface of the mol-
ecule via a salt-bridge interaction with a strictly conserved aspartic
acid residue.
Besides the catalytic domain, active MMPs (except MMP-7
and MMP-26) possess additional modules or domains. These non-
catalytic domains mediate the interaction of MMPs with other
macromolecules, thus determining the cellular location, activation,
and proteolytic activity of the enzymes [6]. These include the
hemopexin-like domain that is present in the majority of multido-
main MMPs (except MMP23), the three fibronectin-like modules
inserted in the catalytic domains of MMP-2 and MMP-9, and the
transmembrane domain characteristic of the membrane-type
MMPs (MT-MMPs). Interestingly, these noncatalytic domains
provide secondary sites for substrate binding that could be tar-
geted by MMP inhibitors. In fact, binding to the exosites (i.e.,
substrate-binding sites located outside the active-site cleft) has
been proposed as a promising new route to more specific MMP
inhibitors. Crystal structures for full-length MMPs are available for
MMP-1, MMP-2, MMP-12, and MMP-13 and these structures
suggest a conserved and well-defined spatial relationship between
the catalytic and noncatalytic domains. However, NMR and spec-
troscopy analyses show that the catalytic and hemopexin-like
domains of MMP-1, MMP-9, and MMP-12 are able to adopt both
compact and significantly extended conformations in solution [7].
Molecular Dynamics Studies of Matrix Metalloporteases 113
Molecular dynamics (MD) is a valuable simulation method for

analyzing the structure and dynamics of complex macromolecules
like MMPs. MD simulations usually employ atomic force field
models to sample the configurational space of systems comprising
many thousands of atoms over time scales of nanoseconds and
beyond. Most MD studies of multidomain MMPs have focused on
the isolated catalytic domains of MMP-1, MMP-2, MMP-3, or
MMP-13 with the aim of computing relative binding free energies
for active site inhibitors (e.g., carboxylate, hydroxamate, phospho-
nate, thiadiazole urea, alloxan, and pyrimidine dicarboxamide
inhibitors) [8–15]. MD simulations have also been used to analyze
the dynamic flexibility of each recognition pocket within the active
site cleft and their substrate specificities [16–18]. The MMP-2 acti-
vation process (i.e., removal of the propeptide from the active site)
and its interaction with collagen-like peptides have also been inves-
tigated with the aid of MD simulations [19, 20]. Our group has
intensively applied MD-based methods to analyze various aspects
of the structure and dynamics of MMPs such as the specific roles of
the noncatalytic Zn2 ion and several calcium ions on the structure
of the catalytic domain [21], the effect of the conformation
adopted by the N-terminal coil [22], interactions with small and
macromolecular peptide substrates [22–26], the influence of non-
catalytic domains [27–29], etc.
In this chapter, we will attempt to distill our previous experi-
ence into technical and methodological guidelines that could be
useful to and encourage other researchers to perform future MD
simulations aimed at addressing unresolved challenges regarding
the molecular mode of action of MMPs. Subheading 2 discusses
the appropriate choice of software and hardware, and highlights
the AMBER package of molecular simulation programs as a refer-
ence tool [30, 31]. In Subheading 3, we review the various tasks
and procedures required to set up, run, and analyze an extended
MD simulation of a biomolecular system, placing emphasis on the
points that are particularly relevant to MMPs. We comment on
how to derive realistic parameters for metal-containing systems and
what specific structural features of MMPs need special attention
during the preparation of the system, differentiating between sim-
ulations of the MMP catalytic domain alone and the full-length
enzyme in the presence of a peptide substrate, inhibitor, or
collagen-like molecule.
2 Materials
The three main components required to run an MD simulation are

the initial spatial coordinates of the system of interest (e.g., a par-
ticular MMP), software for implementing the various methods,
and hardware for performing the calculations.
2.1 Initial Structure To start an MD simulation, initial coordinates for all the atoms in
the desired MMP should be provided to the minimization/simula-
tion packages.
1. Check the Protein Data Bank for available structures and select
a suitable X-ray crystal structure with a high resolution.
Currently, structural information is only available for 12 of the
23 human MMPs: MMP-1 (catalytic domain and full-length
enzyme), MMP-2 (catalytic domain, fibronectin modules,
hemopexin domain, and full-length enzyme), MMP-3 (cata-
lytic domain), MMP-7 (catalytic domain), MMP-8 (catalytic
domain), MMP-9 (catalytic-fibronectin domains, hemopexin
domain, and a model for the full-length enzyme), MMP-10
(catalytic domain), MMP-12 (catalytic domain, hemopexin
domain, and full-length enzyme), MMP-13 (catalytic domain,
hemopexin domain, and full-length enzyme), MMP-14 (cata-
lytic and hemopexin domains), MMP-16 (catalytic domain),
and MMP-20 (catalytic domain). For the remaining MMPs,
homology modeling may provide a convenient initial structure
provided the homology models are carefully validated (see
Note 1).
2. Edit the initial structure by removing additional molecules,
reverse mutations (e.g., the catalytic glutamic acid), and select-
ing appropriate protonation states for the titratable amino-
acids (see Note 2). Take into account that the catalytic Zn1 is
coordinated by three histidines through their side chain Nε
atom, but Zn2 is bound to two Nε and one Nδ atoms of three
different histidines.
3. Refine the structure by adding missing atoms/residues and
optimize the conformational state of some side chains (i.e.,
His, Asn, Gln) (see Note 3).
4. The use of explicit solvent in extended molecular simulations
of MMPs is highly recommended to realistically account for
both entropic and enthalpic solvent effects, which can influ-
ence inter and intra-domain conformational changes as well as
ligand binding. Therefore, the MMP molecule should be
placed in an appropriate solvent box that extends at least 15 Å
from the protein atoms. Neutralize the system by randomly
adding Na+/Cl− counterions.
2.2 Software Fortunately, a variety of programs are readily available for prepar-
ing and running MD simulations of biomolecules assuming an all
atom representation. Our preferred choice for studying MMPs is
the AMBER (Assisted Model Building with Energy Refinement)
package, which integrates a collection of programs to set up, per-
form, and analyze MD simulations [32]. AMBER can be used with
several types of molecular mechanics (MM) methods or force fields
for calculating potential energies and forces for large biomolecules,

small organic molecules, and typical solvents. For extended MD
simulations of MMP systems, we recommend the family of amber
force fields since previous experience shows that these potentials
exhibit a favorable tradeoff between the quality of their structural
and energetic predictions and their computational cost. Later ver-
sions of amber force fields, which use fixed partial charges centered
on atoms, are ff14SB for proteins and the general amber force field
(GAFF) for small organic molecules.
Among the numerous programs included in the AMBER pack-
age, the following tools are particularly important:
1. In the preparation toolbox, the program antechamber gener-
ates parameter files for nonstandard residues (e.g., zinc coordi-
nated histidines) and ligands (e.g., MMP inhibitors), whereas
LEaP creates the coordinate and topology files required for
performing energy calculations and MD simulations.
2. The MD simulation programs comprise sander and pmemd.
Sander implements many simulation techniques and potential-
energy methods including hybrid quantum mechanical/
molecular mechanical methods (QM/MM), while pmemd
maximizes the performance of the most used simulation tech-
niques combined with classical MM force fields.
3. Structure and dynamic analyses of MD trajectories can be car-
ried out using cpptraj, a powerful tool available in the latest
versions of AMBER. In addition, AMBER offers various tools
(e.g., MMPBSA.py) to extract energetic information from MD
trajectories using end-point free energy methods.
4. For those learning to use the software, various aids are avail-
able on the AMBER website (http://ambermd.org/) such as
an easy to read but comprehensive manual, an updated mailing
list, and numerous tutorials.
Other popular packages such as CHARMM (http://www.
charmm.org/), NAMD (http://www.ks.uiuc.edu/Research/
namd/), or GROMACS (http://www.gromacs.org/) are suitable
alternatives in terms of functionality, efficiency, and license cost.
Besides the adoption of a reference package like AMBER, a research
project on molecular modeling of MMPs will likely demand the
use of quantum mechanical (QM) methods and the execution of
other miscellaneous tasks for the preparation/analysis of the simu-
lations that will require further specialized software tools (see
Subheading 3).
2.3 Hardware Computational algorithms and hardware technologies have experi-

enced impressive developments during the last years that bring
high performance computing (HPC) to workstations affordable
for a wide range of researchers. Recent developments in Graphic
Processing Units (GPUs) have led to a dramatic acceleration of

MD simulations. In this respect, it should be noted that pmemd
efficiently runs on GPUs and is currently one of the best programs
for running extended MD simulations in a timely manner.
Nevertheless, efficient editing and parameterization, as well as the
postprocessing of MD trajectories for simulations run over
extended periods, demands more computational resources in terms
of conventional central processing units (CPUs), rapid access
memory (RAM), high performance storage on multiple hard
drives, graphic cards for visualization, etc.
The recommended starting configuration for a workstation or a
computer server dedicated to perform molecular simulations of sys-
tems with the typical size of MMPs, would include a dual socket
motherboard equipped with two multicore CPUs, 4 GB of RAM
per CPU core and a distributed storage capacity greater than 2 TB
constituted by SAS/SSD drives in a RAID device. The same com-
puter should be equipped with at least two state-of-the-art NVIDIA
GPU cards (currently Maxwell GPUs). Assuming that a Linux oper-
ating system (e.g., CENTOS, Scientific Linux, etc.) and the AMBER
package are installed, such a workstation/server machine would be
able to simultaneously carry out production runs on the GPUs with
minimum CPU usage and other tasks for the preparation and analy-
sis of MD trajectories using the remaining RAM and CPU cores.
Further information pertaining to the recommended hardware can
be found on the AMBER website (http://ambermd.org/gpus). Of
course, all major HPC hardware vendors offer workstations and
clusters suitable for performing MD simulations.
3 Methods
In this section, we comment on selected methodological and tech-

nical issues that should be addressed to set up and run a molecular
simulation for a particular MMP either in its unbound form or in
complex with inhibitor molecules, short peptide substrates, or
triple-
helical substrates mimicking collagen molecules.
Consideration of specific points to be discussed here, which have
been primarily chosen on the basis of our previous experience in
the molecular modeling of MMPs, may be helpful in the design of
future studies.
3.1 MM The MMP catalytic domain binds two zinc ions (Zn1 catalytic and
Parametrization Zn2 structural) and a number of calcium ions that are structurally
of the Coordination and functionally relevant. To describe these metal ions in classical
Environment of Metal force fields for extended MD simulations, we assign these metal
Ions ions to either a nonbonded or a bonded representation.
1. In the nonbonded form, only the total charge (+2) and van der
Waals parameters available in the literature are required to
describe these ions. We suggest using the nonbonding param-
eters developed by Li et al. that reproduce experimental ion-
oxygen distance values and coordination number for the first
solvation shell [33]. This works well for calcium ions but it is
not recommended for describing the coordination environ-
ment surrounding the zinc ions.
2. Different coordination spheres have been observed for zinc
ions in MMPs (e.g., tetrahedral for the structural and the pro-
peptide bound catalytic zincs, distorted trigonal bipyramid for
the hydroxamate inhibited catalytic zinc, etc.) which, in prin-
ciple, can be well represented in the bonded metal-ligand con-
figuration (see Note 4).
3. If there is incomplete experimental data to define the proton-
ation state of the zinc-binding groups (ZBGs) and the essential
Glu residue, and/or the number and position of water mole-
cules, preliminary QM/MM studies should be carried out to
determine the most likely configuration of the zinc sites [34].
For example, by initially assigning a nonbonded conformation
for the zinc ion in AMBER, it is relatively simple to define a
QM region comprising the catalytic Zn1 ion, closest residues,
and ligand/water molecules for each configuration. The QM
region is described by either one of the semiempirical QM
methods implemented in sander or a more sophisticated
method using an external QM code (Gaussian, ORCA,
GAMESS; see Note 5). Subsequent energy minimization of
the QM region followed by structural comparisons with the
experimental structures and/or energy comparison allow for
discrimination between various configurations.
4. Equilibrium values and force constants for all zinc-ligand
bonds and angles should be provided in the bound configura-
tion. In order to do this we usually carry out QM calculations
on realistic cluster models of the zinc coordination spheres
(Fig. 1). For example, the selected cluster model for the struc-
tural Zn2 should comprise the side chains of the coordinating
ligands (three histidines and an aspartic acid). On the other
hand, for the active site Zn1, it is advisable to include the cata-
lytic glutamic acid located in the second coordination shell and
other relevant ligands (e.g., water molecules, inhibitor ZBGs,
etc.). Bond and angle reference values involving zinc can be
directly taken from the optimized cluster models. To compute
force constants, perturbed structures should be generated by
selectively modifying all the bond distances and angles involv-
ing the zinc ion and fitting the relative energies to a second-
order polynomial [35]. The torsion parameters associated with
the zinc-ligand interactions can be set to zero.
Fig. 1 Ball-and-stick representation of a cluster model used to parametrize the Zn1 ion following the bonded
approach. Selected atomic charges and MM parameters (r: reference bond distance, θ: reference bond angle,
Kr: bond force constant, Kθ: angle force constant for Zn1–L bonds and the corresponding angles) as obtained
from quantum mechanical (B3LYP/6-31G*) calculations are included. An example of curve fitting to obtain a
force constant is also shown for the His2-Nε⋯Zn1 bond (y: relative energy of the perturbed structure with
respect to the unperturbed one; x: distance perturbation)
5. To properly describe metal-ligand charge transfer effects in the

zinc-bound conformation, atomic charges from standard force
fields should be avoided for the zinc ions and their zinc-bind-
ing groups. New partial charges must be derived by fitting to
the QM electrostatic potential computed at a theoretical level
compatible with the selected force field (e.g., a density func-
tional theory method like B3LYP combined with a double-ζ
basis set as 6-31G+*; HF-based charges are not recommended
because they underestimate Zn-ligand charge-transfer effects).
The electrostatic potential should be computed for the opti-
mized QM cluster model used to derive metal-ligand force-
constants or on the isolated QM region extracted from a QM/
MM model. During the charge fitting procedure, the global
charges of those capping residues (acetyl or N-methylamide)
or H-atoms included in the cluster model to saturate open ter-
mini should be adjusted to zero or to the existing values in the
chosen force field.
In line with the residue-based approach characteristic of MM
methods, partial atomic charges and force constants for metal clus-
ters should be transferable among MMPs. Thus, parameter sets
cited for some MMPs in the literature may be applicable to simula-
tions involving other MMPs. Alternatively, the latest version of
AMBER provides a semi-automated procedure, named MCPB for
Metal Center Parameter Builder, for building a bound representa-

tion of a metal ion center within a metalloprotein (see Note 6).
3.2 Simulation of To date, most MD simulations of MMPs have focused on the cata-
the Catalytic Domain lytic domain, mainly examining the binding of active site inhibi-
tors. We believe that future studies of the catalytic domain should
carefully address the following points:
1. The N-terminal coil in most MMP catalytic domain crystal
structures is either longer or shorter than the biologically rel-
evant form because its length depends on the protocol used for
the activation of the pro-MMP. This disparity prevents the
binding of the N-terminal portion to the surface of the cata-
lytic domain via the formation of a salt bridge with a conserved
aspartic acid residue (Fig. 2) and leads to a reduction in the
catalytic activity of the enzyme. Thus, for MD simulations of
MMPs it is important to select a structure with the correct
N-terminus, usually a tyrosine or a phenylalanine, and a prop-
erly positioned N-terminal coil. If this physiologically relevant
(superactivated) structure is not available for a particular MMP,
it can be easily generated by superposition with structures for
other MMPs.
2. The number of bound calcium ions fluctuates between one
and three, depending on the crystal structure chosen (Fig. 3).
These calcium ions interact with several loops and turns and
contribute toward generating a catalytically competent enzyme.
For MMP-2, previous MD simulations followed by approxi-
mate free energy calculations have found that the binding of
Fig. 2 Ribbon representation of the MMP-3 catalytic domain generated for different crystal structures with PDB
codes 1SLM (a), 1CAQ (b), and 1HFS (c) illustrating the different positions of the N-terminal coil (in orange).
The side chains of the conserved aspartic acid (Asp237) and N-terminal residues are shown as sticks. Zinc and
calcium ions are depicted as spheres (magenta and green, respectively) and the active site histidines as sticks
two calcium ions is energetically favored [21]. However, it is

advisable to carefully determine the number and location of
calcium ions to include in future simulations of other MMP
catalytic domains. This can be done by using a similar approach
to that mentioned above of extended MD simulations and
energetic analyses.
3. The catalytic domain of the gelatinases MMP-2 and MMP-9
has an insertion of around 200 amino-acid residues corre-
sponding to the three fibronectin-like modules. If needed, the
fibronectin insertion can be replaced by a short connecting
peptide to reduce the size of the molecule, without altering the
structural and dynamic properties of the catalytic domain. In
fact, crystal structures are available for both MMP-2 and
MMP-9 with and without the fibronectin modules.
3.3 Binding of Small Numerous MD simulations have previously analyzed the binding
Ligands to the Active of small peptide substrates and inhibitors within the active site of
Site various MMPs. Moreover, MD simulations of the Michaelis com-
plex formed between the MMP catalytic domain and a substrate of
interest must be considered a prerequisite prior to attempting to
study the mechanism of catalysis (see Note 7).
1. The initial MMP/peptide complex can be built taking into
account previously published data showing that small peptide
substrates extend within the active site and make contact via a
number of polar and hydrophobic interactions with the S3–S3′
binding sites. Several crystal structures are available that can
act as a template for initially binding a short peptide around
the catalytic zinc ion (Fig. 4). For this purpose, use a program
Fig. 3 Ribbon representation of the MMP-13 catalytic domain generated using different crystal structures with
PDB codes 1YOU (a), 1XUC (b), and 1ZTQ (c). Zinc and calcium ions are depicted as spheres (magenta and
green, respectively) and the side chains of peptide-binding residues as sticks
for interactive visualization and editing of molecular structures

(e.g., chimera: http://www.cgl.ucsf.edu/chimera/). For
larger substrates, the conformation of the N- and/or C-terminal
portions of the peptide can be properly sampled using a con-
formational search algorithm prior to production MD simula-
tions. For example, the Low MODe (LMOD) algorithm in
sander performs a conformational search by applying structural
perturbations to a molecular system derived from low fre-
quency vibrational modes representing large amplitude con-
certed motions.
2. Automatic docking programs like autodock4.0 (http://
autodock.scripps.edu/wiki/AutoDock4) can be used to bind
novel inhibitors within the active site. The parameters for these
docking programs usually require some adjustment to take
into account interactions between the zinc ion and the inhibi-
tor, which largely control binding to the whole molecule.
Normally, it is easy to recognize the ZBG of the inhibitor, but
it is also important to realize that some ZBGs can modify their
protonation state upon going from solution into the active site
(e.g., the hydroxamate group transfers a proton to the active
site glutamate residue when binding to MMP-2 and MMP-9).
Make sure that the catalytic zinc ion is included in the docking
calculations and select a proper configuration (i.e., atom types
and partial charges) for the ZBG and other active sites. This is
readily accomplished in autodock4.0 since this program allows
for defining different protonation states and it is possible to
modify the partial charges of selected residues. Alternatively,
Fig. 4 Ribbon representation of the catalytic domains of MMP-12 (a, b) and MMP-2 (c) generated from crystal
structures with PDB codes 2OXW, 2OXZ, and 1CK7, respectively. Ligands bound within the active site are illus-
trated as sticks with carbon atoms colored green. The ligand sequence and position within the active site is
also indicated below each structure. Zinc and calcium ions are depicted as spheres (magenta and green,
respectively) and the side chains of the Zn1 binding protein residues as sticks
use the new force field Autodock4Zn, which includes a special-

ized potential that describes the interactions of zinc coordinat-
ing ligands [36].
3. Protein flexibility can be effectively included during docking
calculations by considering different protein configurations. It
is highly recommended to select a set of 25–50 snapshots from
a previous MD simulation of the desired catalytic domain
(preferably in complex with another ligand). Extract the coor-
dinates of the catalytic domain alone and explore the docking
of the inhibitor to the different protein frames. Due to the
inherent plasticity of the Ω-loop and the induced fit effects
observed in the S1′ pocket, the use of different protein con-
figurations can be indispensable for docking inhibitors with
bulky side chains.
3.4 Binding Large and structurally complex biomolecules like those based on
of Collagen-like the collagen triple helix are among the natural substrates of the
Substrates MMPs. For collagen-like substrates the complexity of both the
substrate and the receptor (enzyme) increases the level of difficulty
in the preparation and analysis of molecular simulations involving
MMP complexes. Although such studies are particularly challeng-
ing, they are potentially more interesting. Therefore, in this sub-
section, we give some broad guidelines that will hopefully
complement procedures available in the literature.
1. Initial coordinates for triple helical peptides (THPs) contain-
ing standard amino-acid residues can be obtained from the
amino acid sequence using an interactive triple helical collagen
building script termed THe BuScr. (http://structbio.biochem.
dal.ca/jrainey/THeBuScr.html). Nevertheless, an initial MD
simulation of the solvated helix is required to properly relax
the system, especially if nonstandard amino-acid residues need
to be included in the THP model.
2. The triple helix must locally unfold to accommodate one of its
chains within the narrow active site of a MMP. Targeted MD
simulations can be used to locally force one of the three chains
into a more extended configuration and relax the whole triple
helix. This transition can be smoothly driven along the tar-
geted MD simulation by defining a harmonic penalty based on
the root mean squared deviation of the backbone residues
located around the scissile peptide bond. The reference struc-
ture used for the constraint can be taken from a previous MD
simulation of the complex formed between the MMP and a
small linear peptide substrate with the appropriate sequence.
3. We recommend implementing a semiflexible docking protocol
combining various software tools (e.g., cpptraj, sander,
Mmpbsa.py) using linux shell scripts to dock the distorted triple
helix within the active site of a MMP originally in complex

with a smaller ligand as well as to partially relax the most severe
steric clashes between the MMP and the THP (Fig. 5). First,
MMP and THP coordinates, respectively, are retrieved from
equally spaced snapshots from the targeted MD simulation of
the distorted triple helix and from a previous MMP/peptide
trajectory. The two sets of coordinates are combined by super-
posing the backbone coordinates of the triple helix around the
scissile peptide bond onto the corresponding ones for the lin-
ear peptide substrate, thus generating a large set of MMP/
THP models. Energy minimization and short MD simulations
at high temperature (500 K) are used to relax steric clashes in
the MMP/THP models by promoting uphill moves of bulky
side chains (at this stage only residues involved in steric clashes
are allowed to move). Finally, the relaxed MMP/THP com-
plexes are scored by calculating the MMP⋯THP interaction
energies using the MM-PBSA method.
4. Alternatively, several MMP/THP complexes have previously
been published and can be used as templates in future studies
examining other MMPs and/or triple helices. To date, several
structures are available, including a crystal structure for the
interaction of full-length MMP-1 with collagen in a nonreac-
Fig. 5 Schematic representation of the semiflexible protocol proposed for docking a THP molecule within the
active site of an MMP. Initial coordinates for the MMP/THP complexes (11, 12, …,1N ,…, MN) come from the
superposition of M snapshots obtained from the MD simulation of the distorted THP onto N snapshots extracted
from the trajectory computed for the MMP bound to a short linear peptide. After relaxing the steric clashes and
scoring the complexes with MMPBSA energies, the most stable MMP/THP complexes are structurally evalu-
ated using the whatcheck software tool (http://swift.cmbi.ru.nl/gv/whatcheck/) prior to run extended MD
simulations
tive complex (4AUO) and computational models for several

MMP-2/THP pre-reactive complexes [25, 26, 37]. Starting
from these templates, molecular editing software can also be
used to build initial structures for other complexes.
3.5 Modeling MD simulations of full-length MMPs can be used to gain detailed

of the Full-Length insights into the nature and location of exosites, the relative mobil-
Enzyme ity of the different domains, or the interaction of MMPs with mac-
romolecular substrates. Preparation of simulations to address these
questions is not trivial and the following points should be taken
into account:
1. Crystal structures for full-length enzymes are only available for
MMP-1, MMP-2, MMP-12, and MMP-13. For other MMPs,
molecular models could be generated from crystal structures
that are available for the isolated domains by assuming a rela-
tive orientation similar to that observed for the full-length
enzyme. However, the linker segment connecting the catalytic
and hemopexin-like domains is quite variable in length (from
18 amino-acid residues in MMP-21 to 67 in MMP-15) and
composition. The secondary structure of this linker segment
can be predicted using tools such as BioAssemblyModeler
BAM (http://dunbrack.fccc.edu/BAM/). It is worth noting
that this is also the case for MMP-2 since the linker in the crys-
tal structures for this enzyme has missing residues (8–11
depending on the selected structure).
2. In general, full-length MMPs are assumed to experience some
degree of relative inter-domain motion, so sufficiently large
solvent boxes should be employed for these MD simulations
(see Note 8). Moreover, extensive simulations (at least hun-
dreds of nanoseconds) and/or the use of enhanced sampling
techniques are mandatory to investigate inter-domain confor-
mational freedom.
3. Protein-protein docking calculations may be used to generate
alternative conformations of the full-length enzyme. By con-
sidering the catalytic and hemopexin-like domains as indepen-
dent units, it is possible to use programs like pyDock
implemented in the pyDockWeb (http://life.bsc.es/servlet/
pydock/home/) to provide the best rigid-body docking orien-
tations of two fragments. Alternatively, pyDockTET has been
developed to obtain domain-domain assemblies by scoring the
domain-domain poses in terms of binding energies and a
pseudo-energy term based on restraints derived from linker
end-to-end distances. The energetically most favored poses
should then be sampled using extended MD simulations to
relax the system and analyze the strength of the inter-domain
interactions.
3.6 Running the MD To properly run an MD simulation for MMP systems in explicit
Simulations solvent, the following steps should be sequentially performed to
relax and thermalize the system:
1. Solvent boxes around macromolecules are usually generated
by replicating the coordinates of a small and pre-equilibrated
box in three directions. As a result, the large solvent box sur-
rounding an MMP molecule contains vacuum bubbles at the
corners of the interfaces formed by the smaller boxes. Thus,
before starting the simulation for the entire system we first
relax the large solvent box. This can be done by performing an
initial energy minimization, a short MD simulation at constant
pressure, and a final minimization of the solvent molecules
(counterions can be also included in these steps). In this way,
the solvent molecules are allowed to adapt to the solute (i.e.,
the protein) and the volume of the box changes to the correct
density.
2. The whole system should be minimized to remove potentially
bad contacts in the initial MMP structure. If needed, this can
be done in several steps relaxing first those parts of the system
with more severe steric clashes.
3. The system is then gradually heated from 0 K to the desired
simulation temperature, usually 300 K. To avoid any problem,
heating is best done by running a number of short MD simula-
tions to progressively increase the target temperature of the
system. We usually employ six constant volume simulations of
10 ps each to increase the temperature according to the fol-
lowing sequence 0 K → 50 K → 100 K → 150 K → 200 K →
250 K → 300 K.
4. After thermalization, the actual MD simulation can be started
by choosing either constant volume (NVT) or constant pres-
sure (NPT) conditions. If using NVT conditions, we recom-
mend running a preliminary pressurization job under NPT
conditions (1–2 ns).
5. During the MD simulation of the MMP, it is necessary to dis-
criminate between equilibration and production phases. In
general simulations begin in a state (edited X-ray structure,
homology model, docking model, etc.) that can be structurally
different from the one that you wish to sample and, therefore,
the first part of the MD trajectory is rather sensitive to the ini-
tial conditions and to singular transitions occurring during the
structural relaxation of the system. Thus, during the equilibra-
tion phase, which can extend for tens or even hundreds of ns in
the case of full-length MMPs, the structure and energy of the
system normally evolve to reach a relatively stable plateau.
Sampling during the production phase of the simulation occurs
along this plateau where the system is in equilibrium (Fig. 6).
Fig. 6 Time evolution of (a) RMSD and (b) MM-PB energy for the MD simulation of full-length MMP-2 (PDB ID:
1CK7). The vertical blue line separates the equilibration and production phases. The red and green curves
represent the evolution of the accumulated average and the adjacent average, respectively
6. From our experience MD simulations involving the MMP cat-

alytic domain alone require at least tens of ns to sample the
flexibility of the important Ω-loop, while simulations of full-
length MMPs unbound or in complex with THP molecules
should extend to hundreds of ns to properly explore the con-
formational landscape of slow inter-domain motions.
3.7 Structural Structural analyses of MD simulations can be used to characterize

Analysis of MD different aspects of MMP structure and flexibility. Depending on
Simulations the particular objectives, different kinds of analyses can be per-
formed using a wide variety of tools. But, in general, the following
approaches may be useful for extracting relevant information from
MMP simulations:
1. Along with the conventional structural analysis, inspection of
the time evolution of the root mean squared deviation (RMSD)
with respect to a reference structure (usually the initial struc-
ture) allows for the selection of the production phase for fur-
ther analysis. Besides computing RMSDs for the whole protein
it is advisable to segregate it into secondary (helices, β-sheets,
loops) and tertiary (domains) structural elements.
2. Relevant structural elements within the MMP catalytic domain
include the long Ω-loop and the hydrophobic S1′ binding site.
To characterize the structure and flexibility of the Ω-loop, clus-
tering analysis based on RMSD values for Cα and Cβ Cartesian
coordinates or backbone torsion angles can be performed on
selected snapshots extracted from the production phase of the
MD trajectory. This can be achieved using the cpptraj program
in the AMBER package or other programs like the MMTSB-
tool set (http://blue11.bch.msu.edu/mmtsb/Cluster.pl).
3. Accessibility to different binding sites can be characterized by
computing the radius of accessibility for selected backbone and
side chain groups located within the active site cleft. The radius
of accessibility [21] is defined as the maximum radius of a
spherical ligand that can touch the desired target and its deter-
mination is based on the rapid calculation of the molecular
surface using programs such as MSMS (http://mgl.scripps.
edu/people/sanner/html/msms_home.html).
4. The evolution of the size and shape of cavities in the MMPs
like the S1′ pocket during the MD simulation can be moni-
tored using a program like mdpocket (http://fpocket.source-
forge.net/).
5. Interdomain and enzyme-ligand contacts include H-bonds
and hydrophobic interactions. H-bonds can be characterized
on the basis of geometrical criteria (e.g., X⋯Y distance <3.5 Å
and X–H⋯Y angle >120°; cpptraj has different commands to
search for and characterize H-bond contacts). In addition,
nonpolar interactions can be detected by measuring the dis-
tance between the centers of mass of two interacting groups.
We have also used in-house developed software (available upon
request) to evaluate the empirical dispersion attraction between
pairs of atoms belonging to two hydrophobic groups.
3.8 Energetic Relative free energies can be estimated from conventional MD sim-
Analysis of MD ulations using end-point free energy methods [38, 39] to evaluate
Trajectories the inhibitory ability of structurally dissimilar compounds toward
MMPs and/or to discriminate among different configurations or
inter-domain arrangements of full-length MMPs. In particular, the
MM-PB(GB)SA (molecular mechanics-Poisson Boltzmann or
generalized Born surface area) method offers a good balance
between speed and accuracy and has been widely used to compute
relative binding affinities [11, 40, 41]. However, there are differ-
ent variants of this method and various protocols describing its
application. We comment below on options that particularly apply
to MMPs.
1. Extract a minimum of 500 snapshots separated by 25–50 ps
from the production phase. Compute solvation energies using
the Poisson-Boltzman (PB) method (this method is especially
applicable for large systems). Both linear and nonlinear solvers
of the PB equation are implemented in the program pbsa that
is included in the AMBER package.
2. The calculation of binding energies for ligands bearing ZBGs
with a strong affinity for the catalytic Zn ion can be problem-
atic. First, a nonbonded approach for the Zn environment
should be used for the MM-PBSA calculations. However, the
purely MM representation of the Zn···ZBG interaction can
result in unbalanced interaction energies and, consequently, a
careful validation of the relative MM-PBSA energies would be
necessary and/or alternative approaches using semiempirical

QM methods (e.g., QM-GBSA) could be employed.
3. In the case of small and rigid ligands bound to MMPs, it can
be acceptable to approximate the full binding energy (ΔbindG)
by the interaction energy (ΔintG) evaluated using the dis-
torted geometries of the protein and ligand moieties
extracted from the MD simulation of the complex (i.e., the
so-called one-trajectory approximation). This approach min-
imizes the statistical uncertainty of the average energy differ-
ences and facilitates the estimation of relative binding
affinities for a series of ligand molecules. Entropic correc-
tions may be omitted.
4. For complexes involving the MMP catalytic domain and a
flexible peptide or peptide-mimetic ligands, conventional
MM-PBSA calculations will suffice provided that a mixed
bonded and nonbonded parameterization of the catalytic Zn
site is used (e.g., the direct or water-mediated
Zn···O=C(peptide) can be reasonably described by non-
bonding MM terms; see Note 4). However, due to the
intrinsic flexibility of peptides, it is necessary to estimate the
configurational entropy [42] change associated with the
binding of the ligand. To do this we recommend performing
energy minimizations and normal mode calculations for a
subset of MD snapshots when using the AMBER MM-PBSA
tools. These calculations yield absolute entropy terms based
on the rigid-rotor and harmonic oscillator approximations
that can be improved by adding the purely conformational
entropy of the flexible ligand computed with the program
cencalc (http://sourceforge.net/projects/cencalc/).
Relative binding energies must be obtained by combining
the MM-PBSA energies from independent trajectories of the
MMP/peptide and isolated peptide molecules (e.g., Pep1 +
MMP/Pep2 → Pep2 + MMP/Pep1) [24].
5. The MM-PBSA method is probably the only viable approach
to assess the relative stability and binding affinity of various
configurations of full-length MMPs and MMPs complexes
with large THP substrates [26]. As mentioned earlier, relative
MM-PBSA binding energies must be estimated using indepen-
dent trajectories because interaction energies from the one-
trajectory approximation ignore distortion effects, which are
important in large biomolecular systems. Configurational
entropy can be estimated using the quasi-harmonic method as
implemented in the program cpptraj. This method is very
computationally efficient and facilitates an estimation of the
absolute entropy for large macromolecules. However, it largely
overestimates the configurational entropy and provides only an
insight into the entropy effects at work.
6. For extended MD simulations of full-length MMPs, the MM-

PBSA energies oscillate widely through fast sub-nanosecond
motions and can also exhibit lower-amplitude oscillations on
longer timescales (~ns), as shown by the evolution of the adja-
cent average energies in Fig. 6. In this case, it is recommended
to assess the statistical uncertainty of the MM-PBSA values by
reporting both the standard error and the estimated block-
averaging error.
The application of rigorous free energy methods, such as
Thermodynamic integration (TI) and free energy perturbation
(FEP), is beyond the scope of this chapter. Nevertheless, we note
that these methods are of limited applicability due to their high
computational cost and convergence problems. In principle, they
can compute free energy changes for any chemical transformation
occurring within the system. But in practice, only very limited
changes in the structure of an MMP-bound inhibitor can be toler-
ated to achieve convergence of the relative binding affinities within
an affordable sampling [43]. Furthermore, Ω-loop flexibility and
induced-fit effects that are experimentally observed for different
MMP/inhibitor complexes require extended simulation times for
TI energy calculations [43].
4 Supporting Information
Readers interested in practicing the preparation and execution of

an MD simulation can download a tar file (mini_howto_mmp12_
hae_MD.tar) containing a tutorial for MD simulations of the
catalytic domain of the human MMP-12 enzyme complexed with
acetohydroxamic acid (PDB code: 1Y93).
5 Notes
1. Due to the structural similarity of the catalytic and hemopexin

domains of various MMPs, it is possible to build an initial
model using homology modeling for MMPs with no known
three-dimensional structure. The amino acid sequences for the
MMP of interest can be obtained from the UniProt database
(http://www.uniprot.org/) and aligned against sequences in
the PDB database using BLAST to find proteins with similar
sequences and known structure. Based on sequence similarity,
structure coverage, and crystal structure resolution, three or
four templates should be selected to perform the comparative
modeling using a software tool such as MODELLER (https://
salilab.org/modeller/).
2. Ionization states for titratable residues can be selected based on

approximate pKa calculations using the H++ web server (http://
biophysics.cs.vt.edu/H++). Only reasonably high resolution
(i.e., <2.5 Å) crystal structures should be used in these calcula-
tions. Besides the structure, the user should provide a value for
the dielectric constant within the protein. An interior dielectric
constant of 10 is usually recommended though lower values of
4 and 6 seem to provide better results for buried side chains.
Thus, it may be a good strategy to examine the consistency of
the predicted pKa values by comparing the results obtained with
two or three high quality structures and a few values (i.e., 4, 6,
10, and 20) for the interior dielectric constant. To include the
zinc and calcium ions in the pKa calculations, edit the input
PDB file to change “HETATM” to “ATOM” and use “CA” as
residue name and atom type for calcium, and “ZN” as residue
name and atom type for zinc. Alternatively, “manually” include
the lines for zinc and calcium in the PQR file (a PDB file includ-
ing atomic charges and bondi radii) which is generated by a
preliminary H++ execution retrieving the structure from the
protein data bank. In this case, charges (+2.0) and radii (1.00 for
Zn2+ and 1.70 for Ca2+) should be provided.
3. Building and refinement of side chain conformations in pro-
teins can be easily done using the program SCWRL4 devel-
oped in the Dunbrack Lab (http://dunbrack.fccc.edu/
scwrl4/). This refinement is particularly important for the
asparagine and glutamine side chains because they may be
flipped with respect to their most favored conformation in
crystal structures. The program also evaluates the protonation
of Nδ or Nε atoms for histidine residues.
4. A hybrid bonded and nonbonded representation has been pre-
viously employed to model the catalytic zinc in MMP-2 bound
to different peptide substrates [22]. In this hybrid model,
explicit Zn-L bonds were defined to the three coordinating
histidines and to the hydrolytic water molecule, but no explicit
bond was defined to the carbonyl group of the scissile peptide
bond that acts as the fifth zinc ligand. This hybrid description
results in a stable coordination sphere around the metal and, at
the same time, is flexible enough to allow for some fluctuations
over important reactive distances (e.g., Zn–O⋯C-peptide for
the nucleophilic attack and Glu–COOH⋯N-peptide for the
protonation of the leaving group).
5. The flexible binding environment of the MMP catalytic Zn
ion, which can accommodate a large variety of ZBGs in poten-
tial inhibitors as well as to bind and hydrolyze peptide sub-
strates, explains why both QM and QM/MM methods play a
prominent role in molecular simulation studies of MMPs. As
described in Subheading 3, state-of-the-art QM methods are
used to obtain the reference data necessary for deriving the

MM parameters for the Zn1 site in the presence of different
ligands and ZBGs using the bonded approach. Parameterization
using these QM methods can be computationally costly due to
the levels of theory used in these methods (i.e., the still popu-
lar B3LYP method combined with a double-ζ basis set). Of
course, specialized QM codes like ORCA (https://orcaforum.
cec.mpg.de/), Gaussian (http://www.gaussian.com/), or
GAMESS (http://www.msg.ameslab.gov/gamess/) are
needed. For beginners, we highly recommend ORCA since
this program offers a wide variety of QM methods and is very
efficient. It is also distributed freely for academic purposes and
comes with an excellent manual. Importantly, ORCA and
other programs are readily recognized by the QM interface
implemented in the sander program, thus facilitating the exe-
cution of high level QM/MM calculations (if desired, the
sander interface can be used to drive full QM single-point or
geometry optimizations).
Researchers interested in MMP modeling should be aware of
approximated QM methods like the Density Functional based
Tight Binding method using a self-consistent redistribution of
Mulliken charges (this is the so-called SCC-DFTB method;
http://www.dftb.org), which can yield reasonably accurate
energies and forces much faster than conventional QM meth-
ods provided that pairwise atomic parameters are available for
the system of interest. Over the last decade many SCC-DFTB
applications on Zn enzymes have been reported and perform
remarkably well [44]. This has prompted software developers to
implement SCC-DFTB/MM simulation methods to carry out
energy minimization, molecular dynamics in explicit solvent
and potential of mean force calculations. Thus, latter versions of
the sander program include a second-order SCC-DFTB imple-
mentation that can be useful for studying both ligand binding
and chemical reactivity in MMPs (we recommend, however,
validating the structural and energetic SCC-DFTB data by
comparison with conventional QM/MM calculations). The
current SCC-DFTB implementation can be used to produce
MD trajectories on the ps-ns time scales rather than the ns-μs
scales that can be reached by classical MM jobs on GPUs. In
addition, the ability of SCC-DFTB to model flexible metal
coordination environments is a major achievement that elimi-
nates the need for obtaining MM parameters for the metal sites
using the rather cumbersome bonded approach. Nevertheless,
the choice between SCC-DFTB and purely MM methods still
depends on the required amount of sampling and the goal of
the simulation. Note also that a third-order DFTB methodol-
ogy having improved parameters for biological and organic
molecules containing the following elements, Br-C-Ca-Cl-F-
H-I-K-Mg-N-Na-O-P-S-Zn, is currently available and will

presumably be accessible in upcoming revisions of the AMBER
package.
It is envisaged that with software tools like AMBER, integra-
tion of QM and SCC-DFTB methods and continuous improve-
ment of hardware technologies will lead to even more realistic
and useful simulations of metallo-enzymes including MMPs.
6. The Metal Center Parameter Builder (MCPB) computes bond
and angle parameters in a sidechain cluster model (i.e., includes
the sidechains of the ligating residues terminated with a methyl
group) and charges in an extended cluster model also includ-
ing the backbone (capped with acetyl and N-methyl groups).
It uses the Gaussian program to perform geometry optimiza-
tion and frequency calculations at the B3LYP/6-31G* level of
theory.
7. Although computational studies of enzymatic catalysis are
beyond the scope of this chapter, it may be interesting to note
that to study the molecular details of MMP-mediated peptide
hydrolysis and/or reaction with ZBGs, it is essential to build a
realistic model of the pre-reactive complex and explore its con-
formational landscape by means of MD simulations. However,
QM calculations on cluster models or static QM/MM calcula-
tions starting from a single geometry provide an incomplete
picture of the reaction mechanism. Such calculations can miss
many subtle details of the reaction mechanism that render
them useless for determining specificity for various substrates.
Therefore, QM/MM studies of reaction profiles should be
started from at least several snapshots selected from the refer-
ence MD trajectory using geometrical criteria.
8. For linear systems like THPs or those that may experience
expansion during an MD simulation, octahedral solvent boxes
are preferred over the rectangular ones to avoid any direct con-
tact between different copies of the protein.
References
1. Butler GS, Overall CM (2009) Updated bio- 4. Fabre B, Ramos A, de Pascual-Teresa B (2014)
logical roles for matrix metalloproteinases and Targeting matrix metalloproteinases: exploring
new “intracellular” substrates revealed by the dynamics of the S1′ pocket in the design of
degradomics. Biochemistry 48:10830–10845 selective, small molecule inhibitors. J Med
2. Fingleton B (2008) MMPs as therapeutic tar- Chem 57:10205–10219
gets—still a viable option? Semin Cell Dev Biol 5. Gioia M, Fasciglione GF, Marini S, D’Alessio
19:61–68 S, de Sanctis G, Diekmann O, Pieper M, Politi
3. Bode W, Maskos K (2003) Structural basis of V, Tschesche H, Coletta M (2002) Modulation
the matrix metalloproteinases and their physi- of the catalytic activity of neutrophil collage-
ological inhibitors, the tissue inhibitors of nase MMP-8 on bovine collagen I. J Biol
metalloproteinases. Biol Chem 384:863–872 Chem 277:23123–23130
6. Overall CM (2002) Molecular determinants of tion of Matrix Metalloproteinase 2: fluctua-

metalloproteinase substrate specificity. Mol tions and time evolution of recognition
Biotechnol 22:51–86 pockets. J Comput Aided Mol Des
7. Bertini I, Fragai M, Luchinat C, Melikian M, 17:837–848
Mylonas E, Sarti N, Svergun DI (2009) 18. Manzetti S, McCulloch DR, Herington AC,
Interdomain flexibility in full-length matrix van der Spoel D (2003) Modeling of enzyme–
metalloproteinase-1 (MMP-1). J Biol Chem substrate complexes for the metalloproteases
284:12821–12828 MMP-3, ADAM-9 and ADAM-10. J Comput
8. Donini OAT, Kollman PA (2000) Calculation Aided Mol Des 17:551–565
and prediction of binding free energies for the 19. Azhagiya Singam ER, Rajapandian V,
matrix metalloproteinases. J Med Chem Subramanian V (2014) Molecular dynamics
43:4180–4188 simulation study on the interaction of collagen-
9. Hou TJ, Zhang W, Xu XJ (2001) Binding like peptides with gelatinase-A (MMP-2).
affinities for a series of selective inhibitors of Biopolymers 101:779–794
gelatinase-A using molecular dynamics with a 20. Kotra LP, Cross JB, Shimura Y, Fridman R,
linear interaction energy approach. J Phys Schlegel HB, Mobashery S (2001) Insight into
Chem B 105:5304–5315 the complex and dynamic process of activation
10. Aschi M, Roccatano D, di Nola A, Gallina C, of matrix metalloproteinases. J Am Chem Soc
Gavuzzo E, Pochetti G, Piepere M, Tscheschee 123:3108–3113
H, Mazza F (2002) Computational study of 21. Díaz N, Suárez D (2007) Molecular dynamics
the catalytic domain of human neutrophil col- simulations of matrix metalloproteinase 2: role
lagenase. Specific role of the S3 and S3′ subsites of the structural metal ions. Biochemistry
in the interaction with a phosphonate inhibitor. 46:8943–8952
J Comput Aided Mol Des 16:213–225 22. Díaz N, Suárez D (2008) Molecular dynamics
11. Hou T, Guo S, Xu X (2002) Predictions of simulations of the active matrix metallopro-
binding of a diverse set of ligands to gelatinase- teinase-2: positioning of the N-terminal frag-
A by a combination of molecular dynamics and ment and binding of a small peptide substrate.
continuum solvent models. J Phys Chem B Proteins 72:50–61
106:5527–5535 23. Díaz N, Suárez D (2008) Peptide hydrolysis
12. Rizzo RC, Toba S, Kuntz ID (2004) A molec- catalyzed by matrix metalloproteinase 2: a
ular basis for the selectivity of thiadiazole urea computational study. J Phys Chem B
inhibitors with stromelysin-1 and gelatinase-A 112:8412–8424
from generalized Born molecular dynamics 24. Díaz N, Suárez D, Suárez E (2010) Kinetic
simulations. J Med Chem 47:3065–3074 and binding effects in peptidesubstrate selectiv-
13. Khandelwal A, Lukacova V, Comez D, Kroll ity of matrix metalloproteinase-2: molecular
DM, Raha S, Balaz S (2005) A combination of dynamics and QM/MM calculations. Proteins
docking, QM/MM methods, and MD simula- 78:1–11
tion for binding affinity estimation of metallo- 25. Díaz N, Suárez D, Valdés H (2013) Unraveling
protein ligands. J Med Chem 48:5437–5447 the molecular structure of the catalytic domain
14. Durrant JD, de Oliveira CAF, McCammon JA of matrix metalloproteinase-2 in complex with
(2010) Including receptor flexibility and a triple-helical peptide by means of molecular
induced fit effects into the design of MMP-2 dynamics simulations. Biochemistry
inhibitors. J Mol Recognit 23:173–182 52:8556–8569
15. Giangreco I, Lattanzi G, Nicolotti O, Catto M, 26. Díaz N, Suárez D (2015) Extensive simula-
Laghezza A, Leonetti F, Stefanachi A, Carotti tions of the full-length matrix metalloprotein-
A (2011) Insights into the complex formed by ase-2 enzyme in a prereactive complex with a
matrix metalloproteinase-2 and alloxan inhibi- collagen triple-helical peptide. Biochemistry
tors: molecular dynamics simulations and free 54:1243–1258
energy calculations. PLoS One 6:e25597 27. Díaz N, Suárez D, Valdés H (2008) From the
16. de Oliveira CAF, Zissen M, Mongon J, X-ray compact structure to the elongated form
McCammon JA (2007) Molecular dynamics of the full-length MMP-2 enzyme in solution:
simulations of metalloproteinases types 2 and 3 a molecular dynamics study. J Am Chem Soc
reveal differences in the dynamic behavior of 130:14070–14071
the S1′ binding pocket. Curr Pharm Des 28. Valdés H, Díaz N, Suárez D, Fernández-Recio
13:3471–3475 J (2010) Interdomain conformations in the
17. Falconi M, Altobelli G, Iovino MC, Politi V, full-length MMP-2 enzyme explored by
Desideri A (2003) Molecular dynamics simula- protein-protein docking calculations using
pyDock. J Chem Theory Comput catalytic mechanism. Theor Chem Acc

6:2204–2213 117:171–181
29. Díaz N, Suárez D (2012) Alternative interdo- 36. Santos-Martins D, Forli S, Ramos MJ, Olson
main configurations of the full-length MMP-2 AJ (2014) AutoDock4Zn: an improved
enzyme explored by molecular dynamics simu- AutoDock force field for small-molecule dock-
lations. J Phys Chem B 116:2677–2686 ing to zinc metalloproteins. J Chem Inf Model
30. Salomon-Ferrer R, Case DA, Walker RC 54:2371–2379
(2013) An overview of the Amber biomolecu- 37. Manka SW, Carafoli F, Visse R, Bihan D,
lar simulation package. WIREs Comput Mol Raynal N, Farndale RW, Murphy G, Enghild
Sci 3:198–210 JJ, Hohenester E, Nagase H (2012) Structural
31. Case DA, Berryman JT, Betz RM, Cerutti DS, insights into triple-helical collagen cleavage by
Cheatham TEI, Darden TA, Duke RE, Giese matrix metalloproteinase. Proc Natl Acad Sci
TJ, Gohlke H, Goetz AW, Homeyer N, Izadi U S A 109:12461–12466
S, Janowski P, Kaus J, Kovalenko A, Lee TS, 38. Gohlke H, Case DA (2004) Converging free
LeGrand S, Li P, Luchko T, Luo R, Madej B, energy estimates: MM-PB(GB)SA studies on
Merz KM, Monard G, Needham P, Nguyen H, the protein-protein complex Ras-Raf.
Nguyen HT, Omelyan I, Onufriev A, Roe DR, J Comput Chem 25:238–250
Roitberg A, Salomon-Ferrer R, Simmerling 39. Homeyer N, Gohlke H (2012) Free energy
CL, Smith W, Swails J, Walker RC, Wang J, calculations by the molecular mechanics
Wolf RM, Wu X, Kollman PA (2015) Poisson-Boltzmann surface area method. Mol
AMBER15. University of California, San Inform 31:114–122
Francisco 40. Zhou Z-G, Yao Q-Z, Lei D, Zhang Q-Q,
32. Case DA, Babin V, Berryman JT, Betz RM, Cai Zhang J (2014) Investigations on the mecha-
Q, Cerutti DS, Cheatham TEI, Darden TA, nisms of interactions between matrix metallo-
Duke RE, Gohlke H, Goetz AW, Gusarov S, proteinase 9 and its flavonoid inhibitors using a
Homeyer N, Janowski P, Kaus J, Kolossváry I, combination of molecular docking, hybrid
Kovalenko A, Lee TS, LeGrand S, Luchko T, quantum mechanical/molecular mechanical
Luo R, Madej B, Merz KM, Paesani F, Roe calculations, and molecular dynamics simula-
DR, Roitberg A, Sagui C, Salomon-Ferrer R, tions. Can J Chem 92:821–830
Seabra G, Simmerling CL, Smith W, Swails J, 41. Marcial BL, Sousa SF, Barbosa IL, Dos Santos
Walker RC, Wang J, Wolf RM, Wu X, Kollman HF, Ramos MJ (2012) Chemically modified
PA (2014) AMBER 14. University of tetracyclines as inhibitors of MMP-2 matrix
California, San Francisco metalloproteinase: a molecular and structural
33. Li P, Roberts BP, Chakravorty DK, Merz KM study. J Phys Chem B 116:13644–13654
(2013) Rational design of particle mesh Ewald 42. Suárez D, Díaz N (2015) Direct methods for
compatible Lennard-Jones parameters for +2 computing single-molecule entropies from
metal cations in explicit solvent. J Chem molecular simulations. WIREs Comput Mol
Theory Comput 9:2733–2748 Sci 5:1–26
34. Díaz N, Suárez D, Sordo TL (2006) Quantum 43. Fabre B, Filipiak K, Díaz N, Zapico JM, Suárez
chemical study on the coordination environ- D, Ramos A, de Pascual-Teresa B (2014) An
ment of the catalytic zinc ion in matrix metal- integrated computational and experimental
loproteinases. J Phys Chem B approach to gaining selectivity for MMP-2
110:24222–24230 within the gelatinase subfamily. Chembiochem
35. Sousa SF, Fernandes PA, Ramos MJ (2007) 15:399–412
Effective tailor-made force field parameteriza- 44. Xu D, Cui Q, Guo H (2014) Quantum
tion of the several Zn coordination environ- mechanical/molecular mechanical studies of
ments in the puzzling FTase enzyme: opening zinc hydrolases. Int Rev Phys Chem 33:1–41
the door to the full understanding of its elusive
Part IV
Determining Matrix Metalloprotease Substrate Specificity

Chapter 8
Determining the Substrate Specificity of Matrix

Metalloproteases using Fluorogenic Peptide Substrates
Maciej J. Stawikowski, Anna M. Knapinska, and Gregg B. Fields
Abstract
A continuous assay method, such as the one that utilizes an increase in fluorescence upon hydrolysis, allows
for rapid and convenient kinetic evaluation of proteases. To better understand MMP behaviors toward
native substrates, a variety of fluorescence resonance energy transfer (FRET)/intramolecular fluorescence
energy transfer (IFET) triple-helical substrates have been constructed to examine the collagenolytic activ-
ity of MMP family members. Results of these studies have been valuable for providing insights into (a) the
relative triple-helical peptidase activities of the various collagenolytic MMPs, (b) the collagen preferences
of these MMPs, and (c) the relative roles of MMP domains and specific residues in efficient collagenolysis.
The present chapter provides an overview of MMP FRET triple-helical substrates and describes how to
construct and utilize these substrates.
Key words Collagen, FRET substrates, Matrix metalloproteinases, Triple-helical substrates,

Fluorogenic substrates
1 Introduction
1.1 Development A continuous assay method, such as the one that utilizes an increase
of Peptide Substrates in fluorescence upon hydrolysis, allows for rapid and convenient
for Matrix kinetic evaluation of proteases, both in solution and cell surface
Metalloproteinases bound. For this reason, significant research efforts have focused on
designing substrates for matrix metalloproteinase (MMP) family
members [1]. Sequence specificity studies for MMPs have been
performed with peptides based on protein sequences surrounding
MMP cleavage sites, and the results of these studies have been
comprehensively reviewed [2, 3]. In addition, phage display and
peptide libraries have been utilized to identify MMP-selective sub-
strate sequences [4–10]. Results of the aforementioned studies
were subsequently used to design fluorogenic MMP substrates.
137
138 Maciej J. Stawikowski et al.
1.2 General Fluorogenic substrates provide a particularly convenient enzyme

Characteristics assay method, as they can be monitored continuously and utilized
of Fluorogenic at reasonably low concentration ranges. There are three types of
Substrates fluorogenic substrates: (1) aromatic amines, (2) contact-quenched,
and (3) resonance energy transfer quenched [11]. MMP fluoro-
genic substrates have been developed using resonance energy
transfer quenching, which are sometimes referred to as intramo-
lecular fluorescence energy transfer substrates (IFETS) [12].
1.2.1 Resonance Energy Förster resonance energy transfer (FRET) may occur between a
Transfer fluorescent donor group and a quenching acceptor group when
the fluorophore has a high quantum yield (ΦF) and a fluorescence
emission spectrum that is exactly overlapped by a strongly absorb-
ing acceptor (quencher) [11]. The efficiency of transfer will also
depend upon the distance between the donor and acceptor [11,
12]. Resonance energy transfer quenched fluorogenic substrates
are thus created by incorporating the donor and acceptor on oppo-
site sides of the scissile bond, at a distance that allows for highly
efficient energy transfer. When the substrate is cleaved by an
enzyme, diffusion of the donor-containing substrate fragment
away from the acceptor-containing substrate fragment results in
loss of energy transfer and subsequent fluorescence. The creation
of an optimal MMP fluorogenic substrate thus depends upon (1)
incorporation of a fluorescent donor that has a high quantum yield,
(2) incorporation of an acceptor that absorbs at the donor fluores-
cence emission wavelength (λem), and (3) the number of amino
acid residues between the donor and the acceptor.
1.2.2 Chemical Moieties There are a wide variety of donor and acceptor groups that have
Used for Donor/Acceptor been used in fluorogenic substrates [11, 12]. MMP substrates have
Pairs used one of eight different fluorophores. The first was Trp [13],
which has an ε280 value of 5600 M−1 cm−1 and ΦF of 0.2. The
quencher group was N-2,4-dinitrophenyl (Dnp) [13, 14], which
has an absorption maximum at λ = 363 nm and a prominent shoul-
der at λ = 410 nm [15]. Alternatively, a fluorogenic substrate for
the Leishmania surface metalloproteinase used a dansyl group to
quench Trp fluorescence [16]. It should be noted that dansyl
quenching of Trp decreases rapidly with intervening residue dis-
tance [11]. Knight and colleagues proposed the use of
(7-methoxycoumarin-4-yl)acetyl (Mca; ε325 = 14,500 M−1 cm−1
and ΦF = 0.49) as a fluorophore for MMP substrates [15]. Mca is
efficiently quenched by Dnp moieties [15, 17], as the shoulder in
the Dnp absorption spectrum overlaps the Mca fluorescence emis-
sion spectrum [15]. Dnp is also used as a quencher for MMP sub-
strates containing the fluorophore N-methylanthranilic acid (Nma)
[18]. 5-[(2-Aminoethyl)amino]naphthalene-1-sulfonic acid
(Edans; ε336 = 5400 M cm and ΦF = 0.13) has been used as a
−1 −1
donor group for an MMP substrate [19]. Edans fluorescence is

Fluorogenic Peptide Substrates for Matrix Metalloproteases 139
quenched by the 4-(4-dimethylaminophenylazo)benzoic acid

(Dabcyl) group [19]. MMP substrates have been developed with
Lucifer Yellow (LY) as a fluorophore [20, 21]. LY has an ε420 value
of 12,000 M−1 cm−1 and is quenched by
5-carboxytetramethylrhodamine (Ctmr) [20]. MMP and ADAM
(a disintegrin and metalloproteinase domain) substrates have uti-
lized 5-carboxyfluorescein (5-Fam) as a fluorophore [22, 23]. Fam
has an ε492 of 65,000 M−1 cm−1 and ΦF of 0.92. Dabcyl was used
for quenching 5-Fam in ADAM substrates [23], while QXL 520
was used to quench 5-Fam fluorescence in MMP substrates [22].
2-Aminobenzoyl (Abz)/anthranilamide has been applied as a fluo-
rophore for solution and solid-phase MMP-9 and MMP-12 sub-
strates used for screening potential phosphinic peptide inhibitors
[24, 25]. Tyr(NO2) was used to quench Abz fluorescence that pos-
sesses an ε316 value of 2300 M−1 cm−1 and an ΦF of 0.60 [26–28].
Lastly, Cy3 (ε550 = 150,000 M−1 cm−1 and ΦF = 0.15) was used as
fluorophore and Cy5Q as quencher for an MMP-3 high-throughput
screening (HTS) substrate [29].
1.2.3 Practical Aspects Trp and Tyr(NO2) can be incorporated easily by solid-phase pep-
of Donor/Acceptor Groups tide synthesis and Dnp or dansyl acylated to the N-terminus of the
peptide. Other moieties require additional synthetic steps. For
example, to use the Mca/Dnp pair, a Dnp derivative such as N-3-
(2,4-dinitrophenyl)-l-2,3-diaminopropionic acid (Dpa) [15] or
Lys(Dnp) [17] must be synthesized (see below) and incorporated
into a peptide by solid-phase methods, while Mca is acylated to the
peptide N-terminus. Alternatively, the Mca derivative l-2-amino-
3-(7-methoxy-4-coumaryl)propionic acid (Amp) can be synthe-
sized (see below) and incorporated by solid-phase methods [30].
For Dabcyl/Edans substrates, derivatives of Glu(Edans) must be
prepared (see below). Substrates containing Nma, Fam, or Cy3
can be synthesized using “on-resin” reactions to create Lys(Nma),
Lys(Fam), or Lys(Cy3) moieties (see below).
In contrast to solid-phase assembly of fluorogenic substrates,
“post-synthesis” approaches have been used for the preparation of
peptides containing LY and Ctmr [20, 21] or Cy5Q [29]. Although
they are more laborious, post-synthesis approaches do alleviate one
potential purification problem of fluorogenic peptides. Typically,
fluorophores and quenchers are hydrophobic, and thus the reverse-
phase (RP) HPLC separation of the desired peptide from deletion
peptides containing these groups can be difficult.
1.3 Collagen-Model Collagen is the most abundant protein in animals, and is the major
Triple-Helical structural protein found in the connective tissues such as basement
Substrates membranes, tendons, ligaments, cartilage, bone, and skin. The
most defining feature of collagen is the supersecondary structure,
composed of three parallel extended left-handed polyproline type
II alpha chains of primarily repeating Gly-Xxx-Yyy triplets. Three
left-handed strands intertwine in a right-handed fashioned around

a common axis to form a triple-helix. In the collagen Gly-Xxx-Yyy
triplet, the residue in the Xxx position is often l-proline (Pro) and
the residue in the Yyy position is often 4(R)-hydroxy-l-proline
(Hyp), accounting for 20% of the total amino acid composition in
collagen [31]. The other commonly found amino acids are Ala,
Lys, Arg, Leu, Val, Ser, and Thr [31]. The packing of the triple-
helical coiled-coil structure requires Gly in every third position.
For several decades triple-helical peptides (THPs) consisting of
collagen-model sequences and their three-dimensional folds have
been constructed and studied to fully investigate the structural and
biological roles of collagenous proteins [32–39]. THP models of
the collagenase cleavage site in type I collagen were utilized to
examine the substrate specificity of MMP family members. The
α1(I)772–786 THP was hydrolyzed by MMP-1, MMP-1(Δ243–450),
MMP-2, and MMP-13 but not by MMP-3 or MMP-3(Δ248–460)
[40, 41]. Sequence and mass spectrometric analysis of the products
generated by the MMPs indicated that MMP hydrolysis occurred
primarily at the Gly-Ile bond, which is analogous to the bond
cleaved in native type I collagen. Thus, the α1(I)772–786 THP
contained all necessary information to direct MMP-1, MMP-2,
and MMP-13 binding and proteolysis. Similar results were reported
by Ottl et al. [42–44], who found that a heterotrimeric THP
incorporating the [α1(I)]2α2(I)772–783 sequence was efficiently
hydrolyzed by MMP-8. The α1(I)772–786 THP was subsequently
used to dissect the mechanism of the initial steps of MMP-1 cata-
lyzed collagneolysis [45].
To measure collagenolytic MMP activities, our laboratory has
described a number of FRET triple-helical peptide (fTHP) sub-
strates that are either suitable for most collagenolytic MMPs or
selective for different collagenolytic MMPs (Table 1) [46–52].
These triple-helical substrates utilize FRET primarily via incorpo-
ration of Mca as the fluorophore and the Dnp as the quencher
within the same peptide chain [46, 53, 54]. Typically, the P5’ sub-
site of the fTHP accommodates Lys(Dnp) while the P5 subsite
accommodates Lys(Mca) (Fig. 1).
Designing fTHP substrates poses many challenges. Several key
factors have to be taken into account: (1) the distance between a
donor and acceptor allowing for efficient FRET; (2) suitable loca-
tion of the donor/acceptor pair such that they do not impair sub-
strate–enzyme interactions; and (3) suitable location of the donor/
acceptor pair so as not to significantly influence triple-helical pep-
tide stability. The situation is furthermore complicated by the fact
that each MMP domain makes contact with different fTHP strands.
For example, the leading strand of the triple-helix forms most of
the contact with the catalytic (CAT) domain while the middle
strand forms significant contact with the hemopexin-like (HPX)
domain [45]. Quite often meeting all of the above criteria is not
Table 1
Hydrolysis of fTHPs by MMPs
Substrate
Enzyme abbreviation Substrate sequence kcat (s−1) KM (μM) kcat/kMa (s−1 M−1) Reference
MMP-1 fTHP-15 (Gly-Pro-Hyp)5-Gly-Pro- 0.085 ± 15 ± 2.3 5680b [52]

Lys(Mca)-Gly-Pro-Gln- 0.019
Gly~Leu-Arg-Gly-Gln-
Lys(Dnp)-Gly-Val-Arg-
(Gly-Pro-Hyp)5-NH2
MMP-1(R291A) fTHP-15 (Gly-Pro-Hyp)5-Gly-Pro- 0.13 ± 38 ± 14 3340b [52]

(Gly-Pro-Hyp)5-NH2
MMP- fTHP-15 (Gly-Pro-Hyp)5-Gly-Pro- 0.085 ± 24 ± 7.1 3540b [52]

1(I290A,R291A) Lys(Mca)-Gly-Pro-Gln- 0.017
(Gly-Pro-Hyp)5-NH2
MMP-1(Δ243–450) fTHP-15 (Gly-Pro-Hyp)5-Gly-Pro- 0.02 ± 13 ± 1.9 1500b [52]

(Gly-Pro-Hyp)5-NH2
MMP-1 fTHP-16 (Gly-Pro-Hyp)5-Gly-Pro- 0.314 ± 20 ± 4.3 15,490b [52]

Ser-Gly-Ala-Glu-Gly- 0.012
Pro-L ys(Mca)-Gly-Pro-
Gln-Gly~Leu-Arg-Gly-
Gln-Lys(Dnp)-Gly-Val-
Arg-(Gly-Pro-
Hyp)5-NH2
MMP-1(R291A) fTHP-16 (Gly-Pro-Hyp)5-Gly-Pro- 0.712 ± 166 ± 4273b [52]

Ser-Gly-Ala-Glu-Gly- 0.008 16
Arg-(Gly-Pro-
Hyp)5-NH2
MMP- fTHP-16 (Gly-Pro-Hyp)5-Gly-Pro- 0.524 ± 122 ± 4298b [52]

1(I290A,R291A) Ser-Gly-Ala-Glu-Gly- 0.016 2.9
Arg-(Gly-Pro-
Hyp)5-NH2
MMP-1(Δ243–450) fTHP-16 (Gly-Pro-Hyp)5-Gly-Pro- 0.085 ± 79 ± 17 1073b [52]

Arg-(Gly-Pro-
Hyp)5-NH2
(continued)
Table 1
(continued)
Substrate
MMP-1 fTHP-17 (Gly-Pro-Hyp)4-Gly-Pro- 0.303 ± 6.8 ± 44,350b [52]

Lys(Mca)-Gly-Pro-Gln- 0.003 1.7
Gly-Leu-Hyp-Gly-Gln-
Arg-Gly-Glu-Arg-(Gly-
Pro-Hyp)4-NH2
MMP-1(R291A) fTHP-17 (Gly-Pro-Hyp)4-Gly-Pro- 0.105 ± 14 ± 2.3 7515b [52]

Pro-Hyp)4-NH2
MMP- fTHP-17 (Gly-Pro-Hyp)4-Gly-Pro- 0.144 ± 17 ± 0.8 8527b [52]

1(I290A,R291A) Lys(Mca)-Gly-Pro-Gln- 0.003
Pro-Hyp)4-NH2
MMP-1(Δ243–450) fTHP-17 (Gly-Pro-Hyp)4-Gly-Pro- 0.016 ± 32 ± 9.6 506b [52]

Pro-Hyp)4-NH2
MMP-1 fTHP-15 (Gly-Pro-Hyp)5-Gly-Pro- 0.0071 ± 1.8 ± 4100 ± 800c [66]

(Gly-Pro-Hyp)5-NH2
MMP-1 fTHP-16 (Gly-Pro-Hyp)5-Gly-Pro- 0.046 ± 13 ± 2.2 3500 ± 580c [66]

Arg-(Gly-Pro-
Hyp)5-NH2
MMP-1 fTHP-17 (Gly-Pro-Hyp)4-Gly-Pro- 0.105 ± 2.9 ± 37,000 ± [66]

Lys(Mca)-Gly-Pro-Gln- 0.0046 0.077 4700c
Pro-Hyp)4-NH2
(continued)
Table 1
(continued)
Substrate
MMP-2 fTHP-15 (Gly-Pro-Hyp)5-Gly-Pro- 1.2 ± 0.12 5.9 ± 210,000 ± [66]

Lys(Mca)-Gly-Pro-Gln- 0.57 17,000c
(Gly-Pro-Hyp)5-NH2
MMP-2 fTHP-16 (Gly-Pro-Hyp)5-Gly-Pro- 1.2 ± 0.05 5.2 ± 230,000 ± [66]

Ser-Gly-Ala-Glu-Gly- 0.84 43,000c
Arg-(Gly-Pro-
Hyp)5-NH2

Lys(Mca)-Gly-Pro-Gln- 0.0022 0.062 12,000c
Pro-Hyp)4-NH2
MMP-8 fTHP-15 (Gly-Pro-Hyp)5-Gly-Pro- 0.088 ± 9.2 ± 9600 ± 1200c [66]

(Gly-Pro-Hyp)5-NH2
MMP-8 fTHP-16 (Gly-Pro-Hyp)5-Gly-Pro- 0.24 ± 11 ± 22,000 ± [66]

Ser-Gly-Ala-Glu-Gly- 0.009 0.44 1800c
Arg-(Gly-Pro-
Hyp)5-NH2
MMP-8 fTHP-17 (Gly-Pro-Hyp)4-Gly-Pro- 0.30 ± 29 ± 2.4 10,000 ± 900c [66]

Pro-Hyp)4-NH2
MMP-13 fTHP-15 (Gly-Pro-Hyp)5-Gly-Pro- 4.2 ± 0.25 67 ± 3.9 63,000 ± [66]

Lys(Mca)-Gly-Pro-Gln- 1200c
(Gly-Pro-Hyp)5-NH2
(continued)
Table 1
(continued)
Substrate
MMP-13 fTHP-16 (Gly-Pro-Hyp)5-Gly-Pro- 2.7 ± 0.12 25 ± 2.7 107,000 ± [66]

Ser-Gly-Ala-Glu-Gly- 8400c
Arg-(Gly-Pro-
Hyp)5-NH2

Lys(Mca)-Gly-Pro-Gln- 0.00054 0.022 12,000c
Pro-Hyp)4-NH2
MT1-MMP fTHP-15 (Gly-Pro-Hyp)5-Gly-Pro- 0.034 ± 11 ± 1.9 3,000 ± 120c [66]

(Gly-Pro-Hyp)5-NH2
MT1-MMP fTHP-16 (Gly-Pro-Hyp)5-Gly-Pro- 0.102 ± 20 ± 5.8 5200 ± 120c [66]

Arg-(Gly-Pro-
Hyp)5-NH2
MT1-MMP fTHP-17 (Gly-Pro-Hyp)4-Gly-Pro- 0.017 ± 5.2 ± 3300 ± 510c [66]

Pro-Hyp)4-NH2
MMP-8 fTHP-4d (Gly-Pro-Hyp)5-Gly-Pro- 0.035 7.6 4500e [50]

Lys(Mca)-Gly-Pro-Gln-
(Gly-Pro-Hyp)5-NH2
MMP-8 C10-fTHP- C10-(Gly-Pro-Hyp)5-Gly- 0.089 14.4 6200e [50]

4d Pro-L ys(Mca)-Gly-Pro-
Arg-(Gly-Pro-
Hyp)5-NH2
(continued)
Table 1
(continued)
Substrate

Arg-(Gly-Pro-
Hyp)5-NH2
MMP-8 fTHP-9 (Gly-Pro-Hyp)5-Gly-Pro- 0.94 5.4 175,000e [50]

Gly~Cys(Mob)-Arg-
Gly-Gln-L ys(Dnp)-Gly-
Val-Arg-(Gly-Pro-
Hyp)5-NH2f
MMP-8 C10-fTHP-9 C10-(Gly-Pro-Hyp)5-Gly- 0.64 8.8 73,000e [50]

Gln-Gly~Cys(Mob)-
Arg-Gly-Gln-L ys(Dnp)-
Gly-Val-Arg-(Gly-Pro-
Hyp)5-NH2f

Gln-Gly~Cys(Mob)-
Hyp)5-NH2f
MMP-8 fTHP-10 (Gly-Pro-Hyp)5-Gly-Pro- 0.035 8.15 4300e [50]

Lys(Mca)-Gly-Pro-Orn-
(Gly-Pro-Hyp)5-NH2f

Gly~Cys(Mob)-Arg-
Val-Arg-(Gly-Pro-
Hyp)5-NH2f
MMP-8 C10- C10-(Gly-Pro-Hyp)5-Gly- 0.007 3.60 1900e [50]

fTHP-11 Pro-L ys(Mca)-Gly-Pro-
Orn-Gly~Cys(Mob)-
Hyp)5-NH2f
(continued)
Table 1
(continued)
Substrate

Orn-Gly~Cys(Mob)-
Hyp)5-NH2f
MMP-8 C6-fTHP-12 C6-(Gly-Pro-Hyp)5-Gly- 0.070 14.5 5090e [51]

Leu-Gly~Met-Arg-Gly-
Arg-(Gly-Pro-Hyp)5-
NH2f

Lys(Mca)-Gly-Pro-Val-
Asn~Phe-Arg-Gly-Gln-
(Gly-Pro-Hyp)5-NH2f

Val-Asn~Phe-Arg-Gly-
Arg-(Gly-Pro-Hyp)5-
NH2f
MMP-13 fTHP-4d (Gly-Pro-Hyp)5-Gly-Pro- 0.015 8.6 1600e [50]

(Gly-Pro-Hyp)5-NH2

Arg-(Gly-Pro-
Hyp)5-NH2

Arg-(Gly-Pro-
Hyp)5-NH2

Gly~Cys(Mob)-Arg-
Val-Arg-(Gly-Pro-
Hyp)5-NH2f
(continued)
Table 1
(continued)
Substrate

Gln-Gly~Cys(Mob)-
Hyp)5-NH2f

Gln-Gly~Cys(Mob)-
Hyp)5-NH2e

(Gly-Pro-Hyp)5-NH2f
MMP-13 C10- C10-(Gly-Pro-Hyp)5-Gly- 0.31 17.4 18,000e [50]

Orn-Gly~Leu-Arg-Gly-
Arg-(Gly-Pro-Hyp)5-
NH2f

Gly~Cys(Mob)-Arg-
Val-Arg-(Gly-Pro-
Hyp)5-NH2f

Orn-Gly~Cys(Mob)-
Hyp)5-NH2f

Orn-Gly~Cys(Mob)-
Hyp)5-NH2f

Arg-(Gly-Pro-Hyp)5-
NH2f
(continued)
Table 1
(continued)
Substrate

Lys(Mca)-Gly-Pro-Val-
(Gly-Pro-Hyp)5-NH2f

Arg-(Gly-Pro-Hyp)5-
NH2f
MT1- fTHP-4d (Gly-Pro-Hyp)5-Gly-Pro- 1.37 23.1 59,000e [50]

MMP(Δ279–523) Lys(Mca)-Gly-Pro-Gln-
(Gly-Pro-Hyp)5-NH2
MT1- C10-fTHP- C10-(Gly-Pro-Hyp)5-Gly- 0.0033 0.80 4100e [50]

MMP(Δ279–523) 4d Pro-L ys(Mca)-Gly-Pro-
Arg-(Gly-Pro-
Hyp)5-NH2
MT1- C16-fTHP- C16-(Gly-Pro-Hyp)5-Gly- 0.0021 1.80 1200e [50]

MMP(Δ279–523) 4d Pro-L ys(Mca)-Gly-Pro-
Arg-(Gly-Pro-
Hyp)5-NH2
MT1- fTHP-9 (Gly-Pro-Hyp)5-Gly-Pro- 1.34 7.95 168,000e [50]

MMP(Δ279–523) Lys(Mca)-Gly-Pro-Gln-
Gly~Cys(Mob)-Arg-
Val-Arg-(Gly-Pro-
Hyp)5-NH2f
MT1- C10-fTHP-9 C10-(Gly-Pro-Hyp)5-Gly- 0.080 2.80 29,000e [50]

MMP(Δ279–523) Pro-L ys(Mca)-Gly-Pro-
Gln-Gly~Cys(Mob)-
Hyp)5-NH2f

MMP(Δ279–523) Lys(Mca)-Gly-Pro-Orn-
(Gly-Pro-Hyp)5-NH2f
(continued)
Table 1
(continued)
Substrate
MT1- C10- C10-(Gly-Pro-Hyp)5-Gly- 0.091 4.40 20,000e [50]

MMP(Δ279–523) fTHP-10 Pro-L ys(Mca)-Gly-Pro-
Orn-Gly~Leu-Arg-Gly-
Arg-(Gly-Pro-Hyp)5-
NH2f

MMP(Δ279–523) Lys(Mca)-Gly-Pro-Orn-
Gly~Cys(Mob)-Arg-
Val-Arg-(Gly-Pro-
Hyp)5-NH2f
MT1- C10- C10-(Gly-Pro-Hyp)5-Gly- 0.036 3.50 11,000e [50]

MMP(Δ279–523) fTHP-11 Pro-L ys(Mca)-Gly-Pro-
Orn-Gly~Cys(Mob)-
Hyp)5-NH2f
MT2-MMP/ fTHP-11 (Gly-Pro-Hyp)5-Gly-Pro- 0.41 14.3 29,000e [50]

MMP- Lys(Mca)-Gly-Pro-Orn-
15(Δ268–628) Gly~Cys(Mob)-Arg-
Val-Arg-(Gly-Pro-
Hyp)5-NH2f
MMP-8 fTHP-4d (Gly-Pro-Hyp)5-Gly-Pro- 0.035 ± 7.60 ± 4500 [51]

Lys(Mca)-Gly-Pro-Gln- 0.016a 2.40a
(Gly-Pro-Hyp)5-NH2
MMP-8 C6-fTHP-12 C6-(Gly-Pro-Hyp)5-Gly- 0.070 ± 14.50 ± 5090 [51]

Pro-L ys(Mca)-Gly-Pro- 0.020 4.80
Arg-(Gly-Pro-Hyp)5-
NH2f
MMP-8 fTHP-13 (Gly-Pro-Hyp)5-Gly-Pro- 0.028 ± 9.30 ± 3010 [51]

Lys(Mca)-Gly-Pro-Val- 0.004 1.84
(Gly-Pro-Hyp)5-NH2f

Arg-(Gly-Pro-Hyp)5-
NH2f
(continued)
Table 1
(continued)
Substrate
MMP-13 fTHP-4d (Gly-Pro-Hyp)5-Gly-Pro- 0.015 ± 8.60 ± 1600 [51]

Lys(Mca)-Gly-Pro-Gln- 0.005a 1.70a
(Gly-Pro-Hyp)5-NH2

Arg-(Gly-Pro-Hyp)5-
NH2f
MMP-13 fTHP-13 (Gly-Pro-Hyp)5-Gly-Pro- 0.127 ± 20.50 ± 6116 [51]

Lys(Mca)-Gly-Pro-Val- 0.046 6.22
(Gly-Pro-Hyp)5-NH2f

Arg-(Gly-Pro-Hyp)5-
NH2f
MMP-9 fTHP-15 (Gly-Pro-Hyp)5-Gly-Pro- 0.78 ± 0.13 12.1 ± 64,851 ± [52, 64]

Lys(Mca)-Gly-Pro-Gln- 0.6 10,698
(Gly-Pro-Hyp)5-NH2
MMP-9 CAT fTHP-15 (Gly-Pro-Hyp)5-Gly-Pro- 1.05 ± 0.14 15.1 ± 69,566 ± 9138 [52, 64]
(Gly-Pro-Hyp)5-NH2
MMP-9 CAT-FN fTHP-15 (Gly-Pro-Hyp)5-Gly-Pro- 1.75 ± 0.15 143.0 ± 12,227 ± 1077 [52, 64]
(Gly-Pro-Hyp)5-NH2
MMP-2 fTHP-15 (Gly-Pro-Hyp)5-Gly-Pro- 0.654 5.4 ± 120,423b [67]

(Gly-Pro-Hyp)5-NH2
MMP-2 GDK A Gly-Asp-Lys-(Gly-Pro- 1.485 5.7 ± 262,004b [67]

Hyp)4-Gly-Pro- 0.79
(Gly-Pro-Hyp)4-Gly-
Asp-Lys-NH2
(continued)
Table 1
(continued)
Substrate
MMP-2 GDK B Gly-Pro-Hyp-Gly-Asp-L ys- 1.424 4.5 ± 314,950b [67]

(Gly-Pro-Hyp)3-Gly- 0.60
Pro-Lys(Mca)-Gly-Pro-
Arg-(Gly-Pro-Hyp)3-
Gly-Asp-Lys-Gly-Pro-
Hyp-NH2
MMP-2 GDK C (Gly-Pro-Hyp)2-Gly-Asp- 2.374 3.3 ± 728,309b [67]

Lys-(Gly-Pro-Hyp)2- 0.35
Gly-Pro-L ys(Mca)-Gly-
Pro-Gln-Gly~Leu-Arg-
Gly-Gln-Lys(Dnp)-Gly-
Val-Arg-(Gly-Pro-
Hyp)2-Gly-Asp-Lys-
(Gly-Pro-Hyp)2-NH2
MMP-2 GDK D (Gly-Pro-Hyp)3-Gly-Asp- 2.635 2.7 ± 986,155b [67]

Lys-Gly-Pro-Hyp-Gly- 0.62
Arg-Gly-Pro-Hyp-Gly-
Asp-L ys-(Gly-Pro-
Hyp)3-NH2
MMP-2 GDK E (Gly-Pro-Hyp)4-Gly-Asp- 2.804 2.4 ± 1,162,216b [67]

Lys-Gly-Pro-Lys(Mca)- 0.42
Gly-Pro-Gln-Gly~Leu-
Arg-Gly-Gln-Lys(Dnp)-
Gly-Val-Arg-Gly-Asp-
Lys-(Gly-Pro-
Hyp)4-NH2
MMP-2 KGD A Lys-Gly-Asp-Hyp-(Gly- 1.987 5.9 ± 337,988b [67]

Pro-Hyp)4-Gly-Pro- 0.94
(Gly-Pro-Hyp)3-Gly-
Pro-Lys-Gly-Asp-
Hyp-NH2
MMP-2 KGD B Gly-Pro-Lys-Gly-Asp-Hyp- 2.296 7.1 ± 324,215b [67]

Arg-(Gly-Pro-Hyp)2-
Gly-Pro-Lys-Gly-Asp-
Hyp-Gly-Pro-Hyp-NH2
(continued)
Table 1
(continued)
Substrate
MMP-2 KGD C Gly-Pro-Hyp-Gly-Pro-L ys- 2.627 4.8 ± 543,854b [67]

Gly-Asp-Hyp-(Gly-Pro- 0.49
Hyp)2-Gly-Pro-
Gly-Pro-Hyp-Gly-Pro-
Lys-Gly-Asp-Hyp-(Gly-
Pro-Hyp)2-NH2
MMP-2 KGD D (Gly-Pro-Hyp)2-Gly-Pro- 2.234 9.8 ± 227,927b [67]

Lys-Gly-Asp-Hyp-Gly- 0.97
Pro-Hyp-Gly-Pro-
Gly-Pro-L ys-Gly-Asp-
Hyp-(Gly-Pro-
Hyp)3-NH2
MMP-2 KGD E (Gly-Pro-Hyp)3-Gly-Pro- 1.29 4.6 ± 279,035b [67]

Pro-Hyp)4-NH2

(Gly-Pro-Hyp)5-NH2

Hyp)4-Gly-Pro- 0.75
(Gly-Pro-Hyp)4-Gly-
Asp-Lys-NH2

Arg-(Gly-Pro-Hyp)3-
Hyp-NH2
(continued)
Table 1
(continued)
Substrate

Val-Arg-(Gly-Pro-
Hyp)2-Gly-Asp-Lys-
(Gly-Pro-Hyp)2-NH2

Asp-L ys-(Gly-Pro-
Hyp)3-NH2
MMP-8 GDK E (Gly-Pro-Hyp)4-Gly-Asp- 8.233 58.5 ± 140,727b [67]

Lys-(Gly-Pro-
Hyp)4-NH2

(Gly-Pro-Hyp)3-Gly-
Pro-Lys-Gly-Asp-
Hyp-NH2

Arg-(Gly-Pro-Hyp)2-
Hyp-Gly-Pro-Hyp-NH2

Hyp)2-Gly-Pro-
Pro-Hyp)2-NH2
(continued)
Table 1
(continued)
Substrate

Pro-Hyp-Gly-Pro-
Hyp-(Gly-Pro-
Hyp)3-NH2

Pro-Hyp)4-NH2

(Gly-Pro-Hyp)5-NH2

Hyp)4-Gly-Pro- 0.52
(Gly-Pro-Hyp)4-Gly-
Asp-Lys-NH2
MMP-9 GDK B Gly-Pro-Hyp-Gly-Asp-L ys- 2.7 2.3 ± 1,171,153b [67]

Arg-(Gly-Pro-Hyp)3-
Hyp-NH2
MMP-9 GDK C (Gly-Pro-Hyp)2-Gly-Asp- 2.621 1.3 ± 1,948,364b [67]

Val-Arg-(Gly-Pro-
Hyp)2-Gly-Asp-Lys-
(Gly-Pro-Hyp)2-NH2
(continued)
Table 1
(continued)
Substrate
MMP-9 GDK D (Gly-Pro-Hyp)3-Gly-Asp- 4.071 1.7 ± 2,443,634b [67]

Asp-L ys-(Gly-Pro-
Hyp)3-NH2
MMP-9 GDK E (Gly-Pro-Hyp)4-Gly-Asp- 2.679 1.4 ± 1,982,689b [67]

Lys-(Gly-Pro-
Hyp)4-NH2

(Gly-Pro-Hyp)3-Gly-
Pro-Lys-Gly-Asp-
Hyp-NH2

Arg-(Gly-Pro-Hyp)2-
Hyp-Gly-Pro-Hyp-NH2
MMP-9 KGD C Gly-Pro-Hyp-Gly-Pro-L ys- 5.335 4.5 ± 1,174,706b [67]

Hyp)2-Gly-Pro-
Pro-Hyp)2-NH2
(continued)
Table 1
(continued)
Substrate

Pro-Hyp-Gly-Pro-
Hyp-(Gly-Pro-
Hyp)3-NH2

Pro-Hyp)4-NH2
MMP-12 fTHP-15 (Gly-Pro-Hyp)5-Gly-Pro- 0.111 56.6 ± 1953b [67]

(Gly-Pro-Hyp)5-NH2
MMP-12 GDK A Gly-Asp-Lys-(Gly-Pro- 0.153 26.3 ± 5791b [67]

Hyp)4-Gly-Pro- 2.2
(Gly-Pro-Hyp)4-Gly-
Asp-Lys-NH2
MMP-12 GDK B Gly-Pro-Hyp-Gly-Asp-L ys- 0.554 72.8 ± 7600b [67]

Arg-(Gly-Pro-Hyp)3-
Hyp-NH2

Val-Arg-(Gly-Pro-
Hyp)2-Gly-Asp-Lys-
(Gly-Pro-Hyp)2-NH2
(continued)
Table 1
(continued)
Substrate

Asp-L ys-(Gly-Pro-
Hyp)3-NH2

Lys-(Gly-Pro-
Hyp)4-NH2
MMP-12 KGD A Lys-Gly-Asp-Hyp-(Gly- 0.408 91.3 ± 4467b [67]

(Gly-Pro-Hyp)3-Gly-
Pro-Lys-Gly-Asp-
Hyp-NH2
MMP-12 KGD B Gly-Pro-Lys-Gly-Asp-Hyp- 0.483 99.7 ± 4846b [67]

Arg-(Gly-Pro-Hyp)2-
Hyp-Gly-Pro-Hyp-NH2
MMP-12 KGD C Gly-Pro-Hyp-Gly-Pro-L ys- 0.618 79.0 ± 7827b [67]

Hyp)2-Gly-Pro-
Pro-Hyp)2-NH2
(continued)
Table 1
(continued)
Substrate
MMP-12 KGD D (Gly-Pro-Hyp)2-Gly-Pro- 0.606 148.6 ± 4077b [67]

Pro-Hyp-Gly-Pro-
Hyp-(Gly-Pro-
Hyp)3-NH2
MMP-12 KGD E (Gly-Pro-Hyp)3-Gly-Pro- 0.437 83.0 ± 5258b [67]

Pro-Hyp)4-NH2

(Gly-Pro-Hyp)5-NH2

Hyp)4-Gly-Pro- 1.12
(Gly-Pro-Hyp)4-Gly-
Asp-Lys-NH2

Arg-(Gly-Pro-Hyp)3-
Hyp-NH2

Val-Arg-(Gly-Pro-
Hyp)2-Gly-Asp-Lys-
(Gly-Pro-Hyp)2-NH2
(continued)
Table 1
(continued)
Substrate

Asp-L ys-(Gly-Pro-
Hyp)3-NH2

Lys-(Gly-Pro-
Hyp)4-NH2

(Gly-Pro-Hyp)3-Gly-
Pro-Lys-Gly-Asp-
Hyp-NH2

Arg-(Gly-Pro-Hyp)2-
Hyp-Gly-Pro-Hyp-NH2

Hyp)2-Gly-Pro-
Pro-Hyp)2-NH2
(continued)
Table 1
(continued)
Substrate

Pro-Hyp-Gly-Pro-
Hyp-(Gly-Pro-
Hyp)3-NH2

Pro-Hyp)4-NH2
MT1-MMP fTHP-15 (Gly-Pro-Hyp)5-Gly-Pro- 0.3 13.1 ± 20,784b [67]

(Gly-Pro-Hyp)5-NH2
MT1-MMP GDK A Gly-Asp-Lys-(Gly-Pro- 0.136 6.9 ± 19,779b [67]

Hyp)4-Gly-Pro- 0.89
(Gly-Pro-Hyp)4-Gly-
Asp-Lys-NH2
MT1-MMP GDK B Gly-Pro-Hyp-Gly-Asp-L ys- 0.305 9.1 ± 33,502b [67]

Arg-(Gly-Pro-Hyp)3-
Hyp-NH2
MT1-MMP GDK C (Gly-Pro-Hyp)2-Gly-Asp- 0.589 8.3 ± 71,005b [67]

Val-Arg-(Gly-Pro-
Hyp)2-Gly-Asp-Lys-
(Gly-Pro-Hyp)2-NH2
(continued)
Table 1
(continued)
Substrate
MT1-MMP GDK D (Gly-Pro-Hyp)3-Gly-Asp- 1.026 9.5 ± 108,043b [67]

Asp-L ys-(Gly-Pro-
Hyp)3-NH2
MT1-MMP GDK E (Gly-Pro-Hyp)4-Gly-Asp- 1.291 12.0 ± 107,577b [67]

Lys-(Gly-Pro-
Hyp)4-NH2
MT1-MMP KGD A Lys-Gly-Asp-Hyp-(Gly- 0.3 14.1 ± 24,077b [67]

(Gly-Pro-Hyp)3-Gly-
Pro-Lys-Gly-Asp-
Hyp-NH2
MT1-MMP KGD B Gly-Pro-Lys-Gly-Asp-Hyp- 0.4 17.1 ± 23,486b [67]

Arg-(Gly-Pro-Hyp)2-
Hyp-Gly-Pro-Hyp-NH2
MT1-MMP KGD C Gly-Pro-Hyp-Gly-Pro-L ys- 0.4 15.5 ± 23,012b [67]

Hyp)2-Gly-Pro-
Pro-Hyp)2-NH2
(continued)
Table 1
(continued)
Substrate
MT1-MMP KGD D (Gly-Pro-Hyp)2-Gly-Pro- 0.5 21.2 ± 24,300b [67]

Pro-Hyp-Gly-Pro-
Hyp-(Gly-Pro-
Hyp)3-NH2
MT1-MMP KGD E (Gly-Pro-Hyp)3-Gly-Pro- 0.6 25.1 ± 23,260b [67]

Pro-Hyp)4-NH2
MMP-1 fTHP-15(5- (Gly-Pro-Hyp)5-Gly-Pro- 0.0217 7.3 2968b [55]

Fam/ Lys(5-Fam)-Gly-Pro-
Dabcyl) Gln-Gly~Leu-Arg-Gly-
Gln-Lys(Dabcyl)-Gly-
Val-Arg-(Gly-Pro-
Hyp)5-NH2
MMP-1 fTHP-15 (Gly-Pro-Hyp)5-Gly-Pro- 0.0042 2.31 1833b [55]

(Gly-Pro-Hyp)5-NH2
MMP-13 fTHP-15(5- (Gly-Pro-Hyp)5-Gly-Pro- 1.1 16.76 65,964b [55]

Val-Arg-(Gly-Pro-
Hyp)5-NH2
MMP-13 fTHP-15 (Gly-Pro-Hyp)5-Gly-Pro- 0.948 20.25 46,815b [55]

(Gly-Pro-Hyp)5-NH2
MT1-MMP fTHP-15(5- (Gly-Pro-Hyp)5-Gly-Pro- 0.0055 4.37 1249b [55]

Val-Arg-(Gly-Pro-
Hyp)5-NH2
(continued)
Table 1
(continued)
Substrate
MT1-MMP fTHP-15 (Gly-Pro-Hyp)5-Gly-Pro- 0.0259 13.1 1981b [55]

(Gly-Pro-Hyp)5-NH2
MT6-MMP/ fTHP-15(5- (Gly-Pro-Hyp)5-Gly-Pro- ND ND ND [84]

MMP-25 Fam/ Lys(5-Fam)-Gly-Pro-
Val-Arg-(Gly-Pro-
Hyp)5-NH2
MMP-2 α1(V)436– (Gly-Pro-Hyp)5-Gly-Pro- 0.062 4.4 14,000e [47]

447 Lys(Mca)-Gly-Pro-Pro-
fTHP Gly~Val-Val-Gly-Glu-
Lys(Dnp)-Gly-Glu-Gln-
(Gly-Pro-Hyp)5-NH2
MMP-9 α1(V)436– (Gly-Pro-Hyp)5-Gly-Pro- 0.044 8.1 5500e [47]

(Gly-Pro-Hyp)5-NH2
MMP-9 α1(V)436– (Gly-Pro-Hyp)5-Gly-Pro- 3.99 ± 0.13 77.8 ± 51,360 ± 1676 [52, 64]
447 Lys(Mca)-Gly-Pro-Pro- 5.8
(Gly-Pro-Hyp)5-NH2
MMP-9 CAT α1(V)436– (Gly-Pro-Hyp)5-Gly-Pro- 1.17 ± 0.21 22.5 ± 52,249 ± 9192 [52, 64]
447 Lys(Mca)-Gly-Pro-Pro- 1.7
(Gly-Pro-Hyp)5-NH2
MMP-12 α1(V)436– (Gly-Pro-Hyp)5-Gly-Pro- 2.35 24.8 97,600e [65]

(Gly-Pro-Hyp)5-NH2
MMP-13 pfTHP-1 (Gly-Pro-Hyp)4-Gly-Pro- 0.9 ± 0.05 21.4 ± 42,000e [68]

Lys(Mca)-Gly-Sar-Ala- 1.1
Gly~Gln-Arg-Gly-Val-
Lys(Dnp)-Gly-NLeu-
Hyp-(Gly-Pro-
Hyp)4-NH2
(continued)
Table 1
(continued)
Substrate
MMP-13 Ac-pfTHP-1 Ac-(Gly-Pro-Hyp)4-Gly- 0.6 ± 0.05 83.3 ± 7200e [68]

Pro-L ys(Mca)-Gly-Sar- 4.3
Ala-Gly~Gln-Arg-Gly-
Val-Lys(Dnp)-Gly-
NLeu-Hyp-(Gly-Pro-
Hyp)4-NH2
a
Assays were performed at 37 °C, pH 7.5 except where noted. ND not determined
b
Assay performed at 25 °C
c
d
fTHP-4 and fTHP-15 have the same sequence
e
f
Substitutions relative to fTHP-4 are indicated in bold
possible and optimal substrate design represents a compromise of

the aforementioned factors (Fig. 1).
Additional fTHPs have been constructed incorporating l- or
d-2-amino-3-(7-methoxy-4-coumaryl)propionic acid (Amp) or l-
or d-2-amino-3-(6,7-dimethoxy-4-coumaryl)propionic acid (Adp)
as the fluorophore and Dnp as the quencher [48]. The desired
sequences were C6-(Gly-Pro-Hyp)5-Gly-Pro-[Amp/Adp]-Gly-
Pro-Gln-Gly~Leu-Arg-Gly-Gln-Lys(Dnp)-Gly-Val-Arg-(Gly-Pro-
Hyp)5-NH2. All four fTHPs formed stable triple-helices. MMP-2
rates of hydrolysis for all fTHPs were considerably more rapid than
MMP-1. Evaluation of individual kinetic parameters indicated that
MMP-2 bound to the fTHPs more efficiently than MMP-1.
Comparison to a triple-helical substrate incorporating the same
sequence but with a different fluorophore [Lys(Mca)] demon-
strated that the shorter side-chain of Amp or Adp was better toler-
ated by MMP-1 and MMP-2. Adp may well be the fluorophore of
choice for fTHPs, as (a) fTHPs incorporating Adp were obtained
in significantly higher yields than the Amp-containing fTHPs, (b)
Adp has a larger Stokes shift than either Amp or Lys(Mca), and
thus has less chance of self-quenching, (c) Adp has a relatively high
quantum yield, (d) the Adp/Dnp pair is compatible with multiwell
plate reader formats, and (e) MMPs better tolerate Adp compared
with Lys(Mca).
We have also examined the preparation of fTHP substrates
using 5-Fam as the fluorophore and Dabcyl as the quencher [55].
The Nα-(9-fluorenylmethoxycarbonyl)-Nε-(5-carboxyfluorescein)-
L-lysine [Fmoc-Lys(5-Fam)] building block was synthesized uti-
lizing two distinct synthetic routes. The first involved copper
complexation of Lys while the second utilized Fmoc-Lys with
microwave irradiation. Both approaches allowed convenient
production of a very pure final product at a reasonable cost. Fmoc-

Lys(5-Fam) and Fmoc-Lys(Dabcyl) were incorporated into the
sequence of a THP substrate utilizing automated solid-phase
peptide synthesis protocols. A second substrate was assembled

where Mca was the fluorophore and Dnp was the quencher.
Circular dichroism (CD) spectroscopy was used to determine the
influence of the fluorophore/quencher pair on the stability of the
triple-helix. The activity of the two substrates was examined with
MMP-1, MMP-13, and MT1-MMP. The combination of 5-Fam
as fluorophore and Dabcyl as quencher resulted in a triple-helical
substrate that, compared with the fluorophore/quencher pair of
Mca/Dnp, had a slightly destabilized triple-helix but was hydro-
lyzed more rapidly by MMP-1 and MMP-13 and had greater
sensitivity.
Due to the differing sequences of the fTHPs, their relative
triple-helical thermal stabilities can vary [50, 51]. In order to
obtain fTHPs that are stable under near-physiological conditions,
as well as to examine the effects of substrate thermal stability on
MMP activity, the “peptide-amphiphile” approach has been uti-
lized [56–59]. Peptide-amphiphiles (PAs) mimic defined topologi-
cal structures by incorporating an amino acid sequence with the
propensity to form a triple-helix as the polar head group and a
dialkyl or monoalkyl hydrocarbon chain as the nonpolar tail [56–
59]. Desirable peptide head group melting temperature (Tm) val-
ues can be achieved for in vivo use, as triple-helical PAs have been
constructed with Tm values ranging from 30 to 70 °C [40, 41, 46,
57, 58, 60–62]. fTHP PAs have been created that, based on their
Tm values, are suitable for studying MMP activity under in vivo
conditions.
1.4 Dissecting MMP A series of fTHPs has been assembled that incorporate sequences
Collagenolytic based on a consensus types I-III collagen 769–783 region (Table
Activities 1). The fTHP substrates have been utilized to determine individual
kinetic parameters (Table 1) and activation energies for hydrolysis
of triple-helices [46, 47, 49–53, 63–69]. MMP-1, MMP-2, MMP-
8, MMP-9, MMP-13, and MT1-MMP hydrolysis of the consensus
types I-III collagen sequence (designated fTHP-15) occurred at
the Gly~Leu bond, the analogous bond cleaved by these MMPs in
the native collagen chains [70]. Thus, the incorporation of the
fluorophore/quenching pair of Mca and Dnp did not affect the
ability of the enzyme to recognize or cleave the substrate. A modi-
fied version of the consensus types I-III collagen fTHP, in which
the P1′ subsite incorporated a Cys(4-methoxybenzyl) group
(fTHP-9), was hydrolyzed efficiently by MT1-MMP, MMP-8, and
MMP-13, but poorly by other MMPs [49, 50].
Analysis of MMP-1/THP interactions by hydrogen/deute-
rium exchange mass spectrometry (HDX MS) followed by evalua-
tion of wild-type and mutant MMP-1 kinetics using fTHPs led to
the identification of three noncatalytic regions in MMP-1 (residues

285–295, 302–316, and 437–457) and two specific residues
(Ile290 and Arg291) that participate in collagenolysis [52]. Ile290
and Arg291 contributed to recognition of triple-helical structure,
and facilitated both the binding and catalysis of the triple-helix.
Evidence from this and prior studies indicated that the MMP-1
CAT and HPX domains collaborate in collagen catabolism by
properly aligning the triple-helix and coupling conformational
states to facilitate hydrolysis. This study was the first to document
the roles of specific residues within the MMP-1 HPX domain in
substrate binding and turnover.
To evaluate the role of charged residue clusters on the regula-
tion of MMP collagenolysis, a series of ten fTHPs were constructed
in which either Lys-Gly-Asp or Gly-Asp-Lys motifs replaced Gly-
Pro-Hyp repeats [67]. The stabilities of fTHPs containing the two
different motifs were analyzed, and kinetic parameters for substrate
hydrolysis by six MMPs determined. A general trend for virtually
all enzymes was that, as Gly-Asp-Lys motifs were moved from the
extreme N- and C-termini to the interior next to the cleavage site
sequence, kcat/KM values increased. Additionally, all Gly-Asp-Lys
fTHPs were as good or better substrates than the parent fTHP, in
which Gly-Asp-Lys was not present. In turn, the Lys-Gly-Asp
fTHPs were also always better substrates than the parent THP, but
the magnitude of the difference was considerably less compared
with the Gly-Asp-Lys series. Of the MMPs tested, MMP-2 and
MMP-9 most greatly favored the presence of charged residues,
with preference for the Gly-Asp-Lys series. Lys-Gly-[Asp/Glu]
motifs are more commonly found near potential MMP cleavage
sites than Gly-[Asp/Glu]-Lys motifs. As Lys-Gly-Asp is not as
favored by MMPs as Gly-Asp-Lys, the Lys-Gly-Asp motif appears
advantageous over the Gly-Asp-Lys motif by preventing unwanted
MMP hydrolysis. More specifically, the lack of Gly-Asp-Lys clus-
ters may diminish potential MMP-2 and MMP-9 collagenolytic
activity. This study indicated that MMPs have interactions span-
ning the P23–P23′ subsites of collagenous substrates.
Although collagenolytic MMPs possess common domain orga-
nizations, there are subtle differences in their processing of collag-
enous triple-helical substrates. We incorporated peptoid residues in
THPs and examined MMP activities toward these peptomeric chi-
meras [68].
Several different peptoid residues were incorporated into
triple-helical substrates in subsites P3, P1, P1′, and P10′ individually
or in combination, and the effects of the peptoid residues were
evaluated on the activities of full-length MMP-1, MMP-8, MMP-
13, and MT1-MMP. Most peptomers showed little discrimination
between MMPs. However, a peptomer containing N-methyl Gly
(sarcosine) in the P1′ subsite and N-isobutyl Gly (NLeu) in the P10′
subsite was hydrolyzed efficiently only by MMP-13 (nomenclature
relative to α1(I)772–786 sequence). Cleavage site analysis showed
hydrolysis at the Gly-Gln bond, indicating a shifted binding of the

triple-helix compared to the parent sequence. Favorable hydrolysis
by MMP-13 was not due to sequence specificity or instability of
the substrate triple-helix, but rather was based on the specific inter-
actions of the P7′ peptoid residue with the MMP-13 HPX domain.
A FRET triple-helical peptomer was constructed and found to be
readily processed by MMP-13, not cleaved by MMP-1 and MMP-
8, and weakly hydrolyzed by MT1-MMP. The influence of the
triple-helical structure containing peptoid residues on the interac-
tion between MMP subsites and individual substrate residues may
provide additional information for the mechanism of collagenolysis
and the understanding of collagen specificity.
MMP selectivity has been observed with a type V collagen
derived fTHP [47]. The α1(V)436–447 fTHP was hydrolyzed effi-
ciently by MMP-2, MMP-9, and MMP-12 (Table 1), but was not
hydrolyzed by MMP-1, MMP-3, MMP-13, or MT1-MMP [47,
65]. The type V collagen sequence Gly-Pro-Pro-Gly~Val-Val-Gly-
Glu-Lys-Gly-Glu-Gln, as a single-stranded peptide, was hydro-
lyzed extremely slowly by either MMP-2 or MMP-9 [47].
Therefore, it is the triple-helical structure, along with the sequence,
that imparts substrate specificity among various MMPs [47].
fTHPs have provided insights into (a) the relative triple-helical
peptidase activities of the various collagenolytic MMPs (Fig. 2),
(b) the collagen preferences of these MMPs, and (c) the relative
roles of MMP domains and specific residues in efficient collagen-
olysis [40, 41, 46–53, 63–69, 71, 72]. fTHPs have been noted as
being particularly advantageous for dissecting the mechanism of
collagenolysis [1].
2 Materials
2.1 Manual Peptide Standard peptide reaction vessels (Chemglass Life Sciences or
Synthesis Peptides International) are used for manual peptide synthesis.
2.2 Automated A conventional peptide synthesizer such as the PS3 (Protein

Peptide Synthesis Technologies Inc.) or a microwave assisted peptide synthesizer
such as the Liberty Blue (CEM Corp.).
2.3 Solid Phase Resins for solid-phase synthesis, including Fmoc-Rink amide
for Peptide Synthesis 4-methylbenzhydrylamine (MBHA) (Rapp-Polymere, EMD
Millipore or Advanced ChemTech).
2.4 Amino Acids 1. Fmoc-protected amino acids (Sigma-Aldrich).

and Other Peptide/ 2. O - ( 6 - c h l o r o b e n z o t r i a z o l - 1 - y l ) - N , N , N ′ , N -
Peptoid Synthesis tetramethyluronium hexafluorophosphate (HCTU).
Reagents
3. 1-hydroxybenzotriazole (HOBt).
4. N,N′-diisopropylcarbodiimide (DIC).
Fig. 1 The design of MMP triple-helical substrates. Illustration of a MMP/peptide complex showing that mul-
tiple contacts are formed between a triple-helical collagen-model peptide and the MMP CAT and HPX domains.
The optimal positions for donor/acceptor moieties when designing synthetic fluorescent peptides were found
to be at the P5 subsite (behind the active site) and the P5′ subsite (within the hinge domain). The active site is
marked with a dashed line
5. N,N-diisopropylethylamine (DIEA).
6. 1,8-diazabicyclo[5.4.0]undec-7-ene (DBU).
7. Hydrazine hydrate.
8. [dimethylamino(triazolo[4,5-b]pyridin-3-yloxy)methylidene]-
dimethylazanium hexafluorophosphate (HATU).
9. 4-dimethylaminopyridine (DMAP).
10. Trifluoroacetic acid (TFA, peptide synthesis grade).
11. Piperidine.
12. Ammonia solution.
13. Bromoacetic acid.
14. Acetic acid.
15. Ninhydrin test kit (Anaspec).
16. Amines of choice (commercial sources for these chemical can
be found using an online chemical directory search, such as
ChemExper Chemical Directory, www.chemexper.com).
2.5 Fluorophore/ The syntheses of Fmoc-chemistry compatible fluorophore and

Quencher Containing quencher building blocks are described in detail in the literature
Synthetic Building (Table 2) and are commercially available.
Blocks
2.6 Organic Solvents 1. Dimethylformamide (DMF).

2. N-methylpyrrolidone (NMP).
3. Dimethylsulfoxide (DMSO).
4. Dichloromethane (DCM).
5. Methanol (MeOH).
6. Acetonitrile.
2.6.1 Fmoc Deprotection Mix 20 mL DBU, 100 mL of piperidine, and 880 mL of NMP.
Solution
2.6.2 Dde Deprotection The 1-(4,4-dimethyl-2,6-dioxocyclohexylidene)ethyl (Dde) pro-

Solution tecting groups are removed using 2% hydrazine in DMF. This
solution is prepared by mixing 2 mL hydrazine hydrate and 98 mL
DMF.
2.6.3 Peptide Cleavage To prepare 5 mL of such a solution mix 250 μL of H2O and 4.75
Cocktail mL of TFA. For every 100 mg of resin use 1.5 mL cleavage cock-
tail (see Note 1).
2.7 HPLC 1. Agilent 1200 Infinity series HPLC (Agilent Technologies).

Characterization 2. Analytical HPLC: Vydac C18 column (5 μm, 300 Å, 150 mm
and Purification × 4.6 mm) (Grace Davison).
3. Preparative HPLC: Vydac C18 column (15–20 μm, 300 Å,
250 mm × 22 mm) (Grace Davison).
4. Mobile phase: Solvent A—0.1% TFA in H2O. Solvent B—0.1%
TFA in acetonitrile.
5. Analytical gradient—2 to 100% B over 20 min at a flowrate of
1 mL/min.
6. Preparative gradient—5 to 50% B in 60 min at a flowrate of 10
mL/min.
7. Detection is carried out at λ = 220 and 280 nm.
2.8 Mass 1. Voyager DE-PRO Biospectrometry workstation (Applied

Spectrometry Biosystems) or Bruker Microflex LF mass spectrometer (Bruker
Daltonics).
2. MALDI matrix - α-cyano-4-hydroxycinnamic acid (HCCA) or
2,5-dihydroxybenzoic acid (DHB) (Sigma-Aldrich or Fisher
Scientific).
2.9 Circular 1. Jasco J-810 spectropolarimeter (Jasco).
Dichroism (CD) 2. Record over a wavelength (λ) range of 190–250 nm with a
Spectroscopy sample concentration of 0.1 mg/mL in 0.5% acetic acid solu-
tion (v/v).
3. Thermal transition curves are obtained by recording the molar
ellipticity ([θ]) at λ = 225 nm with an increase in temperature of
20 °C/h over a range of 5–80 °C. The temperature is controlled
by a JASCO PTC-348WI (Jasco) temperature control unit.
2.10 Peptide Labconco FreeZone lyophilizer (freeze dryer) system (Labconco

Lyophilization Corp).
2.11 Absorption 1. NanoDrop 2000c UV-Vis spectrophotometer (Thermo Fisher

Spectrophotometry Scientific).
2. Fluorescent dye-conjugated peptide is measured using the spe-
cific wavelength (λ) and molar absorption coefficient (ε) char-
acteristic for the fluorophore. Triple-helical peptide-conjugate
concentration is calculated according to the Beer-Lambert law,
using the following formula:
C = A / (l × 3 × ε ) .
where C = peptide concentration (M), A = measured absor-

bance, and l = cell path length (cm). Please note that in case of
triple-helical peptides the molar absorption coefficient ε has to be
multiplied by the number of fluorophore groups present in a
molecule.
2.12 Enzymatic 1. Chymotrypsin stock: 1 mg/ml Chymotrypsin in 1 mM HCl.

Assay Buffers, Stock 2. PMSF stock: 0.2 M phenylmethyl sulfonyl fluoride (PMSF) in
Solutions, ethanol.
and Reagents
3. APMA stock: Prepare 100 mM 4-aminophenylmercuric ace-
tate (APMA) (Sigma) in DMSO and store at 4 °C (stable for
several months).
4. Dimethylsulfoxide (DMSO).
5. 0.1 M ZnCl2.
6. Trypsin-3 (R&D Systems).
7. AEBSF.
8. Furin.
9.
SSM 3 trifluoroacetate (rel-N,N″-[[(1R,3S,4S,6R)-4,6-
bis[(aminoiminomethyl)amino]-1,3-cyclohexanedyil]bis(oxy-
4,1-phenylene)]bisguanidine tetratrifluoroacetate salt) (R&D
Systems).
10.
ProMMP-1, proMMP-2, proMMP-3, proMMP-8,
proMMP-9, proMMP-12, proMMP-13, proMT1-MMP, and
proMT2-MMP (R&D Systems).
11. Knight single-stranded peptide (SSP) [Mca-Lys-Pro-Leu-Gly-
Leu-Lys(Dnp)-Ala-Arg-NH2] was synthesized using methods
described previously [17, 73]. Knight SSP can be prepared as
a 20 mM stock solution in water and stored at −20 °C [73].
12. Enzyme assay buffer (EAB): 50 mM Tris–HCl, pH 7.5, 100
mM NaCl, 10 mM CaCl2, 0.05% Brij 35.
13. TS buffer: 50 mM Tris–HCl pH 7.5, 150 mM NaCl, 0.01%
Brij-35, 10 mM CaCl2, 1 μM ZnCl2.
14. fTHP-15 and α1(V)436–447 fTHP peptide stock solutions:

Dissolve fTHP-15 or α1(V)436–447 fTHP in DMSO and
store at 4 °C for up to 1 year.
3 Methods
3.1 Peptide For the incorporation of individual amino acids by Fmoc-solid-

Synthesis Protocols phase methodology HCTU/HOBt activation is recommended
[74]. Reagent amounts including Fmoc amino acids (Fmoc-AA),
HCTU, HOBt, and DIEA are calculated in relation to peptide
synthesis scale (typically expressed in mmol).
3.1.1 General Amino Acid For Fmoc-amino acid couplings the following molar excess is rec-
Coupling Cycles ommended: Fmoc-AA = 3 equiv., HCTU = 2.5 equiv., and DIEA
= 5.5 equiv. Recommended coupling time using a conventional
automated peptide synthesizer is 1 h.
3.1.2 General Fmoc The Fmoc protecting group is removed by gentle treatment of
Deprotection Cycle peptidyl-resin using standard Fmoc deprotection solution. This
step can be performed on an automated peptide synthesizer using
an appropriate deprotection program or manually as follows:
1. Place the resin in a peptide synthesis reaction vessel.
2. Add the appropriate amount of Fmoc deprotection solution
(three to five times the resin volume).
3. Agitate the resin for 5 min. Decant the solution.
4. Repeat step 2.
5. Agitate the resin for another 15 min. Decant the solution.
6. Wash resin (three times) using 5 mL of DMF.
3.1.3 Removal The Dde protecting group is removed from Lys residues according
of the Dde Protecting to the following steps:
Group 1. Place the resin in a peptide synthesis reaction vessel.
2. Add the appropriate amount of Dde deprotection solution
(four times the resin volume).
3. Agitate the resin for 5 min and then decant the solution.
4. Repeat the treatment with Dde deprotection solution two
more times before decanting off the solution.
5. Wash the resin three times using 5 mL of DMF.
3.1.4 “On-Resin” This protocol was adapted from [75].

Approach for Synthesis 1. The peptide is synthesized using an automated peptide synthe-
of THPs Containing 5-FAM sizer and general coupling conditions for the incorporation of
individual amino acids (see above). The Fmoc group is removed
Fig. 2 Flowchart depicting MMP triple-helical peptidase activities. The nine MMPs examined in this study were
initially subdivided based upon their ability to cleave a triple-helical substrate. Triple-helical peptidase MMPs
were further subdivided upon their ability to process more thermally stable substrates and finally, based on
their specificity for residues within the P1′ and P2 subsites. Ultimately, the flowchart illustrates that “collageno-
lytic” MMPs display distinct proteolytic behaviors. Figure modified from [51]
as outlined (see above) (see Note 1) and the 5-FAM moiety is

incorporated at the end of the peptide synthesis.
2. The side chain Dde protecting groups are removed using the
Dde deprotection solution and conditions discussed (see
above).
3. After Dde deprotection, wash the resin in the peptide synthesis
reaction vessel sequentially using 5 mL DMF, 10 mL of 10%
aqueous DMF, 10 mL DCM, and 15 mL MeOH.
4. Place the reaction vessel (no cap) in a desiccator connected to
a vacuum pump and dry the peptide-resin in vacuo for 4–6 h.
5. Conjugate the 5-FAM fluorescent dye to the ε-amino groups
of Lys using the following conditions (for a typical 0.1 mmol
scale synthesis):
(a) Place the resin in a dry reaction vessel.
(b)
Prepare a solution containing 225 mg of 5-FAM dye,
81 mg HOBt (monohydrate), 185 μL of DIC, and 73 mg
of DMAP in 5 mL of DMF.
(c) Add this mixture to the resin and agitate for 16 h. Filter
the resin and sequentially wash with 5 mL of DMF, DCM,
and MeOH.
(d) Perform a ninhydrin test. A negative result indicates com-
plete coupling of 5-FAM.
(e) Cleave the peptide from the resin using a solution contain-
ing 5% thioanisole and 90% TFA in water (use 1.5 mL of
cleavage cocktail per 100 mg of dry resin). Incubate at
room temperature for 3 h to ensure complete cleavage and
side-chain deprotection of the peptide. Perform this step
under the fume hood (see Note 2).
(f) Collect the filtrate and add 10–15 mL of water. Perform
this step under the fume hood.
(g) Freeze and then lyophilize.
(h) Purify the peptide by preparative RP-HPLC. Then analyze
by analytical HPLC and MALDI-TOF mass spectrometry
to check the purity and mass of the peptide.
(i) The concentration of the 5-FAM-containing THP peptide
can be determined by absorption spectroscopy (λ = 495
nm, ε = 75,000 cm−1 M−1) on a NanoDrop
spectrophotometer.
3.1.5 “On-Resin” This protocol is adapted from [76]. DABCYL is fairly soluble in
Approach for Synthesis DMF, but has much greater solubility in NMP.
of THPs Containing 1. Remove side chain Dde protecting groups using the Dde
DABCYL deprotection solution and conditions (see above).
Table 2
Fmoc fluorophore and quencher building block synthesis and commercial availability
Literature citation for synthesis of building

Building block block Commercial sourcea
Fluorophore
Fmoc-Lys(5-Fam)-OH [55] D, F
Fmoc-Lys(Mca)-OH [85, 86] A, B, C, D, E, F, G
Quencher
Fmoc-Lys(Dnp)-OH [87] A, B, C, D, E, F, G
Fmoc-Lys(DABCYL)-OH [88] B, C, E, F
a
A—Sigma-Aldrich (St. Louis, USA); B—EMD Millipore (Darmstadt, Germany); C—Fisher Scientific (Atlanta, USA);
D—Anaspec (Fremont, USA); E—Chemimpex (Wood Dale, USA); F—Chempep, Inc. (Wellington, USA); G—Bachem
(Torrance, USA)
2. Following Dde deprotection, wash the resin in the peptide

synthesis reaction vessel sequentially with 5 mL DMF, 10 mL
of 10% aqueous DMF, 10 mL DCM, and 15 mL MeOH.
3. In a separate flask dissolve 4 equiv. (in relation to synthesis
scale) of DABCYL in 10 mL NMP. Then activate the DABCYL
carboxyl group by adding 4.4 equiv. of DIC and 8 equiv. of
HOBt along with 8 equiv. of DIPEA.
4. Add this solution to the peptide resin and allow the acylation
reaction to continue overnight. (Note: Coupling of the
DABCYL derivative is fairly slow.)
5. After the DABCYL coupling, wash the resin in the peptide
synthesis reaction vessel sequentially with 5 mL DMF, 10 mL
DCM, and 15 mL MeOH. Repeat this washing step until the
wash solution is transparent.
6. Perform a ninhydrin test. A negative result indicates complete
coupling of DABCYL. Otherwise repeat steps 3–6.
3.1.6 Synthesis of THP This protocol (Fig. 3) is adapted from [68]. Peptoid residues can
Substrates Containing be conveniently introduced during a standard peptide synthesis
Peptoid Residues using the submonomeric approach. The incorporation of each
peptoid unit requires two steps: acylation using chloro- or bromo-
acetic acid followed by substitution with the appropriate amine
[77] as outlined below:
1. Prepare all solvents and reagents in a fume hood.
2. Place the resin in a dry reaction vessel.
3. If necessary, remove Fmoc protecting groups by agitating the
resin with 20% piperidine in DMF (3 × 6 min) using four to
five times the resin volume.
4. Wash the resin (3 min each) three times with 10 mL of DMF
followed by three times with 10 mL of DCM.
5. Dissolve bromoacetic acid (10 equiv. relative to resin loading)
and DIC (10 equiv) in an amount of DMF which equals 2–3
volumes of resin and add this solution to the resin.
6. Allow the resin to agitate at room temperature for 30 min.
7. Wash the resin three times with 10 mL of DMF for 3 min
each.
8. Dissolve the amine (25 equiv) in DMSO, NMP, or DMF.
9. Upon addition of the amine solution to the resin, allow the
resin to agitate at room temperature for 2 h. In some cases,
this displacement step may take up to 12 h (overnight).
10. Wash the resin (3 min each) three times with 10 mL of DMF
followed by three times with 10 mL of DCM.
11. Perform the ninhydrin test when necessary.
12. Continue with peptide synthesis until the desired sequence is

obtained.
3.2 Characterization The peptide molecular mass is confirmed by MALDI-TOF mass

of fTHPs spectrometry using HCCA or SA MALDI matrices. While CD
spectroscopy [40, 41, 46, 57, 58, 61, 78–80] is used to determine
the degree of triple-helicity of each peptide. The general protocol
is as follows:
1. Dissolve the peptide in an appropriate buffer to a final concen-
tration of 0.1 mg/mL. Incubate the peptide solution for
24–48 h at 4 °C to allow triple-helix formation. See Note 3.
2. Record a series of spectra (n = 5–10) over the range of λ =
190–250 nm. The molar ellipticity for a typical spectrum of a
triple-helical peptide displays a positive peak at λ =225 nm and
a minimum at λ = 205 nm.
3. Determine the melting temperature (Tm) by monitoring the
change in molar ellipticity at λ = 225 nm as the temperature is
increased at a rate of 20 °C/h from 5 to 80 °C. Data are plot-
ted using GraphPad Prism software (GraphPad) and, where
possible, the Tm value is defined as the inflection point of the
first derivative of the sigmoidal melting curve. Alternatively, Tm
is evaluated from the midpoint of the transition curve.
3.3 Enzyme 1. Dilute proMMP-1 to 100 μg/mL in TS buffer (see Note 4).
Activation 2. Activate proMMP-1 by adding APMA to a final concentration
3.3.1 Activation of 1 mM.
of proMMP-1
3. Incubate at 37 °C for 2–3 h.
4. Immediately after incubation dilute MMP-1 to 20 μg/mL and
store at −80 °C until ready to use.
5. Activity is initially measured with the Knight substrate (MMP-1
exhibits a weak activity toward fTHP-15).
3.3.2 Activation 1. Dilute proMMP-2 to 100 μg/mL in TS buffer (see Note 4).
of proMMP-2 2. Activate proMMP-2 by adding APMA to a final concentration
of 1 mM.
3. Incubate at 37 °C for 24 h.
5. MMP-2 can also be obtained in active form from EMD
Millipore (cat # PF023).
3.3.3 Activation 1. Activate proMMP-3 at 20 μg/mL in EAB containing 5 μg/
of proMMP-3 mL chymotrypsin.
2. Incubate at 37 °C for 30 min.
3. Terminate activation by adding PMSF to a final concentration

of 2 mM. Pre-warm the PMSF to 37 °C just prior to adding to
the sample.
4. Store at −80 °C for future use.
5. Measure activity using the Knight substrate at 10 μM final sub-
strate concentration.
of 1 mM.
of 1 mM.
5. MMP-9 can also be obtained in active form from EMD
Millipore (cat # PF140).
of proMMP-12 2. Activate proMMP-12 by adding APMA to a final concentra-
tion of 1 mM.
4. Immediately after incubation dilute MMP-12 to 20 μg/mL
and store at −80 °C until ready to use.
of proMMP-13 2. Activate proMMP-13 by adding APMA to a final concentra-
tion of 1 mM.
4. Immediately after incubation dilute MMP-13 to 20 μg/mL
and store at −80 °C until ready to use.
3.3.8 Activation Method 1

of proMT1-MMP 1. Prepare activation buffer containing 50 mM Tris pH 9.0, 1
mM CaCl2 and 0.5% (v/v) Brij-35.
2. Activate proMT1-MMP at 40 μg/mL with 0.86 μg/mL
recombinant furin in activation buffer.
3. Incubate at 37 °C for 1.5 h.
4. Stop the reaction by adding SSM 3 trifluoroacetate to a final

concentration of 1 μM.
Method 2
1. Prepare a stock solution of 100 mM 4-(2-aminoethyl-
benzensulfonyl fluoride hydrochloride) (AEBSF) (R&D
Systems, cat # EI001) in deionized water (see Note 5).
2. Activate 100 μg/mL proMT1-MMP by adding recombinant
human active trypsin-3 at a final concentration of 5 μg/mL in
TS buffer.
4. Stop the reaction by adding AEBSF to a final concentration of
1 mM and incubate at room temperature for 15 min.
5. Store the activated enzyme at −80 °C for future use.
3.3.9 Activation 1. Prepare AEBSF 100 mM stock in deionized water.

of proMT2-MMP (MMP-15) 2. Activate proMT2-MMP at 100 μg/mL with 0.1 μg/mL tryp-
sin-3 in TS buffer.
4. Add AEBSF to a final concentration of 1 mM and incubate at
room temperature for 15 min to stop the reaction.
5. Store the activated enzyme at −80 °C for future use.
3.4 Enzymatic Manual assays are performed in black low volume 384-well plates
Assays Using fTHPs (Greiner Bio-One, Monroe, NC) using a microplate reader such as
BioTek Synergy 4 running Gen5 1.11, Gen5 2.00, and Gen5 2.01
software as described previously [64, 81]. MMP-3 should not
cleave fTHP substrates and can be used as a negative control. Its
activity can be tested with the Knight SSP. Enzymatic assays are
performed as follows:
1. Prepare 100 μM fTHP-15 and α1(V)436–447 fTHP stock
solutions in TS buffer and incubate overnight at 4 °C to allow
for the proper folding of the peptide triple-helix. The working
stock is stable for 1 week.
2. Prepare a 12-point serial dilution of fTHP substrate with the
highest concentration being 50 μM.
3. Add 5 μL of each peptide solution to 12 wells of the 384-well
plate.
4. Measure the initial fluorescence in relative fluorescence units
(RFUs).
5. Add 5 μL of a twofold enzyme stock solution to each well.
6. Monitor fluorescence for 15 min at 25 °C to determine the
initial reaction rate.
7. Store the plates at ambient temperature for 24 h before taking

a final reading. All measurements are performed in triplicate.
8. Plot fluorescence (RFU) versus time (s). The rate of hydrolysis
(Corr Vi; μM/s) is determined by dividing the slope of the
plot by the fluorescence change after 24 h, and then multiply-
ing by the substrate concentration (μM) {Corr Vi = (Slope/
Change in RFU) × [peptide]} (see Note 6).
9. Kinetic parameters are calculated from Lineweaver-Burke,
Hanes-Woolf, and Eadee-Hofstee linear plots (substrate con-
centration versus reaction velocity) as well as nonlinear
Michaelis-Menton regression using GraphPad Prism.
4 Notes
1. During peptide synthesis the last, N-terminal amino acid to be

incorporated must be coupled as a Boc and not Fmoc-derivative
to be compatible with Dde group removal.
2. If the yield or purity of crude peptide is low a different cleavage
cocktail should be considered [82].
3. Due to aggregation of triple-helical peptides, use a concentra-
tion of 2–500 μM.
4. ProMMP-1, proMMP-3, proMMP-8, or proMMP-12 (20
ng/μL) can be activated with a chymotrypsin/trypsin mixture
(final concentration of 5 ng/μL for each) for 30 min at 37
°C. In our experience, this mode of activation results in greater
enzymatic activity compared with the manufacturer’s recom-
Fig. 3 Synthesis of fTHP substrates. (a) Peptidic fTHPs and (b) fTHP containing a peptoid residue
mended activation protocol. The reaction is stopped by the

addition of 2 mM PMSF. Enzyme activity is tested with 5 μM
Knight SSP in TS buffer.
5. AEBSF (20 or 100 mM) stock solution prepared in water is
slightly acidic. If a final pH greater than 7 is required, the pH
should be adjusted just prior to use. Stock solutions are stable
for up to 2 months at −20 °C or 1 week at 4 °C. AEBSF is
inactivated by 50% by incubating at 37 °C after 5 h.
6. The inner filter effect (IFE) is often observed when using fluo-
rogenic substrates at high concentrations. Decreases in fluores-
cence due to IFE exceed 10% once the sum of the absorbances
at excitation and emission wavelengths is greater than 0.08
[81]. Correction factors can be applied to take into account
IFEs [83]. Alternatively, fitting multiple progress curves can be
used to obtain accurate reaction kinetics even in the presence
of IFEs [81].
Acknowledgments
We gratefully acknowledge the National Institutes of Health

(EB000289, AR063795, and CA098799), the Multiple Sclerosis
National Research Institute, and the Texas Higher Education
STAR Award Program for support of our laboratory’s research on
MMPs.
References
1. Lombard C, Saulnier J, Wallach J (2005) nase 3 assessed using a phage-displayed peptide

Assays of matrix metalloproteinases (MMPs) library. J Biol Chem 275:31422–31427
activities: a review. Biochimie 87:265–272 6. Turk BE, Huang LL, Piro ET, Cantley LC
2. Nagase H, Fields GB (1996) Human matrix (2001) Determination of protease cleavage site
metalloproteinase specificity studies using col- motifs using mixture-based oriented peptide
lagen sequence-based synthetic peptides. libraries. Nat Biotechnol 19:661–667
Biopolymers 40:399–416 7. Kridel SJ, Sawai H, Ratnikov BI, Chen EI, Li
3. Fields GB (2001) Using fluorogenic peptide W, Godzik A, Strongin AY, Smith JW (2002) A
substrates to assay matrix metalloproteinases. unique substrate binding mode discriminates
In: Clark IM (ed) Methods in Molecular membrane type 1-matrix metalloproteinase
Biology 151: Matrix Metalloproteinase (MT1-MMP) from other matrix metallopro-
Protocols. Humana, Totowa, NJ, pp 495–518 teinases. J Biol Chem 277:23788–23793
4. Ohkubo S, Miyadera K, Sugimoto Y, K-i M, 8. Chen EI, Kridel SJ, Howard EW, Li W, Godzik
Wierzba K, Yamada Y (1999) Identification of A, Smith JW (2002) A unique substrate recog-
substrate sequences for membrane type-1 nition profile for matrix metalloproteinase-2.
matrix metalloproteinase using bacteriophage J Biol Chem 277:4485–4491
peptide display library. Biochem Biophys Res 9. Park HI, Turk BE, Gerkema FE, Cantley LC,
Commun 266:308–313 Sang Q-XA (2002) Peptide substrate specifici-
5. Deng S-J, Bickett DM, Mitchell JL, Lambert ties and protein cleavage sits of human endo-
MH, Blackburn RK, Carter HL III, metase/matrilysin-2/matrix
Neugebauer J, Pahel G, Weiner MP, Moss ML metalloproteinase-26. J Biol Chem
(2000) Substrate specificity of human collage- 277:35168–35175
10. Pan W, Arnone M, Kendall M, Grafstrom RH, 22. Rakhmanova V, Meyer R, Tong X (2007) New
Seitz SP, Wasserman ZR, Albright CF (2003) substrates for FRET-based assays: Use of a
Identification of peptide substrates for human long-wavelength fluorophore to detect the
MMP-11 (stromelysin-3) using phage display. activity of MMPs. Gen Eng Biotechnol News
J Biol Chem 278:27820–27827 15:40
11. Knight CG (1995) Fluorometric assays of pro- 23. Moss ML, Rasmussen FH (2007) Fluorescent
teolytic enzymes. Methods Enzymol substrates for the proteinases ADAM17,
248:18–34 ADAM10, ADAM8, and ADAM12 useful for
12. Gershkovich AA, Kholodovych VV (1996) high-throughut inhibitor screening. Anal
Fluorogenic substrates for proteases based on Biochem 366:144–148
intramolecular fluorescence energy transfer 24. Buchardt J, Ferreras M, Krog-Jensen C,
(IFETS). J Biochem Biophys Methods Delaisse J-M, Foged NT, Meldal M (1999)
33:135–162 Phosphinic peptide matrix metalloproteinase-9
13. Stack MS, Gray RD (1989) Comparison of inhibitors by solid-phase synthesis using a
vertebrate collagenase and gelatinase using a building block approach. Chem A Eur
new fluorogenic substrate peptide. J Biol Chem J 5:2877–2884
264(8):4277–4281 25. Buchardt J, Schiodt CB, Krog-Jensen C,

14. Niedzwiecki L, Teahan J, Harrison RK, Stein Delaissé J-M, Foged NT, Meldal M (2000)
RL (1992) Substrate specificity of the human Solid phase combinatorial library of phos-
matrix metalloproteinase stromelysin and the phinic peptides for discovery of matrix metal-
development of continuous fluorometric loproteinase inhibitors. J Comb Chem
assays. Biochemistry 31:12618–12623 2:624–638
15. Knight CG, Willenbrock F, Murphy G (1992) 26. Ito AS, Turchiello RDF, Hirata IY, Cezari
A novel coumarin-labelled peptide for sensitive MHS, Meldal M, Juliano L (1999) Fluorescent
continuous assays of the matrix metalloprotein- properties of amino acids labeled with ortho-
ases. FEBS Lett 296(3):263–266 aminobenzoic acid. Biospectroscopy
16. Bouvier J, Schneider P, Malcolm B (1993) A 4:395–402
fluorescent peptide substrate for the surface 27. de Souzaa ES, Hiratab IY, Julianob L, Itoa AS
metalloprotease of Leishmania. Exp Parasitol (2000) End-to-end distance distribution in
76:146–155 bradykinin observed by Förster resonance
17. Nagase H, Fields CG, Fields GB (1994) Design energy transfer. Biochim Biophys Acta
and characterization of a fluorogenic substrate 1474:251–261
selectively hydrolyzed by stromelysin 1 (matrix 28. Takara M, Ito AS (2005) General and specific
metalloproteinase-3). J Biol Chem solvent effects in optical spectra of ortho-
269:20952–20957 aminobenzoic acid. J Fluoresc 15:171–177
18. Bickett DM, Green MD, Berman J, Dezube 29. George J, Teear ML, Norey CG, Burns DD
M, Howe AS, Brown PJ, Roth JT, McGeehan (2003) Evaluation of an imaging platform dur-
GM (1993) A high throughput fluorogenic ing the development of a FRET protease assay.
substrate for interstitial collagenase (MMP-1) J Biomol Screen 8:72–80
and gelatinase (MMP-9). Anal Biochem 30. Knight CG (1991) A quenched fluorescent
212:58–64 substrate for thimet peptidase containing a new
19. Beekman B, Wouter J, Bloemhoff W, Ronday fluorescent amino acid, DL-2-amino-3-(7-
K, Tak PP, te Koppele JM (1996) Convenient methoxy-4-coumaryl)propionic acid. Biochem
fluorometric assay for matrix metalloproteinase J 274:45–48
activity and its application in biological media. 31. Woodhead-Galloway J (1980) In: Collagen:
FEBS Lett 390:221–225 The anatomy of a protein. Edward Arnold
20. Geoghegan KF, Emery MJ, Martin WH, Limited, London, pp. 10–19
McColl AS, Daumy GO (1993) Site-directed 32. Shoulders MD, Raines RT (2009) Collagen
double fluorescent tagging of human renin and structure and stability. Annu Rev Biochem
collagenase (MMP-1) substrate peptides using 78:929–958
the periodate oxidation of N-terminal serine. 33. Fields GB, Prockop DJ (1996) Perspectives on
An apparently general strategy for provision of the synthesis and application of triple-helical,
energy-transfer substrates for proteases. collagen-model peptides. Biopolymers (Pept
Bioconjug Chem 4:537–544 Sci) 40:345–357
21. Geoghegan KF (1996) Improved method for 34. Fields GB (2010) Synthesis and biological
converting an unmodified peptide to an energy applications of collagen-model triple-helical
transfer substrate for proteinase. Bioconjug peptides. Org Biomol Chem 8:1237–1258
Chem 7:385–391
35. Jenkins CL, Raines RT (2002) Insights on the teinase-2 and -9. J Biol Chem
conformational stability of collagen. Nat Prod 278:18140–18145
Rep 19:49–59 48. Lauer-Fields JL, Kele P, Sui G, Nagase H,
36. Brodsky B, Shah NK (1995) The triple-helix Leblanc RM, Fields GB (2003) Analysis of
motif in proteins. FASEB J 9:1537–1546 matrix metalloproteinase activity using triple-
37. Koide T (2005) Triple helical collagen-like helical substrates incorporating fluorogenic L-
peptides: engineering and applications in or D-amino acids. Anal Biochem
matrix biology. Connect Tissue Res 321:105–115
46:131–141 49. Minond D, Lauer-Fields JL, Nagase H, Fields
38. Koide T (2007) Designed triple-helical pep- GB (2004) Matrix metalloproteinase triple-
tides as tools for collagen biochemistry and helical peptidase activities are differentially
matrix engineering. Phil Trans R Soc B regulated by substrate stability. Biochemistry
362:1281–1291 43:11474–11481
39. Brodsky B, Thiagarajan G, Madhan B, Kar K 50. Minond D, Lauer-Fields JL, Cudic M, Overall
(2008) Triple-helical peptides: an approach to CM, Pei D, Brew K, Visse R, Nagase H, Fields
collagen conformation, stability, and self- GB (2006) The roles of substrate thermal sta-
association. Biopolymers 89:345–353 bility and P2 and P1' subsite identity on matrix
40. Lauer-Fields JL, Tuzinski KA, Shimokawa K, metalloproteinase triple-helical peptidase activ-
Nagase H, Fields GB (2000) Hydrolysis of ity and collagen specificity. J Biol Chem
triple-helical collagen peptide models by matrix 281:38302–38313
metalloproteinases. J Biol Chem 51. Minond D, Lauer-Fields JL, Cudic M, Overall
275:13282–13290 CM, Pei D, Brew K, Moss ML, Fields GB
41. Lauer-Fields JL, Nagase H, Fields GB (2000) (2007) Differentiation of secreted and
Use of Edman degradation sequence analysis membrane- type matrix metalloproteinase
and matrix-assisted laser desorption/ionization activities based on substitutions and interrup-
mass spectrometry in designing substrates for tions of triple-helical sequences. Biochemistry
matrix metalloproteinases. J Chromatogr A 46:3724–3733
890:117–125 52. Lauer-Fields JL, Chalmers MJ, Busby SA,
42. Ottl J, Battistuta R, Pieper M, Tschesche H, Minond D, Griffin PR, Fields GB (2009)
Bode W, Kühn K, Moroder L (1996) Design Identification of specific hemopexin-like
and synthesis of heterotrimeric collagen pep- domain residues that facilitate matrix metallo-
tides with a built-in cystine-knot. FEBS Lett proteinase collagenolytic activity. J Biol Chem
398:31–36 284:24017–24024
43. Ottl J, Gabriel D, Murphy G, Knäuper V, 53. Lauer-Fields JL, Fields GB (2002) Triple-
Tominaga Y, Nagase H, Kröger M, Tschesche helical peptide analysis of collagenolytic prote-
H, Bode W, Moroder L (2000) Recognition ase activity. Biol Chem 383:1095–1105
and catabolism of synthetic heterotrimeric col- 54. Lauer-Fields JL, Juska D, Fields GB (2002)
lagen peptides by matrix metalloproteinases. Matrix metalloproteinases and collagen catabo-
Chem Biol 7:119–132 lism. Biopolymers (Pept Sci) 66:19–32
44. Ottl J, Moroder L (1999) Disulfide-bridged 55. Tokmina-Roszyk M, Tokmina-Roszyk D,
heterotrimeric collagen peptides containing Fields GB (2013) The synthesis and applica-
the collagenase cleavage site of collagen type I: tion of Fmoc-Lys(FAM) building blocks.
synthesis and conformational properties. J Am Biopolymers (Pept Sci) 100:347–355
Chem Soc 121:653–661 56. Berndt P, Fields GB, Tirrell M (1995) Synthetic
45. Bertini I, Fragai F, Luchinat C, Melikian M, lipidation of peptides and amino acids: mono-
Toccafondi M, Lauer JL, Fields GB (2012) layer structure and properties. J Am Chem Soc
Structural basis for matrix metalloproteinase 1 117:9515–9522
catalyzed collagenolysis. J Am Chem Soc 57. Yu Y-C, Berndt P, Tirrell M, Fields GB (1996)
134:2100–2110 Self-assembling amphiphiles for construction
46. Lauer-Fields JL, Broder T, Sritharan T, Nagase of protein molecular architecture. J Am Chem
H, Fields GB (2001) Kinetic analysis of matrix Soc 118:12515–12520
metalloproteinase triple-helicase activity using 58. Yu Y-C, Tirrell M, Fields GB (1998) Minimal
fluorogenic substrates. Biochemistry lipidation stabilizes protein-like molecular
40:5795–5803 architecture. J Am Chem Soc 120:9979–9987
47. Lauer-Fields JL, Sritharan T, Stack MS, Nagase 59. Yu Y-C, Roontga V, Daragan VA, Mayo KH,
H, Fields GB (2003) Selective hydrolysis of Tirrell M, Fields GB (1999) Structure and
triple-helical substrates by matrix metallopro- dynamics of peptide-amphiphiles incorporating
triple-helical proteinlike molecular architec- of matrix metalloproteinase activity. J Biomol

ture. Biochemistry 38:1659–1668 Tech 15:305–316
60. Fields GB, Lauer JL, Dori Y, Forns P, Yu Y-C, 72. Hurst DR, Schwartz MA, Ghaffari MA, Jin Y,
Tirrell M (1998) Protein-like molecular archi- Tschesche H, Fields GB, Sang Q-XA (2004)
tecture: biomaterial applications for inducing Catalytic- and ecto-domains of membrane type
cellular receptor binding and signal transduc- 1-matrix metalloproteinase have similar inhibi-
tion. Biopolymers 47:143–151 tion profiles but distinct endopeptidase activi-
61. Malkar NB, Lauer-Fields JL, Borgia JA, Fields ties. Biochem J 377:775–779
GB (2002) Modulation of triple-helical stabil- 73. Neumann U, Kubota H, Frei K, Ganu V,
ity and subsequent melanoma cellular responses Leppert D (2004) Characterization of Mca-
by single-site substitution of fluoroproline Lys- Pro-Leu- Gly-Leu-Dpa-Ala-Arg-NH2, a
derivatives. Biochemistry 41:6054–6064 fluorogenic substrate with increased specificity
62. Malkar NB, Lauer-Fields JL, Juska D, Fields constants for collagenases and tumor necrosis
GB (2003) Characterization of peptide- factor converting enzyme. Anal Biochem
amphiphiles possessing cellular activation 328:166–173
sequences. Biomacromolecules 4:518–528 74. Fields GB, Tian Z, Barany G (1993) Principles
63. Lauer-Fields JL, Sritharan T, Kashiwagi M, and practice of solid-phase peptide synthesis.
Nagase H, Fields GB (2007) Substrate confor- In: Grant GA (ed) Synthetic peptides: A user’s
mation modulates aggrecanase (ADAMTS-4) guide. W.H. Freeman & Co., New York, NY,
affinity and sequence specificity: suggestion of pp 77–183
a common topological specificity of function- 75. Zhang X, Bresee J, Cheney PP, Xu B, Bhowmick
ally diverse proteases. J Biol Chem M, Cudic M, Fields GB, Edwards WB (2014)
282:142–150 Evaluation of a triple-helical peptide with
64. Lauer-Fields JL, Whitehead JK, Li S, Hammer quenched fluorophores for optical imaging of
RP, Brew K, Fields GB (2008) Selective modu- MMP-2 and MMP-9 proteolytic activity.
lation of matrix metalloproteinase 9 (MMP-9) Molecules 19:8571–8588
functions via exosite inhibition. J Biol Chem 76. Pennington MW (1994) Site specific chemical
283:20087–20095 modification procedures. In: Pennington MW,
65. Bhaskaran R, Palmier MO, Lauer-Fields JL, Dunn BM (eds) Peptide synthesis protocols:
Fields GB, Van Doren SR (2008) MMP-12 methods in molecular biology 35. Humana
catalytic domain recognizes triple-helical Press, Totowa, NJ, pp 171–186
peptide models of collagen V with exosites 77. Zuckermann RN, Kerr JM, Kent SBH, Moos
and high activity. J Biol Chem WH (1992) Efficient method for the prepara-
283:21779–21788 tion of peptoids [oligo(N-substituted gly-
66. Robichaud TK, Steffensen B, Fields GB (2011) cines)] by submonomer solid phase synthesis.
Exosite interactions impact matrix metallopro- J Am Chem Soc 114:10646–10647
teinase collagen specificities. J Biol Chem 78. Fields CG, Lovdahl CM, Miles AJ, Matthias-
286:37535–37542 Hagen VL, Fields GB (1993) Solid-phase syn-
67. Lauer JL, Bhowmick M, Tokmina-Roszyk D, thesis and stability of triple-helical peptides
Lin Y, Van Doren SR, Fields GB (2014) The incorporating native collagen sequences.
role of collagen charge clusters in the regula- Biopolymers 33:1695–1707
tion of matrix metalloproteinase activity. J Biol 79. Fields CG, Mickelson DJ, Drake SL, McCarthy
Chem 289:1981–1992 JB, Fields GB (1993) Melanoma cell adhesion
68. Stawikowski MJ, Stawikowska R, Fields GB and spreading activities of a synthetic
(2015) Collagenolytic matrix metalloprotein- 124- residue triple-helical “mini-collagen”.
ase activities towards peptomeric triple-helical J Biol Chem 268:14153–14160
substrates. Biochemistry 54:3110–3121 80. Grab B, Miles AJ, Furcht LT, Fields GB (1996)
69. Zhao Y, Marcink T, Gari RRS, Marsh BP, King Promotion of fibroblast adhesion by triple-
GM, Stawikowska R, Fields GB, Van Doren SR helical peptide models of type I collagen-
(2015) Transient collagen triple helix binding derived sequences. J Biol Chem
to a key metalloproteinase in invasion and 271:12234–12240
development. Structure 23:257–269 81. Palmier MO, Van Doren SR (2007) Rapid
70. Woessner JF, Nagase H (2000) Matrix metal- determination of enzyme kinetics from fluores-
loproteinases and TIMPs. Oxford University cence: overcoming the inner filter effect. Anal
Press, Oxford Biochem 371:43–51
71. Lauer-Fields JL, Nagase H, Fields GB (2004) 82. Guy CA, Fields GB (1997) Trifluoroacetic acid
Development of a solid-phase assay for analysis cleavage and deprotection of resin-bound pep-
tides following synthesis by fmoc chemistry. labeling of peptides on solid phase. Org Biomol
Methods Enzymol 289:67–83 Chem 6:4582–4586
83. Liu Y, Kati W, Chen C-M, Tripathi R, Molla A, 86. Malkar NB, Fields GB (2001) Synthesis of Nα-
Kohlbrenner W (1999) Use of a fluorescence (fluoren-9-ylmethoxycarbonyl)-Nε-[(7-
plate reader for measuring kinetic parameters methoxycoumarin-4-yl)acetyl]-L-lysine for use
with inner filter effect correction. Anal Biochem in solid-phase synthesis of fluorogenic sub-
267:331–335 strates. Lett Pept Sci 7:263–267
84. Amar S, Fields GB (2012) Production and 87. Anastasi A, Knight CG, Barrett AJ (1993)
characterization of matrix metalloproteinases Characterization of the bacterial metalloendo-
implicated in multiple sclerosis. In: Kokotos G, peptidase pitrilysin by use of a continuous fluo-
Constantinou-Kokotou V, Matsoukas J (eds) rescence assay. Biochem J 290:601–607
Peptides 2012 (proceedings of the t hirty-second 88. Wang GT, Krafft GA (1992) Automated syn-
european peptide symposium). European thesis of fluorogenic protease substrates: design
Peptide Society, Athens, pp 102–103 of probes for Alzheimer's disease-associated
85. Katritzky AR, Yoshioka M, Narindoshvili T, proteases. Bioorg Med Chem Lett
Chung A, Johnson JV (2008) Fluorescent 2:1665–1668
Chapter 9
Time-Resolved Analysis of Matrix Metalloproteinase

Substrates in Complex Samples
Pascal Schlage, Fabian E. Egli, and Ulrich auf dem Keller
Abstract
Identification of physiological substrates is the key to understanding the pleiotropic functions of matrix
metalloproteinases (MMPs) in health and disease. Quantitative mass spectrometry-based proteomics has
revolutionized current approaches in protease substrate discovery and helped to unravel many new MMP
activities in complex biological systems. Multiplexing further extended the capabilities of these techniques
and facilitated more complicated experimental designs that include multiple proteases or monitoring the
activity of a single protease at more than one concentration or at multiple time points with a complex test
proteome. In this chapter, we provide a protocol for time-resolved iTRAQ-based Terminal Amine Isotopic
Labeling of Substrates (TAILS), with the focus on MMP substrate identification and characterization in
cell culture supernatants and introduce an automated procedure for the interpretation of time-resolved
iTRAQ-TAILS datasets.
Key words Protease, Substrate discovery, Proteomics, iTRAQ-TAILS, Time-resolved degradomics
1 Introduction
In recent years, a variety of proteomics-based high throughput

techniques have been developed that allow for the identification of
matrix metalloproteinase (MMP) substrates and their correspond-
ing cleavage sites [1–3]. These substrate discovery approaches
often use an experimental setup that compares two conditions
(control vs. protease-treated) and apply rather arbitrary and long
incubation times [4]. Although successful, these 2plex experiments
often suffer from high numbers of false-positive identifications and
may lead to long protein lists containing hundreds of potential
substrate candidates, aggravating the selection of high-confidence
hits for follow-up studies. In order to better discriminate between
bystander substrates, i.e., unphysiological cleavages caused by
unnaturally high concentrations of the test protease or too long
incubation times, and physiologically relevant endogenous cleav-
ages, more sophisticated approaches have been introduced that are
185
186 Pascal Schlage et al.
able to quantify cleavages in more than two conditions, e.g., time-

series experiments [5, 6]. One of these, iTRAQ-based Terminal
Amine Isotopic Labeling of Substrates (TAILS) [7], is a substrate
discovery workflow that allows for the monitoring of cleavages in
up to eight experimental conditions and therefore is ideally suited
to test different test protease concentrations or incubation times in
a single experiment [5, 8]. These methods enable the determina-
tion of cleavage efficiencies for different substrates and thus signifi-
cantly improve the selection of candidate proteins for functional
characterization.
Time-resolved iTRAQ-TAILS monitors cleavages of a test pro-
tease at multiple time points of incubation and has been successfully
applied to identify high-confidence substrates of MMP10 in fibro-
blast and keratinocyte culture supernatants [5, 8]. Importantly, clas-
sification of cleavage events using time-dependent abundance
clusters of neo-N-terminal peptides, rather than based on single
time point protease/control ratios, significantly enhanced the speci-
ficity and sensitivity of substrate discovery. Moreover, time-resolved
iTRAQ-TAILS categorized substrates by cleavage site specificities
and correlated cleavage efficiencies with secondary and tertiary
structures of substrate proteins. Finally, this extended multiplexed
substrate discovery workflow narrowed in on highly relevant
MMP10 substrate candidates with known physiological functions,
generating valid testable hypotheses for mechanistic dissection.
In time-resolved iTRAQ-TAILS experiments, a complex pro-
teome is incubated with a test protease, sampled at multiple time
points after incubation, differentially labeled at the proteome level
with iTRAQ reagents and then analyzed by LC-MS/MS both
prior (preTAILS) and after (TAILS) enrichment of protein N ter-
mini (Fig. 1). Including the preTAILS analyses is critical since it
allows for the quantification of internal tryptic peptides via iTRAQ-
labeling of lysine residues. By default, these do not change in abun-
dance over time and serve, together with natural protein N termini,
as a negative dataset for high confidence filtering of cleavage events
using fuzzy clustering algorithms. In addition to assessing the pro-
gressive action of the test protease at multiple time points, pro-
teome samples not treated with protease are analyzed at the
beginning and at the end of the incubation period to control for
basal activity of endogenous proteases in the sample. Finally, iden-
tified cleavage events are assigned with the help of their time-
dependent abundance profiles to groups of substrates that are
processed by the test protease with different efficiencies.
Here, we present a detailed procedure for time-resolved prote-
ase substrate degradomics using an 8plex-iTRAQ-TAILS protocol.
Although in principle the workflow could also be applied to iden-
tify and characterize substrate cleavages of test proteases that act
on intracellular proteins, we focus our discussion on proteomes
Measuring Matrix Metalloprotease Activity in Complex Samples 187
Fig. 1 Overview of the time-resolved iTRAQ-TAILS workflow. Cell culture supernatant (CCS) is collected from
MMP-deficient cells (MMP KO) and incubated with the test MMP. Aliquots are removed at given time points and
subjected to the 8plex-iTRAQ-TAILS protocol. Including control samples at time zero and at the end of the
incubation time controls for basal proteolysis in the sample proteome. Peptides are analyzed by LC-MS/MS
prior to (preTAILS sample) and after (TAILS sample) N-terminal enrichment to enable stringent categorization
of true protease cleavages and non-cleavage events with the help of the automated analysis pipeline
prepared from cell culture supernatants (i.e., the secretome), which

are the prime targets of MMPs. We provide detailed instructions
on how to obtain a suitable secretome from adherent cells; how-
ever, this may also be performed with proteomes from alternative
sources, which have been prepared in buffer conditions compatible
with recombinant protease treatment. Data were analyzed using
fuzzy clustering for stringent filtering of true cleavage events and
subclustering of cleavages by efficiencies. Fuzzy clustering is dis-
cussed in detail and a novel R script is described that takes output
from an updated CLIPPER [9] version and automates interpreta-
tion of time-resolved iTRAQ-TAILS datasets.
2 Materials
2.1 Secretome 1. Cell culture medium (free of protein, serum, phenol red, and
Collection antibiotics).
and Concentration 2. Phosphate-buffered saline (PBS).
3. Filter with 0.22 μm pore size.
4. Phenylmethylsulfonylfluoride (PMSF).
5. Amicon Ultra-15 Centrifugal Filter Units (3 kDa MWCO,
Millipore).
6. Amine-free buffer with optimized pH for MMP activity, e.g.,
50 mM HEPES, pH 7.8.
7. Bradford protein assay (Bio-Rad).
2.2 Test Protease 1. (Optional) 10 mM APMA (4-aminophenylmercuric acetate)

Treatment (see Note 1).
2. Recombinant MMP (R&D Systems).
3. 1 M calcium chloride (CaCl2).
4. 1 M sodium chloride (NaCl).
2.3 8plex-iTRAQ- 1. 8 M Guanidine hydrochloride (GuHCl).

TAILS Protocol 2.
1 M 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid
(HEPES, pH 7.8).
3.
350 mM Tris(2-carboxyethyl)phosphine hydrochloride
(TCEP).
4. 250 mM iodoacetamide (freshly prepared).
5. 8plex iTRAQ® Reagents (AB Sciex).
6. Dimethyl sulfoxide (DMSO).
7. 1 M Ammonium bicarbonate (NH4HCO3).
8. Acetone (−20 °C).
9. Methanol (MeOH; −20 °C).
10. 100 mM Sodium hydroxide (NaOH).

11. 1 M Hydrochloric acid (HCl).
12. 1 μg/μl Trypsin in 50 mM acetic acid (Trypsin Gold, Promega)
13. Hyperbranched polyglycerol-aldehydes (HPG-ALD) polymer.
Store polymer aliquots under inert gas at −80 °C. Polymer is
available without commercial or company restriction from
Flintbox Innovation Network, The Global Intellectual
Exchange and Innovation Network (http://www.flintbox.
com/public/project/1948/).
14. 5 M sodium cyanoborohydride (NaBH3CN) solution in 1 M
NaOH (Sigma).
15. Amicon® Ultra-0.5 Centrifugal Filter Units (30 kDa MWCO,
Millipore).
2.4 Strong Cation 1. Agilent Technologies 1100/1200 series HPLC system.

Exchange (SCX)—High 2. PolySULFOETHYL A column, 200 × 2.1 mm, 5 μm 300-Å
Performance Liquid (PolyLC Inc.).
Chromatography
3. 50% Phosphoric acid (H3PO4).
(HPLC)
4. SCX buffer A: 10 mM KH2PO4 pH 2.7 in 25% acetonitrile.
5. SCX buffer B: 10 mM KH2PO4 pH 2.7, 0.5 M KCl in 25%
acetonitrile.
2.5 Desalting, 1. SpeedVac concentrator.

Pooling and Clean-Up 2. OMIX C18 pipette tips, 10–100 μl (Agilent Technologies).
of Fractionated
3. Equilibration solution: 3% acetonitrile, 0.1% trifluoroacetic
Peptides
acid (TFA).
4. 50% Acetonitrile.
5. Elution solution: 80% acetonitrile, 0.1% trifluoroacetic acid
(TFA).
6. 3% Acetonitrile, 0.1% formic acid (FA).
2.6 LC-MS/MS 1. Hybrid mass spectrometer (e.g., Thermo Scientific: Orbitrap

Analysis XL, Orbitrap Velos or Q Exactive) coupled to a capillary liquid
chromatography system with a tip column packed with C18
material (e.g., 75 μm × 150 mm; AQ, 3 μm 200 Å, Bischoff
GmbH).
2.7 Data Analysis 1. Search engine for spectrum to peptide assignment (e.g.,
Mascot, X! Tandem).
2. Trans-Proteomic Pipeline (TPP) (v7.1 POLAR VORTEX rev 1)
[10] (http://sourceforge.net/projects/sashimi/files/Trans-
Proteomic%20Pipeline%20%28TPP%29/TPP%20v4.7%20
%28polar%20vortex%29%20rev%201/).
3. CLIPPER [9] (version for TPP_v_4_7_1) (http://clipserve.

clip.ubc.ca/tails/).
4. Functional R installation (https://www.r-project.org) and the
Mfuzz Package [11].
5. R script ‘TAILS_time_analyzer.R’ (http://www.mhs.biol.
ethz.ch/research/werner/group-members-werner/aufdem-
keller.html).
3 Methods
3.1 Secretome 1. Wash 70% confluent MMP knockout (KO) cells (see Note 2)
Collection three times with PBS.
and Concentration 2. Incubate MMP KO cells with cell culture media (free of pro-
tein, serum, phenol red, and antibiotics).
3. After 24–48 h (see Note 3) collect the culture media and cen-
trifuge for 5 min at 400 × g at room temperature to remove
dead cells.
4. Collect supernatant and add serine protease inhibitor (0.5 mM
PMSF final) on ice.
5. Centrifuge supernatant for 30 min at 3,000 × g at 4 °C to
remove cell debris.
6. Collect supernatant and filter it through a 0.22 μm filter.
7. Freeze supernatants at −20 °C for storage or continue with
step 8.
8. Add water to Amicon Ultra-15 Centrifugal Filter Units and
wash filter to remove glycerol by centrifugation for 35 min at
4 °C (for optimal speed see manufacturer’s instructions).
9. Concentrate supernatants by multiple centrifugation steps at
4 °C.
10. Exchange buffer by diluting the concentrated supernatants
with 10 volumes of 50 mM HEPES, pH 7.8 and
centrifugation.
11. Repeat step 10 two times.
12. Reduce the volume of the concentrated supernatant to 1 ml by
further centrifugation.
13. If precipitates occur, try to dissolve them or remove them by
centrifugation.
14. Take an aliquot and determine protein concentration using the
Bradford protein assay.
15. Freeze concentrated secretome at −80 °C or continue with
test protease treatment.
3.2 Test Protease 1. Activate recombinant MMP by auto-activation or with APMA

Treatment according to manufacturer’s instructions (see Note 4).
2. Adjust protein concentration to 2 μg/μl with 50 mM HEPES,
pH 7.8 and add 10 mM CaCl2 and 100 mM NaCl.
3. Add activated MMP (protease-treated sample), or 50 mM
HEPES, pH 7.8 without MMP (control sample).
4. Take aliquots (0.25 mg protein for protease-treated and con-
trol samples) at different time points, e.g., 0 and 16 h for con-
trol and 1, 2, 4, 8, 12, 16 h for protease-treated samples (see
Note 5).
5. Stop the reaction by freezing at −20 °C or directly continue
with 8plex-iTRAQ-TAILS protocol.
3.3 8plex-iTRAQ- 1. Denature aliquots (0.25 mg protein at 2 μg/μl) by adding 75

TAILS Protocol μl of 8 M GuHCl and 55 μl of 1 M HEPES buffer (pH 7.8)
(final concentrations: 1 μg/μl protein, 2.5 M GuHCl, 250
mM HEPES, pH 7.8; total volume: 255 μl) and incubation at
65 °C for 15 min.
2. Reduce cysteine residues by adding 1 mM TCEP and incubate
at 65 °C for 45 min.
3. Add 5 mM iodoacetamide and incubate at 65 °C for 30 min to
alkylate cysteine residues.
4. Add 260 μl of DMSO to each vial of iTRAQ reagents (1 mg
each) and mix by pipetting.
5. Differentially label proteins by adding diluted iTRAQ reagents
(protein:iTRAQ weight ratio of 1:4) resulting in a final DMSO
concentration of 50%.
6. Incubate at room temperature for 30 min.
7. Quench labeling reaction by adding 0.1 M NH4HCO3 and
incubate for 15 min at room temperature.
8. Combine all samples.
9. Precipitate proteins by adding eight sample volumes of ice cold
acetone and one sample volume of ice cold MeOH followed by
incubation at −80 °C for 2 h.
10. Centrifuge sample at 10,000 × g for 30 min at 4 °C.
11. Wash pellet with 15 ml cold MeOH and repeat step 10.
12. Air dry pellet.
13. Resuspend dried pellet in 200 μl 0.1 M NaOH.
14. Adjust the protein concentration to 1 μg/ml and 100 mM
HEPES, pH 7.8 by adding the appropriate amount of 1 M
HEPES, pH 7.8 and water.
15. Add Trypsin at a final protease:protein ratio of 1:100 (w/w)
and incubate overnight at 37 °C.
16. (Optional) Test completeness of digest by running 3 μg of pro-

tein before and after tryptic digest by SDS-PAGE followed by
silver staining (see Note 6).
17. Take 10% of peptide solution and store at −20 °C (preTAILS
sample).
18. Adjust the pH of the remaining solution to pH 6–7 with 1 M
HCl.
19. Add a threefold excess (w/w) of HPG-ALD polymer (see Note
7) and 50 mM NaBH3CN solution and incubate overnight at
37 °C.
20. Collect unbound peptides in the flow-through by filtrating the
polymer solution in multiple steps on the same Amicon
Ultra-0.5 ml Centrifugal Filter Units (30 kDa cut-off) at
13,000 × g for 10 at room temperature. Combine filtrates in a
single tube.
21. Wash the polymer by adding 200 μl of 0.1 M NH4HCO3 to
filtration device and centrifuge again.
22. Combine the flow-throughs of steps 20 and 21 and store at
−20 °C (TAILS sample).
3.4 Peptide 1. Adjust the sample volume (preTAILS or TAILS samples) to

Fractionation 1 ml with SCX buffer A and adjust the pH to ≤2.7 by adding
by Strong Cation 50% H3PO4.
Exchange (SCX)—High 2. Centrifuge the sample for 7 min at 13,000 × g and room tem-
Performance Liquid perature and inject supernatant.
Chromatography 3. For fractionation use a linear gradient elution (0 min 0% SCX
(HPLC) buffer B; 60 min 0% SCX buffer B; 65 min 5% SCX buffer B;
100 min 35% SCX buffer B; 110 min 100% SCX buffer B) at a
flow rate of 0.2 ml/min.
4. Monitor eluate absorbance at 214 and 280 nm.
5. Collect ~27 peptide fractions with a volume of 0.5 ml along
the gradient.
6. Pool fractions containing peptides according to their absor-
bance intensities (high absorbance intensity: pool only two
fractions; low absorbance intensity: pool up to 4–5 fractions)
to ~8 final samples and desalt (see below) for LC-MS/MS
analysis.
3.5 Desalting, 1. Reduce volumes of peptide fractions under vacuum to ≤100 μl.
Pooling, and Cleanup 2. (Optional) If samples are dried or their volumes are too low
of Fractionated (≪100 μl), resuspend samples and adjust volumes to 100 μl by
Peptides adding equilibration solution.
3. Wet the C18 OMIX tip by aspirating twice 100 μl 50%
acetonitrile.
4. Equilibrate the C18 OMIX tip twice by aspirating twice 100 μl

equilibration solution.
5. Bind peptides from multiple fractions to the same C18 OMIX
tip by pipetting each fraction solution ten times up and down.
6. Wash the C18 OMIX tip seven times with 100 μl equilibration
solution.
7. Elute peptides by pipetting ten times up and down in 100 μl
elution solution.
8. Dry pooled peptides under vacuum in a SpeedVac.
9. For LC-MS/MS analysis, resuspend peptides in 20 μl 3% ace-
tonitrile containing 0.1% FA.
3.6 LC-MS/MS Analyze pooled peptide fractions for the preTAILS and TAILS
Analysis samples by tandem mass spectrometry on a hybrid mass spectrom-
eter coupled to a C18 in-line chromatography system. Settings will
depend on the mass spectrometer used, but in general we apply the
same parameters as for standard iTRAQ-based analyses. For exam-
ple, results from our test dataset (see Subheading 3.7) were obtained
on a Thermo Scientific Orbitrap Velos instrument coupled to an
Eksigent-Nano-HPLC system utilizing a self-made tip column (75
μm × 150 mm; C18 material: AQ, 3 μm 200 Å) and the following
settings: 1 μg peptide loaded; flowrate of 250 nl/min, gradient 0
to 35% acetonitrile in 62 min; full scan: 300–1700 m/z with reso-
lution of 30,000 at 400 m/z, target value 1E6; collision-induced
dissociation (CID) scan: eight most intense precursors, normalized
collision energy of 35%, activation time of 10 ms; higher energy
collision dissociation (HCD) of same precursors for iTRAQ: nor-
malized collision energy of 42%, resolution of 7500 at 400 m/z,
exclusion time of 45 s, exclusion window of 20 ppm (see Note 8).
The detailed settings for a Thermo Scientific Q Exactive instru-
ment have been published previously [12].
3.7 Data Analysis In general, time-resolved TAILS datasets are analyzed using the
CLIPPER analysis pipeline [9] according to a previously described
procedure [12]. Here, we will describe modifications needed for
time-resolved studies and focus the detailed discussion on steps
required to exploit time-resolved information for effective
discrimination of cleavage from non-cleavage events and the clas-
sification of cleavages by efficiencies. Central to our analysis strat-
egy is the use of soft clustering to stringently filter out true cleavage
events from true non-cleavage events and non-interpretable pep-
tide abundance patterns. In contrast to hard clustering (e.g.,
k-means) soft clustering algorithms (e.g., fuzzy c-means in the
Mfuzz R package [11]) do not definitively assign each object to a
single cluster, but calculate a membership value alpha (0 < alpha <
1) that defines to what extent an object belongs to a cluster.
Consequently, the number of clusters determines the minimum

alpha value that is indicative for assignment to a specific cluster,
i.e., with two clusters an alpha of 0.5 indicates equal probability for
both clusters and thus no discriminative pattern. In our analysis,
we exploit this value to discriminate between peptides whose abun-
dance increases over time and thus are derived from cleavage events
of the test protease from those that remain constant in abundance
and thus represent non-cleavage events or follow a pattern not
suitable for meaningful interpretation. Cleavage events that are
specific to the test protease are then further selected by applying
abundance cutoffs to control conditions. Finally, in a second round
of fuzzy clustering filtered cleavage events are extracted and
assigned to subclusters comprising cleavages of different efficien-
cies. Several critical parameters that can be adjusted to optimize
filtering and subclassification of true cleavages will be discussed in
more detail below.
Fuzzifier value m: This value defines the fuzziness of the clus-
ters and influences the degree to which an element can belong to
multiple clusters. We have found that the function mestimate()
that was introduced by Schwaemmle and Jensen [13] and inte-
grated into the Mfuzz package gives the best results both for filter-
ing and for subclustering of cleavage events.
Number of clusters: Determining the optimal number of clus-
ters can be very difficult. However, in our experience restricting
clusters to only two allows for the most effective filtering of cleav-
ages related to the test protease from time-resolved TAILS datas-
ets. Applying abundance cutoffs to control conditions to filter for
high-confidence cleavages of the test protease appeared to be more
efficient than generating more clusters at this stage of the analysis.
The number of subclusters highly depends on the underlying
abundance profiles of cleavage events. In general, these vary
between two and three [5, 8], but it is advisable to test higher
numbers to determine which gives the best results.
Membership value alpha: Varying the threshold cutoff for the
alpha membership value allows adjusting the stringency of event
classification. A low alpha value indicates strong deviation of the
abundance profile of a specific event from a cluster core. Thus,
using a high alpha value as a cutoff for filtering cleavage events
from non-cleavage events and noise ensures high stringency. We
recommend values of 0.8–0.9 in the first round of fuzzy clustering
with only two clusters (cleavages and non-cleavages). The cutoff
for subclusters of cleavage events will depend on the number of
clusters, but as a rule of thumb, we obtained good results with an
alpha value cutoff of 0.8 for two [5] and 0.7 for three [8]
subclusters.
Below, we give an example of the procedure using a dataset
that was extracted from [5] and that is available at http://www.
mhs.biol.ethz.ch/research/werner/group-members-werner/
aufdemkeller.html. To aid in performing this part of the analysis we

have provided the CLIPPER output file “MMP10_dataset_merge.
annotate.clip.csv” and the underlying R script (“TAILS_time_
analyzer.R”) which can be adapted to your data. In addition, you
will need to install the Trans-Proteomic Pipeline (TPP v7.1
POLAR VORTEX) [10] and an updated version of CLIPPER (for
TPP_v_4_7_1).
1. Use the TPP for secondary validation of search results, com-
bining SCX fractions and quantification of iTRAQ reporter
signals as detailed in [9] and in the CLIPPER Manual. Since
we include internal tryptic peptides and N termini in our analy-
sis, pep.xml files from both preTAILS and TAILS LC-MS/MS
runs have to be combined at the “xinteract” step in a single file.
2. Open “interact.pep.xml” output in PepXML Viewer, apply
probability cutoffs, include all columns and export as a .xls tab-
delimited text file.
3. Process the PepXML Viewer .xls output with CLIPPER and
check the following options: “Leave unlabeled N-termini”
(important to include tryptic peptides from preTAILS data-
set), “Merge multiple spectra for same peptide,” “Annotate
peptides.” This will generate a file named “_merge.annotate.
clip.csv” as input for the “TAILS_time_analyzer.R” script.
4. Open “TAILS_time_analyzer.R” in a text editor or in an inte-
grated R environment (see Note 9) and adjust the following
user-defined variables:
working_directory: full path to the directory containing the
CLIPPER output file.
data_file: file name for the CLIPPER output file (extension
must be “.csv”).
max_expression_start: relative abundance cutoff for zero
time point control (default: 0.25).
max_expression_control: relative abundance cutoff for end
time point control (default:0.25).
alpha_cutoff_main: alpha membership cutoff for cleavage
event filtering (default: 0.9).
alpha_cutoff_sub: alpha membership cutoff for cleavage
event subclustering.
clipper_unique_id: column used as unique identifiers in
CLIPPER output (default: “spectrum”).
time_point_controls: columns for zero and end time point
controls (defaults: “libra1, libra8”).
time_point_labels: time point labels as vector.
time_point_positions: time points in minutes as a numeric
vector; the number for end time point control should be slightly
higher than the maximal time point.
num_subclusters: number of subclusters to be created from
cleavages.
5. Run ‘TAILS_time_analyzer.R’ to generate the following out-

put files for inspection:
“_main_clusters_a[alpha_cutoff_main].pdf”: figure of main
clusters (see Fig. 4C in ([5])).
“_cleavages_subclusters_a[alpha_cutoff_sub].pdf”: figure
of cleavage subclusters (see Fig. 5A in ([5])).
“_clustering.csv”: CLIPPER output file augmented with
columns for alpha values of main and cleavage subclusters; fil-
tering for “main_clusters_[1,2]” > [alpha_cutoff_main] will
display peptides assigned to corresponding clusters in main
clusters; filtering for “main_clusters_1” > [alpha_cutoff_main]
and “sub_cluster_[0-9]” > [alpha_cutoff_sub] will display
cleavage events assigned to corresponding subclusters.
4 Notes
1. Some MMPs can be auto-activated by incubation at 37 °C, but

most need APMA for chemical induction of activation.
2. Cells that are deficient for the MMP (MMP KO) of interest
produce a secretome that has never seen the test protease and
thus is particularly suitable for substrate discovery.
3. Incubation times will highly depend on the ability of the cells
to grow in serum-free conditions that are mandatory to avoid
contamination with exogenous proteins and MMPs. Since
serum-free conditions are often detrimental to cells and cause
the release of intracellular proteins due to the death and lysis of
cells, we recommend optimizing the incubation time by per-
forming test experiments using assays that monitor cell death,
e.g., lactate dehydrogenase (LDH) release or MTT
((3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bro-
mide) assay, to reduce cell death to a minimum.
4. To avoid APMA-induced activation of endogenous MMPs dur-
ing test protease treatment, remove APMA after activation of
the recombinant MMP by buffer exchange using spin columns.
5. Incubation times depend on the activity of the test protease.
For MMPs, we have had good results for monitoring cleavages
up to 16 h.
6. In our experience, tryptic digests are very efficient and after
overnight incubation almost complete. However, monitoring
digestion of proteins by SDS-PAGE should be performed for
troubleshooting and especially, if the protocol is run for the
first time.
7. The actual amount depends on the HPG-ALD concentration
and binding capacity of the specific batch you purchased.
Please check package insert for details.
8. In our experience, recording both CID and HCD spectra on a

Thermo Scientific Orbitrap Velos mass spectrometer gave bet-
ter results for identification and iTRAQ-based quantification
than only HCD. In this case, spectra from both collision types
(CID and HDC) have to be merged prior to data analysis. This
can be achieved using Mascot Distiller (Matrix Science) or any
other appropriate software package.
9. There are many ways to execute R scripts and it depends upon
your particular R installation. For details, refer to the extensive
documentation available at www.r-project.org. Open the direc-
tory that contains the file named “dataset_merge.annotate.
clip.csv” and run the command “R < TAILS_time_analyzer.R
--no-save.” Alternatively, use an integrated R-environment
such as RStudio (www.rstudio.com) [14] that will allow easy
editing and execution.
Acknowledgments
A special thanks goes to S. Werner (ETH Zurich) for continuous

support of our research. We want to thank Paolo Nanni, Tobias
Kockmann, and the whole proteomics team of the Functional
Genomics Center Zurich (FGCZ) for excellent support in mass
spectrometry. This work was funded by grants from the Swiss
National Science Foundation (31003A_140726 and
31003A_163216), the European Commission (Marie Curie
International Reintegration Grant; FP7-PEOPLE- 2010-RG/
SkiNterminomics) and the Novartis Foundation for Medical-
Biological Research to U.a.d.K., and by funds from the ETH
Zurich.
References
1. Schlage P, auf dem Keller U (2015) Proteomic ase 10 substrate degradome. Mol Cell
approaches to uncover MMP function. Matrix Proteomics 13:580–593
Biol 44–46:232–238 6. Plasman K, Van Damme P, Kaiserman D et al
2. Rogers LD, Overall CM (2013) Proteolytic (2011) Probing the efficiency of proteolytic
post-translational modification of proteins: events by positional proteomics. Mol Cell
proteomic tools and methodology. Mol Cell Proteomics 10(M110):003301
Proteomics 12:3532–3542 7. Prudova A, auf dem Keller U, Butler GS et al
3. Plasman K, Van Damme P, Gevaert K (2013) (2010) Multiplex N-terminome analysis of
Contemporary positional proteomics strategies MMP-2 and MMP-9 substrate degradomes by
to study protein processing. Curr Opin Chem iTRAQ-TAILS quantitative proteomics. Mol
Biol 17:66–72 Cell Proteomics 9:894–911
4. auf dem Keller U, Prudova A, Gioia M et al 8. Schlage P, Kockmann T, Sabino F et al (2015)
(2010) A statistics-based platform for quantita- Matrix metalloproteinase 10 degradomics in
tive N-terminome analysis and identification of keratinocytes and epidermal tissue identifies
protease cleavage products. Mol Cell bioactive substrates with pleiotropic functions.
Proteomics 9:912–927 Mol Cell Proteomics 14:3234–3246
5. Schlage P, Egli FE, Nanni P et al (2014) Time- 9. auf dem Keller U, Overall CM (2012)
resolved analysis of the matrix metalloprotein- CLIPPER-An add-on to the Trans-Proteomic
Pipeline for the automated analysis of TAILS proteomes by iTRAQ-TAILS on a Thermo Q

N-terminomics data. Biol Chem 393: Exactive instrument. In: Grant J, Li H (eds)
1477–1483 Analysis of post-translational modifications and
10. Deutsch EW, Mendoza L, Shteynberg D et al proteolysis in neuroscience, Neuromethods.
(2010) A guided tour of the Trans-Proteomic Springer Protocols, New York
Pipeline. Proteomics 10:1150–1159 13. Schwammle V, Jensen ON (2010) A simple
11. Kumar L, Futschik ME (2007) Mfuzz: a soft- and fast method to determine the parameters
ware package for soft clustering of microarray for fuzzy c-means cluster analysis.
data. Bioinformation 2:5–7 Bioinformatics 26:2841–2848
12. Kockmann T, Carte N, Melkko S et al (2015) 14. RStudio Team (2015) RStudio: Integrated
Identification of protease substrates in complex Development for R. RStudio, Inc., Boston, MA
Chapter 10
Identification of Protease Cleavage Sites by Charge-Based

Enrichment of Protein N-Termini
Zon W. Lai and Oliver Schilling
Abstract
Differential proteolytic processing, for example by matrix metalloproteases (MMPs), has been recognized
as an important hallmark in numerous pathological conditions. One crucial challenge in the present studies
of proteases is system-wide identification of endogenous biological substrates. In this chapter, we highlight
a robust method for the identification of bioactive substrates and their sites of MMP cleavage, as well as by
other proteases and peptidases, in a system-wide manner. This approach enriches for putative protein
N-termini by removal of internal peptides using a charge reversal strategy. In addition, this straightforward
method can be used in combination with gel-based pre-separation of proteins to allow better estimation
of the molecular weight of the identified cleavage product of a given bioactive substrate.
Key words Protease, Cleavage, Degradome, Substrate, Proteomics
1 Introduction
Dysregulated proteolysis is a hallmark of numerous pathological

processes. However, very little is known regarding in vivo cleavage
sites and degradation targets for most proteases (including MMPs)
[1]. In recent years, investigations into elucidating endogenous
proteases substrates have gained significant impetus. As such, a
number of proteomic-based approaches have been developed to
allow for system-wide substrate identification [2–6]. These
approaches focus on the enrichment of peptides corresponding to
protein N-termini, which includes stable cleavage fragments stem-
ming from a substrate cleaved by a protease. Analysis of this
N-terminome will unequivocally yield valuable information of pro-
tease cleavage sites and cleavage specificities, ultimately leading to
identification of novel bioactive substrates. However, many of
these strategies often involve laborious sample preparation fol-
lowed by elaborate enrichment steps. To circumvent this, we have
recently proposed a cost-effective, fast, and robust method to
enrich for these N-terminal peptides via direct removal of internal
199
200 Zon W. Lai and Oliver Schilling
peptides using a charge reversal strategy [7]. First, N-termini of

proteins and polypeptides chains, including stable cleavage prod-
ucts stemming from bioactive substrates, are protected using
chemical dimethylation. Next, these proteins and polypeptide
chains are digested using trypsin, resulting in the generation of
internal peptides with unmodified neo-N-termini. These unmodi-
fied neo-N-termini are readily susceptible to coupling (through
Schiff–base reduction) to a commercially available, cost-effective
compound containing two disulfonate groups. These modified
internal peptides are subsequently removed using strong-cation
exchange chromatography, while peptides reflecting native and
protease-generated protein N-termini are retained for analysis
using high resolution mass spectrometry to elucidate precise cleav-
age sites of endogenous bioactive substrates. A general outline of
the workflow is illustrated in Fig. 1.
Fig. 1 Workflow for the enrichment of protein N-termini for protease cleavage site identification. Unmodified
primary amines of protein N-termini and lysine side chains are first protected by chemical dimethylation fol-
lowed by digestion with trypsin. Internal peptides bearing unmodified neo-N-termini are chemically modified
using sodium 4-formylbenzene-1,3-disulfonate. Endogenously modified (e.g., acetylated Protein N-termini)
and dimethylated peptides are subsequently enriched by strong cation exchange chromatography and identi-
fied by mass spectrometry analysis
Identification of Matrix Metalloprotease Cleavage Sites 201
2 Materials
2.1 Chemical 1. Recommended starting material: 1.5–3 mg total protein from

Dimethylation tissue/cell lysates or cell conditioned medium.
of Protein N-Termini 2. Refrigerated benchtop centrifuge with swing-bucket.
and Lysine
3. Water-bath sonicator.
Side Chains
4. 0.5 M iodoacetamide stock solution.
5. 1 M 1,4-dithiothreitol stock solution.
6. 0.1 M sodium hydroxide solution (keep cold on ice).
7. 6 M guanidine hydrochloride stock solution.
8. 2 M HEPES (4-(2-hydroxyethyl)piperazine-1-ethanesulfonic
acid) stock solution.
9. 37% (w/v) formaldehyde stock solution (12.2 M).
10. 1 M sodium cyanoborohydride stock solution (make fresh).
11. Optional, for quantitation: 20% (w/v) d2-13C formaldehyde
stock solution (6.6 M).
12. Acetone Precipitation Buffer: 90% (v/v) ice-cold acetone, 10%
(v/v) ice-cold methanol. Keep cold on ice.
2.2 Digestion 1. Refrigerated benchtop centrifuge with swing-bucket.

of Protein Mixtures 2. Water-bath sonicator.
by Trypsin
3. 0.1 M sodium hydroxide solution (keep cold on ice).
4. 1 M Tris (Tris(hydroxymethyl)aminomethane) stock solution.
acid) stock solution.
6. 1 mg/mL sequencing-grade trypsin stock solution.
7. Bicinchoninic acid assay kit.
8. Acetone Precipitation Buffer: 90% (v/v) ice-cold acetone, 10%
(v/v) ice-cold methanol (keep cold on ice).
2.3 N-Terminal 1. SDS-PAGE apparatus.

Modification 2. 10% SDS-PAGE acrylamide gel.
of Internal Peptides
3. Silver staining kit.
4. Bicinchoninic acid assay kit.
5. Vacuum concentrator (e.g.SpeedVacTM).
acid), pH 7.0 stock solution.
7. 1 M 4-formylbenzene-1,3-disulfonic acid disodium salt hydrate
stock solution (prepare fresh prior to use).
8. 1 M sodium cyanoborohydride stock solution (prepare fresh
prior to use).
9. C18 solid phase extraction columns.

10. C18 Equilibration Buffer: 80% (v/v) acetonitrile, 0.1% (v/v)
trifluoroacetic acid.
11. C18 Wash Buffer: 0.1% (v/v) trifluoroacetic acid.
12. C18 Elution Buffer: 80% (v/v) acetonitrile.
2.4 Removal 1. High Pressure Liquid Chromatography (HPLC) device.

of Internal Peptides 2. High resolution mass spectrometer coupled to a nano-HPLC
Using Strong-Cation (e.g., LC-MS/MS).
Exchange HPLC
3. Software for analysis of LC-MS/MS data.
4. Vacuum evaporator (e.g.SpeedVacTM).
5. 1 M Tris stock solution.
6. C18 solid phase extraction columns.
7. Strong Cation Exchange Chromatography Buffer A: 5 mM
KH2PO4, pH 2.7, 25% (v/v) acetonitrile.
8. Strong Cation Exchange Chromatography Buffer B: 5 mM
KH2PO4, 0.5 M KCl, pH 2.7, 25% (v/v) acetonitrile.
9. C18 Equilibration Buffer: 80% (v/v) acetonitrile, 0.1% (v/v)
trifluoroacetic acid.
10. C18 Wash Buffer: 0.1% (v/v) trifluoroacetic acid.
11. C18 Elution Buffer: 80% (v/v) acetonitrile.
3 Methods
3.1 Chemical 1. Extract proteins from cell/tissue lysate or cell conditioned

Dimethylation media.
of Protein N-Termini 2. Reduce extracted proteins using 5 mM dithiothreitol for 1 h at
70 °C. Adjust pH to 7.0.
3. Alkylate proteins using 15 mM iodoacetamide. Incubate in the
dark at room temp for 30 min.
4. Add another 5 mM dithiothreitol and incubate at room tem-
perature for 30 min to quench excess iodoacetamide. Adjust to
pH 7.0.
5. Precipitate proteins using prechilled Acetone Precipitation
Buffer (10:1 (v/v) buffer to sample ratio). Store at −80 °C for
at least 1.5 h.
6. Pellet precipitated proteins by centrifugation at 4,000 × g,
4 °C for 1 h.
7. Wash protein pellet using 1 mL ice-cold methanol followed by
centrifugation at 4,000 × g, 4 °C for 15 min. Repeat three
times.
8. Dissolve protein pellet in 0.5–1 mL prechilled 0.1 M

NaOH/2 M guanidine hydrochloride. If needed, sonicate in a
water-bath sonicator until proteins are completely dissolved
(keep samples cold by adding ice into water-bath sonicator).
9. Adjust protein mixture to pH 7.5–8.0 by bringing to 0.2 M
HEPES by using 2 M HEPES stock solution.
10. Determine protein concentration using bicinchoninic acid
assay (adjust concentration to approximately 1 mg/mL).
11. Heat the samples at 70 °C for 10 min. Let it cool.
12. Add 40 mM sodium cyanoborohydride and 40 mM formalde-
hyde (see Note 1 for experiments using multiplex labeling for
relative quantitation). Mix and adjust to pH 7.0. Incubate at
37 °C for 16 h.
3.2 Digestion 1. Quench excess formaldehyde by adding 50 mM Tris. Incubate

of Protein Mixtures at 37 °C for 1.5 h.
by Trypsin 2. If different formaldehyde isotopes are used for comparative
study, mix labeled samples in an equal ratio.
3. Precipitate proteins using ice-cold Acetone Precipitation Buffer
(10:1 (v/v) buffer to sample ratio) for 1.5 h at −80 °C.
Centrifuge precipitated proteins at 4,000 × g, 4 °C for 1 h.
4. Wash protein pellet using 1 mL ice-cold methanol followed by
centrifugation at 4,000 × g, 4 °C for 15 min. Repeat three
times, each time removing the fluid completely.
5. Dissolve protein pellet in 0.5–1 mL prechilled 0.1 M NaOH. If
needed, sonicate in a water-bath sonicator until proteins are
completely dissolved (keep samples cold by adding ice into
water-bath sonicator).
6. Adjust protein mixture to pH 7.5–8.0 by bringing to 0.2 M
HEPES using 2 M HEPES stock solution.
7. Determine protein concentration using bicinchoninic acid
assay and save a small aliquot for SDS-PAGE (see Subheading
3.3, step 1).
8. Add 1:100 (w/w) ratio of sequencing-grade trypsin to protein
and incubate at 37 °C for 16 h (see Note 2).
3.3 N-Terminal 1. Check tryptic digest by running 10% SDS-PAGE and follow
Modification silver staining procedure to compare samples from pre- and
of Internal Peptides post-digest. No protein bands (except trypsin) should appear
(i.e., Trypsin in the digested sample.
Cleavage Sites) 2. Add 20 mM (final concentration) sodium cyanoborohydride
and 20 mM (final concentration) 4-formyl-1,3-
benzenedisulfonic acid.
3. Adjust pH to 7.0 and incubate for 1 h at 37 °C. Repeat twice

to achieve final concentration of 60 mM sodium cyanoborohy-
dride and 60 mM 4-formyl-1,3-benzenedisulfonic acid.
Incubate for 16 h at 37 °C.
3.4 Removal 1. Add 0.1 M Tris (final concentration) for 1.5 h at 37 °C.
of Internal (Tryptic) 2. Acidify digested peptides using 0.3% (v/v) trifluoroacetic acid
Peptides Using to achieve pH of less than 3.0. If needed, hydrochloric acid can
Strong-Cation also be used.
Exchange HPLC 3. Desalt digested peptides using C18 solid phase extraction col-
umn (55–105 μm particle size and 125 Å pore size, up to 3 mg
binding capacity, e.g., Waters Sep-Pak C18 Cartridges):
(a) Equilibrate C18 column using one volume of 80% (v/v)
acetonitrile.
(b) Wash C18 column using three volumes of 0.1% (v/v) tri-
fluoroacetic acid.
(c) Apply digested peptides into C18 column. Reapply the
flow-through once to ensure all peptides are captured.
(d) Wash C18 column using three volumes of 0.1% (v/v) tri-
fluoroacetic acid.
(e) Elute captured peptides using one volume of C18 elution
Buffer.
4. Evaporate acetonitrile using vacuum evaporator.
5. Determine peptide concentration using bicinchoninic acid
assay. Adjust peptide concentration to approximately 1 mg/
mL using Strong Cation Exchange Chromatography Buffer A.
6. Resolve peptide mixture by strong cation exchange high per-
formance liquid chromatography (HPLC). Recommended
column specification: 5 μm particle size and 300 Å pore size,
column length 100 mm, inner diameter 4.6 mm, e.g., PolyLC
Polysulfoethyl A. Elute peptides using a linear gradient with
increasing concentration of Strong Cation Exchange
Chromatography Buffer B up to 45% (v/v) over a period of
90 min. (Due to a relatively large volume of samples, a larger
injection loop (>2.0 mL) is recommended. See Note 3.)
7. Collect between 10–14 peptide fractions and evaporate aceto-
nitrile using vacuum evaporator.
8. Desalt fractionated peptides using micro C18 solid phase
extraction column or self-packed C18 STAGE-tips [8].
9. Analyze fractionated peptides using high resolution LC-MS/
MS mass spectrometry analyzer (see Note 4).
3.5 LC-MS/MS Data 1. Convert raw LC-MS/MS data to a format compatible with the
Analysis X! Tandem search engines. Vendor-specific converters as
described in detail at: http://tools.proteomecenter.org/wiki/

index.php?title=Formats:mzXML.
2. Perform spectrum-to-sequence assignment searches using the
following parameters: semi Arg-C for enzyme specificity with
up to one missed cleavage site; static modification including
carboxyamidomethylation of cysteine residues (+57.02 Da),
N-terminal acetylation (+42.01 Da), and chemical dimethyl-
ation of primary amines at protein N-termini and lysine side
chains (+28.03 Da) (see Note 4). If isotopic labeling with d2-
13
C formaldehyde was employed, dimethylation of primary
amines at protein N-termini and lysine side chains (+34.06 Da)
should be included. Exemplary mass tolerances for a Q-Exactive
mass spectrometer are ±10 ppm for precursor ions and ±20
ppm for fragment ions.
3. Perform secondary validation of search results with the
PeptideProphet [9] algorithm provided as part of the Trans
Proteomic Pipeline (TPP) [10]. Convert search result files to
pep.xml format (http://tools.proteomecenter.org/wiki/
index.php?title=Formats:pepXML) and analyze peptides by
PeptideProphet, allowing for an error (false discovery) rate of
0.05.
4. Optional step for quantitation: relative quantitation for each
peptide can be analyzed using the XPRESS [11] algorithm in
TPP, using the following settings: mass tolerance of 0.015 Da
and mass difference between isotopes of N-terminal and lysine
residues of 6.0131 Da.
5. For each identified peptide, derive nonprime-side sequences
(approximately ten amino acids preceding identified peptides)
and generate input files for specificity heatmaps and sequence
logos by processing peptide lists with WebPICS [12].
4 Notes
1. For quantitative analyses of multiple proteomes, multiplex

comparison using different isotopes of formaldehyde (e.g., d0-
12
C, d2-12C, or d2-13C) and sodium cyanoborodeuteride (for
Schiff base reduction) may also be used for chemical dimethyl-
ation of protein N-termini and lysine side chains.
2. For gel-based pre-fractionation of proteins: Separate chemi-
cally dimethylated proteins using SDS-PAGE. Stain gels using
bromophenol blue, excise specific band of interest, and pro-
ceed with in-gel tryptic digestion.
3. Modified internal peptides lose their ability to bind the strong
cation exchange column at acidic pH due to the addition of
two negatively charged disulfonate groups at peptide N-termini.
Hence, these internal peptides readily flow through the col-

umn and are removed, while peptides containing true protein
N-termini bind to the strong cation-exchange resin and are
retained on the column. If insufficient material is available for
separation by HPLC, use micro-SCX tips with several elution
conditions [13].
4. For mass spectrometry analysis, disulfonate-modified peptides
are not detected unless MS acquisition is performed in the
negative ion mode. To analyze these peptides, search parame-
ters should include +247.94 Da for single and +495.89 Da for
double disulfonation of N-terminal residues.
Acknowledgment
This study was funded by a Marie Curie Fellowship for Career

Development (PIIF-GA-2012-329622 GlycoMarker to Z.W.L.),
Deutsche Forschungsgemeinschaft (SCHI 871/2 and SCHI
871/5, SCHI 871/6, GR 1748/6, INST 39/900-1 and
SFB850-Project B8 to O.S.), European Research Council (ERC-
2011- StG 282111-ProteaSys to O.S.), and the Excellence
Initiative of the German Federal and State Governments (EXC
294, BIOSS to O.S.).
References
1. Overall CM, Blobel CP (2007) In search of Arabidopsis thaliana. Proteomics

partners: linking extracellular proteases to sub- 15:2458–2469
strates. Nat Rev Mol Cell Biol 8:245–257 6. Mommen GPM, van de Waterbeemd B,
2. Kleifeld O, Doucet A, Keller UAD, Prudova A, Meiring HD, Kersten G, Heck AJR, de Jong
Schilling O, Kainthan RK, Starr AE, Foster LJ, APJM (2012) Unbiased selective isolation of
Kizhakkedathu JN, Overall CM (2010) protein N-terminal peptides from complex
Isotopic labeling of terminal amines in complex proteome samples using phospho tagging
samples identifies protein N-termini and prote- (PTAG) and TiO2-based depletion. Mol Cell
ase cleavage products. Nat Biotechnol Proteomics 11:832–842
28:281-U1441 7. Lai ZW, Gomez-Auli A, Keller EJ, Mayer B,
3. Staes A, Van Damme P, Helsens K, Demol H, Biniossek ML, Schilling O (2015) Enrichment
Vandekerckhove J, Gevaert K (2008) Improved of protein N-termini by charge reversal of
recovery of proteome-informative, protein internal peptides. Proteomics 15:2470–2478
N-terminal peptides by combined fractional 8. Rappsilber J, Ishihama Y, Mann M (2003)
diagonal chromatography (COFRADIC). Stop and go extraction tips for matrix-assisted
Proteomics 8:1362–1370 laser desorption/ionization, nanoelectrospray,
4. Shen PT, Hsu JL, Chen SH (2007) Dimethyl and LC/MS sample pretreatment in pro-
isotope-coded affinity selection for the analysis teomics. Anal Chem 75:663–670
of free and blocked N-termini of proteins using 9. Keller A, Nesvizhskii AI, Kolker E, Aebersold
LC-MS/MS. Anal Chem 79:9520–9530 R (2002) Empirical statistical model to esti-
5. Venne AS, Solari FA, Faden F, Paretti T, mate the accuracy of peptide identifications
Dissmeyer N, Zahedi RP (2015) An improved made by MS/MS and database search. Anal
workflow for quantitative N-terminal charge- Chem 74:5383–5392
based fractional diagonal chromatography 10. Keller A, Eng J, Zhang N, Li XJ, Aebersold R
(ChaFRADIC) to study proteolytic events in (2005) A uniform proteomics MS/MS analysis
platform utilizing open XML file formats. Mol derived peptide libraries improved using
Syst Biol 1:2005.0017 WebPICS, a resource for proteomic identifica-
11. Han DK, Eng J, Zhou HL, Aebersold R tion of cleavage sites. Biol Chem
(2001) Quantitative profiling of differentiation- 392:1031–1037
induced microsomal proteins using isotope- 13. Rappsilber J, Mann M, Ishihama Y (2007)
coded affinity tags and mass spectrometry. Nat Protocol for micro-purification, enrichment,
Biotechnol 19:946–951 pre-fractionation and storage of peptides for
12. Schilling O, Keller UAD, Overall CM (2011) proteomics using StageTips. Nat Protoc
Factor Xa subsite mapping by proteome- 2:1896–1906
Chapter 11
Mapping the Substrate Recognition Landscapes

of Metalloproteases Using Comprehensive Mutagenesis
Colin A. Kretz
Abstract
Protease specificity is controlled by exosites, which capture and orient the substrate, and the active site,
which binds substrate residues near the P1–P1′ scissile bond and catalyzes peptide hydrolysis. Techniques
used to identify critical contact points between a protease and its substrate can be time consuming and
labor-intensive. Screening tools such as phage display have been revitalized with the emergence of high-
throughput sequencing technology, and can be used to interrogate protease substrate specificity. This
article will outline a method for creating and screening a comprehensive mutagenesis substrate phage
display library. High-throughput sequencing of uncleaved phage at various reaction time points enables
kcat/KM determination for every possible single amino acid substitution at each position of the substrate,
providing unprecedented resolution for the interaction between a protease and its substrate.
Key words High-throughput sequencing, Phage display, Protease, Metalloprotease, Mutagenesis,

Substrate
1 Introduction
Phage display is a powerful screening tool that has been success-

fully applied to antibody development [1], peptide engineering [2,
3], and protease biology [4, 5]. Traditional approaches to phage
display utilize the replicative capacity of bacteriophage to itera-
tively reduce the complexity of a library to a subset of the best
responders. Although phage display holds promise for countless
applications, it has been limited by technical challenges (Table 1).
Despite these limitations, M13 filamentous substrate phage
display has played an important role in understanding metallopro-
tease specificity, with efforts primarily focused on describing sub-
strate residues surrounding the scissile bond. In this method,
random peptides are displayed as fusions to minor coat protein
PIII with an NH2-terminal epitope tag. Phage-displaying peptides
that are cleaved by the protease are separated from the epitope tag,
allowing for purification and selective re-amplification of cleavable
209
210 Colin A. Kretz
Table 1
Traditional vs high-throughput phage display
Traditional phage display High-throughput sequencing phage display

Multiple rounds of selection/amplification Single round selection
Interference of displayed protein with infective No infection step required
domain of phage can bias clone distribution
Toxic protein expression can lead to depletion or loss Enrichment is calculated for each clone
of specific clones individually, correcting for initial abundance
Individual validation of selected clones Replicates performed in parallel allow
statistical interpretation of responders and
non-responders
Sanger sequencing only practical for a limited number High-throughput sequencing > 108 sequences
of individual clones (20–100)
clones (Fig. 1a). Following the final panning reaction, individual

phage clones are isolated and tested in functional assays to identify
cleavable peptides, which are ultimately identified by Sanger
sequencing (Fig. 1b). Through these and other studies, matrix
metalloproteases have been found to have broad and overlapping
substrate recognition preferences [4, 6–8]. Other metalloproteases
in the human genome, such as the Adamalysins [9], have been less
well characterized [5]. Through these studies, it is hypothesized
that metalloproteases utilize exosites (substrate-binding sites dis-
tinct from the active site) to control substrate specificity.
Bacteriophage provides a direct physical link between a protein
and the DNA encoding it. As a result, recent advances in high-
throughput sequencing (HTS) technology [10] have enabled
comprehensive and quantitative profiling of every clone in a phage
display library following a single round of selection (Fig. 1c) [11–
15]. This approach alleviates many shortcomings of the traditional
approach to phage display, and greatly expands its potential uses to
interrogate biological systems. For the first time, selection repli-
cates performed in parallel can be combined for statistical interpre-
tations of enrichment and depletion [16]. Substrate phage display
affords a unique opportunity to leverage HTS technology because
the stringency of the selection procedure can be controlled by
varying incubation periods with the enzyme.
This article outlines a protocol for cloning, screening, and ana-
lyzing a mutagenesis substrate phage display library using HTS
[15]. The size of the peptides can be engineered to include critical
exosite-binding residues, allowing comprehensive profiling of resi-
dues beyond the scissile bond. Furthermore, carrying out the
selection at various reaction time points allows for the quantifica-
tion of the change in mutation burden as a function of time, and
an estimation of the kcat/KM value for every single amino acid resi-
due at each position of the peptide.
Defining the Protease Substrate Recognition Landscape 211
Fig. 1 High-throughput substrate phage display. (a) In substrate phage display, a random peptide library is
displayed as an NH2-terminal fusion to the PIII protein along with an epitope tag (such as FLAG). Phage display-
ing a peptide that is cleaved by the protease is separated from the epitope tag, allowing purification of remain-
ing uncleaved phage particles. (b) This procedure is performed for multiple rounds of selection. The selected
phage is re-amplified in bacteria to iteratively evolve the library. Following multiple rounds of selection and
amplification, individual colonies are picked off agar plates, and a number of clones are identified by Sanger
sequencing. (c) High-throughput sequencing mitigates the need for multiple rounds of selection and amplifica-
tion. Millions of reads are obtained from the starting library and from phage following a single round of
selection
2 Materials
2.1 Designing The design of a mutant substrate phage display library should
Mutagenesis Library account for HTS read length. The Illumina HiSeq platform cur-
rently delivers up to 300 million sequences of 50, 100, or 150 base
pair lengths, whereas the MiSeq platform currently delivers 30 mil-
lion reads up to 300 base pairs in length. Both formats include a
paired-end read option, thus allowing for longer sequence designs.
Because the quality of sequence data decreases toward the end of
longer reads, the library should be designed such that paired-end
212 Colin A. Kretz
reads overlap, to improve confidence in these sequence reads for

downstream data analysis. As a result of these considerations, the
mutant oligonucleotide library should currently be restricted to less
than 250 bases in length, when using the HiSeq platform. Other
HTS platforms suffer from high sequence error rates, lower
throughput, or do not offer paired end read options, which make
them less suitable for this application [17]. Future advances in this
rapidly evolving technology will undoubtedly allow for longer
mutant libraries. Consult with your HTS service provider before
designing the library to better understand current capacities.
The library is constructed using mutations introduced during
oligonucleotide synthesis, such that each nucleotide substitution is
an independent event (Fig. 2). This is achieved by spiking in a
desired amount of each alternative nucleotide into the individual
nucleotide solutions prior to synthesis. For example, to achieve a
6% mutation frequency at each G in a given sequence, 2% of A, C,
and T is added to the G solution. Alternative approaches, such as
mutagenic PCR [18], can also be used but can result in unequal
distribution of mutations and incomplete coverage of the entire
mutation space.
Fig. 2 Comprehensive mutagenesis library. To generate a comprehensive mutagenesis library, a series of

overlapping oligonucleotides are designed spanning the target sequence. The oligonucleotides will contain
both mutagenic sequences (dark shaded), and non-mutagenic sequences. Mutagenized oligonucleotides
should contain not less than 30 nucleotides of overlap. Non-mutagenic oligonucleotides contain unique Bgl1
restriction sites and an NH2-terminal epitope tag for cloning into the FUSE55 phage display vector. Following
assembly and cloning, each phage particle will contain up to five copies of a unique substrate variant displayed
on its surface, with the DNA coding for that mutant peptide contained within the phage body
The required mutation frequency at each nucleotide position is

calculated by the desired frequency of wild-type clones in the pool.
For example, a 3% mutation frequency at each nucleotide position
of a 210 base pair sequence results in 0.167% (0.97210) of clones
carrying no mutations. The resultant amino acid mutation fre-
quency will vary between constructs because of codon usage, but a
general guide is to design the library to contain 0.1–1% wild-type
clones. Because nucleotide substitutions are independent events,
the mutation frequency also dictates the fraction of clones that
contain a single amino acid substitution or multiple amino acid
substitutions. If multiple amino acid substitutions within a single
clone are not desired, then mutation frequency should be decreased.
Other options for creating mutagenized libraries with only a single
amino acid substitution have recently been reported [12, 14], but
will not be discussed here.
Once the library has been designed, contact your oligonucle-
otide synthesis provider for pricing and customization options.
The wild-type sequence is typically indicated in upper case letters
and sequences requiring the desired mutation frequency in lower
case letters.
For example:
Primer 1: AAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
AAAAAAAAAAAA.
Mutant Oligo 1: AAAAAAAAAAAAAAAaaaaaaaaaaaaaaaaa
aaaaaaaaaaaa.
Mutant Oligo 2: aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaa.
Mutant Oligo 3: aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaAAAAA
AAAAAAA.
Primer 2: AAAAAAAAAAAAAAAAAAAAAAAAA.
2.2 Cloning Library 1. Gotaq green master mix (Promega).

2. Herculase II Fusion Enzyme kit (Agilent).
3. Bgl1 (New England Biolabs).
4. T4 DNA ligase, HC (Promega).
5. FUSE55: M13 filamentous phage display vector from Smith
lab [19] (University of Missouri) containing Bgl1 cloning site
5′ to the gIII gene, and a single copy of each phage gene. As
opposed to the phagemid system, FUSE55 does not require
helper phage to supply the remaining phage proteins required
for assembly and secretion. Importantly, all PIII proteins will
display the recombinant protein fragment, allowing more pre-
cise control over the stoichiometry of substrate:enzyme ratios,
and a better estimate of kcat/KM values for each mutant
peptide.
214 Colin A. Kretz
6. Glycogen: UltraPure 20 μg/mL (Invitrogen). Working solu-

tion: 2 μg/mL in RNase-free water.
7. 3.5 M Sodium acetate, pH 3.
8. Isopropanol.
9. 70% Ethanol.
10. TE buffer: 10 mM Tris–HCl (pH 8.0), 1 mM EDTA.
11. LB-agar.
12. Tetracycline: Tetracycline-HCl is soluble in aqueous solutions
whereas tetracycline is not. Prepare 30 mg/mL solution in
water and 50% glycerol, stored at −20 °C and protected from
light. Prolonged exposure to light has anecdotally been
observed to limit growth or kill bacterial cultures.
13. MegaX DH10B electrocompetent E. coli (or other F− E.coli).
These bacteria lack F-pili to prevent phage amplification bias.
F+ bacterial strains are appropriate for helper-phage systems or
for titering phage preps.
14. Cuvettes: 0.1 cm gap (BioRad).
15. Gene Pulser Xcell Electroporation system (BioRad).
16. SOC recovery media: 20 g/L Tryptone, 5 g/L yeast extract,
4.8 g/L MgSO4, 3.603 g/L glucose, 0.5 g/L NaCl, 0.186
g/L KCl. Glucose, and MgSO4 should be added after auto-
claving. Final pH should be 6.8–7.0, and can be adjusted using
sodium hydroxide. Alternatively, entire recipe can be assem-
bled and sterile filtered through 0.22 μM filter.
17. 12 mL culture tubes.
18. Microtiter tubes.
19. Elutrap System (Whatman) which includes BT1 and BT2
membranes, electrophoresis chamber, and Elutrap device.
20. TBS-T: 20 mM Tris–HCl, pH 7.4, 150 mM NaCl, 0.05%
Tw20.
21. PCR and Gel purification column Kits (QIAgen).

22.
F55-S1 primer (5′–3′): CACCTCGAAAGCAAGCTGA
TAAACCG.
23. F55-AS1 primer (5′–3′): CGCCTGTAGCATTCCACAGA
CAGCCC.
2.3 Phage 1. 2800 mL Fernbach flask: aeration during bacterial growth

Preparation improves phage yield as well as plasmid yields.
2. 2× YT media: 16 g/L Tryptone, 10 g/L Yeast extract, 5 g/L
NaCl.
3. PEG/NaCl: 100 g polyethylene glycol-8000, 116.9 g NaCl,
475 mL water, total volume 600 mL (autoclave to sterilize and
fully solubilize).
4. TBS: 20 mM Tris–HCl, 150 mM NaCl, pH 7.4.

5. K91 E. coli, or other F-pilus containing strain. Ensure that
F-plasmid does not require tetracycline for maintenance.
6. LB agar plates containing 30 μg/mL tetracycline.
2.4 Panning 1. Protease.

2. 5× Reaction buffer: TBS, 5% bovine serum albumin, 25 mM
CaCl2, 50 μM ZnCl2 (or other appropriate buffer for your
protease).
3. Stop buffer: Reaction buffer + 50 mM EDTA (AEBSF can be
used in place of EDTA for proteases other than
metalloproteases).
4. M2 anti-FLAG agarose beads (Sigma).
5. 5 mg/mL 3× FLAG peptide (Sigma).
2.5 AlphaLISA 1. AlphaLISA buffer: 25 mM Hepes, 1 mg/mL dextran, 0.5%

Quantification Triton-100, 0.5% bovine serum albumin, 0.5% Proclin-300.
of Phage Proteolysis 2. 96-well half-well plate.
3. Anti-FLAG acceptor beads (Perkin Elmer).
4. Anti-Fd-tet biotinylated antibody (B2661-0.5ML, Sigma).
5. Streptavidin donor beads (Perkin Elmer).
6. EnSpire 2300 Multilabel Plate Reader (Perkin Elmer).
2.6 High-Throughout 1. Proteinase K (10 mg/mL).

Sequencing 2. UltraPure Phenol/chloroform/isoamyl alcohol (25:24:1,v/v)
(Invitrogen).
3. Herculase II Fusion Enzyme Kit (Agilent Technologies, prod-
uct #: 600677).
4. Agarose.
5. Illumina primers Table 2 (see Note 1).
3 Method
Mutagenized substrate phage display libraries have the potential to

identify substrate residues that make important contacts with the
protease. Individual substrate clones harboring amino acid substi-
tutions that impair proteolysis will be consumed less efficiently in
the reaction, and thus be disproportionately represented in the
selected phage population. As a result, this method can not only
identify substrate residues that interact with the protease active
site, but also define residues that interact with the protease at sites
distinct from the active site, called exosites. A well-constructed
216
Table 2
Colin A. Kretz
HTS library preparation primers
Primer Sequence
NGS S1a ACACTCTTTCCCTACACGACGCTCTTCCGATCTNNNNNNgaatcTCAATGACAGGAGGACAACAAATG
NGS S1b ACACTCTTTCCCTACACGACGCTCTTCCGATCTNNNNNNtctgaTCAATGACAGGAGGACAACAAATG
NGS S1c ACACTCTTTCCCTACACGACGCTCTTCCGATCTNNNNNNcagtcTCAATGACAGGAGGACAACAAATG
NGS S1d ACACTCTTTCCCTACACGACGCTCTTCCGATCTNNNNNNagtctTCAATGACAGGAGGACAACAAATG
NGS S1e ACACTCTTTCCCTACACGACGCTCTTCCGATCTNNNNNNtgcaaTCAATGACAGGAGGACAACAAATG
NGS AS1a CTCGGCATTCCTGCTGAACCGCTCTTCCGATCTNNNNNNgtctaACTTTCAACAGTTTCGGCCC
NGS AS1b CTCGGCATTCCTGCTGAACCGCTCTTCCGATCTNNNNNNcagagACTTTCAACAGTTTCGGCCC
NGS AS1c CTCGGCATTCCTGCTGAACCGCTCTTCCGATCTNNNNNNttcacACTTTCAACAGTTTCGGCCC
NGS AS1d CTCGGCATTCCTGCTGAACCGCTCTTCCGATCTNNNNNNagtctACTTTCAACAGTTTCGGCCC
NGS AS1e CTCGGCATTCCTGCTGAACCGCTCTTCCGATCTNNNNNNtcaggACTTTCAACAGTTTCGGCCC
PESeq S1 CAAGCAGAAGACGGCATACGAGATCGGTCTCGGCATTCCTGCTGAACCGCTCTTCCGATCT
PESeq AS1 AATGATACGGCGACCACCGAGATCTACACTCTTTCCCTACACGACGCTCTTCCGATCT
mutagenized substrate phage display library can theoretically

include every possible amino acid substitution at each position,
offering a more comprehensive interrogation of protease/sub-
strate interactions compared with other approaches.
3.1 Design, 1. Design overlapping mutagenized oligonucleotides (contact

Synthesis, supplier for details about oligonucleotide length limits) until
and Assembly entire length of desired target is covered (Fig. 2). Terminal 5′
of Mutagenized and 3′ oligonucleotides should contain non-mutagenic adapter
Substrate Phage sequences for equal amplification following assembly as well as
Display Library appropriate restriction sites for cloning into the vector (see
and Wild-Type Note 2). Internal oligonucleotides can be entirely muta-
Substrate Displaying genized. Provide no less than 30 nucleotides of overlap
Phage between oligonucleotides to allow for mismatches during
assembly.
2. In parallel, order non-mutageneic oligonucleotides for synthe-
sis of a wild-type substrate phage clone.
3. Assemble an analytical PCR reaction: 1 nM of each internal
oligonucleotide, 0.5 μM 5′ and 3′ terminal primers using
Herculase II. Thermal profile: 95 °C (30 s), 55 → 66 °C gradi-
ent (30 s), 72 °C. Repeat for 15–20 cycles.
4. Identify lowest annealing temperature in Subheading 3.1, step
3. Repeat PCR at lowest optimal annealing temperature using
Herculase II (lower annealing temperatures will reduce selec-
tion bias against nucleotide mismatches within the overlap
region) (see Note 3).
5. Separate PCR product on 1.5% agarose gel, stain with ethid-
ium bromide, and excise band.
6. Extract DNA from cut gel using EluTRAP system (see Note 4).
7. Precipitate DNA by adding 10 ng glycogen, 0.1:1 v:v sodium
acetate, and 1:1 isopropanol. Incubate at −80 °C for 2 h.
Centrifuge at 14,000 × g for 30 min at 4 °C. Wash pellet with
70% ethanol. Centrifuge at 14,000 × g for 20 min. Allow the
pellet to airdry. Resuspend in 50 μL nuclease-free water.
8. Digest PCR product and vector with Bgl1, and purify (repeat
Subheading 3.1, steps 5–7).
9. Assemble ligation reaction using DNA Ligation kit HC
(Promega). Use a 3:1 molar ratio of insert to vector and 1 μg
vector in a 50 μL ligation reaction (containing 5 μL 10× Ligase
Buffer and 1 μL ligase) (see Note 5). Assemble the ligation in
PCR tubes, and use a thermocycler to incubate the reaction at
16 °C overnight (see Note 6).
10. Heat-inactivate ligation reaction at 65 °C for 20 min and pre-
cipitate (Subheading 3.1, step 7). Resuspend ligation product
in 10 μL nuclease-free water and use immediately.
218 Colin A. Kretz
11. Electroporate ligation product into DH10B E.coli using

BioRad Gene Pulser. Add 1 μL ligation reaction to 30 μL cells
in microcentrifuge tubes on ice. Add to prechilled 1 cm dis-
posable cuvettes. Electroporation conditions: 2.0 kV, 200 Ω,
and 25 μF (see Note 7). Recover each electroporation using 2
× 1 mL SOC in a 12 mL culture tube and place at 37 °C in
shaker incubator for 1 h.
12. Pool cultures, remove 20 μL for titration, and add remaining
volume to 1 L 2× YT in a Fernbach flask containing 30 μg/
mL tetracycline. Place in 30 °C shaker incubator overnight.
Note total volume prior to addition to Fernbach flask.
13. To assess library depth, perform a 1/10 (20 μL into 180 μL)
serial dilution of culture into TBS-T. Plate 100 μL of each
dilution onto LB Agar plates containing 30 μg/mL tetracy-
cline. Incubate plates at 37 °C overnight.
14. Calculate total number of phage in recovered media: 0.01 ×
(CFU on plate) × (plate dilution factor) × (total electropora-
tion volume in μL from Subheading 3.1, step 12).
15. Identify the number of phage clones with expected insert size.
Pick 20–50 colonies into 50 μL autoclaved distilled deionized
water to serve as a template in PCR reactions. Assemble a PCR
reaction with 25 μL of 2× GoTaq. Green Master mix, 2 μL
template solution, 0.2 μM F55-S1, 0.2 μM F55-AS1, nuclease-
free water to 50 μL final volume. Thermal profile: (1) 95 °C
(5 min), (2) 95 °C (30 s), (3) 60 °C (30 s), (4) 72 °C (30 s),
(5) 72 °C (10 min), repeat 2–4 for 30 cycles.
16. Assess PCR reaction by running 10 μL on a 1–2% agarose gel.
Count the frequency of PCR products with a band of the
expected size for the insert and multiply this frequency by the
phage depth estimate from Subheading 3.1, step 14. This is
the final library depth.
17. Purify PCR products from reactions containing expected band
size using QIAquick PCR purification. Submit for Sanger
Sequencing using F55-S1 primer to validate mutation
frequency.
3.2 Phage 1. From Subheading 3.1, step 12. Pellet bacteria using a two-
Preparation step centrifugation. (Step 1) 10 min at 2400 × g, 4 °C. Transfer
supernatant containing phage to fresh bottle. (Step 2) 10 min
at 6200 × g, 4 °C.
2. Store aliquots of final library. Resuspend bacterial pellet in 50 ×
1 mL aliquots of 2× YT with 20% glycerol, and store in −80 °C.
3. To the phage-containing culture supernatant, add 0.15× vol-
ume (150 mL for 1 L culture) of PEG/NaCl. Gently mix by
inverting bottles several times (up to 100 times). Incubate at
4 °C overnight to precipitate.
4. Pellet phage by centrifugation at 6500 × g for 60 min at 4 °C.

5. Resuspend phage in 30 mL TBS. Incubate in an shaker incu-
bator for 20 min to aid in solubilizing. Centrifuge at 12,000 ×
g for 10 min at 4 °C to pellet debris.
6. Repeat Subheading 3.2, steps 3–4. Resuspend phage in 10
mL TBS-T.
7. Titer phage. Perform a 1/10 serial dilution of phage solution
(20 μL into 180 μL) into log growth phase (OD600nm 0.3–0.5)
K91 E. coli (or equivalent F-pilus expressing strain). Shake
dilutions for 1 h at 37 °C. Plate 100 μL of each dilution onto
LB Agar plates containing 30 μg/mL tetracycline. Incubate
plates at 37 °C overnight.
8. Assess phage concentration:
(a) 0.01 × (CFU on plate) × (plate dilution factor) = C1
(phage/μL).
(b) (C1 × 106)/(6.022 × 1023) = Cf (M).
3.3 Initial Reaction Experiments conducted using phage displaying the wild-type sub-
Time Course strate are important for informing the appropriate conditions for
with Phage Displaying the reaction between the protease and the mutant substrate phage
Wild-Type Substrate display library.
and AlphaLISA 1. Assemble 500 μL reaction containing 400 μL wild-type sub-
strate displaying phage (from Subheading 3.2, step 6), 100 μL
5× reaction buffer. Remove 50 μL of solution into 50 μL stop
buffer (0 min).
2. Initiate the reaction by adding remaining 450 μL of phage
solution into 5 μL of 455 nM protease stock solution, achiev-
ing a final concentration of 5 nM protease (see Note 8).
3. Remove 50 μL reaction time points into 50 μL of 4 °C stop
buffer. Repeat at each of the remaining reaction time points for
a total of eight time points. Recommended initial time points
(min): 0, 1, 5, 10, 50, 100, 500, overnight (see Note 9).
4. During reaction time course, perform a dilution series for a
standard curve. Repeat Subheading 3.3, step 1, omitting pro-
tease. Perform a 1:2 dilution series, by adding 250 μL of phage
to 250 μL 1× Reaction Buffer (1 mL 5× reaction buffer + 4 mL
TBS). Add 250 μL Stop Buffer to each dilution (see Note 10).
5. Test the selected time course for rates of phage cleavage using
AlphaLISA (see Note 11). Prepare a 1100 μL Master Mix
solution of 20 μg/mL anti-FLAG coated Acceptor beads and
4 nM biotinylated anti-fdTET in AlphaLISA Buffer. Add 25
μL of Master Mix solution to 32 wells of a 96 half-well plate.
In duplicate, add 5 μL of each standard curve dilution, or 5 μL
of each reaction time point to each well. Incubate at room
temperature for 1 h. Prepare a 700 μL solution of 40 μg/mL
220 Colin A. Kretz
streptavidin-
coated donor beads in AlphaLISA Buffer, and
add 20 μL to each well. Protect from light, and incubate at
room temperature for 30 min (see Note 12). Read the plates
on an EnSpire 2300 Multilabel Plate Reader at an excitation
wavelength of 680 nm and emission wavelength of 615 nm.
6. Construct a standard curve of AlphaLISA signal as a function
of the natural log of the molar phage concentration. Determine
the concentration of uncleaved phage at each time point from
the protease reacted phage, and calculate kcat/KM value.
3.4 Reaction 1. Repeat Subheading 3.3, steps 1–6 using mutant substrate
of Phage Displaying phage display library.
Mutant Substrate 2. Selection of reaction time points (t0–t5) for analysis by HTS
Phage Display Library should be based on time points exhibiting less than 75% sub-
with Protease strate consumption (see Note 13). A single time point repre-
senting maximal library cleavage (Δ[substrate] → 0) should also be
taken (tfinal).
3. Add 500 μL Stop Buffer and 50 μL anti-FLAG agarose beads
to select mutant substrate phage library reaction time points.
Incubate at 4 °C overnight with end-over-end mixing.
4. Centrifuge at 1,000 × g for 2 min to pellet beads. Wash beads
5× with Stop Buffer.
5. Elute phage from beads at each reaction time point by adding
500 μL of 0.15 mg/mL 3× FLAG peptide in Stop Buffer.
Incubate 30 min at room temperature with end-over-end
mixing.
6. Centrifuge at 1,000 × g for 2 min to pellet beads. Remove
supernatant (containing selected phage) into fresh tube.
7. Add 5 μL of 10 mg/mL proteinase K. Incubate at 50 °C for 4
h (or overnight) to digest phage coat protein, exposing ssDNA.
8. Add 500 μL phenol/chloroform/isoamyl alcohol, and vortex.
Centrifuge at 14,000 × g for 10 min at room temperature.
Remove top aqueous layer into clean microcentrifuge tube.
Add 200 μL chloroform to remove excess phenol, vortex, and
then centrifuge for 10 min at 14,000 × g. Remove top aqueous
layer into a fresh tube.
9. Precipitate and concentrate ssDNA as in Subheading 3.1, step 7.
10. Resuspend pelleted ssDNA in 20 μL TE Buffer.
3.5 Preparation 1. Assemble a 50 μL PCR reaction for each ssDNA sample col-
of High-Throughput lected from reaction time points (Subheading 3.4, step 10)
Sequencing Samples containing 5 μL template, 0.5 μM NGS-S1(a → e), 0.5 μM
NGS-AS1(a → e), and Herculase reaction components accord-
ing to manufacturer’s instructions. The thermal profile is: (1)
98 °C (2 min), (2) 98 °C (20 s), (3) 60 °C (20 s), (4) 72 °C (30
s), (5) 72 °C (5 min), repeat 2–4 for 8–16 cycles (see Note 14).
2. Gel purify on a 2% agarose gel.

3. Assemble a 50 μL PCR reaction for each PCR product of reac-
tion time points, containing 10 ng template (from Subheading
3.5, step 2), 0.5 μM PE1-seq, 0.5 μM PE2-seq, and Herculase
reaction components according to manufacturer’s instruc-
tions. The thermal profile is: (1) 98 °C (2 min), (2) 98 °C (20
s), (3) 60 °C (20 s), (4) 72 °C (30 s), (5) 72 °C (5 min), repeat
2–4 for 8–16 cycles (see Note 15).
4. Repeat Subheading 3.5, step 2.
HTS libraries can now be submitted for sequencing. To assess
library quality (a) samples can be quantified using qPCR (Kapa
BioSystems, product #: KK4835) to ensure proper Illumina
adapter assembly, and (b) samples can be run on a BioAnalyzer to
ensure proper fragment size. Paired-end sequencing is the most
common format for these applications because it allows for cover-
age of longer sequences, and sequence redundancy in overlap
regions can be used to correct for sequencing errors (see Note 16).
Each sample has a unique nucleotide sequence added to it (see
primers for Subheading 3.5, step 1) that acts as a barcode for that
sample. This allows many samples to be pooled and sequenced on
the same HTS run, which saves cost and time. Sequencing data for
each reaction time point is then separated bioinformatically based
on the barcode sequence.
3.6 Data Analysis Data analysis pipelines for high-throughput sequencing of phage
display experiments are still largely dependent on in-house coding
algorithms. Although some formal pipelines are available to calcu-
late mutation frequency and assess enrichment [16, 20], they do
not provide the flexibility for customized analysis. A general strat-
egy is provided for analyzing the performance of mutant substrate
sequences (Fig. 3), as quantified from high-throughput sequenc-
ing data.
3.6.1 Raw Data 1. Parse sequencing data into bins according to barcodes, repre-
sented by nucleotide reads 7–11. Each bin now contains the
sequencing data collected from the uncleaved phage at each
reaction time point. Check for comparable sequence read
depth across reaction time points.
2. Trim sequences of the 6 N wobble, barcode, and primer, leav-
ing only the mutant library.
3. FASTQ files from paired-end reads are aligned to wild-type
sequence using Bowtie2 [21], yielding a paired SAM output
file (http://bowtie-bio.sourceforge.net).
4. Assemble paired end reads into a single sequence. If paired
reads overlap and alignments show discordant base calls, the
base with the higher quality score is kept. If the discordant
222 Colin A. Kretz
Fig. 3 High-throughput sequencing analysis scheme. An outline for the analysis of high-throughput sequencing
data is shown. Following high-throughput sequencing of the uncleaved phage at various reaction time points,
raw sequencing reads are separated into bins. Paired-end (PE) reads are assembled into single sequences, and
the frequency (f) of wild-type sequences is initially analyzed. Next, the frequency of each amino acid (AA) at
each position of the substrate is calculated by dividing the count by the total number of reads assigned to that
bin. Alternatively, the frequency of each amino acid is only calculated from the subset of clones that only con-
tain a single amino acid substitution. This can be used to mitigate the effects of multiple substitutions within a
clone affecting the results. Ultimately, the frequency of each amino acid at each time point is multiplied by the
concentration of uncleaved phage at that time point, before fitting the data to the first order rate law
base calls have identical quality scores, the forward read (read
1) should be preferred.
3.6.2 Analysis of Wild- 1. Assess the frequency of wild-type substrate in each reaction
Type Sequences time point. In the starting library (t0), wild-type should be ~(1
Within the Library − μ)n, where μ is the mutation frequency at each nucleotide
position and n is the number of nucleotides in the sequence
(see Note 17).
2. Multiply the frequency of wild-type sequence, by the molar
concentration of uncleaved phage at the corresponding reac-
tion time point (from Subheading 3.4, step 1).
3. Fit the data to the first order rate law: ln([A]t) = −kt + ln([A]0).
Perform a regression of the natural log of molar concentration
of WT phage as a function of time. The slope, k, divided by the
molar concentration of enzyme yields kcat/KM. Compare this
value to previously determined values (from Subheading 3.3)
to assess the reliability of the data (see Note 18).
4. Determine the effect of multiple substitutions within a clone
on the rate of proteolysis measured for a given mutant at a
position (i.e., epistasis), using the WT amino acid frequency at
each position. These counts are derived from substrate clones
containing multiple amino acid substitutions, but share the
WT amino acid at this position in common. As in Subheading
3.6.2, steps 2–3, calculate the kcat/KM for each wild-type
amino acid at each position. If the value for each WT amino
acid at each position is comparable to each other, it suggests
that the aggregate effect of multiple substitutions on a given
peptide is mostly averaged across sequences sharing a given
amino acid. Thus, strong epistasis is not common within the
dataset. If needed, more formal and rigorous analysis of epista-
sis should be explored [15, 22].
3.6.3 Analysis of Mutant 1. If Subheading 3.6.2, step 4 indicates limited epistasis within
Sequences the library, then the counts for each amino acid at each posi-
Within the Library tion of the substrate (Subheading 3.6.2, step 3, above) can be
used to calculate the kcat/KM for each mutant sequence.
2. If Subheading 3.6.2, step 4 suggests multiple interactions
between mutations at certain positions within the library, then
the mutation frequency calculated from Subheading 3.6.2,
step 3, cannot reliably estimate the effect of a single amino
acid substitution on the rate of cleavage in the library. Instead,
the frequency of each amino acid substitution at each position
of the peptide should be calculated from the subset of sequences
that contain only a single amino acid substitution. This will
only cover a fraction of the potential sequence space of the
peptide. However, this subset of mutations will likely contain
224 Colin A. Kretz
the majority of amino acid changes requiring only a single

nucleotide substitution within a codon, which accounts for the
vast majority of point mutations that cause human disease (see
Subheading 4).
3. Assemble a table for the counts of each 20 amino acid at each
position of the substrate. Divide each count by the total num-
ber of sequence reads assigned to each time point. The wild-
type amino acid should be the highest frequency amino acid at
each position. Calculate the molar concentration of phage con-
taining each amino acid by multiplying each frequency by the
concentration of uncleaved phage at the corresponding reac-
tion time point (from Subheading 3.4, step 1).
4. Exclude amino acid substitutions from the table that are cov-
ered by less than five sequencing reads in any time point.
5. If the slope of a regression line of concentration vs time is >0,
then set kcat/KM value to 0. These values represent amino acid
substitutions that prevent any detectable substrate cleavage in
the conditions of the assay.
6. Calculate kcat/KM for each amino acid substitution at each
position of the substrate to determine the effect of every pos-
sible amino acid substitution on proteolytic cleavage (as in
Subheading 3.6.2, step 3).
7. Use the Benjamini-Hochberg [23] statistical model on reac-
tion replicates to calculate p-values, adjusted for multiple
observations, for every kcat/KM value (also called the False
Discovery Rate, or FDR).
4 Conclusion
The capacity to investigate the entire sequence space of a protein

can provide important insights into function. The method out-
lined in this report should serve as a useful guide for creating and
screening mutagenized substrate phage display libraries. Coupling
this approach with high-throughput sequencing allows the quanti-
fication of the impact of every possible amino acid substitution at
each position of a substrate on the rate of its cleavage by the cog-
nate protease.
In addition to defining important contact points between a
protease and its substrate, this approach may be useful for inter-
preting the impact of genetic variants on protein function [24].
Whole genome and whole exome sequencing routinely identifies
genetic mutations in both healthy individuals and patients. While
frameshifts, stop-gain, and gene deletions can generally be expected
to cause a loss of protein function, interpretation of the much
larger set of mutations that result in an amino acid change is more
difficult. These variants of uncertain significance are often assigned

a functional score based on evolutionary sequence conservation
using predictive computer algorithms. However, these computa-
tional tools exhibit high error rates [25], which limit their utility.
High-throughput mutagenesis profiling methods, like the one
described here, can generate empirical biochemical data for all pos-
sible amino acid substitutions of a given protein. Precise quantifi-
cation of the impact of amino acid changes on protein function
may better predict the link between genetic variation and disease.
Ultimately, these techniques will facilitate the establishment of
comprehensive missense mutation databases with accurate predic-
tions of functional significance that can be used by investigators
and clinicians to improve diagnosis and patient care.
5 Notes
1. These barcoding primers contain sequences that anneal up-

stream or down-stream of the mutagenized library (bold), a
barcode specific to each sample to allow for multiplexing many
samples on a single sequencing lane (lowercase), and the first
half of the Illumina adapter. Illumina sequencers use the first
six nucleotide reads to establish optical coordinates for each
cluster on the chip. As a result, it is often advisable to place a
6 N wobble upstream of the barcode to provide enough
sequence diversity for each cluster to be distinguished from its
neighbors.
2. In the case of FUSE55, these include unique BglI sites. In
addition, the 5′ terminal primer should also add an epitope tag
to the library, such as FLAG.
3. Be careful about the number of PCR cycles. Too many cycles
can introduce biases in the library at an early stage and will not
become evident until the first set of HTS data comes back. We
aim for not more than 15–20 cycles.
4. Gel extraction techniques based on chaotropic salts to melt the
gel and columns to capture DNA (such as QIAgen) can also be
used, but can impair downstream steps. For example, ligation
reactions are very sensitive to salt contamination.
5. To calculate the amount of insert to include in the ligation
reaction, use the following equation: Insert Mass (ng) = Vector
Mass × 3 × (Number of insert base pairs)/(Number of vector
base pairs). It is generally recommended to perform multiple
ligation reactions in parallel to optimize the ratio of insert:vector
and the total volume of the ligation reaction.
6. Avoid freeze thaw cycles of the 10× ligation buffer, as this low-
ers the ATP yield in the buffer and negatively impacts ligation
efficiency.
226 Colin A. Kretz
7. Much of the success of the library build comes from this step.
Poor transformation efficiency can lead to poor library depth,
and ultimately incomplete diversity. Typically, ten electropora-
tions per library are sufficient to yield 1 × 10 [7] independent
clones. However, more may be required. 30–40 electropora-
tions can be needed to yield sufficient diversity. Keep an eye on
the time constant, optimally at 4.2 ms, if drops below 4 ms
may indicate salts in the ligation product that increase conduc-
tivity. Try additional 70% ethanol washes at room temperature
(including a 20 min incubation) to extract excess salts from the
precipitation reaction. Also, too much glycogen can impede
transformation efficiency.
8. It is important to perform these reactions under pseudo-first
order reaction conditions, where the concentration of sub-
strate is ≪KM. Leave lots of room here, in case mutations are
discovered that enhance cleavage relative to WT. But an initial
reaction setup to the conditions for a WT substrate will suffice
for an initial analysis.
9. The reaction time course will need to be determined empiri-
cally if the interaction is not already well defined. Adjust prote-
ase concentration and time points accordingly.
10. Stop buffer contains EDTA, which we have found to interfere
with the AlphaLISA signal. This must be accounted for in the
standard curve.
11. AlphaLISA is a rapid alternative to standard ELISA protocols.
However, if appropriate plate readers are not available, then
standard ELISA can also be used. In this case, precoat plates
with 10 μg/mL anti-fd-TET antibody (Sigma) and detection
with 1/10,000 HRP-anti-FLAG antibody (Sigma). Detection
is achieved using 1 Step Ultra TMB-ELISA substrate (Thermo
Scientific #34028). Filamentous phage exhibits a high degree
of nonspecific binding to plastic surfaces, which can be lever-
aged by capturing with an anti-phage antibody.
12. Donor beads are highly susceptible to photo-bleaching. Turn
off overhead fluorescent lights when working with donor
beads.
13. This may need to be determined empirically, as it depends on
the KM of the interaction. Dogmatically, initial rates of a reac-
tion are determined with less than 10% substrate consumption.
However, fitting the data to a first order rate law accounts for
decreased rates of cleavage at later time points caused by sub-
strate consumption.
14. The number of cycles should be determined empirically to
avoid PCR recombination (product laddering) that can cause
bias in library distribution unrelated to the protease selection.
This phenomenon tends to occur when PCR reactions are
allowed to proceed beyond log-phase, and may be related to

products themselves acting as primers. This is less of an issue
with mutagenic libraries, and may be more relevant with whole
genome sequencing libraries or other samples where products
contain partially complementary overlapping sequences.
15. These are universal primers that can be used regardless of bar-
coding primer.
16. Illumina sequencing is currently limited by the size of DNA
fragments that can reliably cluster on the chip surface. Typically,
total fragment length of less than 700 bp is preferable. Check
with sequencing services for more details. Discuss with
sequencing service provider about your anticipated library
diversity. Low mutational burden within the libraries can affect
sequence quality scores. Increasing PhiX to 20% of total sam-
ple quantity is useful. Alternatively, multiplexing with RNASeq,
ExomeSeq, or Whole genome sequencing can provide the nec-
essary sample diversity for a successful sequencing reaction.
17. Some nucleotide substitutions will not alter the wild-type
amino acid. This fact is not accounted for in this formula.
18. Because of the complexity of the reaction conditions, these
values represent apparent rate constants.
Acknowledgments
I would like to thank David Ginsburg (University of Michigan) for

critical review of the manuscript. Colin A. Kretz holds a McMaster
University Department of Medicine Internal Career Award. This
work was also supported by the Judith Graham Pool Fellowship
from the National Hemophilia Foundation, awarded to Colin
A. Kretz.
References
1. Chan CE, Lim AP, MacAry PA, Hanson BJ 5. Hills R, Mazzarella R, Fok K, Liu M,
(2014) The role of phage display in therapeutic Nemirovskiy O, Leone J, Zack MD, Arner
antibody discovery. Int Immunol 26:649–657 EC, Viswanathan M, Abujoub A,
2. Chen S, Heinis C (2015) Phage selection of Muruganandam A, Sexton DJ, Bassill GJ, Sato
bicyclic peptides based on two disulfide bridges. AK, Malfait AM, Tortorella MD (2007)
Methods Mol Biol 1248:119–137 Identification of an ADAMTS-4 cleavage
3. Whitney M, Crisp JL, Olson ES, Aguilera TA, motif using phage display leads to the devel-
Gross LA, Ellies LG, Tsien RY (2010) Parallel opment of fluorogenic peptide substrates and
in vivo and in vitro selection using phage display reveals matrilin-3 as a novel substrate. J Biol
identifies protease-dependent tumor- targeting Chem 282:11101–11109
peptides. J Biol Chem 285:22532–22541 6. Eckhard U, Huesgen PF, Schilling O, Bellac
4. Kridel SJ, Chen E, Kotra LP, Howard EW, CL, Butler GS, Cox JH, Dufour A, Goebeler V,
Mobashery S, Smith JW (2001) Substrate Kappelhoff R, Keller UA, Klein T, Lange PF,
hydrolysis by matrix metalloproteinase-9. Marino G, Morrison CJ, Prudova A, Rodriguez
J Biol Chem 276:20572–20578 D, Starr AE, Wang Y, Overall CM (2016)
228 Colin A. Kretz
Active site specificity of the matrix metallopro- FA, Meng F, Ginsburg D (2015) Massively
teinase family: proteomic identification of 4300 parallel enzyme kinetics reveals the substrate
cleavage sites by nine MMPs explored with recognition landscape of the metalloprotease
structural and synthetic peptide cleavage analy- ADAMTS13. Proc Natl Acad Sci U S A
ses. Matrix Biol 49:37–60 112:9328–9333
7. Chen EI, Kridel SJ, Howard EW, Li W, Godzik 16. Fowler DM, Araya CL, Gerard W, Fields S
A, Smith JW (2002) A unique substrate recog- (2011) Enrich: software for analysis of protein
nition profile for matrix metalloproteinase-2. function by enrichment and depletion of vari-
J Biol Chem 277:4485–4491 ants. Bioinformatics 27:3430–3431
8. Ratnikov BI, Cieplak P, Gramatikoff K, Pierce 17. Glenn TC (2011) Field guide to next-
J, Eroshkin A, Igarashi Y, Kazanov M, Sun Q, generation DNA sequencers. Mol Ecol Resour
Godzik A, Osterman A, Stec B, Strongin A, 11:759–769
Smith JW (2014) Basis for substrate recogni- 18. Desch KC, Kretz C, Yee A, Gildersleeve R,
tion and distinction by matrix metalloprotein- Metzger K, Agrawal N, Cheng J, Ginsburg D
ases. Proc Natl Acad Sci U S A (2015) Probing ADAMTS13 substrate speci-
111(40):E4148–E4155 ficity using phage display. PLoS One
9. Huxley-Jones J, Clarke TK, Beck C, Toubaris 10:e0122931
G, Robertson DL, Boot-Handford RP (2007) 19. Scott JK, Smith GP (1990) Searching for pep-
The evolution of the vertebrate metzincins; tide ligands with an epitope library. Science
insights from Ciona intestinalis and Danio 249:386–390
rerio. BMC Evol Biol 7:63 20. Love MI, Huber W, and Anders S (2014).
10. Shendure J, Lieberman Aiden E (2012) The Moderated estimation of fold change and dis-
expanding scope of DNA sequencing. Nat persion for RNA-seq data with DESeq2.
Biotechnol 30:1084–1094 Genome Biology 15:550
11. Tinberg CE, Khare SD, Dou J, Doyle L, 21. Langmead B, Salzberg SL (2012) Fast gapped-
Nelson JW, Schena A, Jankowski W, Kalodimos read alignment with Bowtie 2. Nat Methods
CG, Johnsson K, Stoddard BL, Baker D (2013) 9:357–359
Computational design of ligand-binding pro- 22. Araya CL, Fowler DM, Chen W, Muniez I,
teins with high affinity and selectivity. Nature Kelly JW, Fields S (2012) A fundamental pro-
501:212–216 tein property, thermodynamic stability, revealed
12. Kitzman JO, Starita LM, Lo RS, Fields S, solely from large-scale measurements of pro-
Shendure J (2015) Massively parallel single- tein function. Proc Natl Acad Sci U S A
amino- acid mutagenesis. Nat Methods 109:16858–16863
12:203–206 23. Hochberg YBY (1995) Controlling the false
13. Fowler DM, Araya CL, Fleishman SJ, Kellogg discovery rate: a practical and powerful
EH, Stephany JJ, Baker D, Fields S (2010) approach to multiple testing. J R Stat Soc
High-resolution mapping of protein sequence- Series B 57:289–300
function relationships. Nat Methods 7:741–746 24. Shendure J, Akey JM (2015) The origins,
14. Starita LM, Young DL, Islam M, Kitzman JO, determinants, and consequences of human
Gullingsrud J, Hause RJ, Fowler DM, Parvin mutations. Science 349:1478–1483
JD, Shendure J, Fields S (2015) Massively par- 25. Gnad F, Baucom A, Mukhyala K, Manning G,
allel functional analysis of BRCA1 RING Zhang Z (2013) Assessment of computational
domain variants. Genetics 200:413–422 methods for predicting the effects of missense
15. Kretz CA, Dai M, Soylemez O, Yee A, Desch mutations in human cancers. BMC Genomics
KC, Siemieniak D, Tomberg K, Kondrashov 14(Suppl 3):S7
Part V
Detection of Matrix Metalloproteases

Chapter 12
Detection of Matrix Metalloproteinases by Zymography

Rajeev B. Tajhya, Rutvik S. Patel, and Christine Beeton
Abstract
Matrix metalloproteinases (MMPs) represent more than 20 zinc-containing endopeptidases that cleave
internal peptide bonds, leading to protein degradation. They play a critical role in many physiological cell
functions, including tissue remodeling, embryogenesis, and angiogenesis. They are also involved in the
pathogenesis of a vast array of diseases, including but not limited to systemic inflammation, various can-
cers, and cardiovascular, neurological, and autoimmune diseases. Here, we describe gel zymography to
detect MMPs in cell and tissue samples and in cell culture supernatants.
Key words Zymography, Zymogram, Protease, Detection, Semiquantitative
1 Introduction
Matrix metalloproteinases (MMPs) are zinc-containing endopepti-

dases responsible for extracellular matrix degradation via cleavage
of internal peptide bonds of specific proteins. More than 20 MMPs
have been identified. MMPs can be either membrane-bound or
secreted; they can be segregated into six groups based on their
substrate specificity and structure: collagenases, gelatinases,
stromelysins, matrilysins, membrane type, and others.
MMPs are crucial to development and tissue repair and remod-
eling but are also important players in many disease processes.
They have been suggested as either diagnosis tools or therapeutic
targets in a broad range of conditions such as multiple sclerosis [1,
2], periodontal disease [3, 4], cancer [5], rheumatoid arthritis and
osteoarthritis [6, 7], intestinal disease [8, 9], or cardiovascular dis-
eases [10]. The accurate and reproducible detection of different
MMPs is therefore critical for both basic and translational research.
A number of techniques have been developed to detect and
identify MMPs in various samples. Enzyme linked immunosorbent
assays (ELISA) and Western blots require the use of antibodies
targeted to the MMP of interest. Here, we describe the procedure
for gel zymography, which was first described in 1980 by Heussen
231
232 Rajeev B. Tajhya et al.
and Dowdle [11], and continues to be extensively used to detect

MMPs in many cell types and tissues and in most bodily fluids
[2–4, 6].
2 Materials
Prepare all solutions using analytical grade reagents and ultrapure

water. All solutions should be stored at room temperature in closed
bottles unless otherwise specified and must be discarded after use
following all waste disposal regulations in place.
2.1 Sample 1. Phosphate-buffered saline (PBS), pH 7.2: Add 800 mL water

Preparation to a 1 L graduated cylinder. Weigh 8 g NaCl, 0.2 g KCl, 1.44
g Na2HPO4 (dibasic anhydrous) (see Note 1), and 0.24 g
KH2PO4 (monobasic anhydrous) and add to the graduated
cylinder. Mix and adjust pH with HCl. Add water to 1 L (see
Note 2).
2. Protease inhibitor cocktail (100×): Add 16 mL water to a 20
mL graduated cylinder. Weigh 20 mg aprotinin, 4 mg leu-
peptin, 961 mg benzamidine and add to the graduated cylin-
der. Add water to 20 mL and mix. Aliquot into microcentrifuge
tubes and store at −20 °C.
3. NP-40 lysis buffer, pH 7.5 (see Note 3): Add 80 mL water to
a 100 mL graduated cylinder. Weigh 394 mg Tris–HCl and
584 mg NaCl, and add to the graduated cylinder. Add 1 mL
Nonidet P-40 (NP-40) to the graduated cylinder and mix.
Add water to 100 mL. Store at 4 °C. Immediately before use,
add 10 μL of protease inhibitor cocktail (100×) (Subheading
2.1, item 2) to 1 mL of NP-40 lysis buffer.
4. Triton X-100 lysis buffer, pH 8.5 (see Note 3): Add 80 mL
water to a 100 mL graduated cylinder. Weigh 315 mg Tris–
HCl, 730 mg NaCl, 1 mL Triton X-100 (see Note 4) and add
to the graduated cylinder. Add water to 100 mL and mix. Store
at 4 °C. Immediately before use, add 10 μL of protease inhibi-
tor cocktail (100×) to 1 mL of Triton X-100 lysis buffer.
2.2 Zymogram Gel 1. Separating gel buffer: Add 80 mL water to a 100 mL glass
cylinder. Weigh 4.63 g ammediol∙HCl and 0.02 g sodium
azide and transfer to the cylinder. Mix and adjust pH to 8.96
with HCl. Add water to 100 mL. Store at 4 °C for up to a
month.
2. Casein substrate solution (10×) (see Note 5): Add 100 mL
0.1 N NaOH to a glass beaker. Weigh 800 mg casein and
transfer to the beaker. Heat the solution to 37 °C with occa-
sional vortexing until full dissolution (see Note 6). Store at
−20 °C for up to 6 months.
Zymography of Matrix Metalloproteases 233
3. Gelatin substrate solution (10×) (see Note 5): Add 100 mL

water to a glass beaker. Weigh 800 mg porcine skin gelatin.
Heat in the microwave until the solution just boils (see Note
7). Swirl thoroughly to ensure homogeneous distribution (see
Note 8). Store at 4 °C for 1 week or at −20 °C for up to 6
months.
4. Sucrose solution: Add 50 mL water to a 100 mL glass cylinder.
Weigh 50 g sucrose and 0.02 g sodium azide and transfer to
the cylinder. Add 30 μL toluene and make up to 100 mL with
water. Store at 4 °C for up to a month.
5. Separating gel (see Note 9): Mix the components in the appro-
priate proportions (Table 1) for the choice of gel (see Note
10).
6. Stacking gel buffer: Add 80 mL water to a 100 mL glass cylin-
der. Weigh 3.51 g ammediol∙HCl and 0.02 g sodium azide
and transfer to the cylinder. Mix and adjust pH to 8.37 with
HCl. Add water to 100 mL. Store at 4 °C for up to a month.
7. Stacking gel (see Note 9): Mix 142 μL acrylamide/bisacryl-
amide solution (37.5:1), 285 μL stacking buffer, 285 μL
sucrose solution, 428 μL water, 14 μL 10% (weight/volume)
ammonium persulfate, and 3 μL TEMED (N,N,N′,N′-
tetramethylethylenediamine) (see Note 11).
8. 0.5 M Tris–HCl, pH 6.8: Add 400 mL water to a 1 L glass
cylinder (see Note 12). Weigh 39.39 g Tris and transfer to the
cylinder. Mix and adjust pH with HCl. Add water to 500 mL.
9. Sample buffer (2×): In a 10 mL glass cylinder, add 2.5 mL of
0.5 M Tris–HCl pH 6.8, 2.0 mL of glycerol, 10% (weight/
Table 1
Recipes for preparing zymogram separating gels with different percentages of acrylamide
Final acrylamide concentration in gel
Solution 6% 7.5% 10% 12%

Acrylamide/bisacrylamide solution 37.5:1 0.8 mL 1.0 mL 1.32 mL 1.61 mL
Separating gel buffer 1.0 mL 1.0 mL 1.0 mL 1.0 mL
10× casein or gelatin substrate solution 0.4 mL 0.4 mL 0.4 mL 0.4 mL
Sucrose solution 0.86 mL 0.86 mL 0.86 mL 0.86 mL
Water 0.94 mL 0.74 mL 0.42 mL 0.13 mL
10% (w/v) ammonium persulfate 14 μL 14 μL 14 μL 14 μL
TEMEDa 1.5 μL 1.5 μL 1.5 μL 1.5 μL
The volumes are sufficient for a single mini-gel (9 cm × 6 cm × 0.75 cm)
a
TEMED: N,N,N′,N′-tetramethylethylenediamine
volume) sodium dodecyl sulfate (SDS), and 0.1% (weight/vol-

ume) bromophenol blue. Add water to 10 mL.
10. Running buffer (10×): To prepare 1 L of running buffer, add
100 mL water to a 1 L graduated cylinder. Weigh 29 g of Tris
Base, 144 g of glycine, and 10 g of SDS. Add all three to the
graduated cylinder and add water to 1 L. Mix with a magnetic
stir-bar until fully dissolved.
2.3 Protease 1. Renaturing buffer (see Note 13): Add 50 mL of water to a 100
Detection mL graduated cylinder. Add 2.5 mL of Triton X-100 to the
cylinder (see Note 4). Add water to 100 mL and mix thor-
oughly. Do not store for more than a day.
2. Developing buffer (10×): To prepare 1 L of developing buffer,
add 100 mL water to a 1 L graduated cylinder. Weigh 12.1 g
Tris Base, 63 g Tris–HCl, 117 g NaCl, 7.4 g CaCl2 and add all
four to the cylinder. Add 2 mL Brij 35 to the cylinder (see Note
14). Add water to 1 L and mix thoroughly.
3. Coomassie Blue staining solution (see Note 15): Add 800 mL
water to a graduated cylinder. Add 50 mL methanol and 100
mL acetic acid (see Note 16). Weigh 5 g Coomassie Blue
R-250 and add to the graduated cylinder. Add water to 1 L.
4. Coomassie Blue destaining solution (see Note 15): Add 800
mL water to a 1 L beaker. Add 100 mL methanol and 50 mL
acetic acid (see Note 16). Add water to 1 L.
3 Methods
Perform all procedures at room temperature unless otherwise

specified and wear personal protective equipment.
3.1 Sample 1. If the cells to be tested are adherent, plate in complete growth
Preparation media and allow time to adhere (see Note 17).
3.1.1 Preparation of Cell 2. Wash the cells three times with sterile PBS or serum-free cul-
Culture Supernatants ture media.
3. Incubate the cells at 37 °C with serum-free culture media (see
Note 18) for the optimal duration dependent on the cells (see
Note 19).
4. Collect the culture supernatants.
5. Proceed to measuring protein concentration (Subheading 3.2)
or freeze at −80 °C until use (see Note 20).
3.1.2 Preparation of Cell 1. If the cells to be tested are adherent, plate in complete growth
Lysates media and allow time to adhere (see Note 21).
2. Wash the cells three times with sterile PBS or culture media.
3. Incubate the cells at 37 °C with culture media for the optimal

duration dependent on the cells (see Note 19).
4. Place a bottle of PBS and a bottle of NP-40 lysis buffer on ice
for a minimum of 20 min to prepare ice-cold buffers.
5. Wash the cells twice with ice-cold PBS.
6. Add cold NP-40 lysis buffer at a volume of 2 mL per 150 mm
dish.
7. If the cells to be tested are adherent, scrape them using a cell
lifter.
8. Collect the cell lysate and incubate on ice for 15 min.
9. Centrifuge the cell lysate 16,000 × g for 20 min at 4 °C in a
microcentrifuge.
10. Collect the supernatant.
3.1.3 Preparation 1. Place a bottle of NP-40 lysis buffer on ice for a minimum of
of Tissue Extracts— 20 min to prepare ice-cold buffers.
Technique 1: NP-40 Lysis 2. Collect the tissue and process immediately.
Buffer
3. Cut approximately 50 mg of tissue into small pieces and place
in a 1.5 mL microcentrifuge tube.
4. Add 0.5 mL of cold NP-40 lysis buffer (see Note 22).
5. Homogenize the tissue with a pestle over ice.
6. Centrifuge the homogenate at 16,000 × g for 20 min at 4 °C
in a microcentrifuge.
3.1.4 Preparation 1. Collect the tissue and immediately freeze with dry ice.
of Tissue Extracts— 2. Pulverize the frozen tissue with a micro-dismembrator.
Technique 2: Pulverization
3. Weigh 50 mg of pulverized tissue and resuspend in 0.15 mL
Triton X-100 lysis buffer under gentle rotation at 4 °C.
4. Centrifuge the homogenate at 16,000 × g for 20 min at 4 °C
in a microcentrifuge.
3.2 Measuring 1. Remove the bovine serum albumin (BSA) standard reagent
Protein Concentration (see Note 23) from 4 °C storage and let it warm to room tem-
perature (see Note 24). Invert a few times before use.
2. Prepare serial dilutions of 2 mg/mL BSA to create a calibra-

tion curve (linear) in the range of 200–1,000 μg/mL.
3. Pipet an aliquot of each standard or sample into separate clean
cuvettes filled with 500 μL dH2O and 500 μL Bradford reagent
and vortex. Prepare a blank sample in a clean cuvette by mixing
500 μL dH2O and 500 μL Bradford reagent.
4. Incubate at room temperature (see Note 24) for at least 5 min.
Do not incubate longer than 1 h (see Note 25).
5. Warm up the spectrophotometer for 10 min. Zero the instru-
ment with the blank sample. Measure the absorbance of the
standards and the samples at 280 nm.
6. Create a standard curve by plotting the absorbance at 280 nm
on the y-axis and concentration of the serial standard dilutions
on the x-axis. Determine the slope of the standard curve.
7. Using the slope and absorbance at 280 nm calculate the con-
centration of each sample (see Note 26).
3.3 Gel Preparation 1. Assemble the electrophoresis plates (see Note 27).
2. Prepare the separating gel and immediately pour (see Note 28)
between the electrophoresis plates up to 2 cm from the top.
3. Overlay with water-saturated n-butanol and allow the gel to
polymerize for approximately 30 min at room temperature (see
Note 29).
4. Prepare the stacking gel (see Note 28).
5. Decant the n-butanol from the gel and gently rinse with water
(see Note 30).
6. Pour the stacking gel on the top of the separating gel.
7. Immediately insert the well comb (see Note 31).
8. Allow the stacking gel to polymerize for approximately 20 min
at room temperature.
3.4 Loading 1. Remove the well comb from the gel and rinse gently with 1×
and Electrophoresis running buffer.
2. Install the zymography gel in the electrophoresis setup.
3. Fill the inside chamber with 1× running buffer and the outside
chamber with 500 mL 1× running buffer (see Note 32).
4. Place 10 μL of 2× sample buffer on a Parafilm membrane and
add 10 μL of sample. Pipet a few times to mix and load the mix-
ture into each well using gel loading pipet tips (see Note 33).
5. Load a protein molecular weight standard in at least one well
per gel.
6. Load appropriate recombinant MMP or a control known to
contain the MMP of interest to identify the sample MMPs
(optional) (see Note 34).
7. Run the gel at 125 V for 90 min or until the loading dye
reaches the bottom of the gel.
3.5 Gel Renaturation 1. Place 100 mL of 1× renaturing buffer in a container large

enough to fit the gel.
2. Open the gel plates (or cassette in the case of a commercial
precast gel).
3. Make a diagonal cut in a corner of the gel to mark its
orientation.
4. Remove the gel from the plate or cassette (see Note 35).
5. Place the gel in the container containing the renaturing buffer.
6. Incubate for 30 min at room temperature with gentle agitation
on a bench rocker.
3.6 Gel Development 1. Remove the renaturing buffer by gently pouring it into a waste
container (see Note 36).
2. Add 100 mL of 1× developing buffer to the container with the
gel.
on a bench rocker.
4. Remove the developing buffer by gently pouring it into a waste
container.
5. Add 100 mL of 1× developing buffer to the container with the
gel. Close the lid tightly.
6. Incubate overnight (16–18 h) at 37 °C (see Note 37).
3.7 Gel Staining 1. Remove the developing buffer by gently pouring it into a waste
and Destaining container.
2. Add 100 mL of water to the container containing the gel.
on a bench rocker.
4. Remove the water by gently pouring it into a waste container.
5. Repeat steps 2–4 twice.
6. Scan the gel to record the exact position of the protein stan-
dard bands as they will become difficult to see after gel staining
and destaining.
3.7.1 Staining 1. Add 20 mL of Coomassie blue staining solution (see Note 15)
with Coomassie Blue to the container containing the gel.
R-250 2. Incubate for 1 h at room temperature with gentle agitation on
a bench rocker (see Note 38).
3. Remove the staining buffer by gently pouring it into a waste
container.
4. Add 100 mL of 1× destaining solution to the container con-

taining the gel.
5. Incubate at room temperature with gentle agitation on a bench
rocker until areas of proteolytic activity are clearly visible (see
Note 39).
3.7.2 Staining 1. Add 20 mL of ready-to-use commercial buffer with colloidal

with Colloidal Coomassie Coomassie Blue to the container with the gel.
Blue (G-250) 2. Incubate for 1 h at room temperature with gentle agitation on
3. Remove the staining buffer by gently pouring it into a waste
container.
4. Add 100 mL of water to the container with the gel.
5. Incubate for 1 h at room temperature with gentle agitation on
3.8 Data Acquisition 1. Remove the gel from the water, place in a plastic sheet protec-
and Analysis tor, and gently remove air bubbles.
2. Scan the gel with a digital scanner (see Note 41).
3. The integrative intensity of the MMP bands can be quantified
with ImageJ (freely available from the National Institutes of
Health) or similar software.
4 Notes
1. The 1.44 g of Na2HPO4 (dibasic anhydrous) can be replaced

by 1.81 g Na2HPO4∙2H2O (dibasic dihydrate) or by 2.72 g
Na2HPO4∙7H2O (dibasic septahydrate) in the preparation of
PBS.
2. If PBS is to be kept more than a day, it should be sterilized to
prevent bacterial growth by either autoclaving or sterile filtra-
tion (0.22 μm filter pores).
3. Choice of lysis buffer for protein extraction will depend on the
sample and also on the target MMP. For example, MMP-23
can only be efficiently extracted using the nonionic detergent
Triton X-100 [12].
4. Triton X-100 is very viscous. When pipetting Triton X-100,
only place the very tip of the pipet into the liquid to avoid
coating the outside of the pipet with the detergent. Apply suc-
tion very slowly to allow the liquid to reach the correct volume
in the pipette. If Triton X-100 is visible on the outside of the
pipet, remove it with a clean tissue. Empty the pipet very slowly
into a receiving container containing water and wait until the
pipet has completely emptied (can take a few minutes).
5. The choice of substrate in the gel will depend on the MMP of

interest. For example, MMP-2 and MMP-9 can be detected
using gelatin gels, as can MMP-1, MMP-8, and MMP-13
although gelatin is not their preferred substrate. MMP-1 and
MMP-13 are best detected on collagen gels while casein is the
preferred substrate for MMP-11 and also allows for the detec-
tion of MMP-1, MMP-3, MMP-7, MMP-12, and MMP-13
[13].
6. Full dissolution of casein is very important to ensure homoge-
neity of the separating gel during zymography. Lack of homo-
geneity will prevent data interpretation.
7. Do not allow the solution to boil over as that will change its
concentration.
8. Full dissolution of the gelatin is very important to ensure
homogeneity of the separating gel during zymography. Lack of
homogeneity will prevent data interpretation.
9. Precast gels can be purchased from various vendors. These gels
are provided in sealed pouches and have strict expiration dates;
expired gels will not provide expected results as the substrate
will have started degrading in a heterogeneous manner.
10. The correct acrylamide percentage to use will depend on the
molecular weight of the MMP of interest. Proteins with high
molecular masses resolve better on low percentage gels,
whereas proteins with low molecular masses resolve better on
high percentage gels. The most commonly used gels for
zymography are 10% gels as they resolve proteins between 20
and 100 kDa.
11. The volumes given here are sufficient for a single mini-gel
(9 cm × 6 cm × 0.75 cm).
12. Placing water at the bottom of the cylinder will help to rapidly
dissolve the Tris. In addition, warming the water to 37 °C will
hasten the dissolution of Tris. However, the solution should be
brought to room temperature before adjusting the pH.
13. The renaturing buffer should be prepared fresh on the day it
will be used.
14. Brij 35 is very viscous. When pipetting Brij 35, only place the
very tip of the pipet into the liquid to avoid coating the outside
of the pipet with the detergent. Apply suction very slowly to
allow the liquid to reach the correct volume in the pipette. If
Brij 35 is visible on the outside of the pipet, remove it with a
clean tissue. Empty the pipet very slowly into the receiving
container containing water and wait until the pipet has com-
pletely emptied (can take a few minutes).
15. Coomassie Blue R-250 can be replaced by a ready-to-use com-
mercial solution containing colloidal Coomassie blue G-250.
An advantage of the colloidal Coomassie blue is that it only

requires water (no methanol or acetic acid) for destaining.
16. Always add concentrated acid to a large volume of water; never
add water to concentrated acid as this would result in a violent
exothermic reaction.
17. The number of cells to use for preparing cell culture superna-
tants for zymography will vary depending on the type of cells
and must be tested prior to assaying a new cell system. For
example, MMP-2 was detectable in the culture supernatants of
human fibroblast-like synoviocytes isolated from patients with
rheumatoid arthritis when cells were plated at a density of
50,000 cells per well with 0.3 mL culture medium per well in
a 24-well plate [14].
18. Serum contains MMPs (Fig. 1); it is therefore crucial to remove
all serum for the cell culture and then maintain the cells in
serum-free media during MMP production. Indeed, any serum
present in the culture may obscure effects due to differential
treatment of the cells (Fig. 2) [3, 6, 8, 14–16].
19. The duration of incubation will differ depending on the cell
type and must be tested prior to assaying a new cell system. For
example, when assessing MMP-2 levels in the culture superna-
tants of human fibroblast-like synoviocytes isolated from
patients with rheumatoid arthritis or from rats in an animal
model of rheumatoid arthritis, a 24 h incubation gave optimal
results [2, 3, 14, 16]. In contrast, rabbit corneal fibroblasts
were cultured for 72 h before collecting the supernatant [6, 17]
whereas a 1 h incubation was sufficient to detect pro-MMP-9 in
the culture supernatants of human neutrophils [2, 18].
20. It is crucial that MMPs are not denatured to remain functional.
If the samples are to be frozen, they should be frozen immedi-
ately upon collection. In addition, every effort should be made
to avoid repeated cycles of freezing and thawing of the
samples.
Fig. 1 Gelatin zymography gel loaded with different dilutions of horse serum (HS),
goat serum (GS), and human serum (HuS) (1%, 5%, 10%). Stronger bands of
MMP-2 (~65 kDa) and MMP-9 (~85 kDa) are visible with increasing concentra-
tions of the different sera
Fig. 2 Gelatin zymography gel comparing MMP-2 and MMP-9 production in the
supernatant of untreated fibroblast-like synoviocytes isolated from a patient with
rheumatoid arthritis (RA-FLS; control) and paxilline-treated RA-FLS incubated in
the presence of 0%, 1%, 5%, or 10% of fetal bovine serum. Paxilline-treated
RA-FLS exhibit a decrease in MMP-2 production when incubated without serum.
This effect is masked by incubation with 1%, 5%, 10% serum
21. The number of cells to use for preparing cell culture superna-
tants for zymography will vary depending on the type of cells
and must be tested prior to assaying a new cell system. For
example, MMP-2 was detectable in the culture supernatants of
human fibroblast-like synoviocytes isolated from patients with
rheumatoid arthritis when cells were plated at a density of
50,000 cells per well with 0.3 mL culture medium per well in
a 24-well plate [14].
22. The volume of lysis buffer to add will depend on the type of
tissue to be homogenized as protein extractability varies with
specimens.
23. While BSA is commonly used, any known protein standard
can be used as a control.
24. Cold solutions can condense atmospheric moisture on the
outside of the cuvette and scatter light giving erroneous
readings.
25. Absorbance of protein-dye conjugate increases over time and
affects the accuracy of protein concentration measurements.
26. If cell supernatants or tissue extracts contain low levels of gela-
tinases, a phase extraction with Triton X-114 is recommended
[19].
27. The electrophoresis plates must be completely clean and dry at
this time or the gel could incompletely polymerize or polym-
erize in a heterogeneous manner.
28. Once the polymerization agent (TEMED) is added to the gel,
immediately mix and pour as the gel will start polymerizing
rapidly. If the gel polymerizes too fast, reduce the ambient
temperature as this will delay polymerization.
29. Ensure that the electrophoresis system is perfectly horizontal
during gel pouring and polymerization.
30. It is crucial to remove all n-butanol to ensure full contact

between the separating and stacking gels.
31. Place the comb carefully so as to not trap air bubbles.
32. Ensure that there is no leaking of the inside chamber buffer. If
the inside chamber buffer leaks before sample loading, empty
the gel box and unlock the gel. Wet the chamber gasket with
running buffer and install the gel.
33. Volumes given here are for 12-well gels. These volumes can
and should be increased when loading a gel with fewer wells.
34. Using recombinant MMP as a control can be very useful for
semiquantification of MMPs in the samples. In addition,
although MMPs are highly conserved between species in terms
of molecular weight and protease activity, a positive control
containing the MMP of interest in the species of choice would
be beneficial.
35. Removing the gel from the plate or cassette can be difficult as
the gel is very fragile and easy to damage. It is safer to slide the
gel off one side of the cassette rather than lifting it.
36. When removing the buffer after incubating the gel, take pre-
cautions not to damage the gel. Wearing gloves, use two fin-
gers to gently hold the gel to the bottom of the container
while pouring the liquid out from one of the corners of the
container.
37. We recommend an overnight incubation in developing buffer.
If the bands are too strong the developing time can be reduced
to 4 h. In contrast, if the bands are very faint, gels can be incu-
bated in the developing buffer for up to 48 h.
38. The gel should be stained in Coomassie Blue staining solution
for 1 h or until the gel is uniformly dark blue.
39. If no zones of digestion are evident, use a positive control
(e.g., recombinant MMP-2 or MMP-9 for gelatin gels and
MMP-1 for casein gels). If these controls are visible, the test
samples may be devoid of proteases specific to the chosen sub-
strate or the MMP concentration may be lower than expected,
in which case increase the incubation time in developing buffer
or concentrate the test samples. If the positive controls are not
visible, check that (1) the samples were not boiled prior to gel
loading as this denatures the proteins, (2) the sample loading
buffer does not contain anything that inhibits the proteases
(e.g., EDTA), and (3) the renaturing buffer contains sufficient
amounts of Triton X-100.
40. Destaining the gel stained with a ready-to-use commercial
solution containing colloidal Coomassie Blue for an hour may
not be optimal in all situations. Further destaining can be
achieved with a second incubation of 1 h in fresh water. The
gel can be scanned multiple times during the destaining pro-

cess to catch the best contrast in the bands.
41. If the background contains blotches or streaks, this could indi-
cate an uneven distribution of the substrate protein in the gel.
Ensure that the gelatin of casein solution used to prepare the
separating gel contains no undissolved materials.
Acknowledgment
This work was supported in part by funding from Baylor College

of Medicine and National Institutes of Health grants AR059838
and NS073712 to C.B.
References
1. Baranger K, Rivera S, Liechti FD, Grandgirard 8. Chang J, Wehner S, Schäfer N, Sioutis M,

D, Bigas J, Seco J, Tarrago T, Leib SL, Bortscher S, Hirner A, Kalff JC, Bauer AJ,
Khrestchatisky M (2014) Endogenous and Overhaus M (2012) Iatrogenic extracellular
synthetic MMP inhibitors in CNS physiopa- matrix disruption as a local trigger for postop-
thology. Prog Brain Res 214:313–351 erative ileus. J Surg Res 178:632–639
2. Fainardi E, Castellazzi M, Bellini T, 9. Medina C, Radomski MW (2006) Role of
Manfrinato MC, Baldi E, Casetta I, Paolino matrix metalloproteinases in intestinal inflam-
E, Granieri E, Dallocchio F (2006) mation. J Pharmacol Exp Ther 318:933–938
Cerebrospinal fluid and serum levels and 10. Hopps E, Caimi G (2015) Matrix metallopro-
intrathecal production of active matrix metal- teases as a pharmacological target in cardiovas-
loproteinase-9 (MMP-9) as markers of disease cular diseases. Eur Rev Med Pharmacol Sci
activity in patients with multiple sclerosis. 19:2583–2589
Mult Scler 12:294–301 11. Heussen C, Dowdle EB (1980) Electrophoretic
3. Baeza M, Garrido M, Hernández-Ríos P, analysis of plasminogen activators in polyacryl-
Dezerega A, García-Sesnich J, Strauss F, amide gels containing sodium dodecyl sulfate
Aitken JP, Lesaffre E, Vanbelle S, Gamonal J, and copolymerized substrates. Anal Biochem
Brignardello-Petersen R, Tervahartiala T, 102:196–202
Sorsa T, Hernández M (2015) Diagnostic 12. Pei D, Kang T, Qi H (2000) Cysteine array
accuracy for apical and chronic periodontitis matrix metalloproteinase (CA-MMP)/MMP-
biomarkers in gingival crevicular fluid: an 23 is a type II transmembrane matrix metallo-
exploratory study. J Clin Periodontol proteinase regulated by a single cleavage for
43:34–45 both secretion and activation. J Biol Chem
4. Thomadaki K, Bosch J, Oppenheim F, 275(43):33988–33997
Helmerhorst E (2013) The diagnostic poten- 13. Snoek-van Beurden PA, Von den Hoff JW
tial of salivary protease activities in periodontal (2005) Zymographic techniques for the analy-
health and disease. Oral Dis 19:781–788 sis of matrix metalloproteinases and their
5. Cathcart J, Pulkoski-Gross A, Cao J (2015) inhibitors. Biotechniques 37:73–83
Targeting matrix metalloproteinases in cancer: 14. Hu X, Laragione T, Sun L, Koshy S, Jones KR,
bringing new life to old ideas. Genes Dis Ismailov II, Yotnda P, Horrigan FT, Gulko PS,
2:26–34 Beeton C (2012) KCa1.1 potassium channels
6. Cuéllar VG, Cuéllar JM, Kirsch T, Strauss EJ regulate key pro-inflammatory and invasive
(2015) Correlation of synovial fluid biomark- properties of fibroblast-like synoviocytes in
ers with cartilage pathology and associated out- rheumatoid arthritis. J Biol Chem 287:
comes in knee arthroscopy. Arthroscopy 4014–4022
32:475–485 15. Hu X, Beeton C (2010) Detection of func-
7. Murphy G, Nagase H (2008) Reappraising tional matrix metalloproteinases by zymogra-
metalloproteinases in rheumatoid arthritis and phy. J Vis Exp 45:pii 2445
osteoarthritis: destruction or repair? Nat Clin 16. Tanner MR, Hu X, Huq R, Tajhya RB, Sun L,
Pract Rheumatol 4:128–135 Khan FS, Laragione T, Horrigan FT, Gulko
PS, Beeton C (2015) KCa1.1 inhibition atten- 18. Puente J, Jaque M, Carrasco C, Cruz C,
uates fibroblast-like synoviocyte invasiveness Valenzuela M, Wolf M, Mosnaim A (2008)
and ameliorates rat models of rheumatoid Triptan drugs, natural killer cell cytotoxicity,
arthritis. Arthritis Rheumatol 67:96–106 and neutrophils pro-matrix metalloproteinase-
17. Pan H, Chen J, Xu J, Chen M, Ma R (2009) 9 secretion. Headache 48:1482–1489
Antifibrotic effect by activation of peroxisome 19. Toth M, Sohail A, Fridman R (2012) Assessment
proliferator-activated receptor-gamma in cor- of gelatinases (MMP-2 and MMP-9) by gelatin
neal fibroblasts. Mol Vis 15:2279–2286 zymography. Methods Mol Biol 878:121–135
Chapter 13
Imaging Matrix Metalloproteases in Spontaneous Colon

Tumors: Validation by Correlation with Histopathology
Harvey Hensley, Harry S. Cooper, Wen-Chi L. Chang,
and Margie L. Clapper
Abstract
The use of fluorescent probes in conjunction with white-light colonoscopy is a promising strategy for
improving the detection of precancerous colorectal lesions, in particular flat (sessile) lesions that do not
protrude into the lumen of the colon. We describe a method for determining the sensitivity and specificity
of an enzymatically activated near-infrared probe (MMPSense680) for the detection of colon lesions in a
mouse model (APC+/Min-FCCC) of spontaneous colorectal cancer. Fluorescence intensity correlates directly
with the activity of matrix metalloproteinases (MMPs). Overexpression of MMPs is an early event in the
development of colorectal lesions. Although the probe employed serves as a reporter of the activity of
MMPs, our method can be applied to any fluorescent probe that targets an early molecular event in the
development of colorectal tumors.
Key words Colorectal cancer, Mice, Min mice, Imaging, Fluorescence imaging, Matrix metallopro-
teinases, Early cancer detection, Colonoscopy, Activated imaging probes, Near-infrared imaging
1 Introduction
Colonoscopy, the standard procedure for the detection of lesions

in the colon, employs broad-spectrum visible light (“white light”)
for illumination, with lesions identified based on morphological
features. Although effective, it is estimated that this method fails to
detect up to 20% of lesions [1, 2]. Flat, sessile lesions are particu-
larly difficult to detect because they do not protrude from the
colon wall. Efforts are underway to develop optical methods for
detecting these lesions more reliably, including the use of fluores-
cent probes [3]. Fluorescent probes produce optical contrast that
reveals regions of biological activity that are not distinguishable
from adjacent normal tissue on gross examination. Furthermore,
fluorescence imaging can be employed simultaneously with white
light imaging, using a suitably modified endoscope. In a preclinical
setting and following intravenous injection, the distribution of
245
246 Harvey Hensley et al.
fluorescent probes in experimental animals can be determined by

in vivo epifluorescence, fluorescence endoscopy, fluorescence
molecular tomography, or ex vivo examination at necropsy [4].
A unique animal model (Apc+/Min-FCCC mouse) has been estab-
lished by this group that spontaneously develops multiple colorec-
tal adenomas (3.7 ± 0.3, mean ± SEM) [5]. Due to the small size
of the lesions and their location deep within the abdomen, their
detection is not feasible using the external, whole body imaging
protocols that are available currently. While fluorescence endos-
copy is possible, only the distal portion of the colon can be inter-
rogated using currently available Hopkins telescopes. In addition,
the presence of a large lesion in the colon often impedes full inser-
tion of the endoscope, preventing the detection of more caudal
lesions. In contrast, a number of instruments are commercially
available to interrogate the entire colon and acquire accurate
ex vivo measurements. Such experiments permit detailed, high
resolution imaging and provide valuable proof-of-principle infor-
mation. Crucial to the evaluation of any fluorescent imaging probe
is validation of the resulting data. This process entails correlating
regions of probe retention, as measured by fluorescence detection,
with histopathologically confirmed biological features.
Numerous fluorescent probe-based strategies have been pro-
posed and investigated to improve the early detection of colon
lesions [6]. Among these, matrix metalloproteinases (MMPs),
calcium- dependent zinc-containing endopeptidases, represent
attractive targets for fluorescence imaging. MMPs play multiple
important roles in the transformation of the normal human
colorectal mucosa to a neoplastic state, in both the initial progres-
sion of adenomatous polyps to invasive colorectal cancers, and
their eventual transition to metastatic disease [7, 8]. Of particular
interest, overexpression of MMP-7 is found in 85% of colorectal
adenocarcinomas [9], and occurs as a result of mutation of the
adenomatous polyposis coli (APC) gene; the “gatekeeper” of spo-
radic colorectal carcinogenesis [10]. Among the probes available
for detecting MMPs is MMPSense680 (PerkinElmer, Hopkinton,
MA), a bioactivatable probe that consists of a number of fluoro-
phores attached in close proximity to an MMP-cleavable backbone.
In the uncleaved state, the fluorophores quench each other, ren-
dering the probe optically silent. In the presence of specific MMPs,
the backbone is cleaved releasing the fluorophores.
Rigorous validation of results obtained using probes can be
challenging. This is accomplished by: (1) careful identification of
regions of fluorescence where the intensity is above that of the
background mucosa; (2) correlating areas positive for fluorescent
signal with the presence of gross lesions identified at necropsy; and
most importantly, (3) histopathological confirmation of areas that
are both neoplastic and positive for fluorescent signal. Pathological
analyses are performed on the entire colon, following tissue
Imaging MMPs in Colon Tumors 247
xation, processing, embedding, and sectioning, while maintain-

fi
ing the location of each tissue sample as it relates to the colon
structure.
2 Materials
2.1 Animal 1. Apc+/Min-FCCC mice, or other animal model that develops

Treatment, Imaging, colorectal tumors.
and Tissue Processing 2. Fluorescence imaging system (IVIS Spectrum or a similar sys-
tem with an appropriate illumination system, filters, and sensi-
tive, thermo-electrically cooled CCD camera).
3. Image processing system (computer and software—Matlab
with image processing toolbox).
4. Plastic sheet with printed ruler (made by copying a plastic ruler
onto transparency sheets using a standard photo-copy
machine).
5. MMP probe (MMP680) (PerkinELmer, Hopkinton, MA).
6. PBS: Isotonic phosphate buffered saline.
7. Insulin syringe with 30-gauge needle (Kendall Monoject
8881600700) (Tyco Healthcare Group).
8. 10% normal buffered formalin.
9. Standard supplies, reagents, and equipment for paraffin embed-
ding and tissue sectioning.
10. Hematoxylin and eosin stain.
11. Dissecting tools (forceps, Metzenbaum scissors, Mayo scissors,
scalpels).
12. 70% Ethanol in a spray bottle.
13. Histowrap.
14. 50 ml conical centrifuge tubes or other container for fixing
individual colons.
15. Tissue cassettes.
16. Microscope slides.
2.2 Immuno 1. Rat anti-human MMP-7 (clone 338; Vanderbilt University).

histochemical 2. Goat anti-rat IgG (Vector Laboratories).
Validation of MMP 7
3. Rabbit anti-MMP-9 (catalog no. 38898; Abcam).
and MMP 9
4. Goat anti-rabbit IgG (EnVision Kit; Dako).
5. Cell conditioning 1 (CC1) (Ventana Medical Systems).
6. Ventana XT automated immunostainer (Ventana Medical
Systems).
2.3 Probes Although this review focuses on our work using MMPsense680 as
probe, the procedures described can be used to study any fluores-
cent probe in the mouse colon, or any similar organ in small
animals.
We have imaged a variety of probes with fluorophores in the
red or near-infrared spectrum including MMPSense680,
ProSense680, ProSense750, IntegriSense680, and IntegriSense750
(PerkinElmer, Hopkinton, MA), as well as PSVUE794 (Molecular
Targeted Technologies, Malvern, PA) using the following
methodology.
2.4 IVIS Images are captured on an IVIS Spectrum (PerkinElmer). This sys-
Imaging System tem consists of a highly sensitive, cooled (−80 °C) CCD camera,
and a versatile filter set for imaging fluorophores. The system also
contains a set of LEDs for white light illumination, permitting the
acquisition of both fluorescent and conventional (black and white)
images, and displays the fluorescence intensity as an overlay on the
white light image. Exposure time, binning, fstop, light illumina-
tion level, imaging field of view, focal point, and excitation and
emission filters can be adjusted to optimize image acquisition.
2.5 Image IVIS images are processed with routines written in the Matlab
Processing (Mathworks) programming environment with the Image
Processing Toolkit. A large number of free and commercial image
processing software programs are available. However, customized
routines are necessary to implement a convenient method for pre-
cisely pairing each fluorescent image with a corresponding paraffin
embedded section.
3 Methods
3.1 Injection To maximize the use of the probe and ensure the ethical treatment
of MMP Probe of animals, endoscopy, MRI, CT, or other noninvasive imaging
methods should be employed prior to injection to confirm the
tumor-bearing status of the experimental animals. If at least one
colon lesion has been identified, inject the mouse intravenously
(via either the tail vein or retro-orbital sinus) with an appropriate
amount of the fluorescent imaging agent (~2 nM for MMPSense680)
in 100–150 μl of isotonic PBS (see Note 1). The inoculation time
in vivo will differ depending upon the pharmacokinetic properties
(plasma half-life) of the probe to be imaged. Time course analyses
revealed 66 h post-injection as the optimal time to image
MMPSense680 ex vivo in C57BL/6J Apc+/Min-FCCC mice. At the
predetermined time point, euthanize the mice according to the
American Veterinary Medical Association Guidelines for Euthanasia
(slow inhalation of CO2, followed by confirmation of death).
3.2 Preparation 1. Immediately following euthanasia, place the mouse in a supine

of the Mouse Colon position and dampen its abdominal fur with 70% ethanol.
for Ex Vivo Imaging Using steel dissecting scissors, cut a longitudinal slit through
the ventral skin starting near the anus, moving through the
pelvis, and finishing at the sternum to expose the colon in its
entirety. Care should be taken not to cut through the colon,
any internal organs, or major blood vessels.
2. Remove the small intestine (anterior to the colon) from the
abdominal cavity to reveal the colon as a long tubular struc-
ture, approximately 3 mm in diameter, that extends from the
anus to the mouse’s left kidney, and then runs transverse across
the body to connect to the cecum, a large, sac-like structure.
3. Insert Metzenbaum scissors (large), closed, into the pelvis
area, and open scissors to crack through the pelvic bone.
Remove the Metzenbaum scissors.
4. Gently lift the colon with forceps, and with the smaller Mayo
scissors, dissect it free by cutting any attached mesenteries or
other connective tissues that are securing it to other portions
of the colon or other organs within the peritoneum.
5. Make a further cut between the proximal end of the colon and
the cecum, completely separating the two structures.
6. With the Mayo scissors, cut the distal colon as close as possible
to the anus to separate the distal section of the colon from the
body cavity. This will permit the separation of the distal end of
the colon from the base of the tail and surrounding skin.
7. Once the colon has been completely cut free, remove it from
the abdominal cavity, place the entire colon in a weigh boat
containing 1× PBS, and rinse. The colon must be handled with
forceps at all times after its removal from the abdominal
cavity.
8. Place the colon on a paper towel and make a longitudinal inci-
sion along its entire length with fine dissecting (Mayo)
scissors.
9. Return the colon to the weigh boat and vigorously wash it in
1× PBS to remove all fecal matter.
10. Lay the colon flat (luminal side up) on a preprinted plastic
ruler made by copying a plastic ruler onto transparency sheets
using a standard copy machine. Place the distal end of the
colon to the left or right of the origin of the ruler, with the
gradations and numbers visible (see Note 2). Write the mouse
ID on the plastic sheet with a black permanent marker (see
Note 3), at a peripheral location that will not mask the image.
3.3 Ex Vivo Imaging 1. Place the transparent plastic sheet containing the excised colon
in the imaging system on a black plastic mat (see Note 4).
Image settings for our work [4] employing the IVIS Spectrum
system were: excitation wavelength = 640 nm, detection wave-

length = 700 nm, exposure time = 5 s, F-stop = 2, binning = 4
(“small” setting), field of view = 13.6 cm (setting “C”). (see
Note 5). The “small” binning section corresponds to a bin-
ning factor of 4 and producs images with 230,400 pixels (480
× 480), making each pixel (28 μm)2. For higher resolution
images , binning factors of 1 can be used to successfully create
images with pixels of (7 μm2), with appropriate increases in
image acquisition time applied. Adjust the camera binning set-
ting for the white light photographic image (acquired immedi-
ately prior to the acquisition of the fluorescence image in the
IVIS Spectrum) to be the same as that for the fluorescence
image (see Note 6).
2. Place the colon, while attached to the transparent plastic sheet,
in a 50 ml conical centrifuge tube with 10% normal buffered
formalin for fixation at 4 °C overnight.
3.4 Tissue 1. Following fixation overnight on the transparent plastic sheet,

Processing process the tissue for paraffin embedding under a fume hood.
and Scoring of Images 2. Using the white-light image as a guide, cut the entire colon at
3.4.1 Preparation 2 mm increments (see Note 7).
of the Colons 3. Assign each piece a consecutive identifier and place it in a
for Embedding unique tissue cassette after wrapping it in histowrap dampened
and Documentation with formalin.
of the Position of Each 4. Record the precise location of each excised piece relative to the
Tissue Segment ruler markings by drawing a boxed area on a hard copy of the
fluorescence image/white light image overlay (Fig. 1) that
corresponds to the position of each excised piece of colon on
the ruler.
5. Embed all tissues in paraffin, cut 5 μm sections and stain them
with hematoxylin and eosin for histopathological review.
3.4.2 Identification 1. Using the locations marked on the hard copy printout as a
of Tissue Sections guide (see step 4 of Subheading 3.4.1), identify the position of
with Fluorescence Image each 2 mm segment of the colon on both white light and
Features and Image fluorescence/white light image overlays as displayed on a com-
Display puter screen.
2. Mark a rectangular region of interest on the images as dis-
played on the computer with an image processing software
program.
3. Since up to 60 separate regions of interest (ROIs) must be
placed on the images to cover the entire colon, we have found
it convenient to write routines to help do this using the
MATLAB Imaging Processing Toolbox (MathWorks, Inc.,
Natick, MA). 30–60 ROIs (2 mm in width; the number
depending on the length of the particular colon) are placed on
Fig. 1 Images of MMPSense680 fluorescence within the mouse colon, coded

according to intensity with a “hot” colormap (far right). Each image represents the
same colon, with the fluorescence intensity scaled to display a different color
range. In the fluorescence overlay image on the left, a distinct region of strong
signal intensity can be discerned (section 47), relative to local background, while
regions of the distal colon exhibit more intense signal (sections 1–8), but display
no discernable contrast. After adjusting the contrast (see the fluorescence overlay
image at the far right), a discrete lesion becomes apparent in section 5. The gray-
scale white-light image, acquired immediately preceding the acquisition of the
fluorescence images, is shown on the far left. Note the presence of other lesions
(sections 11, 13, 18, and 26–27) visible on all fluorescence overlay images
the image (Fig. 1). The ROI positions are then adjusted indi-
vidually according to the precise position of each segment on
the ruler, which were recorded when the colons were cut into
segments for embedding into paraffin.
4. Display the fluorescence images as overlays on the white light
images, with the white light intensities set to a grayscale colort-
able, while the fluorescence image is set to a different color
lookup table. We employed a “hot colortable” for fluorescence
intensities (Fig. 1), with the most intense signals correspond-
ing to a display color of white. The least intense is set to a mini-
mum indicative of the signal background level in the colon and
presented as dark red, while the intermediate values display as
a range of shades of red to yellow (Fig. 1). Images were
acquired with a 16 bit CCD camera, producing a total of
65,536 intensity levels. In order to take advantage of this large
dynamic range, and given the large range of fluorescence inten-

sities in the colon, display the fluorescence intensities in at least
two different color mappings (Fig. 1), so that both low and
high intensity regions can be visualized. This method is prefer-
able to the alternative of a nonlinear (for example logarithmic)
mapping of fluorescence intensity to color (see Note 8).
5. Estimate lesion size by measuring the number of pixels that
span the lesion on either the white light or fluorescence image,
and converting the region to millimeters using the image
acquisition field of view. Both images are useful for determin-
ing the lesion edges; however, they tend to be more distinct in
the fluorescence images.
3.4.3 Use Use of a subset of the data as a training set is essential to: (1) develop
of a Training Set experience identifying lesions; (2) establish rules and guidelines for
scoring images; and (3) identify features on fluorescence images
that correspond to histologically confirmed adenomas.
1. Set aside 10 of the annotated fluorescence/white light images
to be used as a training set.
2. Identify regions of high fluorescence intensity relative to the
local background, and compare these regions with results from
H&E stained sections as reviewed and annotated by a trained
pathologist (Fig. 2).
3. Conversely, determine which features on the fluorescence
images correspond to the presence of a pathologically con-
firmed lesion.
4. Identify which areas have pathologically confirmed lesions and
determine whether or not there is an identifiable fluorescence
signal on the image in that location.
Fig. 2 Micrographs of colorectal lesions in an Apc+/Min-FCCC mouse. (a) 40× view of a peduncular adenoma that
extends above the surface of the normal mucosa compared with a nonpolypoid lesion (b). (b) 40× view of a
sessile adenoma (blue arrow) with a height less than twice that of the adjacent normal mucosa (yellow arrow)
Fig. 3 Immunohistochemical staining of MMP-7 and -9 in colorectal lesions of Apc+/Min-FCCC mice. (a) 400×
view of an adenoma stained with anti-MMP-7 antibodies. MMP-7 is strongly and diffusely expressed in the
adenomatous epithelium. (b) 400× view of an adenoma stained with anti-MMP-9 antibodies. MMP-9 is
strongly expressed in the polymorphonuclear leukocytes (PMNLs) in the stroma/lamina propria of the adenoma
but not in the epithelial cells
5. After training is complete, at least two independent observers

can conduct an analysis of an independent set of fluorescence
images in a blinded manner to identify regions most likely con-
taining atypical pathology or early lesions.
3.5 MMP-7 1. Employ standard immunohistochemical methods, as per the

and MMP-9 manufacturer’s instructions, to determine the expression of
Immuno MMP-7 and MMP-9 in paraffin embedded sections (Fig. 3).
histochemistry 2. Use rat anti-human MMP-7 antibody clone 338, at a dilution
of 1:400 with sections incubated for 1 h at 37 °C, followed by
goat anti-rat secondary antibody as the secondary antibody.
3. For MMP-9 staining, use rabbit anti MMP-9 at a dilution of
1:200, with sections incubated for 1 h at room temperature. A
goat anti-rabbit secondary antibody is used with this primary
MMP-9 antibody. For best results, sections should be stained
using an automated immunostainer, such as the Ventana XT.
4 Notes
1. The presence of fluorescent substances in the most commonly

used animal chows can lead to false-positive areas of strong
fluorescence. The fecal matter of the mice will generally be
fluorescent if all of these contaminants are not fully removed.
The use of a non-fluorescent chow (Teklad 2019s, Envigo,
East Millstone, NJ) for 1 week prior to imaging can be
advantageous.
2. When laying the colon flat on the plastic ruler, be sure to keep
track of the proximal and distal ends. This can be done by not-
ing the distinct three-dimensional architecture (V-shaped) of
the proximal colonic mucosa, and placing the colon so it is
always oriented in the same direction. Always lay the colon
with the luminal surface facing up, with no folds or twists.
3. Use a plain black colored marker when recording the animal
number on the plastic sheets, as the dyes in some colored
markers may have some fluorescent signal.
4. The black plastic mat is usually supplied by the manufacturer of
the imaging system.
5. The focal distance on the IVIS Spectrum is set to a default of
1.5 cm for in vivo imaging. When making high spatial resolu-
tion images with a low f-number setting in excised organs, this
setting must be decreased to approximately 1.5 mm and
adjusted carefully.
6. Imaging times (from dissection to image acquisition) should
not exceed 5 min to prevent dehydration.
7. The ruler markings photocopied onto the transparent sheet
can be used as a guide for cutting the colon.
8. To increase the dynamic range of the images, an alternative to
displaying multiple images with different contrast settings
would be to employ a nonlinear mapping (e.g., logarithmic) of
intensity to color. In practice, we find it preferable to examine
multiple representations of the same data with a linear map-
ping of intensity to color.
Acknowledgments
This work was supported by grants CA-006927 and CA-124693

from the National Cancer Institute, by Fox Chase Cancer Center
Keystone Initiative in Personalized Risk and Prevention, and by an
appropriation from the Commonwealth of Pennsylvania. The
authors have no conflicts of interest or financial disclosures to
report.
References
1. Cornett D, Barancin C, Roeder B et al (2008) 3. Burggraaf J, Kamerling IM, Gordon PB et al
Findings on optical colonoscopy after positive (2015) Detection of colorectal polyps in
CT colonography exam. Am J Gastroenterol humans using an intravenously administered
103:2068–2074 fluorescent peptide targeted against c-Met. Nat
2. Heresbach D, Barrioz T, Lapalus MG et al Med 21:955–961
(2008) Miss rate for colorectal neoplastic pol- 4. Clapper ML, Hensley HH, Chang WC et al
yps: a prospective multicenter study of back-to- (2011) Detection of colorectal adenomas using
back video colonoscopies. Endoscopy a bioactivatable probe specific for matrix metal-
40:284–290 loproteinase activity. Neoplasia 13:685–691
5. Cooper HS, Chang WC, Coudry R et al cancer: is it worth talking about? Cancer
(2005) Generation of a unique strain of mul- Metastasis Rev 23:119–135
tiple intestinal neoplasia (Apc(+/Min-FCCC)) 8. Zucker S, Vacirca J (2004) Role of matrix
mice with significantly increased numbers of metalloproteinases (MMPs) in colorectal can-
colorectal adenomas. Mol Carcinog 44: cer. Cancer Metastasis Rev 23:101–117
31–41 9. Newell KJ, Witty JP, Rodgers WH et al (1994)
6. Weissleder R, Ntziachristos V (2003) Shedding Expression and localization of matrix-
light onto live molecular targets. Nat Med degrading metalloproteinases during colorectal
9:123–128 tumorigenesis. Mol Carcinog 10:199–206
7. Wagenaar-Miller RA, Gorden L, Matrisian LM 10. Kinzler KW, Vogelstein B (1996) Lessons from
(2004) Matrix metalloproteinases in colorectal hereditary colorectal cancer. Cell 87:159–170
Part VI
Matrix Metalloprotease Inhibitors

Chapter 14
Virtual High-Throughput Screening for Matrix

Metalloproteinase Inhibitors
Jun Yong Choi and Rita Fuerst
Abstract
Structure-based virtual screening (SBVS) is a common method for the fast identification of hit structures
at the beginning of a medicinal chemistry program in drug discovery. The SBVS, described in this manu-
script, is focused on finding small molecule hits that can be further utilized as a starting point for the
development of inhibitors of matrix metalloproteinase 13 (MMP-13) via structure-based molecular design.
We intended to identify a set of structurally diverse hits, which occupy all subsites (S1′–S3′, S2, and S3)
centering the zinc containing binding site of MMP-13, by the virtual screening of a chemical library com-
prising more than ten million commercially available compounds. In total, 23 compounds were found as
potential MMP-13 inhibitors using Glide docking followed by the analysis of the structural interaction
fingerprints (SIFt) of the docked structures.
Key words Matrix metalloproteinase, Structure-based virtual screening, Docking, Structural interac-
tion fingerprints, Zn-chelating inhibitor
1 Introduction
The identification of a lead structure usually initiates a medicinal

chemistry program in drug discovery. Traditionally, high-
throughput screenings (HTS) of large compound libraries are
used to deliver lead compounds for a certain protein target. The
resultant small molecule leads need to be optimized through iter-
ative analog synthesis efforts. The high costs and low hit rates
associated with HTS screening campaigns as well as a steadily
increasing number of new drug targets accelerated the develop-
ment of cheaper and faster computational alternatives starting in
the early 1990s [1, 2]. Nowadays, virtual screening (VS) methods
are broadly used in early-stage drug discovery for hit identification
by analyzing chemical databases.
There are two different approaches, ligand- and receptor-based
VS, used to prioritize compounds for synthesis and evaluation. The
ligand-based approach tries to find molecules with similar physical
259
260 Jun Yong Choi and Rita Fuerst
and chemical properties to already known protein ligands. On the

other hand, the receptor-based approach, also known as structure-
based VS (SBVS), is not biased by known protein interaction part-
ners. The screening starts with the 3-D structure of a target protein
and a 3-D database of ligands. After virtual filtering of the com-
pound library, the predictive binding mode of the lead structures is
found by docking and the hits are scored by evaluating their bind-
ing affinities to the protein target [3].
Within this manuscript we describe a SBVS aimed at finding
new lead compounds capable of inhibiting the zinc-dependent
matrix metalloproteinase 13 (MMP-13). MMP-13 is an extremely
promising drug target for the treatment of osteoarthritis (OA).
This protease is mainly responsible for the cleavage of collagen
type II, which leads to the destruction of articular cartilage, one of
the main symptoms of OA [4]. MMP-13 is also involved in cancer
progression. The proteolytic degradation of the extracellular envi-
ronment in melanoma invasion and metastasis is dependent on the
stromal expression of MMP-13 [5]. In breast cancer, metastasis to
the bone is a very common mechanism for secondary tumor
growth. Breast cancer cells manipulate signaling pathways in osteo-
blasts, leading to increased MMP-13 release, which promotes met-
astatic invasion to bone tissue [6].
1.1 Matrix MMP activity was described for the first time in 1962 as an enzy-
Metalloproteinases matic player in the metamorphosis process of tadpoles. Since then,
as Drug Targets 24 structurally and functionally related MMPs have been found in
mammals. MMPs are characterized by a Zn2+-cation in their
enzyme active site, which is coordinated by three histidines fol-
lowed by a conserved methionine residue. Under healthy condi-
tions, MMPs are very important regulators of cellular activities and
physiological processes including reproduction, tissue remodeling,
embryogenesis, and angiogenesis via the degradation of extracel-
lular matrix (ECM) components. MMPs can also influence the
immune system by directly cleaving signaling molecules including
the tumor necrosis factor (TNF) and other cytokines. MMP activ-
ity is very low under healthy conditions but can be detected during
repair or remodeling processes including angiogenesis, bone devel-
opment, or wound healing, as well as in inflamed tissue [7].
Originally, MMPs attracted the attention of numerous research
groups due to their involvement in tumor angiogenesis and metas-
tasis. First-generation MMP inhibitors, developed in the early
1990s, were generally not selective for any particular MMP. Drug
candidates resultant from these early programs were designed to
mimic peptide sequences of collagen. These peptidomimetics also
consisted of a functional group capable of chelating the active site
zinc within the MMP active site in a competitive manner. Zinc che-
lating functionalities include the carboxylate-, carbamate-, thiol-,
and hydroxamate group as well as phosphoric acid derivatives. The
Virtual High-Throughput Screening for Matrix Metalloproteinase Inhibitors 261
broad-spectrum MMP inhibitors and hydroxamic acid containing

peptidomimetics, Batimastat (1) and Marimastat (2), were the first
MMP inhibitors that entered clinical testing for cancer therapy.
Clinical trials were suspended in phase III due to musculoskeletal
toxicities resultant from nonselective binding of the inhibitors
within the MMP family [8].
Broad-spectrum MMP inhibition is common among MMP
inhibitors due to the high structural homology among the active
sites between MMP isoforms (Fig. 1). To circumvent this diffi-
culty, recent research has switched to target alternative, less con-
served allosteric sites with non-zinc chelating small molecules. The
reports of co-crystal structures of MMPs with bound ligands could
demonstrate that the significant structural differences within this
highly conserved protein family lie in the conformation of surface
loops (S1–S3 and S1′–S3′ subsites), which surround the catalyti-
cally active zinc ion [9]. Additionally, the development of isoform
selective MMP inhibitors was aided by the discovery of the S1′*
specificity loop, which is a hydrophobic pocket located adjacent to
the S1′ subsite that shows the highest sequence variability within
the MMP family. In the case of MMP-13, the S1′/S1′* subsite
forms a very deep hydrophobic pocket that is enclosed at one end
by a lysine residue [10, 11].
1.2 MMP-13 in Drug During the past 15 years, several highly potent and selective non-
Discovery zinc chelating inhibitors targeting the S1′/S1′* subsite of MMP-
13 were developed and several have entered clinical trials for the
treatment of osteoarthritis. In the early 2000s, the first highly
selective non-zinc chelating MMP-13 inhibitors, 3 and 4 (Fig. 2),
were identified as part of a HTS campaign followed by an extensive
structure activity relationship (SAR) study [12]. X-ray co-crystal
structures revealed that the dynamic movement of the S1′ pocket
O O
H
HO N CH3 O O
N N H
H H HO N CH3
O N N
S H H
OH O
S
1, IC50 2, IC50
MMP-1 = 3 nM MMP-1 = 5 nM
MMP-3 = 20 nM MMP-3 = 230 nM
MMP-7 = 6 nM MMP-7 = 16 nM
Fig. 1 Broad-spectrum MMP inhibitors

O
Ph
O
N
O N N OH
N O
-
O S O CH3 O
3, IC50 4, IC50
MMP-13 = 30 nM MMP-13 = 0.67 nM
MMP-1, -2, -3, -7, -9, -12, -14 MMP-1, -2, -3, -7, -9, -12, -14
>100,000 nM >30,000 nM
COOH
H O O
CH3
O N O O
N N O
H H HN
N N H
O N OH N S
MeO N
O
5, IC50 6, IC50
MMP-13 = 0.03 nM MMP-13 = 3.9 pm
MMP-1, -2, -3, -7, -8, -9, -12, -14 > 20,000 nM MMP-8 = 720 nM
MMP-10 = 160 nM
MMP-1, -2, -3, -7, -9, -14 > 4000 nM
Fig. 2 Non-zinc chelating MMP-13 inhibitors
seems essential for enzymatic turnover and the two compounds

block enzymatic activity simply by inducing rigidity in the S1′ sub-
site and its adjacent S1′* specificity loop. The highly potent inhibi-
tors 3 and 4 entered preclinical trials but had to be discontinued
unexpectedly because both compounds induced renal toxicity in
cynomolgus monkeys [13, 14].
In 2011, Baragi et al. from Alantos Pharmaceuticals reported
the discovery and evaluation of compound 5 for intra-articular
(IA) treatment of OA. In a manner analogous to compounds 3 and
4, this compound binds deep within the S1′ pocket and addition-
ally interacts with the S1′* specificity loop. An impressive 20,000-
fold selectivity for 5 was observed after assaying the activity of the
compound against a broad panel of MMPs (Fig. 2). Compound 5
also exhibited a remarkable half-life in rat knee joints after IA injec-
tion. The concentration of 5 in cartilage was >2 μM after 8 weeks
following a single treatment [15].
Kori et al. from Takeda published the activity of compound 6,
which binds in the same binding mode as the compounds described
before (3–5) and exhibits excellent potency and selectivity for
MMP-13 relative to other MMPs. The oral bioavailability of com-
pound 6 in different species (rats, guinea pigs, rabbits, beagle dogs,
and cynomolgus monkeys) could be significantly improved by
using the monosodium salt of the carboxylic acid functionality of
this compound. Furthermore, the authors demonstrated that oral
administration of the monosodium salt induced a significant reduc-
tion of a cartilage marker (C-telopeptide of type II collagen
(CTX-II)) of OA, after inducing OA symptoms by injecting

monoiodoacetic acid (MIA) into the rat knee joint [16].
Despite these extremely promising preclinical results, no fur-
ther improvements of compounds 5 and 6 or clinical trials using
these compounds have been published in the literature and no
MMP-13 inhibitor has achieved FDA approval thus far. The com-
munity continues to search for a disease-modifying anti-OA agent.
However, the promising anti-cancer activities of MMP-13 inhibi-
tors merit further investigation.
In the following section, a SBVS for MMP-13 is described.
The SBVS is focused on the identification of small molecule hits
that can be further utilized as starting points for the development
of MMP inhibitors via structure-based molecular design. Therefore,
we sought to identify structurally diverse hits, which occupy all
subsites (S1′–S3′, S2, and S3) centering on the zinc containing
binding site of MMP-13. We attempted to search for Zn-chelating
structures within this SBVS, which could be additionally subjected
to the structure-based design of selective MMP-inhibitors target-
ing different MMP isoforms. Although compounds interacting
with the catalytically active Zn-ion lacked selectivity, the identified
hits may be a useful resource for the structure-based design of non-
Zn-chelating inhibitors targeting individual MMP isozymes.
2 Materials
2.1 Software Two different programs were used in the SBVS focused on finding
new inhibitors for MMP-13. The program Glide from the
Schrödinger Small Molecule Drug Discovery Suite [17] was uti-
lized for ligand docking and the program MOE developed by
Chemical Computing Group [18] was used for chemical library
composition and ligand-based search. We chose these programs
due to their availability in our research laboratory and their popu-
larity within the research community reflected by a myriad of cited
references. The use of Glide and MOE requires the purchase of
their commercial licenses from Schrödinger Inc. and Chemical
Computing Group, respectively. Alternatively, other docking pro-
grams are freely available such as AutoDock [19], UCSF-DOCK
[20], GOLD [21], and SwissDock [22], which have been devel-
oped and maintained by academic institutions.
2.2 Database The virtual chemical library collection was obtained from the
ZINC12 database [23, 24], which comprises around ten million
commercially available compounds filtered for drug-like physical
properties [25]: 150 ≤ molecular weight ≤ 500, xlogP ≤ 5, num-
ber of rotatable bonds ≤ 7, polar surface area < 150 Å2, number of
hydrogen bond donors ≤ 5, and number of hydrogen bond accep-
tors ≤ 10. The structures were downloaded in an SDF file format
and converted to molecular database (mdb) files using the MOE

program. The chemical structures were cleaned up to add/remove
hydrogen atoms, remove salt and solvent molecules, and generate
relevant protonation states at the physiological pH 7.4. The collec-
tion of chemical structures was then saved in an MOE database file
format for future use.
3 Methods
3.1 Workflow The virtual screening comprised five steps as outlined below (Fig. 3):
1. Ligand-based search.
2. Analysis of X-ray structures and preparation of a target
structure.
3. Grid generation.
4. Glide docking.
5. Data analysis.
3.2 Ligand- Initially, all chemical structures possessing one or more carboxylic
Based Search acid functionalities, which can chelate the catalytically active Zn-ion
of MMPs, were selected from the MOE database collection gener-
ated in Subheading 2.2. In the Select panel of the MOE database,
ZINC Drug-like Chemical Collection

(ca. 10 million)
Ligand-based Searching
(ca. 1.3 million hits)
Glide HTVS docking

(1,325 docked structures)
Glide SP docking
(604 docked structures)
SIFt, Gscore analyses

Visual inspection
Hit Selection
Fig. 3 Virtual screening process and outcomes

the keyword, mol=“O=CO” was used to search for chemical struc-

tures containing at least one carboxylic acid moiety. This search
identified 1,387,623 compounds out of ca. ten million drug-like
small molecules (see Note 1). These structures were assembled and
saved as SDF files for structure-based virtual screening.
3.3 Analysis Structure coordinates were downloaded for all MMP isozymes
of X-Ray Structures found in the Protein Data Bank (PDB). The sequence alignment
and Preparation for these MMP X-ray crystal structures showed about 50% sequence
of the Target Structure identity between each isozyme. Superimposition of the MMP
structures, based on the sequence alignment, also showed that the
catalytic domain of different isozymes is highly conserved, and the
binding sites of co-crystallized Zn-chelating inhibitors are located
around the metal ion and extended toward all subsites (S1′–S3′,
S2, and S3). Furthermore, the chelating moieties of inhibitors that
interact with the Zn-ion are located around the metal ion at a dis-
tance of approximately 2.1 Å (Fig. 4).
The X-ray co-crystal structure of MMP-13 (PDB ID: 3ELM)
[26] was used as the target for the SBVS within this study. First,
the protein structure was prepared for docking by refining the tar-
get protein structure. This step included: assigning bond orders,
adding hydrogen atoms, creating zero-order bonds to metals, cre-
ating disulfide bonds, filling in missing side chains and loops, and
deleting water molecules beyond 5 Å from hetero groups. Chain A
of the target protein was used for further refinement that com-
prised assigning hydrogen bond interactions using the PROPKA
module (pH = 7.4). Finally, the structure was subjected for
restrained minimization using the OPLS3 force field.
Fig. 4 Superimposed X-ray co-crystal structures of MMP isozymes (a) and Zn-chelating inhibitors (b). Substrate
mimetic (decapeptide) is represented as a surface (PDB ID: 3AYU) (a). Each subsite in the substrate-binding
cleft is annotated (a) [29]
3.4 Receptor Grid A receptor grid was generated for the ligand-binding site of the
Generation target protein. The grid was set to a 25 Å3 box centered on the
Zn-chelating reference ligand of the MMP-13 co-crystal structure
(3ELM) and the metal coordination to the zinc ion was assigned
as a constraint.
3.5 Glide Docking The Virtual Screening Workflow module was used for Glide dock-
ing. The compound collection, containing at least one carboxylic
acid moiety from the ligand-based search described in Subheading
3.2, was directly used without further ligand filtering or prepara-
tion. Ligand docking was restricted by the comparison of core pat-
terns, which required the carboxylic acid unit of docked ligands to
align with that of the reference compound with a tolerance of
0.7 Å. Initially, the ligand structures were sequentially docked into
the target grid in the high-throughput virtual screening (HTVS)
mode, which reduced about 1.3 million carboxylic acid-containing
compounds to 1325 docked structures. This process was followed
by Glide standard precision (SP) mode docking of the remaining
1325 structures, which generated 604 ligands that occupied the
zinc-binding site of MMP-13. All 604 docked structures were
scored and saved for data analysis and hit triage (see Note 2).
3.6 Data Analysis The 604 docked structures occupied the entire area surrounding
and Hit Triage the Zn-ion. Since the ligand interactions with the subsites S1′–S3′,
S2, and S3 are known to improve selectivity within the MMP fam-
ily, the goal of this virtual screen was to identify diverse sets of
Zn-chelating molecules that occupied these regions centered
around the Zn-containing active site of MMP-13. Docking post-
processing procedures were applied in Maestro to calculate struc-
tural interaction fingerprints (SIFt) to analyze intermolecular
binding interactions between the protein and the docked struc-
tures. After the analysis of the SIFt matrix, a number of structurally
diverse compounds, which show interaction with the five subsites,
were selected as potential MMP-13 inhibitors [27, 28].
The computed matrix map shows the clustering of SIFts cor-
related to the docking score of the screened compounds (Fig. 5).
Various positions on the map represent ligand occupancy of differ-
ent binding sites. Ligand binding poses in each cluster of SIFts
were visually inspected and potential hits were selected using the
following selection criteria: the Glide docking score, the conforma-
tional stability of the bound inhibitor, ligand rigidity, and unique-
ness. Furthermore, compounds were triaged by comparing their
binding poses to currently known inhibitors exhibiting interactions
considered to be important for potency and selectivity for MMP-
13, such as a π-stacking interaction with His222 in the S1′ site,
hydrophobic contacts with Tyr176, His187, and Phe189 in the S3
site, and a hydrogen bond interaction with the protein backbone in
the S2′ site [26, 29–31].
Fig. 5 Structural interaction fingerprints and binding poses of ligands clustered

based on their interactions with the target structure. Blue circle represents the
Zn-binding site. The color code in the map represents the average of the Glide
docking score of different ligands clustered in each cell. It ranges from −6.6
(white, highest score) to −1.8 (dark, lowest score). The top left cell (a) of the map
represents ligands occupying the S2 site, the top middle cell (b) represents
ligands interacting with the S1′ site, the top right cell (c) represents ligands
occupying the S2′ and S3′ sits, the bottom left cell (d) represents ligands binding
the area between the subsites S2 and S3, and the middle bottom cell (e) repre-
sents the ligands occupying the S3 site
Finally, 23 ligands were selected as potential Zn-chelating MMP-

13 inhibitors (see Note 3). The ZINC ID of the selected ligands were
ZINC39475452, ZINC37417840, ZINC82555302,
ZINC02860431, ZINC62430943, ZINC41498653,
ZINC83306707, ZINC76441060, ZINC71632004,
ZINC57989822, ZINC71694581, ZINC82401153,
ZINC82867972, ZINC22145620, ZINC00229666,
ZINC76186843, ZINC20231914, ZINC71726611,
ZINC33510136, ZINC82973537, ZINC82305218,

ZINC19476737, and ZINC19438618. The structure and purchas-
ing information for each of these hits can be found at the ZINC data-
base website (www.zinc15.docking.org) using these ID numbers. The
new updated version of ZINC15 was released in early 2016.
Binding poses of four representative hits are shown in Fig. 6.
ZINC62430943 forms a metal-π interaction between the Zn-ion
and the pyridine ring and the sulfonamide moiety can form a
Fig. 6 Representative hits and their binding poses in MMP-13. (a) ZINC62430943, (b) ZINC82305218, (c)
ZINC41498653, and (d) ZINC57989822 are shown in green sticks. Amino acids in the active site of MMP-13
are shown in gray sticks, and protein structures are shown in a gray ribbon diagram. Figures were generated
using the program PyMol [31]
hydrogen bond interaction with the amide backbone of Leu185

(Fig. 6a). ZINC82305218 interacts with the amide backbone of
Leu185 in the substrate binding site via a hydrogen bond and is
oriented toward the S2′–S3′ subsite (Fig. 6b). ZINC41498653
forms hydrophobic contacts with Tyr179, His187, and Phe189 in
the S3 subsite (Fig. 6c). ZINC57989822 occupies the substrate
binding-site near the Zn-ion by forming a hydrogen bond interac-
tion with the amide backbone of Ala188 (Fig. 6d).
4 Notes
1. We assembled a drug-like chemical library of structures obtained

from the ZINC database and performed a ligand-based search
using MOE. These procedures could also be conducted using
LigPrep [32], Canvas [33], and Knime [34] in the Schrödinger
program suite. In our study, the analysis of the database to
search for potential ligands was faster in MOE compared to
using other modules of the Schrödinger program suite.
2. We identified pan-assay interference compounds (PAINS) [35]
in the 604 docked ligands obtained from Glide SP, which implies
that the ZINC drug-like compound set includes PAINS. Thus,
it is necessary to filter for PAINS during the chemical library
composition process. Current MOE and Schrödinger programs
have additional substructure filters for the removal of PAINS.
3. Coordination to the Zn-ion and the highly conserved catalytic
domain of MMP isozymes (Fig. 4a) suggests that most of the
identified hits will also inhibit other MMP isoforms. Currently,
199 X-ray structures of MMP isozymes are available in the PDB
database: 63 for MMP-12 (macrophase metalloelastase), 43 for
MMP-13 (collagenase 3), 31 for MMP-3 (Stromelysin-1), 19 for
MMP-9, 18 for MMP-8 (Neutrophil collagenase), 10 for MMP-2
(72 kDa type IV collagenase), 9 for MMP-14, 3 for MMP-7, 1
for MMP-10, 1 for MMP-11, 1 for MMP-16, and a solution
structure for MMP-20. Therefore, the hits identified from this
virtual screen may form the basis for the structure-based design of
inhibitors specifically targeting each MMP isoform.
References
1. Cheng T, Li Q, Zhou Z, Wang Y, Bryant SH 3. Ghosh S, Nie A, An J, Huang Z (2006)
(2012) Structure-based virtual screening for Structure-based virtual screening of chemical
drug discovery: a problem- centric review. libraries for drug discovery. Curr Opin Chem
AAPS J 14(1):133–141 Biol 10(3):194–202
2. Ripphausen P, Nisius B, Peltason L, Bajorath J 4. Takaishi H, Kimura T, Dalal S, Okada Y,
(2010) Quo vadis, virtual screening? A compre- D’Armiento J (2008) Joint diseases and matrix
hensive survey of prospective applications. J Med metalloproteinases: a role for MMP-13. Curr
Chem 53(24):8461–8467 Pharm Biotechnol 9(1):47–54
5. Zigrino P, Kuhn I, Bauerle T, Zamek J, Fox Taveras AG, Baragi VM (2011) Discovery and
JW, Neumann S, Licht A, Schorpp-Kistner evaluation of a non-Zn chelating, selective
M, Angel P, Mauch C (2009) Stromal expres- Matrix Metalloproteinase 13 (MMP-13) inhib-
sion of MMP-13 is required for melanoma itor for potential intra-articular treatment of
invasion and metastasis. J Invest Dermatol osteoarthritis. J Med Chem 55(2):709–716
129(11):2686–2693 16. Nara H, Sato K, Naito T, Mototani H, Oki H,
6. Morrison C, Mancini S, Cipollone J, Yamamoto Y, Kuno H, Santou T, Kanzaki N,
Kappelhoff R, Roskelley C, Overall C (2011) Terauchi J, Uchikawa O, Kori M (2014)
Microarray and proteomic analysis of breast Discovery of novel, highly potent, and selective
cancer cell and osteoblast co-cultures: role of quinazoline-2-carboxamide-based Matrix
osteoblast Matrix Metalloproteinase (MMP)- Metalloproteinase (MMP)-13 inhibitors with-
13 in bone metastasis. J Biol Chem 286(39): out a zinc binding group using a structure-
34271–34285 based design approach. J Med Chem
7. Vandenbroucke RE, Libert C (2014) Is there 57(21):8886–8902
new hope for therapeutic matrix metallopro- 17. Friesner RA, Banks JL, Murphy RB, Halgren
teinase inhibition? Nat Rev Drug Discov TA, Klicic JJ, Mainz DT, Repasky MP, Knoll
13(12):904–927 EH, Shelley M, Perry JK, Shaw DE, Francis P,
8. Rothenberg ML, Nelson AR, Hande KR (1999) Shenkin PS (2004) Glide: a new approach for
New drugs on the horizon: matrix metallopro- rapid, accurate docking and scoring: 1. Method
teinase inhibitors. Stem Cells 17(4):237–240 and assessment of docking accuracy. J Med
9. Engel CK, Pirard B, Schimanski S, Kirsch R, Chem 47(7):1739–1749
Habermann J, Klingler O, Schlotte V, 18. Molecular Operating Environment (MOE),
Weithmann KU, Wendt KU (2005) Structural 2013.08 (2016). Chemical Computing Group
basis for the highly selective inhibition of Inc., 1010 Sherbooke St. West, Suite #910,
MMP-13. Chem Biol 12(2):181–189 Montreal, QC, Canada H3A 2R7
10. Li JJ, Johnson AR (2011) Selective MMP13 19. Morris GM, Huey R, Lindstrom W, Sanner
inhibitors. Med Res Rev 31(6):863–894 MF, Belew RK, Goodsell DS, Olson AJ (2009)
11. Dormán G, Cseh S, Hajdu I, Barna L, Konya AutoDock4 and AutoDockTools4: automated
D, Kupai K, Kovacs L, Ferdinandy P (2010) docking with selective receptor flexibility.
Matrix metalloproteinase inhibitors: A critical J Comput Chem 30(16):2785–2791
appraisal of design principles and proposed 20. Allen WJ, Balius TE, Mukherjee S, Brozell SR,
therapeutic utility. Drugs 70(8):949–964 Moustakas DT, Lang PT, Case DA, Kuntz ID,
12. Johnson AR, Pavlovsky AG, Ortwine DF, Prior Rizzo RC (2015) DOCK 6: impact of new fea-
F, Man C-F, Bornemeier DA, Banotai CA, tures and current docking performance.
Mueller WT, McConnell P, Yan C, Baragi V, J Comput Chem 36(15):1132–1156
Lesch C, Roark WH, Wilson M, Datta K, 21. Jones G, Willett P, Glen RC, Leach AR, Taylor
Guzman R, Han H-K, Dyer RD (2007) R (1997) Development and validation of a
Discovery and characterization of a novel genetic algorithm for flexible docking. J Mol
inhibitor of matrix metalloprotease- 13 that Biol 267(3):727–748
reduces cartilage damage in vivo without joint 22. Grosdidier A, Zoete V, Michielin O (2011)
fibroplasia side effects. J Biol Chem SwissDock, a protein-small molecule docking
282(38):27781–27791 web service based on EADock DSS. Nucleic
13. Li JJ, Nahra J, Johnson AR, Bunker A, O’Brien Acids Res 39(Web Server issue):W270–W277
P, Yue W-S, Ortwine DF, Man C-F, Baragi V, 23. Irwin JJ, Sterling T, Mysinger MM, Bolstad
Kilgore K, Dyer RD, Han H-K (2008) ES, Coleman RG (2012) ZINC: a free tool to
Quinazolinones and pyrido[3,4-d]pyrimidin- discover chemistry for biology. J Chem Inf
4-ones as orally active and specific matrix Model 52(7):1757–1768
metalloproteinase-13 inhibitors for the treat- 24. Irwin JJ, Shoichet BK (2005) ZINC—a free
ment of osteoarthritis. J Med Chem database of commercially available compounds
51(4):835–841 for virtual screening. J Chem Inf Model
14. Cai H, Agrawai AK, Putt DA, Hashim M, 45(1):177–182
Reddy A, Brodfuehrer J, Surendran N, Lash 25. Lipinski CA (2000) Drug-like properties and
LH (2009) Assessment of the renal toxicity of the causes of poor solubility and poor permea-
novel anti-inflammatory compounds using bility. J Pharmacol Toxicol Methods
cynomolgus monkey and human kidney cells. 44(1):235–249
Toxicology 258(1):56–63 26. Monovich LG, Tommasi RA, Fujimoto RA,
15. Gege C, Bao B, Bluhm H, Boer J, Gallagher Blancuzzi V, Clark K, Cornell WD, Doti R,
BM, Korniski B, Powers TS, Steeneck C, Doughty J, Fang J, Farley D, Fitt J, Ganu V,
Goldberg R, Goldstein R, Lavoie S, Kulathila 30. Gall AL, Ruff M, Kannan R, Cuniasse P, Yiotakis
R, Macchia W, Parker DT, Melton R, O’Byrne A, Dive V, Rio MC, Basset P, Moras D (2001)
E, Pastor G, Pellas T, Quadros E, Reel N, Crystal structure of the stromelysin-3 (MMP-11)
Roland DM, Sakane Y, Singh H, Skiles J, catalytic domain complexed with a phosphinic
Somers J, Toscano K, Wigg A, Zhou S, Zhu L, inhibitor mimicking the transition-state. J Mol
Shieh W-C, Xue S, McQuire LW (2009) Biol 307(2):577–586
Discovery of potent, selective, and orally active 31. Browner MF, Smith WW, Castelhano AL
carboxylic acid based inhibitors of matrix (1995) Matrilysin-inhibitor complexes: com-
metalloproteinase-13. J Med Chem mon themes among metalloproteases.
52(11):3523–3538 Biochemistry 34(20):6602–6610
27. Singh J, Deng Z, Narale G, Chuaqui C (2006) 32. Schrödinger Release 2016-2: LigPrep, version
Structural interaction fingerprints: a new 3.8 (2016) Schrödinger, LLC, New York, NY
approach to organizing, mining, analyzing, 33. Sastry M, Lowrie JF, Dixon SL, Sherman W
and designing protein-small molecule com- (2010) Large-scale systematic analysis of 2D
plexes. Chem Biol Drug Des 67(1):5–12 fingerprint methods and parameters to improve
28. Deng Z, Chuaqui C, Singh J (2004) Structural virtual screening enrichments. J Chem Inf
interaction fingerprint (SIFt): a novel method Model 50(5):771–784
for analyzing three-dimensional protein-ligand 34. Berthold MR, Cebron N, Dill F, Gabriel TR,
binding interactions. J Med Chem Kötter T, Meinl T, Ohl P, Sieb C, Thiel K,
47(2):337–344 Wiswedel B (2007) Knime: the Konstanz

29. Hashimoto H, Takeuchi T, Komatsu K, information miner. In: Studies in classification,
Miyazaki K, Sato M, Higashi S (2011) data analysis, and knowledge organization.
Structural basis for matrix metalloprotein- Springer, New York
ase-2 (MMP-2)-selective inhibitory action of 35. Baell J, Walters MA (2014) Chemistry: chemi-
beta-amyloid precursor protein-derived inhib- cal con artists foil drug discovery. Nature
itor. J Biol Chem 286(38):33236–33243 513(7519):481–483
Chapter 15
Computational Approaches to Matrix Metalloprotease Drug

Design
Tanya Singh, B. Jayaram, and Olayiwola Adedotun Adekoya
Abstract
Matrix metalloproteinases (MMPs) are a family of zinc-containing enzymes required for homeostasis.
These enzymes are an important class of drug targets as their over expression is associated with many dis-
ease states. Most of the inhibitors reported against this class of proteins have failed in clinical trials due to
lack of specificity. In order to assist in drug design endeavors for MMP targets, a computationally tractable
pathway is presented, comprising, (1) docking of small molecule inhibitors against the target MMPs, (2)
derivation of quantum mechanical charges on the zinc ion in the active site and the amino acids coordinat-
ing with zinc including the inhibitor molecule, (3) molecular dynamics simulations on the docked ligand–
MMP complexes, and (4) evaluation of binding affinities of the ligand–MMP complexes via an accurate
scoring function for zinc containing metalloprotein–ligand complexes. The above pathway was applied to
study the interaction of the inhibitor Batimastat with MMPs, which resulted in a high correlation between
the predicted and experimental binding free energies, suggesting the potential applicability of the
pathway.
Key words Matrix metalloprotease, Computer-aided drug design, Docking and scoring, Molecular
dynamics simulations
1 Introduction
Matrix metalloproteinases constitute a family of Ca2+ containing

and Zn2+ dependent endopeptidases that are involved in cleavage
of extracellular matrix proteins. The family consists of more than
26 proteinases in mammals classified as collagenases (MMP-1, 8,
13, and 18), gelatinases (MMP-2, and 9), stromelysins (MMP-3,
10, 11, and 27), matrilysins (MMP-7 and 26), and membrane-type
(MT-MMP) based on their substrate specificity [1]. MMPs are
secreted as inactive zymogens and are activated through cleavage
by other enzymes [1]. Their enzymatic activity is regulated by their
natural inhibitors viz. tissue inhibitors of metalloproteinases
(TIMPs) [2, 3]. These enzymes are involved in the regulation of
several biological processes such as embryonic development, signal
273
274 Tanya Singh et al.
regulation, wound healing, angiogenesis, ovulation, uterine

involution, bone resorption, and nerve growth [4–11].

Dysregulation of MMP activity has been linked to a number of
diseases. For example, over-expression of MMPs can result in
accelerated matrix degradation mediating a number of pathologies
including cancer, loss of cartilage in osteoarthritis, rheumatoid
arthritis, cardiovascular diseases, acute lung injury, peridontitis,
and many others [5, 12–18]. Thus, MMPs have been a pharma-
ceutical target for more than 20 years and many MMP inhibitors
have been reported in the literature [1]. Most of these MMP
inhibitors bind to the zinc at the active site, thereby blocking its
activity. However, despite many research efforts, only one MMP
inhibitor periostat has been approved by the FDA [1].
The number of available high-resolution X-ray crystal structures
of MMP/inhibitor complexes has increased dramatically in recent
years aiding in the design of potential inhibitors at an early lead gen-
eration stage [19]. Molecules exhibiting high affinity toward Zn2+
can effectively prevent the binding of polypeptides to MMPs, thus
acting as MMP inhibitors [20]. Several Zn2+ binding groups (ZBGs)
have been reported: the hydroxamates, reverse hydroxamates, car-
boxylates, hydroxyureas, hydrazides, phosphinates, sulfones, and sul-
fonylhydrazides of which the hydroxamates appear to be the most
potent [1]. Many broad-spectrum ZBG-containing small molecule
inhibitors from different pharmaceutical companies have also entered
clinical trials for cancer, rheumatoid arthritis, and osteoarthritis [18,
21–26]. These broad-spectrum MMP inhibitors include hydroxa-
mate-based Marimastat, Batimastat, Ilomastat, Prinomastat,
Solimastat, Tanomastat, MMI- 270, Trocade, Periomastat, and
Metamastat. Nearly all of these MMP inhibitors have failed in clinical
trials due to lack of specificity against a given class of MMPs [1], pos-
ing a challenge to rational design of specific MMP inhibitors.
Computational design of MMP inhibitors is limited by several issues
including an appropriate force field to model the metal-ligand inter-
actions. Zinc is commonly found to be four-coordinated with a tetra-
hedral geometry; five and six coordinated geometries are also
observed in zinc metalloproteinases and play an important role in
metal/ligand binding [27]. Computational docking and prediction
of binding affinities of metalloproteinase inhibitors to MMPs remains
a challenge due to the multiple coordination geometries of zinc [28].
Addressing these challenges, we combine here molecular docking,
quantum mechanical charge derivation of the Zn ion in the binding
site followed by molecular dynamics (MD) simulations of an MMP-
inhibitor complex, and post facto analyses of the MD trajectories to
evaluate the binding free energies associated with MMP-inhibitor
complexation. The ability to computationally predict effectively the
binding modes and affinities of small molecule inhibitors [29] for
these zinc containing enzymes can be of utmost importance in
designing very selective clinically relevant inhibitors.
In Silico Design of MMP Inhibitors 275
2 Methods
The over expression of various MMPs (i.e., MMP-1, MMP-2,

MMP-3, MMP-8, MMP-9, and MMP-13) has been implicated in
several important diseases such as breast [30–39] and gastric can-
cer [40–43], peripheral nerve injury [44–50], neuropathic pain
[51], spinal cord damage [52–56], brain Injury [57–64], colorec-
tal cancer [65], pathologic bone resorption [66], chronic wound
[67], inflammation of skin [68] and pulmonary tract [69], hyper-
sensitivity [70], Alzheimer’s disease [71, 72], Crohn’s disease
[72], and peridontitis [73]. Considering the importance of MMPs
in these and various other diseases, we focus here on describing
computational approaches to design inhibitors against MMPs.
Batimastat, a potent, broad-spectrum MMP inhibitor was the
first to enter clinical trials against cancer [1]. Experimental pIC50
values (in nM) for Batimastat against various MMPs have previ-
ously been reported [1] (Table 1). However, there are no crystal
structures available for a Batimastat-MMP complex [1]. Thus, to
determine the molecular basis for the binding of Batimastat to vari-
ous MMPs, a docking and scoring study was performed using an in
silico drug design software suite developed in house called
Sanjeevini [74] (see Note 1). The web server incorporates several
modules that include, (1) a module for the detection of binding
sites for a target protein, scanning against a million compound
library [75] to identify hit molecules against a target protein, (2)
an all atom-based docking, and (3) a scoring module to design
molecules with desired affinity and specificity against the drug tar-
get. The docking [76, 77] and scoring [28, 78, 79] modules of
Sanjeevini have been previously validated on a large dataset of 335
protein/DNA drug targets for inhibitors with known crystal
structures and known experimental binding free energies (see Note 2).
The binding free energies of the top ranked docked structure pre-
dicted by Sanjeevini when compared with the experimental bind-
ing free energies gave a correlation coefficient of 0.83 for 335
Table 1
Experimental IC50 values in nM reported for the binding of Batimastat
with MMP-1, MMP-2, MMP-3, MMP-8, MMP-9, and MMP-13 [1]
Serial number MMPs IC50 values (nM)

1 MMP-1 10
3 MMP-2 4
4 MMP-3 20
5 MMP-8 10
6 MMP-9 1
protein/DNA drug targets. The root mean square deviation of the

top ranked docked structure against the crystal structure lies within
2 Ǻ with 90% accuracy. Further, to test the accuracy of the docking
and scoring modules of Sanjeevini on Matrix metalloproteinases,
we analyzed a test set of known crystal structures of matrix
metalloproteinase-inhibitor complexes with known experimental
binding free energies. A correlation coefficient (R2) of 0.68 was
obtained between experimental and Sanjeevini predicted binding
energies for the top ranked docked structure (Fig. 1).
Here, we demonstrate how it is possible to model the coordi-
nates of Batimastat-MMP complexes using the docking and scor-
ing modules integrated into Sanjeevini in conjunction with
molecular dynamics simulations. The computational pathway
designed for matrix metalloprotease drug design is illustrated in
Fig. 2 and outlined in detail below.
2.1 Protein Three-dimensional structures of target MMPs are downloaded

and Inhibitor from the RCSB (http://www.rcsb.org/pdb/home/home.do).
(Batimastat) PDB ids corresponding to MMP-1, MMP-2, MMP-3, MMP-8,
Preparation MMP-9, and MMP-13 were 2TCL, 1HOV, 1BIW, 1MNC,
1GKC, and 456C respectively [19]. Crystallographic waters are
removed from the protein [28] and hydrogen atoms are added
along with the correct AMBER force field parameters [28]. The
MMP catalytic site consisted of several histidine residues along
with a zinc ion. The protonation state of these histidine residues is
adjusted according to the catalytic site hydrogen bond network
[80]. Basic amino acid residues are protonated and carboxylic
groups are left deprotonated. The ionization state of the inhibitor
Fig. 1 Correlation plot between experimental binding energies (kcal/mol) and

Sanjeevini predicted binding energy (in kcal/mol) of known MMP inhibitors
Fig. 2 Computational pathway showing the steps followed in docking and scoring study of Batimastat binding
to different MMPs
molecule Batimastat is modeled as reported previously in the litera-

ture [1] and its overall formal charge [28] is maintained at −1. The
molecule is further geometry optimized utilizing the AM1 proce-
dure followed by calculation of partial charges by the AM1-BCC
procedure [81]. GAFF force field parameters [82] are used to
assign atom types, bond angle, dihedral, and van der Waals param-
eters [83] for the inhibitor molecule.
2.2 Molecular The target protein-ligand complex and the inhibitor molecule
Docking and Scoring are provided as input to Sanjeevini software suite for docking
and scoring studies. The docking module [76, 77] of Sanjeevini
docks the ligand molecule to the binding site and generates sev-
eral (i.e., 103) configurations via a six-dimensional rigid body
Monte Carlo methodology resulting in many ligand configura-
tions that are scored based on the scoring function built into
Sanjeevini. Four docked structures representing favorable poses

of the inhibitor molecule in the binding site along with the asso-
ciated predicted binding free energies in kcal/mol are provided
as output by the server.
2.3 Derivation In zinc containing metalloprotein-ligand complexes, the ligand

of Partial Atomic replaces a zinc-bound water molecule and forms monodentate or
Charges bidentate coordinate bond(s) with the zinc ion. Thus, due to a
on the Docked Ligand, charge transfer between amino acid residues and the zinc-bound
Protein and the Zinc ligand molecule, the total formal charge on the zinc ion is always
Ion less than +2. HF/6-31G* ab initio calculations were performed on
a region encompassing the zinc ion, the ligand and zinc-bonded
histidine residues using the Gaussian software [84] suite, followed
by RESP fitting on the resultant electrostatic potentials to obtain
partial atomic charges on the ligand and the zinc ion. For these
calculations protein residues within the coordinate bond distance
(<2.7 Å) of the zinc ion are deprotonated. The net charge on the
zinc binding motif comprising the zinc ion, amino acid residues
within a distance of 2.7 Å of the zinc ion and the ligand molecule
is equal to the sum of the formal charge on each amino acid resi-
due, the formal charge on the ligand, and the +2 charge of the zinc
ion [28]. Parameters for the zinc ion are adopted from the work of
Stote and Karplus [85] [σ = 1.95 Å, ε = 0.25 (kcal/mol)], while
those for the ligand atoms are from the GAFF force field [82]. The
AntechAMBER [86] module of AMBER is used to assign the
bonded and the nonbonded parameters to the ligand atoms.
Assignment of force field parameters for protein atoms (RESP
derived partial atomic charges, van der Waals and bonded param-
eters) is carried out using the AMBER force field.
2.4 Molecular The docked protein-inhibitor complexes are subjected to molecu-

Dynamics Simulations lar dynamics simulations to account for flexibility/dynamics of the
of the Docked ligand and the active site residues of the target, explicit solvent, and
Complex small ion effects [87]. These molecular dynamics simulations are
performed under periodic boundary conditions within the AMBER
suite of programs [86]. Prior to molecular dynamics simulations
11 Na+ ions are added to MMP-2 and MMP-3 inhibitor com-
plexes, 12 Na+ ions to MMP-1, MMP-8, and MMP-9 inhibitor
complexes and seven Na+ ions to the MMP-13 inhibitor complex
to ensure that there is a zero net charge on the protein-ligand com-
plex. The complexes are solvated with an 8 Å thick layer of water
modeled using TIP4PEW parameters [88]. Distances defining the
zinc ion and nitrogen atoms of the zinc-bound histidine residues
are restrained to prevent the zinc ion from escaping into the sol-
vent and to maintain the orientation of the zinc-chelating histidine
residues. Once the docked complexes have been prepared for
molecular dynamics simulations, an initial minimization of the sol-
vent molecules is performed followed by minimization of the
solute-solvent system. Slow heating to 300 K, while keeping the

volume constant, is performed over a period of 100 ps on the sol-
ute atoms using harmonic restraints of 25 kcal/mol/Å2. Slow
relaxation from 5 to 1 kcal/mol/Å2 is applied to these restraints,
in five segments consisting of 1,000 steps of energy minimization
and 50 ps of equilibration, with a constant temperature of 300 K
and a pressure of 1 bar via the Berendsen algorithm [89] with a
coupling constant of 0.2 ps for both parameters. Further, we also
apply 50 ps of equilibration with a restraint of 0.5 kcal/mol/Å2
and 50 ps of unrestrained equilibration. This is followed by a
molecular dynamics simulation for 100 ns under constant tempera-
ture and pressure using the Berendsen algorithm with a coupling
constant of 5 ps. Molecular dynamics simulations are performed
on all unbound MMPs studied and their inhibitor-bound com-
plexes [90] and plots of RMSD versus time are used to check the
stability of the docked complexes.
2.5 Post facto Structures at equal intervals over the last 40 ns of the molecular
Analyses of Molecular dynamics simulation run trajectory are extracted (i.e., approxi-
Dynamics Simulations mately 100 structures in total) for each system (see Note 3). For
each structure the binding free energy is estimated using the
Bappl-Z scoring function [28]. For these calculations, the system
is parameterized within the additivity approximation where the net
free energy change is treated as a sum of contributions from indi-
vidual energy components. The equation for the estimation of the
free energy change upon binding is:
22
DG ° = a Eel + b Evdw + å s AD ALSA + l ( DSCR ) + d (1)
( A =1)
The Bappl-Z scoring function used for calculating binding free
energy estimates has been thoroughly validated earlier [28]. The
scoring function captures the theoretical rigor of the MMGBSA/
MMPBSA [91–97] methodologies as well as the rapidity of empiri-
cal/knowledge-based methods [28, 78, 79]. The individual terms
in Eq. 1 are described below.
Eel: Electrostatic interactions including interactions between
protein and ligand atoms and the zinc ion with rest of the complex.
These electrostatic interactions are calculated based on Coulomb’s
law with a sigmoidal dielectric function for solvent screening effects
[28]. To model the electrostatic interactions of zinc with the rest
of the complex, we have adopted the nonbonded model described
by Stote and Karplus [85].
Evdw: Direct van der Waals interactions between protein and
ligand atoms and the zinc ion with the rest of the complex. Van der
Waals interactions are modeled using the Lennard-Jones potential
[28] while interactions with the zinc ion are modeled on the lines
of Stote and Karplus [85].
σAΔALSA: Hydrophobic contributions (nonelectrostatic

components of desolvation) are captured using a modified version
of the Eisenberg-Mclachlan model, where atom types for proteins
and small molecules in the AMBER force field have been com-
bined into a set of 22 atom types [28], to ensure transferability of
parameters to other biological systems. In Eq. 1 ΔALSA represents
the net loss in surface area for an atom type.
ΔSCR: Loss in conformational entropy of protein side chains
upon binding of the ligand to the protein [28].
ΔGo is calculated for each structural snapshot obtained over
the last 40 ns of the trajectory file and block averaging is performed
to obtain average binding free energy values (see Note 4).
2.6 Correlation In zinc containing metalloprotein-ligand complexes, the ligand is

between Experimental bonded to the zinc ion with one or two coordinate bonds. The
and Predicted Binding output of Sanjeevini is comprised of four docked structures repre-
Free Energies senting energetically favorable poses of the ligand molecule in the
binding site. The pose in which the hydroxamate group of the
ligand molecule is chelated to the zinc ion is chosen for further
analysis. Structures obtained from the molecular dynamics trajec-
tory are processed using the BapplZ scoring function and the aver-
age binding energy is calculated for Batimastat binding against
different MMPs. In Fig. 3, we present plots of the correlation
between the experimental pIC50 and predicted binding free ener-
gies for Batimastat binding to different MMPs. These results show
that the computational pathway outlined in this chapter, comprised
Fig. 3 Correlation between experimental activities (pIC50) and Sanjeevini pre-

dicted binding energies (kcal/mol) of Batimastat binding to MMP-1, MMP-2,
MMP-3, MMP-8, MMP-9, and MMP-13
of (1) an efficient docking algorithm, (2) a correct charge assign-

ment protocol, (3) a rigorous molecular dynamics simulation
study, and (3) an accurate scoring function, can contribute toward
structural, dynamic, and thermodynamic rationalization of the
experimental inhibition data. This is evident from the high correla-
tion coefficient [98] obtained between experimental and predicted
binding free energies (Fig. 3).
3 Notes
1. Sanjeevini is freely available as a web server at http://www.

scfbio-iitd.res.in/sanjeevini/sanjeevini.jsp for protein and
DNA targeted lead molecule discovery.
2. Protein targets consisting of zinc containing metalloprotein-
ases have also been tested earlier.
3. This approach subjects the systems to configurational averag-
ing via molecular dynamics simulations, thereby overcoming
the limitations inherent in single point calculations (i.e., stud-
ies on energy minimized structures).
4. The Bappl-Z scoring function is freely accessible as a web tool
at http://www.scfbio-iitd.res.in/software/drugdesign/bap-
plz.jsp.
Acknowledgments
This work is supported by grants from the Department of

Biotechnology, Govt. of India. Tanya Singh is a recipient of Senior
Research Fellowship from Council of Scientific & Industrial
Research, Govt. of India.
References
1. Skiles JW, Gonnella NC, Jeng AY (2004) The 5. Hu J, Van den Steen PE, Sang QX, Opdenakker
design, structure, and clinical update of small G (2007) Matrix metalloproteinase inhibitors
molecular weight matrix metalloproteinase as therapy for inflammatory and vascular dis-
inhibitors. Curr Med Chem 11:2911–2977 eases. Nat Rev Drug Discov 6:480
2. Cawston T (1998) Matrix metalloproteinases 6. Ray JM, Stetler-Stevenson WG (1995)
and TIMPs: properties and implications for Gelatinase A activity directly modulates mela-
the rheumatic diseases. Mol Med Today noma cell adhesion and spreading. EMBO
4:130–137 J 14:908–917
3. Blavier L, Henriet P, Imren S, Declerck YA 7. Chambers AF, Matrisian LM (1997) Changing
(1999) Tissue inhibitors of matrix metallopro- views of the role of matrix metalloproteinases
teinases in cancer. Ann N Y Acad Sci 878: in metastasis. J Natl Cancer Inst 89:
108–119 1260–1270
4. Chang C, Werb Z (2001) The many faces of 8. Kahari VM, Saarialho-Kere U (1993) Matrix
metalloproteases: cell growth, invasion, angio- metalloproteinases and their inhibitors in
genesis and metastasis. Trends Cell Biol tumour growth and invasion. Ann Med
11:S37–S43 31:34–45
9. Kleiner DE, Stetler-Stevenson WG (1999) inhibitors: selectively targeting membrane-

Matrix metalloproteinases and metastasis. anchored MMPs with therapeutic antibodies.
Cancer Chemother Pharmacol 43:S42–S51 Biochem Res Int. doi:10.1155/2011/191670
10. Yong VW (2005) Metalloproteinases: media- Article ID 191670: 11p
tors of pathology and regeneration in the 24. Close DR (2001) Matrix metalloproteinase
CNS. Nat Rev Neurosci 6:931–944 inhibitors in rheumatic diseases. Ann Rheum
11. Whittaker M, Floyd CD, Brown P, Gearing AJ Dis 60:62–67
(1999) Design and therapeutic application of 25. Baragi VM, Becher G, Bendele AM, Biesinger
matrix metalloproteinase inhibitors. Chem Rev R, Bluhm H, Boer J, Deng H, Dodd R, Essers
99:2735 M, Feuerstein T, Gallagher BM, JrGege C,
12. Weinstat-Saslow DL, Zabrenetzky VS, van Hochgürtel M, Hofmann M, Jaworski A, Jin
Houtte K, Frazier WA, Roberts DD, Steeg PS L, Kiely A, Korniski B, Kroth H, Nix D, Nolte
(1994) Transfection of thrombospondin 1 B, Piecha D, Powers TS, Richter F, Schneider
complementary DNA into a human breast car- M, Steeneck C, Sucholeiki I, Taveras A,
cinoma cell line reduces primary tumor growth, Timmermann A, Van Veldhuizen J, Weik J, Wu
metastatic potential, and angiogenesis. Cancer X, Xia B (2009) A new class of potent matrix
Res 54:6504–6511 metalloproteinase 13 inhibitors for potential
13. Kessenbrock K, Plaks V, Werb Z (2010) Matrix treatment of osteoarthritis: evidence of histo-
metalloproteinases: regulators of the tumor logic and clinical efficacy without musculoskel-
microenviroment. Cell 141:52–67 etal toxicity in rat models. Arthritis Rheum
60:2008–2018
14. Ardi VC, Kupriyanova TA, Deryugina EI,
Quigley JP (2007) Human neutrophils 26. Li NG, Shi ZH, Tang YP, Wang ZJ, Song SL,
uniquely release TIMP-free MMP-9 to provide Qian LH, Qian DW, Duan JA (2011) New
a potent catalytic stimulator of angiogenesis. hope for the treatment of osteoarthritis
Proc Natl Acad Sci U S A 104:20262–20267 through selective inhibition of MMP-13. Curr
Med Chem 18:977–1001
15. Egeblad M, Werb Z (2002) New functions for
the matrix metalloproteinases in cancer pro- 27. Kawai K, Nagata N (2012) Metal-ligand inter-
gression. Nat Rev Cancer 2:161–174 actions: an analysis of zinc binding groups
using the Protein Data Bank. Eur J Med Chem
16. Elkington P, Shiomi T, Breen R, Nuttall RK, 51:271–276
Ugarte-Gil CA, Walker NF, Saraiva L, Pedersen
B, Mauri F, Lipman M, Edwards DR, 28. Jain T, Jayaram B (2007) A computational
Robertson BD, D’Armiento J, Friedland JS protocol for predicting the binding affinities of
(2011) MMP-1 drives immunopathology in zinc containing metalloprotein-ligand com-
human tuberculosis and transgenic mice. J Clin plexes. Proteins 67:1167–1178
Invest 121:1827–1833 29. Langley DR, Walsh AW, Baldick CJ, Eggers BJ,
17. Murphy G, Knäuper V, Atkinson S, Butler G, Rose RE, Levine SM, Kapur AJ, Colonno RJ,
English W, Hutton M, Stracke J, Clark I Tenney DJ (2007) Inhibition of hepatitis B
(2002) Matrix metalloproteinases in arthritic virus polymerase by entecavir. J Virol
disease. Arthritis Res 4(Suppl 3):S39–S49 81:3992–4001
18. Fingleton B (2007) Matrix metalloproteinases 30. Liu H, Kato Y, Erzinger SA, Kiriakova GM,
as valid clinical targets. Curr Pharm Des Qian Y, Palmieri D, Steeg PS, Price JE (2012)
13:333–346 The role of MMP-1 in breast cancer growth
and metastasis to the brain in a xenograft
19. Westbrook J, Feng Z, Gilliland G, Bhat TN, model. BMC Cancer 12:583
Weissig H, Shindyalov IN, Bourne PE (2000)
The Protein Data Bank. Nucleic Acids Res 31. Fink-Retter A, Gschwantler-Kaulich D,
28:235–242 Hudelist G, Walter K, Czerwenka C (2007)
Differential spatial expression and activation
20. Tu G, Xu W, Huang H, Li S (2008) Progress in pattern of EGFR and HER2 in human breast
the development of matrix metalloproteinase cancer. Oncol Rep 18:299–304
inhibitors. Curr Med Chem 15:1388–1395
32. Poola I, DeWitty RL, Marshalleck JJ, Bhatnagar
21. Overall CM, López-Otín C (2002) Strategies R, Abraham J, Leffall LD (2005) Identification
for MMP inhibition in cancer: innovations for of MMP-1 as a putative breast cancer predic-
the post-trial era. Nat Rev Cancer tive marker by global gene expression analysis.
2:657–672 Nat Med 11:481–483
22. Fisher JF, Mobashery S (2006) Recent 33. Monteagudo C, Merino MJ, San-Juan J, Liotta
advances in MMP inhibitor design. Cancer LA, Stetler-Stevenson WG (1990)
Metastasis Rev 25:115–136 Immunohistochemical distribution of type IV
23. Dransfield DT (2011) New strategies for the collagenase in normal, benign, and malignant
next generation of matrix-metalloproteinase breast tissue. Am J Pathol 136:585–592
34. Onisto M, Riccio MP, Scannapieco P, Caenazzo 45. Shubayev VI, Angert M, Dolkas J, Campana
C, Griggio L, Spina M, Stetler-Stevenson WG, WM, Palenscar K, Myers RR (2006) TNF
Garbisa S (1995) Gelatinase A/TIMP-2 imbal- alpha-induced MMP-9 promotes macrophage
ance in lymph-node-positive breast carcino- recruitment into injured peripheral nerve. Mol
mas, as measured by RT-PCR. Int J Cancer Cell Neurosci 31:407–415
63:621–626 46. Chattopadhyay S, Myers RR, Janes J, Shubayev
35. Engel G, Heselmeyer K, Auer G, Bäckdahl M, V (2007) Cytokine regulation of MMP-9 in
Eriksson E, Linder S (1994) Correlation peripheral glia: implications for pathological
between stromelysin-3 mRNA level and out- processes and pain in injured nerve. Brain
come of human breast cancer. Int J Cancer Behav Immun 21:561–568
58:830–835 47. Liu LY, Zheng H, Xiao HL, She ZJ, Zhao SM,
36. Ahmad A, Hanby A, Dublin E, Poulsom R, Chen ZL, Zhou GM (2008) Comparison of
Smith P, Barnes D, Rubens R, Anglard P, Hart blood-nerve barrier disruption and matrix
I (1998) Stromelysin 3: an independent prog- metalloprotease-9 expression in injured central
nostic factor for relapse-free survival in node- and peripheral nerves in mice. Neurosci Lett
positive breast cancer and demonstration of 434:155–159
novel breast carcinoma cell expression. Am 48. Platt CI, Krekoski CA, Ward RV, Edwards DR,
J Pathol 152:721–728 Gavrilovic J (2003) Extracellular matrix and
37. Freije JM, Diez-Itza I, Balbin M, Sánchez LM, matrix metalloproteinases in sciatic nerve.
Blasco R, Tolivia J, López-Otín C (1994) J Neurosci Res 74:417–429
Molecular cloning and expression of collage- 49. Shubayev VI, Myers RR (2008) Upregulation
nase-3, a novel human matrix metalloprotein- and interaction of TNFalpha and gelatinases A
ase produced by breast carcinomas. J Biol and B in painful peripheral nerve injury. Brain
Chem 269:16766–16773 Res 855:83–89
38. Ueno H, Nakamura H, Inoue M, Imai K, 50. Ramer R, Hinz B (2008) Inhibition of cancer
Noguchi M, Sato H, Seiki M, Okada Y (1997) cell invasion by cannabinoids via increased
Expression and tissue localization of expression of tissue inhibitor of matrix metallo-
membrane-types 1, 2, and 3 matrix metalloproteinases-1. J Natl Cancer Inst 2:59–69
proteinases in human invasive breast carcino- 51. Sommer C, Schmidt C, George A, Toyka KV
mas. Cancer Res 57:2055–2060 (1997) A metalloprotease-inhibitor reduces
39. Jones JL, Glynn P, Walker RA (1999) Expression pain associated behavior in mice with experi-
of MMP-2 and MMP-9, their inhibitors, and mental neuropathy. Neurosci Lett 237:45–48
the activator MT1-MMP in primary breast car- 52. Noble LJ, Donovan F, Igarash T, Goussev S,
cinomas. J Pathol 189:161–168 Werb Z (2002) Matrix metalloproteinases limit
40. Cai QW, Li J, Li X, Wang J, Huang Y (2012) functional recovery after spinal cord injury by
Expression of STAT3, MMP-1 and TIMP-1 in modulation of early vascular events. J Neurosci
gastric cancer and correlation with pathological 1:7526–7535
features. Mol Med Rep 5:1438–1442 53. Pannu R, Christie DK, Barbosa E, Singh I,
41. Murray GI, Duncan ME, Arbuckle E, Melvin Singh AK (2007) Post-trauma Lipitor treat-
WT, Fothergill JE (1998) Matrix metallopro- ment prevents endothelial dysfunction, facili-
teinases and their inhibitors in gastric cancer. tates neuroprotection, and promotes locomotor
Gut 43:791–797 recovery following spinal cord injury.
42. Kemik O, Kemik AS, Sümer A, Dulger AC, J Neurochem 101:182–200
Adas M, Begenik H, Hasirci I, Yilmaz O, 54. Fleming JC, Norenberg MD, Ramsay DA,
Purisa S, Kisli E, Tuzun S, Kotan C (2011) Dekaban GA, Marcillo AE, Saenz AD, Pasqale-
Levels of matrix metalloproteinase-1 and tissue Styles M, Dietrich WD, Weaver LC (2008) The
inhibitors of metalloproteinase-1 in gastric can- cellular inflammatory response in human spinal
cer. World J Gastroenterol 17:2109–2112 cords after injury. Brain 129:3249–3269
43. Fujimoto D, Hirono Y, Goi T, Katayama K, 55. Amantea D, Corasaniti MT, Mercuri NB,
Vamaguchi A (2008) Prognostic value of Bernardi G, Bagetta G (2008) Brain regional
Protease-activated Receptor-1 (PAR-1) and and cellular localization of gelatinase activity in
Matrix Metalloproteinase-1 (MMP-1) in gas- rat that have undergone transient middle cere-
tric cancer. Anticancer Res 28:847–854 bral artery occlusion. Neuroscience 152:8–17
44. Kawasaki Y, Xu ZZ, Wang X, Park JY, Zhuang 56. Nagel S, Su Y, Horstmann S, Heiland S,
ZY, Tan PH, Gao YJ, Roy K, Corfas G, Lo EH, Gardner H, Koziol J, Martinez-Torres FJ,
Ji RR (2008) Distinct roles of matrix metallopro- Wagner S (2008) Minocycline and hypother-
teases in the early- and late-phase development of mia for reperfusion injury after focal cerebral
neuropathic pain. Nat Med 14:331–336 ischemia in the rat: effects on BBB breakdown
and MMP expression in the acute and subacute minogen activator, plasminogen activator
phase. Brain Res 1188:198–206 inhibitor and gelatinase-B in chronic wound
57. Ding YH, Li J, Rafols JA, Ding Y (2004) fluid switches from a chronic to acute wound
Reduced brain edema and matrix metallopro- profile with progression to healing. Wound
teinase (MMP) expression by pre-reperfusion Repair Regen 7:154–165
infusion into ischemic territory in rat. Neurosci 68. Warner RL, Bhagavathula N, Nerusu KC,
Lett 372:35–39 Lateef H, Younkin E, Johnson KJ, Varani
58. Fujimoto M, Takagi Y, Aoki T, Hayase M, J (2004) Matrix metalloproteinases in acute
Marumo T, Gomi M, Nishimura M, Kataoka H, inflammation: induction of MMP-3 and
Hashimoto N, Nozaki K (2008) Tissue inhibi- MMP-9 in fibroblasts and epithelial cells fol-
tor of metalloproteinases protect blood- brain lowing exposure to pro-inflammatory media-
barrier disruption in focal cerebral ischemia. tors in vitro. Exp Mol Pathol 76:189–195
J Cereb Blood Flow Metab 28:1674–1685 69. Sagel, S.D., Kapsner, R.K, and Osberg, I.
59. Vilalta A, Sahuquillo J, Rosell A, Poca MA, (2005) Induced sputum matrix metallopro-
Riveiro M, Montaner J (2008) Moderate and teinase-9 correlates with lung function and air-
severe traumatic brain injury induce early over- way inflammation in children with cystic
expression of systemic and brain gelatinases. fibrosis. Pediatr Pulmonol 39, 224–232.
Intensive Care Med 34:1384–1392 70. Yang H, Dai Y, Dong H, Zang D, Liu Q, Duan
60. Jiang X, Namura S, Nagata I (2001) Matrix H, Niu Y, Bin P, Zheng Y (2011)
metalloproteinase inhibitor KB-R7785 attenu- Trichloroethanol up-regulates matrix metallo-
ates brain damage resulting from permanent proteinase-9 and tissue inhibitor of metallopro-
focal cerebral ischemia in mice. Neurosci Lett teinase-1 in HaCaT cells. Toxicol In Vitro
305:41–44 25:1638–1643
61. Lee JE, Yoon YJ, Moseley ME, Yenari MA 71. Shibataa N, Ohnumaa T, Higashia S, Usuia
(2005) Reduction in levels of matrix metallo- C, Ohkuboa T, Kitajimaa A, Uekib A, Nagaoc
proteinases and increased expression of tissue M, Araia H (2005) Genetic association
inhibitor of metalloproteinase-2 in response to between matrix metalloproteinase MMP-9
mild hypothermia therapy in experimental and MMP-3 polymorphisms and Japanese
stroke. J Neurosurg 103:289–297 sporadic Alzheimer’s disease. Neurobiol
62. Yang Y, Estrada EY, Thompson JF, Liu W, Aging 26:1011–1014
Rosenberg GA (2007) Matrix metalloproteinase- 72. Kofla-Dlubacz A, Matusiewicz M, Krzystek-
mediated disruption of tight junction proteins in Korpacka M, Iwanczak B (2012) Correlation
cerebral vessels is reversed by synthetic matrix of MMP-3 and MMP-9 with Crohn’s disease
metalloproteinase inhibitor in focal ischemia in activity in children. Dig Dis Sci 57:706–712
rat. J Cereb Blood Flow Metab 4:697–709 73. Kumar MS, Vamsi G, Sripriya R, Sehgal PK
63. Wang Y, Deng XL, Xiao XH, Yuan BX (2007) A (2006) Expression of matrix metalloprotein-
non-steroidal anti-inflammatory agent provides ases (MMP-8 and -9) in chronic periodontitis
significant protection during focal ischemic patients with and without diabetes mellitus.
stroke with decreased expression of matrix metal- J Periodontol 77:1803–1808
loproteinases. Curr Neurovasc Res 3:176–183 74. Jayaram B, Singh T, Mukherjee G, Mathur A,
64. Truettner JS, Alonso OF, Dalton DW (2005) Shekhar S, Shekhar V (2012) Sanjeevini: a
Influence of therapeutic hypothermia on matrix freely accessible web-server for target directed
metalloproteinase activity after traumatic brain lead molecule discovery. BMC Bioinformatics
injury in rats. J Cereb Blood Flow Metab 13:S7
11:1505–1516 75. Mukherjee G, Jayaram B (2013) A rapid iden-
65. Sunami E, Tsuno N, Osada T, Saito S, Kitayama tification of hit molecules for target proteins via
J, Tomozawa S, Tsuruo T, Shibata Y, Muto T, physico-chemical descriptors. Phys Chem
Nagawa H (2000) MMP-1 is a prognostic Chem Phys 15:9107–9116
marker for hematogenous metastasis of 76. Gupta A, Gandhimathi A, Sharma P, Jayaram B
colorectal cancer. Oncologist 5:108–114 (2007) ParDOCK: an all atom energy-based
66. Rodrigues WF, Madeira MF, da Silva TA, Monte Carlo docking protocol for protein-ligand
Clemente-Napimoga JT, Miguel CB, Dias-da- complexes. Protein Pept Lett 14:632–646
Silva VJ, Barbosa-Neto O, Lopes AH, Napimoga 77. Singh T, Biswas D, Jayaram B (2011) AADS—
MH (2012) Low dose of propranolol down- An automated active site identification, dock-
modulates bone resorption by inhibiting inflam- ing and scoring protocol for protein targets
mation and osteoclast differentiation. Br based on physico-chemical descriptors. J Chem
J Pharmacol 165:2140–2151 Inf Model 51:2515–2527
67. Wysocki AB, Kusakabe AO, Chang S, Tuan TL 78. Jain T, Jayaram B (2005) An all atom energy-
(1997) Temporal expression of urokinase plas- based computational protocol for predicting
binding affinities of protein-ligand complexes. 87. Kalra P, Reddy TV, Jayaram B (2011) Free
FEBS Lett 579:6659–6666 energy component analysis for drug design: a
79. Shaikh SA, Jayaram B (2007) A Swift all-atom case study of HIV-1 protease-inhibitor bind-
energy-based computational protocol to pre- ing. J Med Chem 44:4325–4338
dict DNA ligand binding affinity and ΔTm. 88. Horn HW, Swope WC, Pitera JW, Madura JD,
J Med Chem 50:2240–2244 Dick TJ, Hura GL, Head-Gordon T (2004)
80. Giangreco I, Lattanzi G, Nicolotti O, Catto M, Development of an improved four-site water
Laghezza A, Leonetti F, Stefanachi A, Carotti model for biomolecular simulations:
A (2011) Insights into the complex formed by TIP4P-Ew. J Chem Phys 120:9665–9678
matrix metalloproteinase-2 and alloxan inhibi- 89. Berendsen HJC, Postma JPM, van Gunsteren
tors: molecular dynamics simulations and free WF, DiNola A, Haak JR (1984) Molecular
energy calculations. PLoS One 6:1 dynamics with coupling to an external bath.
81. Jakalian A, Bush BL, Jack DB, Bayly CI (2004) J Chem Phys 81:3684–3690
Fast, efficient generation of high-quality atomic 90. Lavery R, Zakrzewska K, Beveridge DL,
charges. AM1-BCC model: I. Method. Bishop TC, Case DA, Cheatham T III, Dixit S,
J Comput Chem 21:132–146 Jayaram B, Lankas F, Laughton C, Maddocks
82. Wang J, Wolf RM, Caldwell JW, Kollman PA, JH, Michon A, Osman R, Orozco M, Perez A,
Case DA (2004) Development and testing of a Singh T, Spackova N, Sponer J (2009) A sys-
general amber force field. J Comput Chem tematic molecular dynamics study of nearest
25:1157–1174 neighbor effects on base pair and base pair step
83. Cornell WD, Cieplak P, Bayly CI, Gould IR, conformations and fluctuations in
Merz KM (1995) A second generation force B-DNA. Nucleic Acids Res 38:299–313
field for the simulation of proteins, nucleic 91. Rizzo RC, Toba S, Kuntz ID (2004) A molec-
acids, and organic molecules. J Am Chem Soc ular basis for the selectivity of thiadiazole urea
117:5179–5197 inhibitors with stromelysin-1 and gelatinase-A
84. Frisch MJ, Trucks GW, Schlegel HB, Scuseria from generalized born molecular dynamics
GE, Robb MA, Cheeseman JR, Scalmani G, simulations. J Med Chem 47:3065–3074
Barone V, Mennucci B, Petersson GA, 92. Chong S, Ham S (2013) Assessing the influ-
Nakatsuji H, Caricato M, Li X, Hratchian HP, ence of solvation models on structural
Izmaylov AF, Bloino J, Zheng G, Sonnenberg characteristics of intrinsically disordered pro-
JL, Hada M, Ehara M, Toyota K, Fukuda R, tein. Comput Theor Chem 1017:194–199
Hasegawa J, Ishida M, Nakajima T, Honda Y, 93. Shaikh SA, Ahmed SR, Jayaram B (2004) A
Kitao O, Nakai H, Vreven T, Montgomery JA molecular thermodynamic view of DNA-drug
Jr, Peralta JE, Ogliaro F, Bearpark M, Heyd JJ, interaction: a case study of 25 minor groove
Brothers E, Kudin KN, Staroverov VN, binders. Arch Biochem Biophys 429:81–99
Kobayashi R, Normand J, Raghavachari K, 94. Jayaram B, Mcconnell K, Dixit SB, Beveridge
Rendell A, Burant JC, Iyengar SS, Tomasi J, DL (2002) Free energy component analysis of
Cossi M, Rega N, Millam NJ, Klene M, Knox 40 protein-DNA complexes: a consensus view
JE, Cross JB, Bakken V, Adamo C, Jaramillo J, on the thermodynamics of binding at the
Gomperts R, Stratmann RE, Yazyev O, Austin molecular level. J Comput Chem 23:1–14
AJ, Cammi R, Pomelli C, Ochterski JW, Martin 95. Jayaram B, McConnell KJ, Dixit SB, Beveridge
RL, Morokuma K, Zakrzewski VG, Voth GA, DL (1999) Free energy analysis of protein-
Salvador P, Dannenberg JJ, Dapprich S, DNA binding: the EcoRI endonuclease—DNA
Daniels AD, Farkas Ö, Foresman JB, Ortiz JV, complex. J Comput Phys 151:333–357
Cioslowski J, Fox DJ (2009) Gaussian, Inc.,
Wallingford, CT 96. Fenley MO, Harris R, Jayaram B, Boschitsch
AH (2010) Revisiting the association of cationic
85. Stote RH, Karplus M (1995) Zinc binding in groove-binding drugs to DNA using a Poisson-
proteins and solution: a simple but accurate Boltzmann approach. Biophys J 99:879–886
nonbonded representation. Proteins Struct
Funct Genet 23:12–31 97. Wong S, Amaro RE, McCammon JA (2009)
MM-PBSA captures key role of intercalating
86. Pearlman DA, Case DA, Caldwell JW, Ross water molecules at a protein−protein interface.
WS, Cheathem JE III, DeBolt S, Ferguson D, J Chem Theory Comput 5:422–449
Seibel G, Kollman P (1995) AMBER, a pack-
age of computer programs for applying molec- 98. Singh T, Adekoya OA, Jayaram B (2015)
ular mechanics, normal mode analysis, Understanding the binding of inhibitors of
molecular dynamics and free energy calcula- matrix metalloproteinases by molecular docking,
tions to simulate the structural and energetic quantum mechanical calculations, molecular
properties of molecules. Comput Phys dynamics simulations, and a MMGBSA/
Commun 91:1–41 MMBappl study. Mol Biosyst 11:1041–1051
Chapter 16
A Simple Adaptable Blood-Brain Barrier Cell Model

for Screening Matrix Metalloproteinase Inhibitor
Functionality
Jennifer S. Myers, Joan Hare, and Qing-Xiang Amy Sang
Abstract
The blood-brain barrier is a multicellular and basement membrane unit that regulates molecular transport
between the blood and central nervous system. Many cerebral pathologies, such as acute stroke and
chronic vascular dementia, result in a disrupted blood-brain barrier, increasing its permeability and allow-
ing the entry of potentially neurotoxic molecules. The activation of matrix metalloproteinases mediates
further blood-brain barrier damage. The inhibition of matrix metalloproteinases is a potential strategy for
stroke therapy. As inhibitors are developed, efficient context-specific screening methods will be required.
Models of the blood-brain barrier have been extensively used to study neuropathologies and the effect of
various treatment options.
Herein, we describe a co-culture model of the blood-brain barrier composed of brain microvascular
endothelial cells and astrocytes grown on an artificial basement membrane-coated membrane insert. Our
cell model forms a barrier and is a simple first approximation of blood-brain barrier integrity. As currently
developed, the model may be applied to testing the effect of matrix metalloproteinases and matrix metal-
loproteinase inhibitors on blood-brain barrier physiology and pathophysiology. The model is a quick and
effective evaluation tool for generating nonclinical data in a living cell system before proceeding to animal
models.
Key words Blood-brain barrier, Matrix-metalloproteinases, Matrix-metalloproteinase inhibitors,

Tissue plasminogen activator, Stroke, BBB co-culture model, BBB permeability, Cell culture mem-
brane inserts, Normal human astrocytes, Bovine brain microvascular endothelial cells
1 Introduction
Cerebral vascular endothelial cells, a basal lamina, pericytes, and

astrocytes or glia combine to form the blood-brain barrier (BBB)
neurovascular unit, a selectively permeable physical barrier between
blood vessels and the central nervous system. Each component of
the neurovascular unit plays a particular role in BBB function [1].
The BBB is a complex entity that ensures normal brain function by
controlling the entry of molecules and white blood cells from
287
288 Jennifer S. Myers et al.
s ystemic circulation and by maintaining homeostasis of the brain’s

extracellular fluid. Tight junctions between the cells of the BBB
maintain integrity of the barrier and restrict paracellular entry.
Cerebral pathologies such as strokes and chronic vascular dementia
alter the structural integrity of brain tissues and blood vessels. The
resulting injury eventually degrades the BBB through the release
and activation of matrix metalloproteinases (MMPs) [2].
Tissue plasminogen activator (tPA) is therapeutic for dissolv-
ing clots resulting from stroke and works by activating plasmino-
gen to plasmin that then digests the fibrin in clots. However,
plasmin has a damaging secondary side effect of activating MMPs
and may thus disrupt the BBB [3]. MMPs have been shown to
mediate stroke injury by destabilizing the BBB and aggravating
cerebral damage through stimulation of inflammatory responses
[4–6]. Loss of a properly functioning BBB leads to edema, hemor-
rhaging, and the introduction of neurotoxic elements to the brain
environment. MMP inhibition has been shown to be a promising
strategy in stroke treatment and recovery [2].
To study stroke and other cerebral pathologies, a vast array of
BBB models have been developed. BBB cellular models have been
employed for drug discovery and development [7, 8]. The com-
plexity of the BBB presents a plethora of options for constructing
a BBB model to varying degrees of physiological accuracy. BBB
models range in complexity from simple monocultures of brain
endothelial cells to three-dimensional co-cultures [9–11]. Early
models used only monocultures and proved to be inadequate in
replicating the BBB. Most current models are a variation of brain
endothelial cells grown in culture with astrocytes or grown in
astrocyte-conditioned medium. More sophisticated models may
even incorporate additional components of the neurovascular unit,
including pericytes and/or neurons [12–14]. More complex mod-
els incorporating multiple cell types more closely mimic in vivo
BBB and achieve better performance in terms of reduced mono-
layer permeability. However, the complexity of such systems makes
them more inconvenient, especially for screening applications.
The underlying challenge of BBB model design is striking a
balance between simplicity and performance- producing a model
that is physiologically accurate and relevant, easy to construct and
adapt, and is economically feasible. We have created a simple model
that reasonably approximates BBB physiology, is suitable as a
screening tool, and may be used to screen libraries of MMP inhibi-
tors for initial phases of drug development. The model described
within is formed from bovine brain microvascular endothelial cells
(bBMVEC) and normal human astrocytes (NHA) grown on oppo-
site sides of an artificial basement membrane-coated porous mem-
brane insert. In the model presented, tPA is used to induce BBB
disruption. The integrity of the BBB model was assessed by
measuring the passage of the fluorescent tracer Lucifer yellow

Cellular BBB Model for MMP Inhibitor Screening 289
between the cell layer, building on the examples of Culot et al. and
Megard et al. [15, 16]. We selected bBMVEC for relative ease of
culture and included astrocytes (see Note 1) because of their sig-
nificant contribution to barrier properties in endothelial cells such
as BBB formation and integrity [1, 17, 18]. The economic benefit
of our model is twofold. First, the permeability of cells can be sim-
ply quantified by the use of Lucifer yellow and a fluorescence plate
reader, a piece of equipment that is readily available and thus avoids
the use of equipment dedicated to measurement of transendothe-
lial electric resistance. Second, our choice of using primary cell
lines instead of freshly harvested cells avoids continual animal sac-
rifice for each experiment and the attendant difficulties and costs.
2 Materials
2.1 Cell Culture 1. Type 1 rat tail collagen.

2. 0.02 N acetic acid.
3. Tris-buffered saline (TBS): 6.0 g Trizma-base, 16 g NaCl,
0.76 g KCl, 0.2 g Na2HPO4, 0.6 mL phenol red (0.5% solu-
tion in H2O) in 2 L of water. Add HCl to pH 7.4.
4. Clonetics® bovine brain microvascular endothelial cells (bBM-
VEC) (Lonza, Walkersville, MD, USA).
5. EBM-2 medium supplemented with ascorbic acid, endothelial
cell growth factor beta, platelet-poor horse serum, heparin,
and penicillin/streptomycin/fungizone.
6. Clonetics® normal human astrocytes (NHA) (Lonza,
Walkersville, MD, USA).
7. ABM medium supplemented with human recombinant epider-
mal growth factor, insulin, ascorbic acid, l-glutamine, gentami-
cin/amphotericin, and fetal bovine serum (FBS).
8. Transparent, 0.4 μm pore polyethylene terephthalate (PET) or
0.4 μm pore polytetrafluoroethylene (PTFE) membrane inserts
(see Note 2).
9. Sterile box with loose fitting lid.
10. Sterile multi-well culture plates corresponding to size of mem-
brane inserts with lids.
11. Human plasma fibronectin.
2.2 Permeability 1. Ringer’s-HEPES buffer: 8.77 g (150 mM) NaCl, 392.9 mg

Testing (5.2 mM) KCl, 323.4 mg (2.2 mM) CaCl2, 0.0406 mg (0.2
mM) MgCl2, 0.504 mg (6 mM) NaHCO3, 0.5045 mg (2.8
mM) glucose, 1.192 g (5 mM) HEPES in 1 L of water. Adjust
pH to 7.4.
2. Lucifer yellow dye.

3. Human tPA.
4. MMP inhibitor of choice.
5. Multi-well cell culture plate.
6. 96-well fluorescence assay plate with black walls and a clear
bottom.
7. Aluminum foil.
8. Microtiter plate reader.
3 Methods
3.1 Cell Culture All experiments were performed in a biological safety cabinet at
room temperature unless otherwise stated. Inserts are easily punc-
tured by pipet tips; extreme care must be taken to avoid contact
with the membrane at each handling.
1. Prepare a stock solution of Type 1 rat tail collagen dissolved in
0.02 N acetic acid. Using this stock makes a dilution of suffi-
cient concentration to provide 15 μg collagen/cm2 of flask
surface. Add the collagen solution to the culture flask and
incubate at 37°C for 1 h (see Note 3). Remove the collagen
solution and rinse the flask with TBS and then with warm
medium.
2. Expand bBMVEC culture by plating at 5,000 cells/cm2 in
collagen-treated flask. Culture cells until 75–80% confluence
and then harvest and freeze using method of choice. This will
be a stock for future experiments. Freeze in multiple small
batches such that each aliquot provides enough cells for a sin-
gle experiment (see Note 4).
3. Expand NHA culture by plating between 5,000 cells/cm2 in
an untreated flask. Culture cells until 75–80% confluence and
then harvest and freeze using method of choice. This will be a
stock for future experiments. Freeze in multiple small batches
such that each aliquot provides enough cells for a single experi-
ment (see Note 4).
4. Three to four days prior to starting the co-culture, thaw an
aliquot of bBMVEC into a collagen-coated culture flask and
allow to recover with regular medium changes.
5. One day prior to starting the co-culture, prepare a dilution of
Type 1 rat tail collagen dissolved in sterile H2O to a sufficient
concentration to deliver 15 μg collagen/cm2 of membrane
insert surface. Apply collagen solution to the apical surface
inside the well of the membrane insert and allow to dry uncov-
ered overnight (see Note 5). After drying, rinse with TBS (see
Note 6).
6. When ready to begin co-culture, prepare a stock solution of

human plasma fibronectin dissolved in unsupplemented EBM
medium. Make a diluted fibronectin solution in unsupple-
mented EBM medium of sufficient concentration to provide 5
μg fibronectin/cm2 of membrane insert surface (see Note 7).
Apply the fibronectin solution to the apical side of the mem-
brane insert and incubate partially covered at 37°C for 6 h and
then rinse the surface with TBS (see Note 8).
7. Place membrane inserts in the sterilized box so that they are
inverted with the basal side exposed. Seed NHA directly from
thaw on basal side of membrane insert at 50,000 cells/cm2.
Slowly add the cells in small volumes to form a droplet on the
surface (Fig. 1). If the surface tension bursts, collect the cell
solution and reapply. Close the lid of the box and carefully
Fig. 1 Construction of co-culture on opposite sides of cell culture membrane

inserts. (a) Schematic illustrating the application of normal human astrocytes
(NHA) on the basal side and bovine brain microvascular endothelial cells (bBM-
VEC) on the apical side of porous polyethylene terephthalate (PET) membrane
inserts. Membrane inserts were initially inverted before depositing NHA. NHA
were allowed to attach to the membrane for 45 min in an incubator. The mem-
brane inserts were then flipped into a multi-well plate. After allowing two to 3
days for NHA growth in an incubator, bBMVEC were seeded onto the apical mem-
brane surface. (b) Picture showing the NHA cell solution droplets on the basal
side of the membrane inserts
place it in an incubator at 37°C and 5% CO2. Cells will attach

in approximately 45 min.
8. After 45 min, use forceps to carefully flip membrane inserts
into a multi-well plate containing warmed NHA-specific ABM
supplemented medium. Allow to grow for 48–72 h with
medium changes every 48 h (see Note 9).
9. After the prescribed NHA growing period, transfer membrane
inserts to new wells containing fresh NHA-specific medium.
Harvest bBMVEC from recovery flask and seed at 50,000
cells/cm2 on the apical membrane surface in bBMVEC-specific
EBM supplemented medium. Every 48 h, exchange EBM sup-
plemented medium in the membrane insert and ABM supple-
mented medium in the well underneath.
10. Check the co-culture to see if it is fit for assay by observing the
endothelial cells under a microscope (see Note 10). Endothelial
cells display a characteristic cobblestone morphology and
approach a confluent monolayer within 48–72 h of seeding.
3.2 Permeability All experiments were performed in a biological safety cabinet at

Testing room temperature unless otherwise stated. The experimental
workflow consisted of treatment with tPA (with or without MMP
inhibitor) followed by the addition of tracer dye to measure per-
meability (Fig. 2).
1. Make a stock tPA solution by diluting tPA in serum-free
medium to around 1 μM. Keep in mind that the final tPA con-
centration in the assay will be 100-fold lower and you will need
to pipet 1/100 the volume.
2. Dilute inhibitor in recommended buffer to 200 μM. Final con-
centration in assay will be 10 μM (see Note 11).
3. Remove medium from membrane inserts to be tested, rinse
once with TBS, and replace with serum-free medium, DMSO
to 2.5%, and enough stock tPA for a final tPA concentration of
10 nM. Add the appropriate amount of MMP inhibitor.
Incubate at 37°C and 5% CO2 for 6 h (see Note 12).
4. During the last 30 min of incubation, dissolve the Lucifer yel-
low in Ringer’s-HEPES buffer at pH 7.4 to make standard
solutions of 200, 100, 50, 20, 10, 1, 0.5, 0.25, 0.1, 0.05,
0.01, and 0 μM from a stock concentration of 40 mM. Make a
sufficient quantity of 100 μM Lucifer yellow to use as both
tracer and for the standards (see Note 13).
5. Remove the tPA solution from the membrane inserts and
replace with 400 μL of 100 μM Lucifer yellow solution (see
Note 14).
6. Move the co-culture membrane inserts to new, sterile multi-
well plates containing a volume of transport buffer in the well
Fig. 2 Experimental workflow for testing the effect of MMP inhibition on monolayer permeability. The integrity
of the monolayer formed in the co-culture was assessed by its permeability to Lucifer yellow fluorescent
tracer. The effect of MMP inhibitors of the integrity of the monolayer was assessed after disrupting the mono-
layer with tissue plasminogen activator (tPA). First, cells were incubated with a treatment solution (tPA, tPA +
inhibitor, inhibitor, etc.) for 6 h. Then, the treatment solution was replaced with a solution of Lucifer yellow. After
the introduction of the tracer, the membrane insert and the well below were sampled at various time points up
to 1 h. The passage of Lucifer yellow between the cells was determined by fluorescence spectroscopy
that ensures that the liquid level inside and outside the insert is
roughly equal (see Note 15).
7. Immediately remove 100 μL aliquots from the well and the
membrane insert and place in the 96-well assay plate. Then
add 100 μL of fresh transport buffer to the membrane
compartment. Repeat sampling at desired time points, keep-
ing the culture in the incubator between time points (see
Note 16).
8. Measure the fluorescence of the samples using a microtiter
plate reader. Excite at 485 nm and record emission at 538 nm.
Calculate the concentration of Lucifer yellow in the sample
using a calibration curve.
9. Calculate the permeability of the cell model as follows using

the % leakage and permeability coefficient, Pe, and mass bal-
ance (see Notes 17 and 18).
(a) % Leakage = moles of dye in well ÷ moles of dye intro-
duced into membrane insert.
(b) Pe = (moles of dye in well ÷ molarity of dye introduced
into membrane insert) ÷ (surface area × time).
(c) Mass balance = (moles of dye in well + moles of dye in
membrane insert) ÷ moles of dye introduced into membrane
insert.
4 Notes
1. Bovine astrocytes are not commercially available therefore

human astrocytes were substituted.
2. Membrane inserts can be 6-well, 12-well, or 24-well format.
All have been tested, but the 24-well format was found to be
easier to work with and more consistent in producing conflu-
ent monolayers. The multi-well format of this assay also allows
us to scale up the experiments with relative ease so that the
proportion of available membrane inserts for testing increases.
Furthermore, using smaller culture membrane inserts enables
experiments to be scaled down to conserve precious samples.
3. Make a sufficient volume of diluted collagen so that the entire
surface of the flask is in contact with the solution. For a T75
flask, this was about 7 mL.
4. Expanding and freezing cells ensures that early passage cells
(no higher than passage 4) are used for each experiment. Cells
passaged more than four times are not ideal for co-culture.
Only early passage bBMVEC can be seeded at this density that
is lower than recommended. Cells were frozen in a 1:1 mixture
of a 50% serum solution in medium with a 50% serum, 5%
DMSO solution in medium, accounting for the amount of
serum already added to the supplemented medium.
5. For a membrane insert with 0.3 cm2 surface area, 50 μL of col-
lagen solution was applied.
6. Coated inserts may be stored until needed if they have not been
rinsed with TBS. Rinse stored inserts with TBS before use.
7. The importance of collagen is documented [19]. The attach-
ment and growth of the bBMVEC on cell culture membrane
inserts was optimized when both 15 μg collagen/cm2 and 5 μg
fibronectin/cm2 were applied.
8. The fibronectin solution may be applied in advance of seeding
NHA. Store the membrane inserts with the fibronectin solu-
tion at 4°C until ready for use.
9. If using a transparent membrane insert, the NHA can be seen

on the basal surface with phase-contrast microscopy. If using
translucent membrane inserts, the NHA will not be visible.
10. Visual observations of the attachment and proliferation of cells
can be used to determine when to proceed with the co-culture.
In co-culture, monolayer formation and integrity was maximal
at 5–6 days with visible monolayer disruption and shedding
beginning 8–11 days after seeding.
11. The inhibitor tested was dissolved in DMSO. After dilution
the DMSO concentration was 2.5%. Inhibitor was delivered in
25 μL aliquots.
12. Consider the amount of solvent introduced in the diluted
inhibitor so that an equivalent amount is added to wells that
will not be treated with the inhibitor. If the recommended
insert volume is 500 μL, then 3.25 μL of 1.54 μM tPA + 0.625
μL of solvent will be added to 496.1 μL of serum-free medium.
For studies on the activation of pro-MMPs, investigators may
wish to add plasminogen together with tPA [3].
13. Making the standards by serial dilution is the easiest and most
precise method. The amount of each standard is based on the
quantity needed during fluorescence measurements plus the
amount needed to make the serial dilution. Since 100 μM
Lucifer yellow is also used as the tracer, the number of mem-
brane inserts that will be tested must also be considered.
14. 100 μM Lucifer yellow gave the most consistent results
15. For a 24-well plate, 600 μL of buffer was placed in the well and
400 μL delivered to the membrane insert. Keeping the liquid
levels balanced between the well and insert reduces hydrostatic
pressure.
16. When testing multiple membrane inserts it is easier to stagger
the timing so that each individual membrane insert does not
start its time course at the same moment. Protect the collected
samples from light and keep chilled until taking the fluores-
cence readings.
17. The permeability coefficient, Pe, provides a way to quantify the
integrity of the co-culture monolayer and the calculated mass
balance provides a check on potential losses of tracer to the
interior of cells or to the membrane insert itself. Inserts with a
mass balance value below a user-determined threshold were
discarded. The Pe equation was modified from [16].
18. Achieving uniformity in any cell culture assay is challenging.
The variability in the growth and surface coverage of cells
within a set of membrane inserts was a significant difficulty.
This prevented the membrane inserts from reaching confluence
at the same time. To compensate for this problem, co-cultures
used for fitting were selected from a pool of membrane inserts
based on their growth and monolayer progression.
Acknowledgments
This work was completed at the Protein Expression Facility of the

Institute of Molecular Biophysics at Florida State University and
supported by the James and Esther King Biomedical Research
Program grant (1KF04) and a Florida State University Research
Foundation GAP grant to Dr. Sang.
References
1. Abbott NJ, Ronnback L, Hansson E (2006) 10. Naik P, Cucullo L (2012) In vitro blood-brain
Astrocyte-endothelial interactions at the blood- barrier models: current and perspective tech-
brain barrier. Nat Rev Neurosci 7(1):41–53 nologies. J Pharm Sci 101(4):1337–1354
2. Roycik MD, Myers JS, Newcomer RG, et al. 11. He Y, Yao Y, Tsirka SE, et al. (2014) Cell-
(2013) Matrix metalloproteinase inhibition in culture models of the blood-brain barrier.
atherosclerosis and stroke. Curr Mol Med Stroke 45(8):2514–2526
13(8):1299–1313 12. Nakagawa S, Deli MA, Kawaguchi H, et al.
3. Niego B, Medcalf RL (2014) Plasmin- (2009) A new blood-brain barrier model using
dependent modulation of the blood-brain bar- primary rat brain endothelial cells, pericytes and
rier: a major consideration during tPA-induced astrocytes. Neurochem Int 54(3–4):253–263
thrombolysis? J Cereb Blood Flow Metab 13. Hatherell K, Couraud P, Romero IA, et al.
34(8):1283–1296 (2011) Development of a three-dimensional,
4. Gasche Y, Soccal PM, Kanemitsu M, et al. all-human in vitro model of the blood–brain
(2006) Matrix metalloproteinases and diseases barrier using mono-, co-, and tri-cultivation
of the central nervous system with a special Transwell models. J Neurosci Methods
emphasis on ischemic brain. Front Biosci 199(2):223–229
11:1289–1301 14. Xue Q, Liu Y, Qi H, et al. (2013) A novel brain
5. Copin J, Bengualid DJ, Da Silva RF, et al. neurovascular unit model with neurons, astro-
(2011) Recombinant tissue plasminogen cytes and microvascular endothelial cells of rat.
activator induces blood–brain barrier break- Int J Biol Sci 9(2):174–189
down by a matrix metalloproteinase-9-inde- 15. Megard I, Garrigues A, Orlowski S, et al.
pendent pathway after transient focal cerebral (2002) A co-culture-based model of human
ischemia in mouse. Eur J Neurosci 34(7): blood-brain barrier: application to active trans-
1085–1092 port of indinavir and in vivo-in vitro correla-
6. Candelario-Jalil E, Yang Y, Rosenberg GA tion. Brain Res 927(2):153–167
(2009) Diverse roles of matrix metalloprotein- 16. Culot M, Lundquist S, Vanuxeem D, et al.
ases and tissue inhibitors of metalloproteinases (2008) An in vitro blood-brain barrier model
in neuroinflammation and cerebral ischemia. for high throughput (HTS) toxicological
Neuroscience 158(3):983–994 screening. Toxicol In Vitro 22(3):799–811
7. Cecchelli R, Berezowski V, Lundquist S, et al. 17. Siddharthan V, Kim YV, Liu S, et al. (2007)
(2007) Modelling of the blood-brain barrier in Human astrocytes/astrocyte-conditioned
drug discovery and development. Nat Rev medium and shear stress enhance the barrier
Drug Discov 6(8):650–661 properties of human brain microvascular endo-
8. Cucullo L, Hossain M, Rapp E, et al. (2007) thelial cells. Brain Res 1147:39–50
Development of a humanized in vitro blood- 18. Kuo YC, Lu CH (2011) Effect of human astro-
brain barrier model to screen for brain penetra- cytes on the characteristics of human brain-
tion of antiepileptic drugs. Epilepsia microvascular endothelial cells in the
48(3):505–516 blood-brain barrier. Colloids Surf B
9. Deli MA, Abraham CS, Kataoka Y, et al. (2005) Biointerfaces 86(1):225–231
Permeability studies on in vitro blood-brain 19. Cecchelli R, Dehouck B, Descamps L, et al.
barrier models: physiology, pathology, and (1999) In vitro model for evaluating drug
pharmacology. Cell Mol Neurobiol transport across the blood-brain barrier. Adv
25(1):59–127 Drug Deliv Rev 36(2–3):165–178
Part VII
Matrix Metalloproteases as Biomarkers

Chapter 17
Matrix Metalloproteases as Biomarkers of Disease

Fernando Luiz Affonso Fonseca, Beatriz da Costa Aguiar Alves,
Ligia Ajaime Azzalis, and Thaís Moura Gáscon Belardo
Abstract
Matrix metalloprotease play a vital role in many cellular processes. Dysfunction in activity of these enzymes
has been implicated in the pathogenesis of a number of diseases. Factors that affect the balanced interac-
tion between MMPs and their inhibitors, such as genetic mutations of extracellular matrix components or
dysregulation of MMP expression, can lead to various diseases. Due to their essential role in ECM remod-
eling, MMPs have become targets of interest as biomarkers for the diagnosis and prognosis of diseases
associated with alterations of the ECM.
Key words Matrix metalloproteases, Disease, Cancer, Inflammatory disease
1 Introduction
1.1 Definition, The space between the cells that form all tissues and organs is filled
Classification, with a series of compounds and macromolecules that constitute the
and Roles of MMPs extracellular matrix (ECM). Therefore, the ECM represents the
noncellular components of all tissues and organs. The ECM has tis-
sue-specific characteristics; however, it is basically composed of water
and relatively stable macromolecules, which are not only produced,
exported, and bound by cells but also responsible for the modula-
tion of their structure, physiology, biochemical interactions, and the
biomechanics of tissues. Such molecules, essential for morphogene-
sis, differentiation, and tissue homeostasis, are classified into three
categories: glycosaminoglycan polysaccharide chains, fibrous pro-
teins with structural roles (collagens and elastins), and non-fibrous
proteins with adhesion properties (laminins and fibronectins) [1].
Extracellular matrix remodeling is a dynamic and continuous
process in which the cell deposits new components and degrades
old ones by proteolytic action [2]. ECM degradation is essential
not only for processes like embryonic development, morphogene-
sis, reproduction and tissue reabsorption and remodeling [3] but
299
300 Fernando Luiz Affonso Fonseca et al.
also for the regulation of its quantity, composition, and structure.

Moreover, ECM degradation plays an important role in the libera-
tion of various biologically active molecules, like growth factors
and hormones. The main enzymatic family that takes part in this
degradation mechanism is the matrix metalloproteinases (MMPs).
MMPs form part of a larger family, the metzincin enzymes, along
with the adamalysins, which encompass the ADAMs disintegrin
and metalloproteinase, the ADAMTS (ADAMs with thrombos-
pondin motifs), and the astacins (meprins) [4].
Metalloproteinases are zinc-dependent enzymes secreted by
cells or bound to the cell membrane. They all have many functional
domains in common, such as the signal peptide domain, the pro-
peptide domain, the hemopexin-like domain (excluding MMP-7,
MMP-23, and MMP-26), and the catalytic domain [3, 4]. The
catalytic domains of MMPs from a wide range of species display a
high degree of sequence homology [5]. So far, a total of 25 MMPs,
including 24 vertebrate MMPs, have been identified. This includes
two alternatively spliced isoforms of MMP-23, namely MMP23A
and MMP23B.
In the human genome, the gene cluster that contains MMP-1,
-3, -7, -8, -10, -12, -13, -20, and -27 is found on chromosome 11
(11q22). The remaining MMPs are located on different chromo-
somes, such as chromosomes 1, 8, 12, 14, 16, 19, and 20 [5].
Based on characteristics like substrate specificity (Table 1),
sequence similarity, and domain organization, MMPs can be classi-
fied into six groups: (a) collagenases (MMP-1, -8, -13, -14, and
-18), which are capable of cleaving triple helical collagen types I,
II, and III. Collagenases are commonly found in tissues, at cell-cell
and cell-matrix contact sites, and they disappear in the event of
evolving fibrosis; (b) gelatinases (MMP-2 and -9), which degrade
collagen fragments released by collagenases and also cleave colla-
gen IV and elastin; (c) stromelysins (MMP-3, -7, -10, -11, -19,
and -20), which act on the proteoglycans, binding proteins, fibro-
nectins, laminins, and collagen types III and IV; (d) matrilysins
(MMP-7 and -16), which degrade versican, elastin, fibronectin,
collagen type IV among others; (e) membrane-bound MMPs
(MT-MMPs), exemplified by MMP-1, -3, -2, -5, -15, -16, -17,
-24, -25; and (f) other MMPs [5].
Most MMPs are secreted in an inactive form (zymogens), and
are subsequently activated in the extracellular space [6]. The main
MMP activation mechanism is mediated by cell surface associated
plasmin, which is generated from plasminogen by the activity of
either tissue-type plasminogen activator (t-PA) or urokinase-type
plasminogen activator (u-PA). The main MMPs activated by this
pathway are pro-MMP-1, pro-MMP-3, pro-MMP-7, pro-MMP-9,
pro-MMP-10, and pro-MMP-13. Once activated, these MMPs
can activate other MMPs [7].
Matrix Metalloproteases as Biomarkers of Disease 301
Table 1
Substrates cleaved by various MMPs and their chromosomal location
MMP Substrate Chromosome

1 Collagen types I, II and III 11
2 Collagen IV, V, VII, X, fibronectin, proteoglycans, elastin 16
3 Procollagen, proteoglycans, collagen X and XI, laminin, elastin, 11
fibronectin, plasminogen, pro-MMP-9
7 Collagen IV, gelatin, fibronectins, laminins, elastin, transferrin, casein, 11
plasminogen
8 Proteoglycans, fibronectin 11
9 Plasminogen, collagen IV, V, VII, and X, gelatin, elastin, fibronectin 20
10 Stromelysin, gelatin, casein, fibronectin 11
11 IGFPB-1, collagen, laminin, elastin, fibronectin, casein, proteoglycans
12 Plasminogen, elastin, collagen IV, fibronectin, gelatin, casein, laminin, 11
fibrinogen
13 Collagen I, plasminogen, fibronectin, proteoglycan 11
14 Collagen I, CD44, fibronectin, laminin, pro-MMP-2 14
15 Cell-surface transglutaminase, fibronectin, gelatin, laminin, 16
pro-MMP-2
16 Cell-surface transglutaminase, collagen III, gelatin, casein, fibronectin, 8
pro-MMP-2
17 Fibrinogen, gelatin, pro-MMP-2 12
18 Collagen I 12
19 Collagen I and IV, fibronectin, gelatin, laminin 12
20 Amelogenin 11
24 Fibronectin, pro-MMP-2, proteoglycans, gelatin 20
25 Collagen IV, gelatin, fibronectin, pro-MMP-2
23 Gelatin 1
(A and B)
26 Collagen IV, fibronectin, fibrinogen, casein, insulin-like growth 11
factor-binding protein 1, pro-MMP-9
Metalloproteinases can collectively cleave all components of

the ECM [4]. The proteolytic activity of MMPs is controlled at
different levels by regulation of gene expression (transcriptional
regulation), activation of pro-enzymes, and inhibition by tissue
inhibitors of metalloproteinases (TIMPs) or other inhibitors, such
as α2-macrobulin. MMPs are not constitutively expressed but
expressed after stimulation by cytokines and growth factors [3].
Normal tissue structure partly results from the balanced interac-

tion between MMPs, α2-macrobulin, and TIMPs [8]. To date,
four types of TIMPs have been identified (TIMP-1 to TIMP-4),
while TIMP-3 is the main regulator of MMP activity (including
ADAMs and ADAM-TSs) [3].
Genetic mutations of extracellular matrix components that
affect the balanced interaction between MMPs and their inhibitors
can cause various diseases, such as inflammation, ulceration, arthri-
tis, periodontitis, vascular disease, diabetes, fibrosis, emphysema,
cancer (onset, vascularization, and metastasis), apoptosis, sepsis,
intestinal defense activation, and nervous system diseases like
stroke and Alzheimer’s [3, 4]. Due to their essential role in ECM
remodeling, MMPs have become targets of interest as biomarkers
for the diagnosis and prognosis of diseases associated with altera-
tions of the ECM.
2 Metalloproteinases and Malignant Hematological Diseases
The action of metalloproteinases in various malignant hematologi-

cal diseases sheds light on the role of the ECM in different blood
pathologies.
Stromal cell activity plays a role in malignant transformation,
the onset and progression of different kinds of malignant hemato-
logical diseases (clonal disorders resulting from the neoplastic
transformation of blood progenitor cells), and solid tumors. While
studies of solid and hematologic cancers focus on the analysis of
gene mutations and malignant lesions due to a gain or loss of bio-
logical function it is important to study the role of MMPs in the
malignant blood pathologies [9].
Two bone marrow niches play an essential role in several malig-
nancies: the endosteal niche with its resident osteoblasts that are
derived from mesenchymal stem cells, and the vascular niche, home
for hematopoietic stem cells. Furthermore, the extracellular matrix
is fundamentally involved in tumorigenesis not only by directing
the differentiation of these pluripotent stem cells but also due to
the participation of metalloproteinases [10].
Metalloproteinases are synthetized by different types of cells
and tissues. Dermal fibroblasts and leukocytes are the main sources
of metalloproteinases, especially MMP-2 while platelets are known
to stimulate the production of MMP-1, -2, -3, and -14. MMPs are
also expressed in vascular cells and tissues, such as smooth muscle
cells. Alterations in the expression, distribution, or activity of
MMPs may impair cell or tissue functionality, thus contributing to
the development of malignant hematological diseases [11].
The bone marrow microenvironment is of paramount impor-
tance for the regulation and maintenance of hematopoietic stem
cells so as to prevent the onset of hematological neoplasias.
Different cell types reside within the bone marrow microenviron-
ment, including macrophages, endothelial progenitors, and osteo-

blasts, and these are responsible for the maintenance of
hematopoietic pluripotent cells that are located close to endothe-
lial cells and the vascularized area. During hematopoietic stem cell
transplantation, proteolytic enzymes released from neutrophils,
including several metalloproteinases, cleave granulocyte colony
stimulating factor (G-CSF) that then stimulates CD34+ cells. This
also results in metalloprotease-mediated extensive remodeling of
the bone marrow environment [12].
Adult stem cells are characterized by their capacity for self-
renewal, maintaining bone marrow homeostasis through cellular
proliferation and differentiation via processes that are under strict
control. Deletion of genes involved in the maintenance of quies-
cent cell cycle status and dysregulated proliferation of hematopoi-
etic stem cells results in the loss of cell functionality. It is important
to point out that quiescent and proliferative cells can be found in
the same niche [13].
Angiogenesis comprises the formation of new vascularized net-
works that contribute to the growth of tumors through the trans-
port of neoplastic cells to secondary sites via lymph and blood
vessels, which is of great importance in various onco-hematological
diseases. Angiogenesis involves the continuous proteolytic modifi-
cation of the ECM and cleavage of various cell-surface molecules
that are released into the microenvironment of the tumor through
tumor infiltrating leukocytes. This process is mediated by various
proteases including several metalloproteinases, predominantly
MMP-9 [14].
Precursor cells of megakaryocytes, platelets, and their mic-
roparticules (MP) are involved in various physiological and path-
ological events that contribute to the aggressiveness and
persistence of tumors. This leads to complications in the treat-
ment of inflammation, vascular complications, and various malig-
nant hematological diseases. Platelet apoptosis exceeding normal
physiological levels promotes the formation of microparticules
(MP). Platelet-derived MPs (PMPs), which constitute around
80% of circulating MPs, express adhesion molecules and are
responsible for the promotion of angiogenesis and dissemination
of neoplasias. These PMPs along with growth factor and metal-
loproteinases show a combined action with vascular endothelial
growth factor (VEGF) [15].
Acute lymphoblastic leukemias are the most common (99%)
form of lymphoid leukemia and for this reason are commonly diag-
nosed as acute lymphoid leukemia. In lymphoblastic leukemias
there is an increase in the density of microvessels in the bone mar-
row. While in acute leukemias, VEGF-A is considered the main
dominant inducer of blood vessel growth. Matrix-bound VEGF-A
is cleaved and released by several MMPs, especially gelatinases A
and B (denoted as MMP-2 and MMP-9, respectively), which
hydrolyze basement membrane collagens of blood vessels, stimu-

lating extramedullary infiltration by lymphoblasts. VEGF-A stimu-
lated secretion of MMPs in Pre-B (B) and pre-T (T) lineage cells is
currently used for disease prognosis in patients [16].
Metalloproteinases also play an important role in chronic lym-
phocytic leukemias, where there is an increase of CD5 B lympho-
cytes in the peripheral blood. Gelatinases B/(pro-) MMP-9, which
are involved in the migration and survival of these malignant
hematologic cells, have been shown to speed up disease progres-
sion. (Pro-) MMP-9, whose expression is regulated by cytokines
and other factors, is detected on the surface of LLC cells and accu-
mulates in these cells [17].
Mutations in Notch 1–4 genes may cause malignancy in
mature B cells (via inhibition of the maturation of common lym-
phoid progenitor cells), chronic lymphocytic leukemia, Hodgkin’s
lymphoma, Burkitt’s lymphoma, and multiple myeloma. The lat-
ter is a peculiar onco-hematologic disease that depends on inter-
action with the bone marrow microenvironment, where
metalloproteinases are responsible for the interrelation of the
stromal tumor, tumor invasion, and metastasis. This interrelation
is characterized by the production of high levels of MMP-1 and
MMP-2, which in turn generate active MMP-9 and MMP-2 that
degrade collagen IV, the major structural component of the base-
ment membrane [18].
Chronic myeloid leukemia (CML) is a myeloproliferative dis-
order that is characterized by the Philadelphia (Ph) chromosome,
a BCR-ABL fusion gene resulting from a specific translocation t
[19, 20], which leads to the expansion of immature leukemic cells.
MMP-9 is known for its ability to cleave specific components of
the extracellular matrix, including collagen (types IV, V, and XI)
and elastin. MMP-9 is also involved in angiogenesis, demonstrat-
ing that MMPs and TIMPs play a key role in hematopoiesis once
they are secreted by stromal and tumor cells [9, 20, 21].
Chronic myeloproliferative disorders are characterized by the
progressive remodeling of the bone marrow stroma, highlighted
by an increase in the deposition of proteins in the extracellular
matrix, neoangiogenesis and the substitution of hematopoietic
cells for fibrous tissue [22]. However, in CML the participation of
MMPs is not limited to the extracellular matrix. They are also
involved in the stimulation of various cytokines in different cell
types and tissues through a specific MMP9-induced pathway [21].
MMPs are involved in coagulation in sepsis and inflammatory
responses, and in malignant hematological diseases, like multiple
myeloma and may play a role in disseminated intravascular coagu-
lation (DIC). It has been suggested that impairment of the interac-
tion of MMP-8 with TIMP-1, which are involved in the activation
of the tissue factor pathway inhibitor (TFPI), is essential for trig-
gering DIC. MMP-8, also known as collagenase-2, is a protein
stored in neutrophil granules (as well as in smaller quantities in

activated macrophages and endothelial cells) that acts via neutro-
phil chemotaxis in the inflammation process [23].
Studies are currently underway to better understand the rela-
tionship between MMP expression and polymorphisms associated
with various malignant hematological diseases, to properly target
medications for the treatment of these diseases. Detection of the
deregulation of MMPs is of utmost importance from the clinical
perspective, since it can be used as a diagnostic and prognostic
tool, or as a new oncologic therapy. In addition, monitoring MMP
regulation may be useful in hematopoietic stem cell transplantation
where MMPs induce hematopoietic cell mobilization [9].
3 Metalloproteinases and Inflammatory Disorders
Whenever tissue injury occurs, modifications ensue which elimi-

nate the offending agent, limit tissue damage, and regenerate the
site of the lesion. These modifications depend on the regulation of
proteins (inflammatory markers) involved in the inflammatory
response [24, 25]. Matrix metalloproteinases (MMPs) figure
among these many types of inflammatory markers. MMPs are a
family of zinc-dependent endopeptidases capable of degrading
extracellular matrix components and proteins of the basement
membrane, which actively participate in tissue remodeling, both
physiologically and pathologically [26].
The analysis of MMPs as inflammatory biomarkers can help
not only in the diagnosis but also in the treatment of some inflam-
matory diseases [27]. Some of the diseases in which MMPs could
be used as biomarkers are described below.
3.1 Ankylosing Ankylosing Spondylitis (AS) is a chronic rheumatic disease that

Spondylitis (AS) first affects the spine and sacroiliac joints. Rheumatoid arthritis
and Rheumatoid (RA) is an autoimmune disease that most commonly affects joints;
Arthritis (RA) however, it is actually a systemic disease that may involve other
organs [28].
MMP-2 and MMP-9 levels are increased in the synovial liquid
of patients with RA [6, 7]. It has lately been observed that high
serum levels of active MMP-3, both in animal and human models,
is also a good marker for AS and RA owing to the fact that MMP-3
is more specifically localized in joints and suffers less from systemic
factors [28–30].
3.2 Sepsis Some studies show a direct correlation between increased levels of
MMPs and low survival rates of patients with sepsis [31]. Aerts
et al. [32] have recently demonstrated that a hydroxypyrone-based
inhibitor selective for MMP-12 displays anti-inflammatory activity
in mice.
3.3 Chronic COPD is characterized by progressive obstruction of the airway

Obstructive Pulmonary associated with chronic inflammation. MMP-1, MMP-8, MMP-
Disease (COPD) 12, TIMP-1 (tissue inhibitor of metalloproteinase 1), and espe-
cially MMP-9 have been shown to be associated with COPD [33].
3.4 Conjuncti Despite the fact the causes of CCh are still uncertain, it is very
vochalasis (CCh) likely that the accumulation of MMPs in the tears of these patients
may contribute to the progression of the disease [34]. In addition,
the effectiveness of surgery to treat CCh has been shown to be
associated with a postoperative decrease in MMP-9 levels in tears.
In fact, many authors suggest that the quantification of MMP-9
could be important for the evaluation of many eye diseases [34].
3.5 Coronary Heart The inflammatory process seems to play a role in the pathogenesis
Disease (CHD) of vascular diseases. Markers of inflammation like C-reactive pro-
tein, fibrinogen, and interleukin-6 are associated with the risk of
developing CHD [35]. However, other markers, such as MMP-9,
also need to be accessed [36, 37].
3.6 Inflammatory Ulcerative colitis (UC) and Crohn’s disease (CD) are autoimmune
Bowel Disease (IBD) diseases characterized by inflammation of the gastrointestinal sys-
tem. Although MMPs are present in IBD, their role is not clear
and further studies are required [38]. The colon of patients with
IBD exhibit high levels of MMP-1, -3, -7, -9, -10, and -12 [39],
and intestinal biopsies of patients with UC demonstrate that vascu-
lar endothelial cells and leukocytic infiltrate produce MMP-7 and
-13 [40]. MMP-9 serum levels are correlated to both CD and UC
disease, with higher levels detected in the latter. This difference in
levels between these MMPs could help to differentiate between
UC disease and CD [41].
Future studies should elucidate the roles of MMPs and tissue
inhibitors of metalloproteinases (TIMPs), thus contributing to the
diagnosis and treatment of inflammatory diseases [42, 43].
4 Metalloproteinases and Solid Tumors
Interactions between tumor cells and their microenvironment

reveal the important role played by matrix metalloproteinases
(MMPs) during carcinogenesis. Tumor growth and metastasis
depend on cell-cell and cell-matrix interactions, as well as on tis-
sue modifications through the action of proteolytic enzymes. The
MMP family is composed of 25 endopeptidases in humans and
these enzymes have an important role in cancer invasion, metas-
tasis, and angiogenesis. They also exert significant impact on cell
behavior, like the growth of metastatic tumor cells and the
increase in motility of epithelial cells. MMPs not only act on the
extracellular matrix (ECM) via protein degradation, cleavage of
cell surface receptors, and the release of apoptotic signals but

they are also associated with advanced stages and clinical progno-
sis of many types of cancer [44].
Various methods can be used to study the relation between
MMPs and cancer. Studies examining various aspects of tumor
biology will shed light on which methodologies should be used to
examine this relationship. Some examples of potential methods
that may be used are the enzyme-linked immunosorbent assay
(ELISA), immunohistochemical (IHC) staining, and various
molecular biology methods, which range from the detection of
genetic polymorphisms to measuring gene expression. The impor-
tance of MMPs as potential biomarkers is highlighted by their
known involvement in many types of cancers.
4.1 Breast Cancer Overexpression of MMPs in the serum of breast cancer patients is
associated with a poor prognosis. For example, overexpression of
MMP-9 and MMP-2 correlates with a higher risk and poor prog-
nosis for this type of cancer [45].
MMP-13 has also been studied and characterized with tumor
progression. Overexpression of MMP-13 has been correlated with
the expression of HER2/neu, more aggressive tumor phenotypes
and decreased life expectancy [46]. A recent study examining the
relationship between MMP and TIMP expression in tumors (i.e.,
tumor center or tumor invasive front) and distant metastasis in
mononuclear inflammatory cells (MICs) and cancer associated
fibroblasts (CAFs) of patients with primary breast cancer, showed
that the expression of MMP-11 by MICs and TIMP-2 by CAFs in
the tumor center or tumor invasive front were the most potent
independent prognostic factors for predicting clinical outcome for
patients [47].
4.2 Prostate Cancer It has recently been shown that genotypic variations of MMP-3
and TIMP-3 (polymorphisms) may affect the balance between
these proteins, thus increasing the risk of developing prostate can-
cer [48].
The relationship between testosterone and MMP-13 levels in
prostate cancer patients has been previously reported [8]. Low lev-
els of testosterone have been linked to a poor prognosis [49]. This
relationship suggests an important biochemical interaction between
this hormone and these endopeptidases. Immunohistochemical
analysis of MMP-3, MMP-13 and TIMP-3, MMP-9, and TIMP-1
levels in prostate cancer also suggests that an imbalance between
MMP-9 and TIMP-1 may explain the invasive potential of pros-
tatic adenocarcinoma [50].
4.3 Lung Cancer MMP-13 is not only associated with breast and prostate cancers
but may also be the most important target among the MMPs in
lung cancer. In a recent study, MMP-2, -3, -9, and -13 were
a nalyzed by fluorescence molecular tomography (FMT) and only

MMP-13 was detected early in the development of pulmonary
invasive adenocarcinomas [51]. It has also been reported that
MMP-2 polymorphisms may be a potential prognostic marker for
inoperable non-small cell lung cancers (NSCLC) treated with
radio- or chemotherapy [52].
4.4 Colorectal Studies of advanced stage colorectal adenocarcinoma found that

Cancer MMP-9 was expressed at high but inconstant levels in tumor cells.
Highest expression levels were found in little or moderately dif-
ferentiated carcinomas, and surprisingly, the lowest expression lev-
els were observed in well-differentiated colorectal cancers. MMP-9
expression has also been observed in peritumoral macrophages and
stroma cells. While high levels of expression were shown in metas-
tasis-free lymphnodes in both macrophages and lymphocytes in
areas of tumor necrosis, angiogenesis was in general found to be
correlated with the intensity of MMP-9 expression [53, 54].
Besides MMP-9, MMP-7 also appears to have an important
role in the development and progression of colorectal cancer. A
recent meta-analysis involving 1631 colorectal cancer patients
showed that overexpression of MMP-7 unfavorably impacted on
the overall survival rate. Moreover, this study demonstrated the
potential of MMP-7 as a prognostic predictor. In summary, over-
expression of MMP-7, detected by immunohistochemical staining,
indicated a poor prognosis for colorectal cancer patients, and could
be used as a guide for the clinical-therapeutic management of this
neoplasia [55].
Studies have also reported that the expression of MMP-13 is
related to pathogenesis, metastasis, and tumor relapse, and for this
reason it may be a valuable predictor of metastasis and relapse of
colorectal cancer [56].
4.5 Bladder Cancer Studies have investigated the role of MMPs in bladder cancer by
identifying metalloproteinase polymorphisms and gene expression
levels as well as measuring MMP levels in urine. These analyses of
neoplasia have looked at various biological samples, including
urine, and the results establish MMPs as potential diagnostic bio-
markers for this neoplasia.
Both MMP-3 and MMP-9 have recently been proposed as
possible early markers for bladder cancer. Using ELISA-based
assays the authors were able to show that levels of both MMPs
were elevated in the urine of patients with bladder cancer when
compared with a control group. Furthermore, MMP-3 levels were
evaluated in urine samples from patients with early stage schisto-
somal and non-schistosomal bladder cancer. Remarkably, MMP-3
levels remained elevated at advanced stages of these two histologi-
cal types of cancer. On the other hand, MMP-9 levels were elevated
only at advanced stages of schistosomal and non-schistosomal
bladder cancer. Assays displayed improved sensitivity and specific-

ity for MMP-3 compared to MMP-9 at diagnosis [57].
A more recent study has examined the role of gene polymor-
phisms of MMP-1, -2, -3, -7, and -9 in bladder cancer. The pres-
ence of MMP-1 gene polymorphisms was detected in all studied
populations. However, MMP-2 and MMP-7 polymorphisms were
only detected in Asian populations, and it was proposed that these
polymorphisms may be protective for the development of bladder
cancer [58].
In conclusion, many studies have shown the involvement of
MMPs at all stages of carcinogenesis. Nevertheless, the diagnostic
and prognostic roles of MMPs are difficult to define not only
because most studies do not examine the same MMPs or use meth-
ods with similar sensitivity and specificity, but also because of the
complexity of carcinogenesis. Nevertheless, MMP-3 and -13 have
been identified as potential biomarkers in a range of cancers includ-
ing breast, prostate, lung, colorectal and bladder cancers, regard-
less of the biological matrix examined or the method used.
References
1. Frantz C, Stewart KM, Weaver VM (2010) 9. Chaudhary KA, Pandya S, Ghosh K, Nadkarni
The extracellular matrix at a glance. J Cell Sci A (2013) Matrix metalloproteinase and its
123:4195–4200 drug targets therapy in solid and hematologi-
2. Harjanto D, Zaman MH (2013) Modeling cal malignances: an overview. Mutat Res
extracellular matrix reorganization in 3D envi- 753:7–23
ronments. PLoS One 8:e52509 10. Klein G, Schmal O, Aicher KW (2015) Matrix
3. Tallant C, Marrero A, Gomis-Rüth FX (2010) metalloproteinases in stem cell mobilization.
Matrix metalloproteinases: fold and function of Matrix Biol 44–46:175–183
their catalytic domains. Biochim Biophys Acta 11. MacColl E, Khalil RA (2015) Matrix metallo-
1803:20–28 proteinases as regulators of vein structure and
4. Bonnans C, Chou J, Werb Z (2014) function: implications in chronic venous dis-
Remodelling the extracellular matrix in devel- ease. J Pharmacol Exp Ther 355:410–428
opment and disease. Nat Rev Mol Cell Biol 12. Szmigielska-Kaplon A, Szemraj J, Hamara K,
15:786–801 et al. (2014) Polymorphism of CD44 influences
5. Theocharis AD, Skandalis SS, Gialeli C, the efficacy of CD34(+) cells mobilization in
Karamanos NK (2015) Extracellular matrix patients with hematological malignancies. Biol
structure. Adv Drug Deliv Rev 97:4–27 Blood Marrow Transplant 20:986–991
6. Giannandrea M, Parks WC (2014) Diverse 13. Kunisaki Y, Bruns I, Scheiermann C (2013)
functions of matrix metalloproteinases during Arteriolar niches maintain haematopoietic stem
fibrosis. Dis Model Mech 7:193–203 cell quiescence. Nature 502:637–643
7. Tenório PP, Duque MAA, Araújo MM, Pontes 14. Deryugina IE, Quigley PJ (2015) Tumor
Filho NT, Melo MR Jr (2012) The role of angiogenesis: MMP-mediated induction of
matrix metalloproteinases in the development intravasation- and metastasis-sustaining neo-
of aortic aneurysm. Avaliable at: http://files. vasculature. Matrix Biol 44–46:94–112
bvs.br/upload/S/0101-5907/2012/v26n4/ 15. Thushara MR, Hemshekhar M, Kemparaju B,
a3488.pdf [Portuguese] et al. (2015) Biologicals, platelet apoptosis and
8. Bonaldi CM, Azzalis LA, Junqueira VBC, human diseases: an outlook. Crit Rev Oncol
Oliveira CGB, Vilas Boas VA, Gáscon TM, Hematol 93:149–158
Gehrke FS, Kuniyoshi RK, Alves BCA, Fonseca 16. Poyer F, Coquerel B, Pegahi R, et al. (2009)
FLA (2015) Plasmatic levels of E-Cadherin Secretion of MMP-2 and MMP-9 induced by
and MMP-13 in prostate cancer patients: cor- VEGF autocrine loop correlates with clinical
relation with PSA, testosterone and pathologi- features in childhood acute lymphoblastic
cal parameters. Tumori 101:185–188 leukemia. Leuk Res 33:407–417
17. Bailón E, Berzal-Ugarte E, Amigo-Jimenes I Biomarkers predict radiographic progression in

(2014) Overexpression of progelatinase B/ early rheumatoid arthritis and perform well
proMMP-9 affects migration regulatory path- compared with traditional markers. Arthritis
ways and impairs chronic lymphocytic leukemia Rheum 56:3236–3247
cell homing to bone marrow and spleen. 31. Vandenbroucke RE, Dejager L, Libert C
J Leukoc Biol 96:185–199 (2011) The first MMP in sepsis. EMBO Mol
18. Mirandola L, Comi P, Cobos E, et al. (2011) Med 3:367–369
Notch-ing from T-cell to B-cell lymphoid 32. Aerts J, Vandenbroucke RE, Dera R, Balusu S,
malignancies. Cancer Lett 308:1–13 Vanwonterghem E, Moons L, Libert C,
19. Lu P, Takai K, Weaver VM, Werb Z (2011) Dehaen W, Arckens L (2015) Synthesis and
Extracellular matrix degradation and remodel- validation of a hydroxypyrone-based, potent,
ing in development and disease. Cold Spring and specific matrix metalloproteinase-12 inhib-
Harb Perspect Biol 3:a005058 itor with anti-inflammatory activity in vitro and
20. Arpino V, Brock M, Gill ES (2015) The role of in vivo. Mediators Inflamm 2015:510679
TIMPs in regulation of extracellular matrix 33. Navratilova Z, Kolek V, Petrek M (2016)
proteolysis. Matrix Biol 44–46:247–254 Matrix metalloproteinases and their inhibitors
21. Nakahara F, Kitaura J, Uchida T, et al. (2014) in chronic obstructive pulmonary disease. Arch
Hes1 promotes blast crisis in chronic myelog- Immunol Ther Exp (Warsz) 64:177–193
enous leukemia through MMP-9 upregulation 34. Acera A, Vecino E, Duran JA (2013) Tear
in leukemic cells. Blood 123:3932–3942 MMP-9 levels as a marker of ocular surface
22. Maral S, Acar M, Balcik SO, et al. (2015) inflammation in conjunctivochalasis. Invest
Matrix metalloproteinase 2 and 9 polymor- Ophthalmol Vis Sci 54:8285–8291
phism in patients with myeloproliferative dis- 35. Libby P, Ridker PM, Hansson GK (2011)
eases. Medicine 94:e732 Progress and challenges in translating the biol-
23. Sivula M, Hästbacka J, Kuitunen A, et al. ogy of atherosclerosis. Nature 473:317–325
(2015) Systemic matrix metalloproteinase-8 36. Kaptoge S, Seshasai SRK, Gao P, Freitag DF,
and tissue inhibitor ofmetalloproteinases-1 lev- Butterworth AS, Borglykke A, Angelantonio
els in severe sepsis-associated coagulopathy. ED, Gudnason V, Rumley A, Lowe GDO,
Acta Anaesthesiol Scand 59:176–184 Jørgensen T, Danesh J (2014) Inflammatory
24. McGuire JK, Manicorne AM (2008) Matrix cytokines and risk of coronary heart disease:
metalloproteinases as modulators of inflamma- new prospective study and updated meta-
tion. Semin Cell Dev Biol 19:34–41 analysis. Eur Heart J 35:578–589
25. Rosa Neto NS, de Carvalho JF (2009) O uso 37. Alfakry H, Malle E, Koyani CN, Pussinen PJ,
de provas de atividade inflamatória em reuma- Sorsa T (2015) Neutrophil proteolytic acti-
tologia. Rev Bras Reumatol 49:413–430 vation cascades: a possible mechanistic link
26. Amar S, Fields GB (2015) Potential clinical between chronic periodontitis and coronary
implications of recent matrix metalloproteinase heart disease. Innate Immun 22:85–99
inhibitor design strategies. Expert Rev 38. O’Sullivan S, Gilmer JF, Medina C (2015) Matrix
Proteomics 12:445–447 metalloproteinases in inflammatory bowel dis-
27. Peng WJ, Yan JW, Wan YN, Wang BX, Tao JH, ease: an update. Mediators Inflamm.
Yang GJ, Pan HF, Wang J (2012) Matrix doi:10.1155/2015/964131 Article ID 964131
metalloproteinases: a review of their structure 39. Pedersen G, Saermark T, Kirkegaard T,
and role in systemic sclerosis. J Clin Immunol Brynskov J (2009) Spontaneous and cytokine
32:1409–1414 induced expression and activity of matrix
28. Sun S, Bay-Jensen AC, Karsdal MA, Siebuhr metalloproteinases in human colonic epithe-
AS, Zheng Q, Maksymowych WP, Christiansen lium. Clin Exp Immunol 155:257–265
TG, Henriksen K (2014) The active form of 40. Rath T, Roderfeld M, Halwe JM, Tschuschner
MMP-3 is a marker of synovial inflammation A, Roeb E, Graf J (2010) Cellular sources of
and cartilage turnover in inflammatory joint MMP-7, MMP-13 and MMP-28 in ulcerative
diseases. BMC Musculoskelet Disord 15:93 colitis. Scand J Gastroenterol 45:1186–1196
29. Ateş A, Türkçapar N, Olmez U, Tiryaki O, 41. Matusiewicz M, Neubauer K, Mierzchala-
Düzgün N, Uğuz E, et al. (2007) Serum Pasierb M, Gamian A, Krzystek-Korpacka M
pro-matrix metalloproteinase-3 as an Indicator (2014) Matrix metalloproteinase-9: its interplay
of disease activity and severity in rheumatoid with angiogenic factors in inflammatory bowel
arthritis: comparison with traditional markers. diseases. Dis Markers doi:10.1155/2014/
Rheumatol Int 27:715–722 643645 Article ID 643645
30. Young-Min S, Cawston T, Marshall N, Coady 42. Khokha R, Murthy A, Weiss A (2013)
D, Christgau S, Saxne T, et al. (2007) Metalloproteinases and their natural inhibitors
in inflammation and immunity. Nat Rev Roels O, Sabourin JC, Thiberville L, Clapper
Immunol 13:649–665 ML (2015) MMP-13 in-vivo molecular imag-
43. Vandenbroucke RE, Libert C (2014) Is there ing reveals early expression in lung adenocarci-
new hope for therapeutic matrix metallopro- noma. PLoS One 10:e0132960
teinase inhibition? Nat Rev Drug Discov 52. Butkiewicz D, Krześniak M, Drosik A, Giglok
13:904–927 M, Gdowicz-Kłosok A, Kosarewicz A, Rusin M,
44. Wieczorek E, Jablonska E, Wasowicz W, Masłyk B, Gawkowska-Suwińska M, Suwiński R
Reszka E (2015) Matrix metalloproteinases (2015) The VEGFR2, COX-2 and MMP-2
and genetic mouse models in cancer research: a polymorphisms are associated with clinical out-
mini-review. Tumour Biol 36:163–175 come of patients with inoperable non-small cell
45. Ren F, Tang R, Zhang X, Madushi WM, Luo lung cancer. Int J Cancer 137:2332–2342
D, Dang Y, Li Z, Wei K, Chen G (2015) 53. Georgescu EF, Mogoantă SŞ, Costache A,
Overexpression of MMP family members func- Pârvănescu V, Totolici BD, Pătraşcu Ş,
tions as prognostic biomarker for breast cancer Stănescu C (2015) The assessment of matrix
patients: a systematic review and meta-analysis. metalloproteinase-9 expression and angiogen-
PLoS One 10(8):e0135544 esis in colorectal cancer. Rom J Morphol
46. Rosseti C, Reis B, Delgado PO, et al. (2015) Embryol 56:1137–1144
Adhesion molecules in breast carcinoma: a 54. Zheng CG, Chen R, Xie JB, Liu CB, Jin Z, Jin
challenge to the pathologist. Rev Assoc Med C (2015) Immunohistochemical expression of
Bras 61:81–85 Notch1, Jagged1, NF-κB and MMP-9 in
47. Eiró N, Fernandez-Garcia B, Vázquez J, Del Casar colorectal cancer patients and the relationship
JM, González LO, Vizoso FJ (2015) A phenotype to clinicopathological parameters. Cancer
from tumor stroma based on the expression of Biomark 15:889–897
metalloproteases and their inhibitors, associated 55. Chen H, Hu Y, Xiang W, Cai Y, Wang Z, Xiao
with prognosis in breast cancer. Oncoimmunology Q, Liu Y, Li Q, Ding K (2015) Prognostic sig-
4(7):e992222 eCollection 2015 nificance of matrix metalloproteinase 7 immu-
48. Srivastava P, Kapoor R, Mittal RD (2013) nohistochemical expression in colorectal
Impact of MMP-3 and TIMP-3 gene polymor- cancer: a meta-analysis. Int J Clin Exp Med
phisms on prostate cancer susceptibility in 8:3281–3290
North Indian cohort. Gene 530:273–277 56. Yan Q, Yuan Y, Yankui L, Jingjie F, Linfang
49. García-Cruz E, Piqueras M, Huguet J, Peri L, J, Yong P, Dong H, Xiaowei Q (2015) The
Izquierdo L, Musquera M, Franco A, Alvarez- expression and significance of CXCR5 and
Vijande R, Ribal MJ, Alcaraz A (2012) Low MMP-13 in colorectal cancer. Cell Biochem
testosterone levels are related to poor progno- Biophys 73:253–259
sis factors in men with prostate cancer prior to 57. El-Sharkawi F, El Sabah M, Hassan Z, Khaled
treatment. BJU Int 110(11 Pt B):E541–E546 H (2014) The biochemical value of urinary
50. Babichenko II, Andriukhin MI, Pulbere S, metalloproteinases 3 and 9 in diagnosis and
Loktev A (2014) Immunohistochemical prognosis of bladder cancer in Egypt. J Biomed
expression of matrix metalloproteinase-9 and Sci 21:72
inhibitor of matrix metalloproteinase-1 in pros- 58. Tao L, Li Z, Lin L, Lei Y, Hongyuan Y,
tate adenocarcinoma. Int J Clin Exp Pathol Hongwei J, Yang L, Chuize K (2015) MMP1,
7:9090–9098 2, 3, 7, and 9 gene polymorphisms and urinary
51. Salaün M, Peng J, Hensley HH, Roder N, cancer risk: a meta-analysis. Genet Test Mol
Flieder DB, Houlle-Crépin S, Abramovici- Biomarkers 19:548–555
Index
A F
Fluorescence imaging��245–247
Activated imaging probes�� 246, 248
Fluorogenic substrates�� 51, 138–139, 179
Affinity chromatography��10, 22, 28, 29, 71
Förster resonance energy transfer (FRET)
substrates�� 58, 138, 140, 167
B
Free energy calculations��119
BBB. See Blood-brain barrier (BBB)
bBMVEC. See Bovine brain microvascular endothelial cells H
(bBMVEC) High-throughput sequencing��210, 211, 220–222, 224
Blood-brain barrier (BBB)��285–293
co-culture model��286 I
permeability��285–287
Imaging�� 64, 78–79, 245–254
Bovine brain microvascular endothelial cells
Inclusion bodies (IBs)�� 4, 6–8, 10, 13, 50, 53, 67–68
(bBMVEC)�� 286–290, 292
Inflammatory disease�� 303, 304
C Isotopically labeled�� 20–25, 67–68
iTRAQ-TAILS�� 186–189, 191–192
Cancer��3, 4, 17, 18, 36, 49, 87, 111, 231,
246, 260, 261, 263, 272, 273, 300, 304–307 L
Cell culture membrane inserts�� 289, 292
Ligand binding�� 114, 131, 166, 272
Cleavage�� 4, 9–12, 15, 28, 31, 38, 53, 62, 137,
140, 166, 169, 173, 178, 185–188, 193–196, 199–206, M
219, 220, 223, 224, 226, 231, 260, 271, 301, 304
Collagen��3, 18, 36, 88, 90, 111, 113, 116, 122–124, Matrix metalloprotease (MMP)�� 3–15, 17–31,
139–168, 239, 260, 262, 287, 288, 292, 297–299, 302 35, 36, 49–59, 61, 62, 64–68, 75, 79, 82, 87–106,
Colonoscopy��245 111–132, 137–179, 185–197, 199, 210, 231–243,
Colorectal cancer�� 273, 306 245–254, 259–269, 271–279, 285–293, 297–307
Computer-aided drug design��271–279 inhibitors��111, 113, 115, 126, 260,
Crystallization��50, 51, 53–55, 58, 88, 89 261, 263, 272–274, 286, 288, 290, 291
membrane-anchored�� 18–20, 23, 50
D MMP2�� 4–15, 18
MMP-14�� 49, 68, 91, 93, 105, 114, 269
Degradome��186
MD. See Molecular dynamics (MD)
Detection�� 14, 27, 79, 169, 226, 231–243,
Meprin�� 35–44, 298
245, 246, 250, 273, 303, 305
Metalloprotease��3–15, 17–31, 35, 36,
Diffraction�� 49–59, 90
49–59, 61–82, 87–106, 111–132, 137–179, 185–197,
Disease�� 17, 36, 87, 224, 225, 231, 246,
209–227, 231–243, 245–254, 259–269, 271–279,
263, 273, 297–307
285–293, 297–307
Docking��65, 66, 76, 121–125, 260,
Metalloproteinase. See Metalloprotease
263–267, 272–275, 279
Metzincin�� 35, 298
and scoring��273–276
Mice�� 247, 248, 253, 303
Min mice�� 246–248, 252, 253
E
Molecular dynamics (MD)�� 76, 111–132
Early cancer detection��246 simulations��113–116, 119–129, 132,
Endoprotease��35 272, 274, 276–279
DOI 10.1007/978-1-4939-6863-3, © Springer Science+Business Media LLC 2017
313
Matrix Metalloproteases: Methods and Protocols
314 Index

Molecular modeling�� 115, 116 Protein–lipid interactions��74
MT-LOOP-dependent localization of membrane type I Proteomics�� 185, 189, 195, 199
matrix metalloproteinase (MT1-MMP)��18–20,
40, 50, 52–58, 144, 148, 149, 160–163, 165–167, 170 R
Multi-domain�� 90, 112 Refolding��4–9, 13, 14, 50, 51, 53, 54, 58, 68
Multidomain enzymes�� 112, 113, 119
Mutagenesis�� 62, 81, 209–227 S
N Semiquantitative��242
Stroke�� 286, 300
Near-infrared imaging��248
Structural interaction fingerprints (SIFt)�� 266, 267
Normal human astrocytes
Structure-based prediction�� 62, 65, 124
Nuclear magnetic resonance (NMR) spectroscopy��20,
Structure-based virtual screening (SBVS)�� 260, 263, 265
23, 26–29, 31, 62–65, 67–77, 79–81, 88–90, 92, 101,
Structure-function relationship��89
102, 106, 112
Substrate�� 4, 9–12, 15, 36, 37, 39, 40, 43,
O 53, 88, 112, 113, 116, 120–124, 128, 130, 132, 137–179,
185–197, 199, 200, 209–227, 231–233, 239, 242, 243,
Oligomer��36 265, 269, 271, 298, 299
discovery�� 185, 186, 196
P
Paramagnetic relaxation enhancement (PRE)�� 62, 68, T
71, 75 Time-resolved degradomics��186
Peptide hydrolysis�� 131, 132 Tissue plasminogen activator (tPA)�� 286, 288,
Peripheral membrane protein��61–82 290, 291, 293
Phage display��137, 209–213, 215, Transmembrane domain�� 17–31, 112
217–221, 224 Triple-helical substrates��116, 139–166, 168, 172
Pichia pastoris�� 35–44
Protease�� 19, 20, 22, 25, 27, 28, 30, 31, Y
35–44, 64, 71, 89, 91, 185–188, 190, 191, 194, 196,
Yeast�� 5, 20, 36, 37, 50, 80, 214
199–206, 209, 211, 215, 217, 219, 220, 224, 226, 232,
234, 242, 260 Z
Protein
dynamics��88 Zn-chelating inhibitor��263–265
expression�� 8, 13, 24, 29, 30, 63–64, 67, 210 Zymogram�� 14, 232–234
flexibility��122 Zymography��6, 9–11, 14, 15, 231–243

(Methods in Molecular Biology 1579) Charles A. Galea (Eds.) - Matrix Metalloproteases - Methods and Protocols-Humana Press (2017)

Uploaded by

Copyright:

Available Formats

(Methods in Molecular Biology 1579) Charles A. Galea (Eds.) - Matrix Metalloproteases - Methods and Protocols-Humana Press (2017)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

(Methods in Molecular Biology 1579) Charles A. Galea (Eds.) - Matrix Metalloproteases - Methods and Protocols-Humana Press (2017)

Uploaded by

Copyright:

Available Formats

Methods in

Molecular Biology 1579

Charles A. Galea Editor

For further volumes:

Methods and Protocols

ISSN 1064-3745 ISSN 1940-6029 (electronic)

Library of Congress Control Number: 2017930809

© Springer Science+Business Media LLC 2017

Printed on acid-free paper

This Humana Press imprint is published by Springer Nature

Parkville, VIC, Australia Charles A. Galea

Part I Expression and Purification of Matrix Metalloproteases

Part II Structural Characterization of Matrix Metalloproteases

Part III Computational Simulations of Matrix Metalloproteases

Part IV Determining Matrix Metalloprotease

10 Identification of Protease Cleavage Sites by Charge-Based Enrichment

Part V Detection of Matrix Metalloproteases

Part VI Matrix Metalloprotease Inhibitors

Part VII Matrix Metalloproteases as Biomarkers

Olayiwola Adedotun Adekoya • Department of Pharmacy, University of Tromsø, Tromso,

Stephan Schilling • Department of Drug Design and Target Validation (IZI-­IMWT),

Expression and Purification of Matrix Metalloproteases

Expression and Purification of Matrix Metalloproteinases

Key words MMP2, Inclusion bodies, Affinity chromatography, Refolding, Zymography

MMPs (matrix metalloproteinases) comprise a family of 23 pro-

substrates in different biological conditions. It is hence imperative

2.2 Protein 1. Terrific broth media (pH ~ 7.4): Dissolve 12 g tryptone, 24 g

6. Inclusion bodies solubilization buffer: 20 mM Tris–HCl pH

4. After loading, incubate the column for 2 h at room tempera-

10. Fractions containing purified and refolded MMP2 protein are

5. Stain the gel by submerging it in staining solution for 30 min

Fig. 3 Gelatin zymography for analysis of refolded MMP2 protein activity.

3. Add 5× loading dye and heat denature at 95 °C for 10 min.

1. For amplification of MMP2 CDS, cDNA prepared using RNA

over the random hexamer to ensure the amplification of full-

chaperonins from O. antarctica which assists improved pro-

11. To facilitate efficient binding of the protein on the Ni+2-NTA

which is evident by the presence of two bands, one of high

Financial assistance to DKS from Department of Science and

Expression and Purification of a Matrix Metalloprotease

Key words Membrane-anchored MMP, Matrix metalloproteases, NMR spectroscopy, Transmembrane

Matrix metalloproteases comprise a large family of zinc-dependent

physiology and may play a role in homeostasis [7, 9, 17–20], while

Previous studies have highlighted the importance of the cyto-

shown to interact with p27RF-Rho (LAMTOR1) which regulates

2.1 Expression 1. E. coli BL21(DE3) or E. coli BL21(DE3)pLysS chemically

6. 13C-labeled d-glucose (99% 13C): This can be obtained from a

2.2 Expression This method was originally described by Studier [50].

4. N-5052 medium: 922 ml of sterile water, 6 ml of vitamins,

2.3 Immobilized 1. BugBuster protein extraction reagent (Merck Millipore).

2.4 HPLC Purification 1. 0.2 μm PTFE syringe filters.

2.5 NMR Sample 1. 5 mm Shigemi tube (Shigemi Inc.).

The truncated membrane-anchored MMP containing the TM and

3.2 Protein Autoinduction typically produces significantly higher cell masses

5. Collect the supernatant and load onto a 5 ml nickel affinity

Fig. 3 Ni2+-affinity (NTA) purification of the His6-tagged TRX-TM/CPD protein. A

5. Dissolve the dried powder for each fraction in 0.5 ml of sam-

3.6 NMR Screening 1. Record a 1D 1H spectrum and a 2D 1H-15 N-TROSY spectrum

1. The pET32a expression vector contains an ampicillin resistance

Parkville, VIC, Australia Charles A. Galea

Part IV Determining Matrix Metalloprotease

Stephan Schilling • Department of Drug Design and Target Validation (IZI-IMWT),

2.4 HPLC Purification 1. 0.2 μm PTFE syringe filters.

2.5 NMR Sample 1. 5 mm Shigemi tube (Shigemi Inc.).

3.6 NMR Screening 1. Record a 1D 1H spectrum and a 2D 1H-15 N-TROSY spectrum

at this site during passage through the secretory p athway, prior to

10. Due to the low amount of expressed protein, process-controlled