Biomarkers PDF

Free ebooks ==> www.Ebook777.
com
Methods in
Molecular Biology 1546
Paul C. Guest Editor
Multiplex
Biomarker
Techniques
Methods and Applications
www.Ebook777.com
Free ebooks ==> www.Ebook777.com
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
www.Ebook777.com
Multiplex Biomarker Techniques
Methods and Applications
Edited by
Paul C. Guest
Laboratory of Neuroproteomics, University of Campinas (UNICAMP), Campinas, Brazil
Editor
Paul C. Guest
Laboratory of Neuroproteomics
University of Campinas (UNICAMP)
Campinas, Brazil
ISSN 1064-3745 ISSN 1940-6029 (electronic)

Methods in Molecular Biology
ISBN 978-1-4939-6729-2 ISBN 978-1-4939-6730-8 (eBook)
DOI 10.1007/978-1-4939-6730-8
Library of Congress Control Number: 2016958277
© 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.
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
Due to continuous technical developments and new insights into the high complexity of
many diseases, there is an increasing need for multiplex biomarker readouts for improved
clinical management and to support the development of new drugs by pharmaceutical com-
panies. The initial rollout of these techniques has led to promising results by helping to read
patients as deeply as possible and provide clinicians with information relevant for a person-
alized medicine approach. This book describes the basic technology platforms being applied
in the fields of genomics, proteomics, transcriptomics, metabolomics, and imaging, which
are currently the methods of choice in multiplex biomarker research. It also describes the
chief medical areas in which the greatest progress has been made and highlights areas where
further resources are required.
More than 1000 biomarker candidates for various diseases have been described in the
scientific literature over the last 20 years. However, the rate of introduction of new bio-
marker tests into the clinical arena is much lower with less than 100 such tests actually
receiving approval and appearing on the marketplace. This disconnect is most likely due to
inconsistencies at the discovery end, such as technical variations within and between plat-
forms, a lack of validation of biomarker candidates, as well as a lack of awareness within the
research community of the criteria and regulatory matters for integrating biomarkers into
the pipeline [1]. Another reason relates to the fact that many diseases are heterogeneous in
nature and comprised of different subtypes. This can cause difficulties in studies attempting
to identify biomarkers since different investigators may analyze cohorts comprised of unique
or even mixed subtypes of a particular disease, and this can make comparisons both within
and across studies invalid. Furthermore, the use of patient and control groups in clinical
studies which have not been properly stratified according to biomarker profiling is one of
the biggest causes of failure in the development of new drugs [2–9].
One way of addressing these issues is through the increasing use of multiplex biomarker
tests which can provide a more complete picture of a disease. Multiplex biomarker assays
can simultaneously measure multiple analytes in one run on a single instrument as opposed
to methods that measure only one analyte at a time or multiple analytes at different times.
The simultaneous measurement of different biomarkers in a multiplex format allows for
lower sample and reagent requirements along with reduced processing times on a per assay
basis (Table 1). In contrast, testing for single analytes can be laborious, time-consuming,
and expensive in cases where multiple assays for different molecules are required.
So how does multiplexing improve classification of diseases?
Multiplexing allows for higher sample throughput with greater cross-comparability
within and across experiments since each of the component assays are processed, read, and
analyzed under identical conditions and at the same time. This obviates traditional prob-
lems of comparing the results of single assays within a given study, which may be subject to
procedural inconsistencies in sampling, methodology, or data analysis. Most importantly,
the use of multiple biomarkers allows for greater accuracy in the diagnosis of complex dis-
eases by providing more complete information about the perturbed physiological pathways
in a shorter time period. This includes in-depth attempts to decipher pathological changes
www.Ebook777.com
vi Preface
Table 1
Characteristics of single versus multiplex immunoassays
Advantages Disadvantages
Single assays Greater sensitivity because there is no Requires prior knowledge to target
competition of different analytes for specific analytes
reagents
Useful as a validation test after identification Requires greater amounts of sample
of biomarker candidates per analyte
Requires greater amounts of reagents
per analyte
Greater amount of time required for
analysis of multiple analytes (in
proportion to analyte number)
Low cross-comparability of multiple assays
as each one is run under different
conditions and at a different time
Multiplex No prior knowledge required as it can be Requires more complex and stringent
assays used for screening statistical analyses
Greater cross-comparability across analytes Often requires bioinformatic analyses to
as all are run simultaneously under the identify over-represented pathways
same conditions
More understanding of physiological Requires validation of analytes identified
pathways affected in disease due to higher as significant during screens using an
number of simultaneous analyte alternate technology
measurements
Lower amounts of sample required
per analyte
Lower amounts of reagents required
per analyte
Lower time required for analysis of multiple
analytes
at the level of the DNA sequence [10], epigenome [11], transcriptome [12], proteome [2],
and metabolome [13]. Thus, we are now moving away from single biomarker tests to more
comprehensive multiplex biomarker analyses in order to better classify and combat these
disorders. This works in the same way that a complete fingerprint allows for more accurate
identification of a suspect in a criminal investigation as opposed to a partial print which may
not be resolvable across multiple suspects.
However, there are still challenges ahead. While some diseases are increasingly being
treated according to biomarker profiling patterns, the “one-size-fits-all” approach is still the
standard treatment for most diseases. Many diseases such as cancers [14–16], heart disease
[17], diabetes and neurological disorders [18–20] present difficult problems when it comes
to deciding on treatment options since multiple molecular pathways of complex network
signaling cascades can be affected. In addition, as these disorders can affect all age groups
and both sexes, even more variables can occur, leading to even greater variability. In order to
deal with this issue, collaborative research networks should be established for multiplexing
efforts to better integrate biomarker discovery in real time to targeted therapeutics.
Preface vii
In the United States, the Clinical Laboratory Improved Amendments (CLIA) act was
passed by Congress in 1988 as a means of integrating quality testing for all laboratories and
to ensure accuracy, reproducibility, and speed of patient testing results [21]. The Food and
Drug Administration (FDA) is the responsible agency for applying these regulations for the
purpose of categorizing biomarker assays based on technological complexity and ease of oper-
ation. Laboratory-developed tests have not necessarily received automatic approval and have
traditionally been endorsed only at the FDA’s discretion. This is because the clinical validation
of multiplex biomarker tests will require the participation of multiple laboratories, and the
resulting platforms are likely to need simplification stages and demonstration of increased
robustness to merit extensive clinical applications. Multiplex tests may also require the use of
an algorithm to derive a composite “score” representing the multiple values of each compo-
nent assay for a classification or diagnosis. For example, scores of 100 and 0 would mean a
100 % and 0 % chance respectively that the disease is present. Of course, scores in the middle
range would be less precise. Besides the multiplexing of analytes, another level of multiplexing
can be achieved by running both patient and control samples in the same assay. For example,
both cDNA arrays and two-dimensional gel electrophoresis (2D-DIGE) enable the analysis
of hundreds of analytes simultaneously for up to three samples at the same time through the
prelabeling of sample extracts with different spectrally resolvable fluorescent dyes.
The multiplex platforms for carrying out screening typically have medium to large foot-
prints and require considerable expertise to operate. For transcriptomic or RNA-based pro-
filing, these include quantitative PCR, cDNA microarray, and microRNA approaches. For
proteomics, there are two-dimensional difference gel electrophoresis, multiplex immunoas-
say, label-free shot-gun mass spectrometry, selective reaction mass spectrometry, and
labeled-based mass spectrometry platforms. For metabolomic screening, the main plat-
forms in use are either mass spectrometry or proton nuclear magnetic resonance-based. For
clinical applications and rollout of biomarker assays, it is becoming increasingly important
that the platforms are small, user friendly, and fast so they can be used in a point-of-care
setting. The latest developments along these lines include lab-on-a-chip and mobile phone
applications. Detailed protocols describing both the discovery and point-of-care devices
incorporating multiplexed assays are described in this book.
Campinas, Brazil Paul C. Guest
References
1. Boja ES, Jortani SA, Ritchie J, Hoofnagle AN, Težak Ž, Mansfield E et al (2011) The journey to regu-
lation of protein-based multiplex quantitative assays. Clin Chem 57:560–567.
2. Lee JM, Kohn EC (2010) Proteomics as a guiding tool for more effective personalized therapy. Ann
Oncol 21(Suppl 7):vii205–vii10. doi: 10.1093/annonc/mdq375. Review
3. Lee JM, Han JJ, Altwerger G, Kohn EC (2011) Proteomics and biomarkers in clinical trials for drug devel-
opment. J Proteomics 74:2632–2641.
4. Kelloff GJ, Sigman CC (2012) Cancer biomarkers: selecting the right drug for the right patient. Nat
Rev Drug Discov 11:201–214.
5. Begg CB, Zabor EC, Bernstein JL, Bernstein L, Press MF, Seshan VE (2013) A conceptual and meth-
odological framework for investigating etiologic heterogeneity. Stat Med 32:5039–5052.
6. Henriksen K, O’Bryant SE, Hampel H, Trojanowski JQ, Montine TJ, Jeromin A et al (2014) Blood-
based biomarker interest group. Alzheimers Dement 10:115–131.
7. Guest PC, Chan MK, Gottschalk MG, Bahn S (2014) The use of proteomic biomarkers for improved
diagnosis and stratification of schizophrenia patients. Biomark Med 8:15–27.
viii Preface
8. Wang X, Adjei AA (2015) Lung cancer and metastasis: new opportunities and challenges. Cancer
Metastasis Rev 34:169–171.
9. Hudler P (2015) Challenges of deciphering gastric cancer heterogeneity. World J Gastroenterol
21:10510–10527.
10. Hudson TJ (2013) Genome variation and personalized cancer medicine. J Intern Med
274:440–450.
11. Mummaneni P, Shord SS (2014) Epigenetics and oncology. Pharmacotherapy 34:495–505.
12. Hu YF, Kaplow J, He Y (2005) From traditional biomarkers to transcriptome analysis in drug devel-
opment. Curr Mol Med 5:29–38.
13. Wishart DS (2008) Applications of metabolomics in drug discovery and development. Drugs R D
9:307–322.
14. Füzéry AK, Levin J, Chan MM, Chan DW (2013) Translation of proteomic biomarkers into FDA
approved cancer diagnostics: issues and challenges. Clin Proteomics 10:13. doi:
10.1186/1559-0275-10-13.
15. Nolen BM, Lokshin AE (2013) Biomarker testing for ovarian cancer: clinical utility of multiplex assays.
Mol Diagn Ther 17:139–146.
16. Ploussard G, de la Taille A (2010) Urine biomarkers in prostate cancer. Nat Rev Urol 7:101–109.
17. Vistnes M, Christensen G, Omland T (2010) Multiple cytokine biomarkers in heart failure. Expert
Rev Mol Diagn 10:147–157.
18. Lee KS, Chung JH, Choi TK, Suh SY, Oh BH, Hong CH (2009) Peripheral cytokines and chemo-
kines in Alzheimer’s disease. Dement Geriatr Cogn Disord 28:281–287.
19. Kang JH, Vanderstichele H, Trojanowski JQ, Shaw LM (2012) Simultaneous analysis of cerebrospinal fluid
biomarkers using microsphere-based xMAP multiplex technology for early detection of Alzheimer’s disease.
Methods 56:484–493.
20. Chan MK, Krebs MO, Cox D, Guest PC, Yolken RH, Rahmoune H et al (2015) Development of a
blood-based molecular biomarker test for identification of schizophrenia before disease onset. Transl
Psychiatry 5:e601. doi: 10.1038/tp.2015.91.
21. http://www.cms.gov/clia.
Contents
Preface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v
Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii
PART I REVIEWS
1 Application of Multiplex Biomarker Approaches to Accelerate
Drug Discovery and Development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Hassan Rahmoune and Paul C. Guest
2 The Application of Multiplex Biomarker Techniques
for Improved Stratification and Treatment of Schizophrenia Patients . . . . . . . . 19
Johann Steiner, Paul C. Guest, Hassan Rahmoune,
and Daniel Martins-de-Souza
3 Multiplex Biomarker Approaches in Type 2 Diabetes Mellitus Research. . . . . . 37
Susan E. Ozanne, Hassan Rahmoune, and Paul C. Guest
4 LC-MSE, Multiplex MS/MS, Ion Mobility, and Label-Free Quantitation
in Clinical Proteomics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
Gustavo Henrique Martins Ferreira Souza, Paul C. Guest,
5 Phenotyping Multiple Subsets of Immune Cells In Situ
in FFPE Tissue Sections: An Overview of Methodologies . . . . . . . . . . . . . . . . 75
James R. Mansfield
PART II STATISTICAL CONSIDERATIONS

6 Identification and Clinical Translation of Biomarker Signatures:
Statistical Considerations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
Emanuel Schwarz
7 Opportunities and Challenges of Multiplex Assays:
A Machine Learning Perspective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
Junfang Chen and Emanuel Schwarz
PART III PROTOCOLS

8 Multiplex Analyses Using Real-Time Quantitative PCR . . . . . . . . . . . . . . . . . . 125
Steve F.C. Hawkins and Paul C. Guest
9 Multiplex Analysis Using cDNA Transcriptomic Profiling . . . . . . . . . . . . . . . . 135
10 Multiplex Single Nucleotide Polymorphism Analyses . . . . . . . . . . . . . . . . . . . . 143
ix
x Contents
11 Pulsed SILAC as a Approach for miRNA Targets Identification

in Cell Culture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
Daniella E. Duque-Guimarães, Juliana de Almeida-Faria,
Thomas Prates Ong, and Susan E. Ozanne
12 Blood Bio-Sampling Procedures for Multiplex Biomarkers Studies. . . . . . . . . . 161
Paul C. Guest and Hassan Rahmoune
13 Multiplex Immunoassay Profiling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
Laurie Stephen
14 Multiplex Sequential Immunoprecipitation of Insulin Secretory
Granule Proteins from Radiolabeled Pancreatic Islets. . . . . . . . . . . . . . . . . . . . 177
Paul C. Guest
15 Two Dimensional Gel Electrophoresis of Insulin Secretory Granule Proteins
from Biosynthetically-Labeled Pancreatic Islets . . . . . . . . . . . . . . . . . . . . . . . . 187
Paul C. Guest
16 Depletion of Highly Abundant Proteins of the Human Blood Plasma:
Applications in Proteomics Studies of Psychiatric Disorders . . . . . . . . . . . . . . . 195
Sheila Garcia, Paulo A. Baldasso, Paul C. Guest,
17 Simultaneous Two-Dimensional Difference Gel Electrophoresis (2D-DIGE)
Analysis of Two Distinct Proteomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
Adriano Aquino, Paul C. Guest, and Daniel Martins-de-Souza
18 Selective Reaction Monitoring for Quantitation of Cellular Proteins . . . . . . . . 213
Vitor M. Faça
19 Characterization of a Protein Interactome by Co-Immunoprecipitation
and Shotgun Mass Spectrometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
Giuseppina Maccarrone, Juan Jose Bonfiglio, Susana Silberstein,
Christoph W. Turck, and Daniel Martins-de-Souza
20 Using 15N-Metabolic Labeling for Quantitative Proteomic Analyses. . . . . . . . . 235
Giuseppina Maccarrone, Alon Chen, and Michaela D. Filiou
21 Multiplex Measurement of Serum Folate Vitamers by UPLC-MS/MS. . . . . . . 245
Sarah Meadows
22 UPLC-MS/MS Determination of Deuterated 25-Hydroxyvitamin D
(d3-25OHD3) and Other Vitamin D Metabolites for the Measurement
of 25OHD Half-Life. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
Shima Assar, Inez Schoenmakers, Albert Koulman, Ann Prentice,
and Kerry S. Jones
23 iTRAQ-Based Shotgun Proteomics Approach for Relative
Protein Quantification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
Erika Velásquez Núñez, Gilberto Barbosa Domont, and Fábio César Sousa
Nogueira
1
24 H NMR Metabolomic Profiling of Human and Animal Blood
Serum Samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275
João G.M. Pontes, Antonio J.M. Brasil, Guilherme C.F. Cruz,
Rafael N. de Souza, and Ljubica Tasic
www.Ebook777.com
Contents xi
25 Lab-on-a-Chip Multiplex Assays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283

Harald Peter, Julia Wienke, and Frank F. Bier
26 Multiplex Smartphone Diagnostics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295
Juan L. Martinez-Hurtado, Ali K. Yetisen, and Seok-Hyun Yun
27 Development of a User-Friendly App for Assisting
Anticoagulation Treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303
Johannes Vegt
PART IV FUTURE DIRECTIONS

28 Multiplex Biomarker Approaches to Enable Point-of-Care Testing
and Personalized Medicine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311
Paul C. Guest
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 317
Contributors
JULIANA DE ALMEIDA-FARIA • University of Cambridge Metabolic Research Laboratories

and MRC Metabolic Diseases Unit, Wellcome Trust-MRC Institute of Metabolic Science,
Addenbrooke’s Hospital, Cambridge, UK; Faculty of Medical Sciences, Department of
Pharmacology, State University of Campinas, Campinas, Brazil
ADRIANO AQUINO • Laboratory of Neuroproteomics, Department of Biochemistry and Tissue
Biology, Institute of Biology, University of Campinas (UNICAMP), Campinas, SP,
Brazil
SHIMA ASSAR • MRC Elsie Widdowson Laboratory, Cambridge, UK
PAULO A. BALDASSO • Laboratory of Neuroproteomics, Department of Biochemistry and
Tissue Biology, Institute of Biology, University of Campinas (UNICAMP), Campinas, SP,
Brazil
FRANK F. BIER • Fraunhofer Institute for Cell Therapy and Immunology, Branch
Bioanalytics and Bioprocesses (IZI-BB), Potsdam, Germany
JUAN JOSE BONFIGLIO • Instituto de Investigación en Biomedicina de Buenos Aires
(IBioBA), CONICET, Partner Institute of the Max Planck Society, Buenos Aires,
Argentina
ANTONIO J.M. BRASIL • Chemical Biology Laboratory, Institute of Chemistry, Organic
Chemistry Department, State University of Campinas, Campinas, SP, Brazil
ALON CHEN • Department Stress Neurobiology and Neurogenetics, Max Planck Institute of
Psychiatry, Munich, Germany
JUNFANG CHEN • Department of Psychiatry and Psychotherapy, Central Institute of Mental
Health, Medical Faculty Mannheim, Heidelberg University, Mannheim, Germany
GUILHERME C.F. CRUZ • Chemical Biology Laboratory, Institute of Chemistry, Organic
GILBERTO BARBOSA DOMONT • Laboratory of Protein Chemistry–Proteomics Unit,
Chemistry Institute, Federal University of Rio de Janeiro, Rio de Janeiro, Brazil
DANIELLA E. DUQUE-GUIMARÃES • University of Cambridge Metabolic Research
Laboratories and MRC Metabolic Diseases Unit, Wellcome Trust-MRC Institute of
Metabolic Science, Addenbrooke’s Hospital, Cambridge, UK; Department of Physiology
and Biophysics, Institute of Biomedical Sciences, University of São Paulo, SP, Brazil
VITOR M. FAÇA • Department of Biochemistry and Immunology, Ribeirão Preto Medical
School, University of São Paulo, RibeirãoPreto, SP, Brazil; Center for Cell Based
Therapy - Hemotherapy Center of Ribeirão Preto, Ribeirão Preto Medical School,
University of São Paulo, Ribeirão Preto, SP, Brazil
MICHAELA D. FILIOU • Department Stress Neurobiology and Neurogenetics, Max Planck
Institute of Psychiatry, Munich, Germany
SHEILA GARCIA • Laboratory of Neuroproteomics, Department of Biochemistry and Tissue
Brazil
PAUL C. GUEST • Laboratory of Neuroproteomics, Department of Biochemistry and Tissue
Brazil
xiii
xiv Contributors
STEVE F.C. HAWKINS • Bioline Reagents Limited, Unit 16, The Edge Business Centre,
London, UK
KERRY S. JONES • MRC Elsie Widdowson Laboratory, Cambridge, UK
ALBERT KOULMAN • MRC Elsie Widdowson Laboratory, Cambridge, UK
GIUSEPPINA MACCARRONE • Department of Translational Research in Psychiatry, Max
Planck Institute of Psychiatry, Munich, Germany
JAMES R. MANSFIELD • Andor Technology, Concord, MA, USA
JUAN L. MARTINEZ-HURTADO • Department of Chemical Engineering and Biotechnology,
University of Cambridge, Cambridge, UK
DANIEL MARTINS-DE-SOUZA • Department of Biochemistry and Tissue Biology, Neurobiology
Center and Laboratory of Neuroproteomics, Institute of Biology, University of Campinas
(UNICAMP), Campinas, SP, Brazil
SARAH MEADOWS • MRC Human Nutrition Research, Cambridge, UK
FÁBIO CÉSAR SOUSA NOGUEIRA • Laboratory of Protein Chemistry—Proteomics Unit,
Chemistry Institute, Federal University of Rio de Janeiro, Rio de Janeiro, Brazil
ERIKA VELÁSQUEZ NÚÑEZ • Laboratory of Protein Chemistry— Proteomics Unit, Chemistry
Institute, Federal University of Rio de Janeiro, Rio de Janeiro, Brazil
THOMAS PRATES ONG • University of Cambridge Metabolic Research Laboratories and
MRC Metabolic Diseases Unit, Wellcome Trust-MRC Institute of Metabolic Science,
Addenbrooke’s Hospital, Cambridge, UK; Food Research Center (FoRC) and Faculty of
Pharmaceutical Sciences, University of São Paulo, São Paulo, Brazil
SUSAN E. OZANNE • University of Cambridge Metabolic Research Laboratories and MRC
Metabolic Diseases Unit, Wellcome Trust-MRC Institute of Metabolic Science,
Addenbrooke’s Hospital, Cambridge, UK
HARALD PETER • Fraunhofer Institute for Cell Therapy and Immunology, Branch
JOÃO G.M. PONTES • Chemical Biology Laboratory, Institute of Chemistry, Organic
ANN PRENTICE • MRC Elsie Widdowson Laboratory, Cambridge, UK
HASSAN RAHMOUNE • Department of Chemical Engineering and Biotechnology, University
of Cambridge, Cambridge, UK
INEZ SCHOENMAKERS • MRC Elsie Widdowson Laboratory, Cambridge, UK
EMANUEL SCHWARZ • Department of Psychiatry and Psychotherapy, Medical Faculty
Mannheim, Central Institute of Mental Health, Heidelberg University, Mannheim,
Germany
SUSANA SILBERSTEIN • Instituto de Investigación en Biomedicina de Buenos Aires (IBioBA),
CONICET, Partner Institute of the Max Planck Society, Buenos Aires, Argentina
GUSTAVO HENRIQUE MARTINS FERREIRA SOUZA • Mass Spectrometry Applications and
Development Laboratory, Waters Corporation, São Paulo, SP, Brazil
RAFAEL N. DE SOUZA • Chemical Biology Laboratory, Institute of Chemistry, Organic
JOHANN STEINER • Department of Psychiatry, University of Magdeburg, Magdeburg,
Germany
LAURIE STEPHEN • Ampersand Biosciences, Saranac Lake, NY, USA
LJUBICA TASIC • Chemical Biology Laboratory, Institute of Chemistry, Organic Chemistry
Department, State University of Campinas, Campinas, SP, Brazil
Contributors xv
CHRISTOPH W. TURCK • Department of Translational Research in Psychiatry, Max Planck

Institute of Psychiatry, Munich, Germany
JOHANNES VEGT • Appamedix UG i.gr, Innovations-Centrum CHIC, Berlin, Germany
JULIA WIENKE • Fraunhofer Institute for Cell Therapy and Immunology, Branch
ALI K. YETISEN • Harvard Medical School and Wellman Center for Photomedicine,
Massachusetts General Hospital, Cambridge, MA, USA; Harvard-MIT Division of
Health Sciences and Technology, Massachusetts Institute of Technology, Cambridge,
MA, USA
SEOK H. YUN • Harvard Medical School and Wellman Center for Photomedicine,
Massachusetts General Hospital, Cambridge, MA, USA; Harvard-MIT Division
of Health Sciences and Technology, Massachusetts Institute of Technology, Cambridge,
MA, USA
Part I
Reviews
Chapter 1
Application of Multiplex Biomarker Approaches

to Accelerate Drug Discovery and Development
Abstract
Multiplex biomarker tests are becoming an essential part of the drug development process. This chapter
explores the role of biomarker-based tests as effective tools in improving preclinical research and clinical
development, and the challenges that this presents. The potential of incorporating biomarkers in the clini-
cal pipeline to improve decision making, accelerate drug development, improve translation, and reduce
development costs is discussed. This chapter also discusses the latest biomarker technologies in use to make
this possible and details the next steps that must undertaken to keep driving this process forwards.
Key words Pharmaceutical company, Drug discovery, Biomarker, Genomics, Transcriptomics,

Proteomics, Metabolomics
1 Introduction
Pharmaceutical companies are under pressure to improve their

returns on existing and novel drug discovery efforts. This is an
almost impossible task, considering that the average drug costs
approximately one billion US dollars to develop and takes 10–15
years from initial discovery to the marketing phase [1]. This prob-
lem is compounded by the fact that around 70 % of drugs do not
recover their research and development costs and approximately
90 % fail to yield an adequate return on investment. In addition,
fewer than one in ten new drugs entering clinical trials make it to
the market and some of those that do make it experience with-
drawal and/or litigation [2–5]. Therefore, in order for the phar-
maceutical companies to survive, minimizing these risks has
become one of the most important objectives in drug discovery
projects in recent years. For example, there has been considerable
effort aimed at establishing standard operating procedures to plot
a course through these problems and to help meet the intimidating
regulatory demands. But the regulatory agencies have not just
been standing by idling watching. In order to assist pharmaceutical
Paul C. Guest (ed.), Multiplex Biomarker Techniques: Methods and Applications, Methods in Molecular Biology, vol. 1546,
DOI 10.1007/978-1-4939-6730-8_1, © Springer Science+Business Media LLC 2017
3
4 Hassan Rahmoune and Paul C. Guest
companies in this process, they have encouraged the incorporation

of biomarker-based tests into the drug discovery pipeline and the
Food and Drug Administration (FDA) has initiated efforts to
modernize and standardize all involved procedures to facilitate
delivery of more effective and safer drugs [6]. The FDA has esti-
mated even a 10 % improvement in the ability to predict failure of
drug before it enters the clinical trial phases could save as much as
one hundred million US dollars in development costs per drug [7].
2 A Brief History of Failed Drugs
The need for biomarker tests to guide drug development is per-

haps best seen by recent major failures in this process. Over the last
two decades more than 30 drugs have been withdrawn, mainly as
a result of hepatotoxic or cardiotoxic effects [8]. In 1997, the FDA
recommended that the antihistamine drug Terfendine be with-
drawn from the market due to an association with heart arrhyth-
mia, which could increase the risk of heart attacks and death [9].
In the year 2000, the antidiabetic and anti-inflammatory drug
Troglitazone was withdrawn due to reports of liver toxicity [10].
In 2001, the 3-hydroxy-3-methylglutaryl-coenzyme A reductase
inhibitor Cerivastatin, developed to treat high cholesterol levels,
was withdrawn due to an increased risk of rhabdomyolysis, a severe
condition that causes muscle pain and weakness and can sometimes
result in renal failure and death [11, 12]. In 2003, the antidepres-
sant drug Nefazodone was withdrawn due to liver toxicity [13].
One of the most infamous cases was the withdrawal of the anti-
inflammatory drug Vioxx by Merck, due to reports about its
increased risk of heart attack and stroke [14]. Merck agreed to pay
4.85 billion US dollars in damages three years after the withdrawal
and also had to pay a further 285 million US dollars four years later
in the face of charges that it “duped customers” into buying the
drug [15]. Another infamous case was the serious adverse effects
seen in phase I clinical studies of the monoclonal antibody
TGN1412, produced by TeGenero [16]. TGN1412 was originally
tested as a potential treatment for B cell chronic lymphocytic leu-
kemia and rheumatoid arthritis and had shown no toxic effects in
preclinical studies. The compound was withdrawn in 2006 after six
healthy male volunteers who took part in the phase I trial experi-
enced the beginnings of a “cytokine storm” within 90 min after
receiving it. The phrase cytokine storm describes a proinflamma-
tory effect, resulting in fever, pain, and organ failure. All of the
volunteers required weeks of hospitalization. This and the other
cases stated above indicate that these disasters may not have
occurred if procedures been adopted using safety biomarkers to
guide dosing and/or predict toxicities at an early stage in the drug
discovery process.
Multiplex Biomarkers for Drug Discovery 5
3 Biomarker Impact in the Drug Discovery Process
Estimates are that the total number of potential biomarkers is

higher than one million, which is clearly an overwhelming number.
However, many researchers and pharmaceutical companies have
been investing in multiplex “omics” technologies to assist in sort-
ing through this mass of analytes and help in understanding dis-
eases at a deeper level than ever before. These platforms consist of
genomics, transcriptomics, proteomics, and metabonomics, along
with others. All of these approaches involve identification of molec-
ular fingerprints from clinical samples and convert this into infor-
mation about physiological status. With the help of these multiplex
biomarker approaches, we are just now beginning to able to better
categorize diseases at the molecular level, rather than on symptoms
alone. By finding molecular biomarkers of a disease, early detection
and diagnosis could be improved by simply testing for the presence
of the disease fingerprint. Biomarkers would also assist pharmaceu-
tical companies who could now look for drugs which help to nor-
malize disease-like signatures. These could be used in early
preclinical stages of drug development such as by looking at the
effect of test compounds in disease models. They could also assist
in looking for markers of toxicity prior to entry of the drug into
clinical trials if representative models could be developed. In the
later stages, biomarker tests could be used to help stratify patient
groups in order to identify those who are most likely to benefit
from treatment. This is critical as too many trials may have failed
simply due to the fact that the wrong patients were included in the
study. This alone could save millions in costs since the phase II and
phase III stages of clinical trials are normally the most expensive
steps in the drug discovery process.
4 Multiplex Biomarker Techniques
Biomarkers are physical characteristics that can be measured in bio-

samples and used as an indication of physiological states such as
good health, disease, or toxicity or to predict or monitor drug
response [17]. For practical purposes, it is becoming increasing
important that biomarkers can be measured with high accuracy
and reproducibility, within a short timeframe and at an affordable
cost. A biomarker should also reflect the underlying nature of the
disorder or condition under investigation, at least to some extent.
Many types of biomarker tests have emerged such as DNA sequenc-
ing for genomic studies and DNA microarrays and quantitative
polymerase chain reaction (qPCR) for transcriptomic analyses. The
sections below focus on methods that deliver proteomic and
metabolomic profiles or “fingerprints” in blood samples. Since
changes in physiological states are dynamic in nature, these are

likely to introduce alterations in numerous proteins and metabo-
lites that converge on similar pathways. Most researchers now con-
sider proteomics and metabolomic methods to be among the most
informative regarding physiological status, considering that pro-
teins and metabolites actually carry out, respond to, or are reflec-
tive of most processes of the body. Furthermore, most of the drugs
in use today are designed to turn on or turn off proteins such as
receptors or enzymes, or induce metabolic changes which can be
seen as turnover of various proteins and small molecules.
4.1 Multiplex Blood serum and plasma contain several vital bioactive and regula-
Immunoassay tory molecules, including hormones, growth factors, and cyto-
Analysis of Proteins kines. The problem is that most of these molecules are present only
and Metabolites at exceeding low concentrations and, therefore, measurement of
these requires highly sensitive detection methods. One way of
achieving this is through the use of multiplex immunoassay
approaches [18]. These assays can target both proteins and metab-
olites. They are constructed and carried out as follows: (1) micro-
sphere are loaded with different ratios of red and infrared dyes to
give unique fluorescent signatures; (2) specific capture antibodies
are attached to the surface of micro-spheres with specific signa-
tures; (3) The antibody–sphere conjugates are mixed together to
form the multiplex; (4) the sample is added and the target mole-
cules bind to their respective antibody–sphere conjugates; (5) fluo-
rescently labeled detection antibodies are added in a mixture and
each of these binds to their target molecules in a sandwich format;
and (6) the samples are streamed though a reader and the micro-
spheres analyzed by two lasers for identification and quantification
of the analyte present. In this final step, the lasers identify which
analytes are present by measuring the unique signature of each
dye-loaded micro-sphere and determine the amount of analyte
bound by measuring the fluorescence associated with the fluores-
cent tags on the secondary antibodies (this is proportional to the
analyte concentration).
4.2 Two-Dimensional Two-dimensional gel electrophoresis (2DE) works as follows: (1)

Gel Electrophoresis protein mixtures in bio-samples are first applied to a strip gel allow-
of Proteins ing their separation according to their isoelectric points (this is pH
at which no net charge occurs on the protein) using isoelectric
focusing; (2) next the proteins are separated according to their
apparent molecular weights by application of the strip to the top of
a sodium dodecyl sulfate–polyacrylamide gel and second electro-
phoresis step; and (3) the protein spots in the gels can be visualized
with any number of stains (e.g., Coomassie Blue or Sypro Ruby)
and then quantitated using an imaging software. The 2DE tech-
nique allows the study of many tissue types but there are some
problems with analysis of blood serum or plasma samples. This
occurs mainly due to the fact that blood contains a massive concen-
tration range of proteins spanning at least 14 orders of magnitude
[19]. This means that very abundant proteins such as albumin and
the immunoglobulin chains would appear as large blobs on the
gels and eclipse less abundant proteins such as the cytokines.
However, a key advantage of 2DE is that it can generate informa-
tion on intact proteins including any effects on posttranslational
modifications, such as phosphorylation or glycosylation changes.
This is not as simple using other proteomic methods such as shot-
gun mass spectrometry (below).
4.3 Mass Just about the time that the human genome project was ending, a
Spectrometry revolution in shotgun mass spectrometry began as this was devel-
oped as a sensitive and medium throughput approach for pro-
teomic biomarker identification [20]. The term “shotgun” derives
from the fact that the protein in bio-samples are cleaved with pro-
teolytic enzymes to generate smaller peptides (analogous to shot-
gun pellets), which are the actual analytes. This is performed as
most intact proteins are too large and complex in their structure
to be ionized or analyzed directly in a mass spectrometer. After
proteolysis, the resulting peptides are separated according to phys-
iochemical properties, such as hydrophobicity, using liquid chro-
matography so that they enter the mass spectrometer more or less
one at a time. As the peptides enter the mass spectrometer, they
are ionized by a process such as electro-spray, which is basically
application of an electric charge to evaporate the fluids, leaving
the peptides in a charged plasma state. After this, the peptide ions
are accelerated by magnets in the mass spectrometer towards a
detector at a rate that is inversely proportional to their mass over
charge ratios (m/z). Quantitation can be carried out since the
amount of each peptide hitting the detector per second is directly
proportional to the quantity of the peptide and, by derivation, to
that of the corresponding parent protein. At the same time, the
sequence of the peptide can be determined by streaming in a gas
such as nitrogen, which breaks the peptides into smaller pieces.
The mass of each piece can then be used to derive the amino acid
sequences that make up the peptides and these are used to search
protein databases for identification of the parent proteins. The
main advantage of this method is the ability to detect more diffi-
cult classes of proteins which are not detectable by 2DE approaches.
The disadvantages include the loss of intact protein information
since the proteins are enzymatically digested prior to analysis. This
method is also used for metabolomic analysis although there is no
need for enzymatic cleavage of the molecules as metabolites are
normally of a manageable size and structure. In this case, the sam-
ple can be infused directly into the mass spectrometer, the quan-
tity calculated as described above and the identity determined by
comparisons to known standards in metabolomic databases.
www.Ebook777.com
4.4 1H (Proton)- As with mass spectrometry, proton NMR spectroscopy can be used
Nuclear Magnetic for metabolomic and small molecule analyses, although in this
Resonance (NMR) method no separation or pre-fractionation of the molecules is
Spectroscopy required. Major advantages of this approach include the points
that the sample preparation step is direct and simple and that it is
highly reproducible at the analytical level. One drawback is that it
is less sensitive compared to the mass spectrometry-based metabo-
lomic techniques [21, 22]. The 1H-NMR method can yield infor-
mation about the structural properties of metabolites and is
therefore highly used for identification purposes. This works since
the method can track the behavior of protons as the nuclei of each
proton on the molecule of interest lines up in a strong magnetic
field. The procedure begins with the application of radio frequency
pulses to the sample. This induces the nuclei to change their rota-
tion away from their equilibrium position in line with the axis of
the magnetic field and the frequency of rotation is directly related
to its physiochemical environment within the parent metabolite.
Therefore, by using different combinations of radio pulses, one
can determine how each atom interacts with the other atoms,
yielding the structure and, consequently, the metabolite identity.
Proton NMR spectroscopy can be used to monitor relative changes
in the levels of small molecules and metabolites such as amino
acids, vitamins, neurotransmitters, and neurotransmitter precur-
sors, making it useful for biomarker studies of multiple disorders.
5 Use of Multiplex Biomarker Profiling in the Drug Discovery Process
Biomarker profiling can be used at multiple stages of drug discov-

ery process as shown in Fig. 1 and described below in further
detail. In the discovery phase, multiplex biomarker profiling could
positively impact on target identification, target validation, lead
compound prioritization, and efficacy screening of suitable pre-
clinical models. In addition, applications in the development phase
include the production of surrogate biomarkers for drug efficacy
and for the validation of preclinical models of human diseases.
Perhaps most importantly, any biomarker tests that arise from these
earlier phase can be translated into user friendly point-of-care
devices that can be used to identify disease signatures and for
monitoring drug efficacy or toxicity in the clinical trial phases. In
this way, mechanistic or targeted biomarkers can be used in pre-
clinical or clinical development to validate the suitability of pre-
clinical models and establish and facilitate translational medicine by
providing pharmacological and biological biomarkers to predict
clinical outcome (Fig. 2).
5.1 Target Validation Most existing drug targets are components of a limited number of
molecular networks that have been validated at the biological and
DRUG DISCOVERY PIPELINE
Target Lead Drug toxicity Clinical studies Approval and

idenficaon opmizaon studies Phase I Phase II Phase III launch
Biomarker identification Development of Clinical validation and development of

and validation prototype assay user friendly hand held diagnostic device
Time
Fig. 1 Co-development of drugs with biomarker tests over the stages of drug discovery. The scenario is expected to lead to the development of more efficacious
and safer drugs and reduce the overall process time, leading to greater returns on investment
Multiplex Biomarkers for Drug Discovery
9
10
Translaon
Pre-clinical studies Clinical studies
Genomics Genomics
Biomarkers Biomarkers
Point of
and/or targets care device and/or targets
Transcriptomics Transcriptomics
Proteomics Proteomics
Metabolomics Safety Metabolomics Safety

biomarkers biomarkers
Efficacy Efficacy
Disease Disease
Reverse translaon
Fig. 2 Translation and reverse translation of safety, efficacy, and disease biomarkers to characterize preclinical models and enable clinical development based on
personalized medicine strategies
physiological levels [23]. However, many of these networks have

not been fully elucidated in terms of the interacting transcript, pro-
tein and small molecule components and their relationship to other
biological pathways. Therefore, there is a need for “mining” cell
signaling and whole body networks further in order to identify
novel tractable drug targets. Identification of molecules that oper-
ate as switch factors in the disease process is usually the first stage
of target validation. This can be tested by manipulating the expres-
sion of the target molecules using gain or loss of function methods
in an attempt to induce or reverse the disease phenotype [24].
Increasing the function of the molecule of interest could be
achieved using agonist-type small molecules or by over-expression
technologies. Alternatively, function could be knocked down using
antagonist-like small molecules, ribozymes, small interfering
RNAs, or genetic approaches. In each case, a multiplex molecular
signature could be obtained for monitoring the resulting pheno-
type. Such approaches would give confidence that small molecules
under drug development would have a similar effect and this would
help to drive the project forward.
For successful validation and prioritization of novel drug targets,
it is important to establish the molecular context or interaction path-
ways associated with potential drug targets. This involves under-
standing the disease at the functional level and confirming that the
therapeutic concept works in preclinical models as well as in clinical
proof-of-principle experiments. Genomic, transcriptomic, pro-
teomic, and metabolomic profiling studies can provide this informa-
tion by identifying components of cellular networks that could be
targeted for possible therapeutic intervention. A single-cell tran-
scriptomic profiling approach was used to validate the involvement
of brown adipocyte tissue to protect against obesity and metabolic
disease [25]. This study confirmed the presence of mRNAs encod-
ing brown adipose tissue proteins such as uncoupling protein 1 and
adrenergic receptor-beta 3 at both the mRNA and protein levels,
and identified mRNAs encoding novel proteins such as orphan
g-protein coupled receptors and other receptors regulated by neu-
rotransmitters, cytokines and hormones. One study demonstrated
that multiple methods are essential, including identification of can-
cer cell membrane proteins by mass spectrometry and phenotypic
antibody screening, for the identification and validation of antibody
tractable targets, which can significantly accelerate the therapeutic
discovery process in cancer research [26]. In another investigation,
a stable isotope-mass spectrometry metabolomic profiling approach
was used to interrogate the mechanism of antibiotic action of
d-cycloserine, a second line antibiotic used in the treatment of mul-
tidrug resistant Mycobacterium tuberculosis infections [27]. The
authors used labeled 13C α-carbon-2H-l-alanine for simultaneous
tracking of alanine racemase and d-alanine:d-alanine ligase in
Mycobacterium tuberculosis and found that the latter was more
strongly inhibited than the former by d-cycloserine.
5.2 Lead Many compounds fail in the later stages of drug development
Optimization because of an unanticipated toxicity or poor efficacy [28]. This
calls for a greater understanding of drug properties at an earlier
stage in the development pipeline to help overcome these prob-
lems. One approach would be through the incorporation of appro-
priate multiplex biomarker tests into this stage of the pipeline.
Such tests can be used to generate expression signatures from cells
or tissues treated with new drugs for target identification and vali-
dation, and for delineating mechanism of action. Biomarker signa-
tures can also be used in the identification and optimization of lead
compounds by looking for correlations of specific molecular pat-
terns with efficacy or specific toxicities. For example, monitoring
the effects of developmental compounds on molecular patterns in
the appropriate models might provide an early prediction of effi-
cacy or toxicity [29]. Compounds which induce the same signature
of protein expression changes are presumed to share the same
mode of action and toxicity effects. Recently, Tang et al. reported
on the development of a miniaturized Luminex assay consisting of
a panel of multiplexed assays for measuring cytokines [30]. This
assay facilitates high-throughput screening of compounds in cell
models using cytokines as physiologically relevant molecular read-
outs. In addition, this multiplexed cytokine test can be used for
profiling of bio-fluids such as blood serum and plasma for transla-
tional research. Cell models can provide useful screening platforms
for drug profiling, using reporter systems for activation of receptor
signaling or enzymatic cascades. This approach has been denoted
as cytomics. By screening cell models with drug libraries, func-
tional responses such as calcium flux, phosphorylation signaling
cascades, mitochondrial membrane potential changes, receptor
expression and/or internalization, ligand binding, apoptosis, oxi-
dative stress, proliferation, and cell cycle status can be measured
[31]. The ultimate aim is to use changes in molecular biomarker
patterns to understand how drugs exert their effects at the molecu-
lar level.
5.3 Drug Toxicology Successful drugs should be potent, specific for their targets and
Studies bio-available with good pharmacokinetic profiles and low toxicity.
In the ideal scenario, compounds lacking one or more of these
traits should be identified during the early stages of the drug dis-
covery pipeline so that only the most promising compounds are
taken through to the clinical trial stages. However, toxicities usu-
ally become apparent only during the preclinical or clinical devel-
opment stages when compound testing occurs in animal or cellular
models or in humans. In some cases, toxicities may not even be
detected until widespread distribution of the drug to the general
public [32]. The reasons for this can be complex and on some
occasions attributed to metabolism of the parent compound to
toxic metabolites or to poor clearance. Multiplex omics methods
can accelerate development of the best lead compounds by facili-

tating screening methods for toxicity based signatures.
As a prime example, the EU Framework 6 Project: Predictive
Toxicology (PredTox), studied the effects of 16 test compounds
using both conventional toxicological parameters and multiplex
biomarker approaches technologies [33]. They found three main
classes of toxicity which were: (1) liver hypertrophy, (2) bile duct
necrosis/cholestasis and (3) kidney proximal tubular damage. The
results demonstrated that that the multiplex approaches can help
drug companies to make better informed decisions during early
phase toxicological studies. Toxicogenomics is the term applied to
investigation of drug responses at gene expression level [34]. The
liver is the main tissue targeted in this approach. DNA microarray
profiling studies are now being carried out with known classes of
toxicity inducing compounds with the objective that these can be
referenced against novel compounds [35, 36]. In addition, the
potential mechanisms of hepatotoxicity of doxorubicin-loaded
microspheres in chemoembolization have been investigated by
DNA microarray analyses combined with histological examinations
[37]. This showed that doxorubicin caused lesions to the liver and
disturbed liver metabolism-related enzymes. Another DNA micro-
array study investigated liver toxicity of rotodrine, a compound
that has been used in preterm labor [38]. They found a specific
increase in the levels of serum amyloid A, which was not induced
by other hepatotoxic drugs like acetaminophen, valproic acid, or
metformin. The increase in serum amyloid A was also more sensi-
tive as a biomarker compared to the commonly measured liver
enzymes aspartate aminotransferase and alanine aminotransferase.
Accessible bio-fluids such as spittle, serum, plasma, and urine hold
the most immediate promise for preclinical assessment in terms of
better biomarkers. Andersson and colleagues analyzed plasma sam-
ples from 134 patients using proteomic and metabolomic
approaches, with the aim of finding predictive biomarkers to
explain the liver toxicity induced by ximelagatran, a compound
developed for the prevention and treatment of thromboembolic
conditions [39]. They found changes in 3-hydroxybutyrate, pyru-
vic acid, colony stimulating factor 1 receptor, l-glutamine, protein
S, and alanine, as well as changes in other molecules. This approach
helped to generate a new hypothesis for an unknown mechanism
of toxicity.
In addition, cellular models could be useful in preclinical toxic-
ity screening. Meneses-Lorente and coworkers used a two-
dimensional differential in-gel electrophoresis and mass spectrometry
profiling approach to identify a proteomic signature associated with
hepatocellular steatosis in rats after dosing with a compound in pre-
clinical development [40]. Within 6 h of dosing, livers showed
hepatocellular vacuolation, which increased in extent and severity
over time. Although alterations in alanine aminotransferase and
aspartate aminotransferase were not detected until day three, pro-

teomic profiling changes were observed at the earliest time point
and many were associated with liver steatosis. The proteins which
showed increased levels were pyruvate dehydrogenase, phenylala-
nine hydroxylase and 2-oxoisovalerate dehydrogenase, which are all
involved in acetyl-CoA production. One of the decreased proteins
was sulfite oxidase, which is involved in triglyceride accumulation.
5.4 Clinical Studies Clinical applications of multiplex biomarker approaches include

early detection of the disease using molecular signatures in bio-
fluids as a complement to other measures carried out which spe-
cifically target the affected biological pathway in patients. Once a
particular signature for compound efficacy has been established,
this can be applied in a high throughput format such as screening
with multiplexed immunoassays to help in the identification of
compounds with optimum profiles. Also, prognostic biomarkers
can be used to help predict drug efficacy in patients and to poten-
tially identify which individuals are likely to benefit from treat-
ment with a specific drug [41, 42]. Such approaches can help
pave the way for development of more individualized therapies,
using personalized medicine approaches [43]. The ultimate appli-
cation of proteomics in drug discovery would be to identify bio-
markers in a readily accessible body fluid, such as serum or plasma,
and which can be correlated with the initiation of efficacy or
severity of toxicity. Such biomarker signatures could also be used
as surrogate markers or secondary endpoints to help predict the
responses of individuals to treatment and allow adjustments of
the therapy to achieve highest possible efficacy without reaching
a level which elicits toxic side effects. Likewise, this approach
could be used to facilitate identification of patient classes who
will respond favorably to the drug in clinical trials as part of the
personalized medicine strategy.
Kopetz et al. investigated the efficacy of fluorouracil (FU),
leucovorin, irinotecan, and bevacizumab in a phase II trial in
patients who were previously untreated for metastatic colorectal
cancer, and measured changes in plasma cytokines and angiogenic
factors as potential markers of treatment response using a multi-
plex immunoassay platform [44]. They found that elevated levels
of interleukin 8 at baseline were associated with a shorter progres-
sion-free survival period and changes in basic fibroblast growth
factor, hepatocyte growth factor, placental growth factor, stromal-
derived factor-1, and macrophage chemoattractant protein-3 were
associated with angiogenesis and myeloid recruitment in the pro-
gressive disease phase. Another clinical study investigated the
effects of daily coffee consumption as potential risk factors for
type II diabetes using gas chromatography mass spectrometry and
multiplex immunoassay analysis approaches [45]. More recently,
analysis of the biomarker results from the AVADO phase III trial
of first-line bevacizumab plus docetaxel for HER2-negative meta-
static breast cancer showed that plasma levels of vascular endothe-
lial growth factor (VEGF)-A and VEGF receptor-2 are potential
predictive markers for bevacizumab efficacy [46].
6 Conclusions and Future Perspectives
This chapter describe the emerging use of multiplex biomarker

profiling techniques as enabling platforms for use in all aspects of
the drug discovery process. This is critical as current diagnostic
procedures and strategies for developing novel medicines are in
need of a paradigm change [47]. The regulatory health authorities
now consider the incorporation of biomarkers into clinical plat-
forms to be of high importance for the future of drug discovery
and have introduced schemes to modernize methods, tools and
techniques to achieve this goal.
Multiplex biomarker tests have now been available for more
than two decades. In general most of the platforms have medium
to large footprints and require expert technicians in order to
operate them. Another drawback is that each of these methods
has a typical turnaround time of approximately one day from the
sample preparation stages to the final results presentation for a
given sample. Within the last 5 years, multiplex biomarker tests
have been miniaturized using microfluidic approaches to yield
devices which are approximately the size of a small pamphlet or
a credit card [48]. Most importantly, these devices are user
friendly since no expertise is required for operation and the
results can be returned in less than 15 min from a single drop of
blood. There have also been recent developments which allow
connection of these biomarker cards with smartphone technol-
ogy using cleverly designed apps. For example, multiplex immu-
noassay-based tests have been developed on a handheld
smartphone-based colorimetric reader using an optomechanical
interface and this has been tested successfully in the case of
detecting mumps, measles and herpes simplex I and II viruses
[49]. It is not hard to imagine that similar tests for other diseases
will be available in the not so distant future. Such devices would
also meet the requirements of clinical studies and slot nicely into
the pipeline in phase I–III clinical studies, considering their
robustness, rapidity, and user friendly operation. This should
help to inject renewed vigor into the pharmaceutical industry
and most importantly help to improve the lives of the patients by
enabling personalized medicine approaches.
References
1. Hughes JP, Rees S, Kalindjian SB, Philpott KL 16. Sheridan C (2006) TeGenero fiasco prompts
(2011) Principles of early drug discovery. Br regulatory rethink. Nat Biotechnol 24:475–476
J Pharmacol 162:1239–1249 17. Atkinson AJ, Colburn WA, DeGruttola VG,
2. Giezen TJ, Mantel-Teeuwisse AK, Straus SM, DeMets DL, Downing GJ, Hoth DF,
Schellekens H, Leufkens HG, Egberts AC Biomarkers Definitions Working Group et al
(2008) Safety-related regulatory actions for (2001) Biomarkers and surrogate endpoints:
biologicals approved in the United States and preferred definitions and conceptual frame-
the European Union. JAMA 300:1887–1896 work. Clin Pharmacol Ther 69:89–95
3. McNaughton R, Huet G, Shakir S (2013) An 18. Fulton RJ, McDade RL, Smith PL, Kienker LJ,
investigation into drug products withdrawn Kettman JR Jr (1997) Advanced multiplexed
from the EU market between 2002 and 2011 for analysis with the FlowMetrix system. Clin
safety reasons and the evidence used to support Chem 43:1749–1756
the decision-making. BMJ Open 4:e004221 19. Anderson NL, Anderson NG (2002) The
4. Onakpoya IJ, Heneghan CJ, Aronson JK human plasma proteome: history, character,
(2016) Post-marketing withdrawal of 462 and diagnostic prospects. Mol Cell Proteomics
medicinal products because of adverse drug 1:845–867
reactions: a systematic review of the world lit- 20. Link AJ, Eng J, Schieltz DM, Carmack E, Mize
erature. BMC Med 14:10 GJ, Morris DR et al (1999) Direct analysis of
5. Rawson NS (2016) Drug safety: withdrawn protein complexes using mass spectrometry.
medications are only part of the picture. BMC Nat Biotechnol 17:676–682
Med 14:28 21. Griffin JL (2003) Metabonomics: NMR spec-
6. Ovens J (2006) Funding for accelerating drug troscopy and pattern recognition analysis of
development initiative critical. Nat Rev Drug body fluids and tissues for characterisation of
Discov 5:271 xenobiotic toxicity and disease diagnosis. Curr
7. http://www.fda.gov/oc/initiatives/critical- Opin Chem Biol 7:648–654
path/whitepaper.html 22. Beckonert O, Keun HC, Ebbels TM, Bundy J,
8. Need AC, Motulsky AG, Goldstein DB (2005) Holmes E, Lindon JC et al (2007) Metabolic
Priorities and standards in pharmacogenetic profiling, metabolomic and metabonomic pro-
research. Nat Genet 37:671–681 cedures for NMR spectroscopy of urine,
9. h t t p : / / w w w. f d a . g o v / o h r m s / d o c k e t s / plasma, serum and tissue extracts. Nat Protoc
ac/98/briefingbook/1998- 3454B1_03_ 2:2692–2703
WL50.pdf 23. Imming P, Sinning C, Meyer A (2006) Drugs,
10. Cohen JS (2006) Risks of troglitazone appar- their targets and the nature and number of drug
ent before approval in USA. Diabetologia targets. Nat Rev Drug Discov 5:821–834
49:1454–5145 24. Sioud M (2007) Main approaches to target dis-
11. Onakpoya IJ, Heneghan CJ, Aronson JK covery and validation. Methods Mol Biol
(2016) Worldwide withdrawal of medicinal 360:1–12
products because of adverse drug reactions: a 25. Spaethling JM, Sanchez-Alavez M, Lee J, Xia
systematic review and analysis. Crit Rev Toxicol FC, Dueck H, Wang W (2016) Single-cell
3:1–13 [Epub ahead of print] transcriptomics and functional target validation
12. Marciante KD, Durda JP, Heckbert SR, of brown adipocytes show their complex roles
Lumley T, Rice K, McKnight B et al (2011) in metabolic homeostasis. FASEB J 30:81–92
Cerivastatin, genetic variants, and the risk of 26. Rust S, Guillard S, Sachsenmeier K, Hay C,
rhabdomyolysis. Pharmacogenet Genomics Davidson M, Karlsson A et al (2013) Combining
21:280–288 phenotypic and proteomic approaches to iden-
13. Choi S (2003) Nefazodone (Serzone) with- tify membrane targets in a ‘triple negative’ breast
drawn because of hepatotoxicity. CMAJ cancer cell type. Mol Cancer 12:11
169:1187 27. Prosser GA, de Carvalho LP (2013) Metabolomics
14. Dogné JM, Hanson J, Supuran C, Pratico D reveal d-alanine:d-alanine ligase as the target of
(2006) Coxibs and cardiovascular side-effects: d-cycloserine in Mycobacterium tuberculosis.
from light to shadow. Curr Pharm Des ACS Med Chem Lett 4:1233–1237
12:971–975 28. Amacher DE (2010) The discovery and devel-
15. http://www.bloomberg.com/news/arti- opment of proteomic safety biomarkers for the
cles/2013-07-18/merck-pays- detection of drug-induced liver toxicity.
23-million-to-end-vioxx-drug-purchase-suits Toxicol Appl Pharmacol 245:134–142
29. Meneses-Lorente G, Watt A, Salim K, approach to understanding elevated serum ala-

Gaskell SJ, Muniappa N, Lawrence J et al nine transaminase levels in a clinical trial with
(2006) Identification of early proteomic ximelagatran. Biomarkers 14:572–586
markers for hepatic steatosis. Chem Res 40. Meneses-Lorente G, Guest PC, Lawrence J,
Toxicol 19:986–998 Muniappa N, Knowles MR, Skynner HA et al
30. Tang H, Panemangalore R, Yarde M, Zhang L, (2004) A proteomic investigation of drug-
Cvijic ME (2016) 384-well multiplexed induced steatosis in rat liver. Chem Res Toxicol
luminex cytokine assays for lead optimization. 17:605–612
J Biomol Screen. pii: 1087057116644164. 41. Killestein J, Polman CH (2011) Determinants
(Epub ahead of print) of interferon β efficacy in patients with multiple
31. Valet G (2006) Cytomics as a new potential for sclerosis. Nat Rev Neurol 7:221–228
drug discovery. Drug Discov Today 42. Gerger A, Labonte M, Lenz HJ (2011)
11:785–791 Molecular predictors of response to antiangio-
32. Gale EA (2001) Lessons from the glitazones: a genesis therapies. Cancer J 17:134–141
story of drug development. Lancet 43. Flood DG, Marek GJ, Williams M (2011)
357:1870–1875 Developing predictive CSF biomarkers-A chal-
33. Suter L, Schroeder S, Meyer K, Gautier JC, lenge critical to success in Alzheimer’s disease
Amberg A, Wendt M et al (2010) EU frame- and neuropsychiatric translational medicine.
work 6 project: predictive toxicology Biochem Pharmacol 81:1422–1434
(PredTox)--overview and outcome. Toxicol 44. Kopetz S, Hoff PM, Morris JS, Wolff RA, Eng
Appl Pharmacol 252:73–84 C, Glover KY et al (2010) Phase II trial of infu-
34. Afshari CA, Hamadeh HK, Bushel PR (2011) sional fluorouracil, irinotecan, and bevaci-
The evolution of bioinformatics in toxicology: zumab for metastatic colorectal cancer: efficacy
advancing toxicogenomics. Toxicol Sci and circulating angiogenic biomarkers associ-
120(S1):S225–S237 ated with therapeutic resistance. J Clin Oncol
35. Bulera SJ, Eddy SM, Ferguson E, Jatkoe TA, 28:453–459
Reindel JF, Bleavins MR et al (2001) RNA 45. Kempf K, Herder C, Erlund I, Kolb H, Martin
expression in the early characterization of hep- S, Carstensen M et al (2010) Effects of coffee
atotoxicants in Wistar rats by high-density consumption on subclinical inflammation and
DNA microarrays. Hepatology 33:1239–1258 other risk factors for type 2 diabetes: a clinical
36. Waring JF, Ciurlionis R, Jolly RA, Heindel M, trial. Am J Clin Nutr 91:950–907
Ulrich RG (2001) Microarray analysis of hepa- 46. Miles DW, de Haas SL, Dirix LY, Romieu G,
totoxins in vitro reveals a correlation between Chan A, Pivot X et al (2013) Biomarker results
geneexpression profiles and mechanisms of from the AVADO phase 3 trial of first-line bev-
toxicity. Toxicol Lett 120:359–368 acizumab plus docetaxel for HER2-negative
37. Verret V, Namur J, Ghegediban SH, Wassef M, metastatic breast cancer. Br J Cancer
Moine L, Bonneau M et al (2013) Toxicity of 108:1052–1060
doxorubicin on pig liver after chemoemboliza- 47. Poste G (2011) Bring on the biomarkers.
tion with doxorubicin-loaded microspheres: a Nature 469:156–157
pilot DNA-microarrays and histology study. 48. Schumacher S, Nestler J, Otto T, Wegener M,
Cardiovasc Intervent Radiol 36:204–312 Ehrentreich-Förster E, Michel D et al (2012)
38. Tsuchiya H, Sato J, Tsuda H, Fujiwara Y, Highly-integrated lab-on-chip system for
Yamada T, Fujimura A et al (2013) Serum point-of-care multiparameter analysis. Lab
amyloid A upsurge precedes standard biomark- Chip 12:464–473
ers of hepatotoxicity in ritodrine-injected mice. 49. Berg B, Cortazar B, Tseng D, Ozkan H, Feng
Toxicology 305:79–88 S, Wei Q et al (2015) Cellphone-based hand-
39. Andersson U, Lindberg J, Wang S, held micro-plate reader for point-of-care test-
Balasubramanian R, Marcusson-Ståhl M, ing of enzyme-linked immunosorbent assays.
Hannula M et al (2009) A systems biology ACS Nano 9:7857–7866
Chapter 2
The Application of Multiplex Biomarker Techniques

for Improved Stratification and Treatment of Schizophrenia
Patients
Johann Steiner, Paul C. Guest, Hassan Rahmoune,
Abstract
In the case of major psychiatric disorders such as schizophrenia, shortcomings in the conversion of scientific
discoveries into newer and safer treatment options has led to a loss of confidence and precipitated a crisis
for large pharmaceutical companies. This chapter describes how incorporation of multiplex biomarker
approaches into the clinical pipeline can lead to better patient characterization, delivery of novel treatment
approaches and help to renew efforts in this important area. The development of specific biomarker test
panels for disease prediction should facilitate early intervention strategies, which may help to slow disease
development or progression. Furthermore, the development of such tests using lab-on-a-chip and smart-
phone platforms will help to shift diagnosis and treatment of this major disorder into a point-of-care
setting for improved patient outcomes.
Key words Schizophrenia, Blood-based biomarkers, Proteomics, Multiplex immunoassay, Mass

spectrometry, Point-of-care, Lab-on-a-chip, Smartphone apps
1 Introduction
Schizophrenia is a debilitating, mental health disorder which can

strike individuals in their late teens or early adulthood and seriously
impair medical health, quality of life, social well-being and produc-
tivity [1]. Clinical presentation usually occurs with symptoms such
as hallucinations, delusions, anhedonia, social retreat, disorganized
thinking and cognition impairment. At present, diagnosis is still
based on expression of symptoms and is dependent on communi-
cations between the affected individual and the attending clinician
or psychiatrist. This is usually achieved in an interview-like format
using the Diagnostic and Statistical Manual of Mental Disorders
(DSM) [2] or the International Classification of Diseases (ICD-
10) [3] criteria as guidelines. However, these texts can only detail
19
20 Johann Steiner et al.
Treatment of schizophrenia
Old paradigm New paradigm
Reacve care Personalised care
Disease severity
Disease severity
Switch drug again
Switch drug
Select drug Treat with right drug

Monitoring
Diagnosis Diagnosis/prognosis
Predisposion screening
Time Time
Diagnose disease Health management
Treat symptoms Molecular screening
Costly trial & error treatment Early detection
Rapid effective treatment
Improved quality of care
Fig. 1 Comparison of the old and new treatment paradigms involving schizophrenia patients, distinguished by
the use of biomarkers for improved stratification
the symptoms of schizophrenia without pointing to the underlying

molecular physiological pathways that may be affected. Further-
more, classification of a person as having schizophrenia can be con-
founded by the fact that individuals with other psychiatric disorders
can share many of the same symptoms. For this reason, there are
now concerted efforts to identify specific multiplex biomarker fin-
gerprints that can potentially predict the onset of schizophrenia,
improve diagnostic accuracy, monitor disease progression, and
guide treatment options. The availability of such tests for use in
blood serum or plasma would be ideal as this would facilitate use
in clinical settings. This is because blood-based biomarkers would
have high accessibility in clinical practice due to the low invasive-
ness of the sampling procedure and the low associated costs.
The application of biomarker-based diagnostic tests that can
accurately classify patients according to the type of disorder or even
disease subtype will help to reduce duration of untreated mental
illness and improve individual responses by placing the right
patients on the right treatments as early as possible. This is because
there is a direct correlation between longer periods without treat-
ment and poor outcomes [4]. It is thought that this will change
the overall paradigm from reactive psychiatric care to a more opti-
mized personalized treatment approach in the field of psychiatry as
well as in other areas of medicine (Fig. 1). Also, implementation of
earlier effective treatment should help to reduce patient referral to
Biomarkers for Schizophrenia 21
secondary services such as hospitals, community groups, and crisis

teams. Any reduction in the use of these expensive services will
help to reduce the overall financial burden of psychiatric disorders,
which surpassed 60 billion dollars per year in the 1990s in the USA
alone [5]. More importantly, an early successful intervention will
help to curb symptom severity. This is because schizophrenia may
lead to decades of life disability [6], more than double that of car-
diovascular disorders [7].
The discovery of validated biomarker tests that reflect the cor-
relation between the patient clinical and molecular readouts, would
also enhance future mental healthcare significantly if the resulting
tests can be incorporated into standard operating systems and clini-
cal decision making, as well as being deployed as fast, cost-effective,
user friendly, point-of-care devices. The strictest classification of
newly developed biomarker tests requires that results must be rep-
licated in different laboratories and in different sites. In the case of
psychiatric disorders like schizophrenia, this will be difficult to
achieve. The major reason for this is that these conditions are
poorly understood at the molecular level and there is high hetero-
geneity in the way that they are manifested in the affected persons
[8]. In this chapter, we discuss the challenges and requirements of
developing and rolling out molecular biomarker tests for schizo-
phrenia. In addition, the chapter focuses on the use of biomarkers
for improved classification and management of patients with
schizophrenia for improved point-of-care treatment and as a means
of rekindling drug discovery efforts within the pharmaceutical
industry in the area of psychiatric disorders.
2 Current Diagnosis of Schizophrenia
Most psychiatrists and clinicians agree that schizophrenia is a gen-

eral term for a mixture of mental conditions that present with simi-
lar symptoms, in the same way that most people with acute
infectious disorders present with an elevated body temperature [8].
It should be emphasized that the enormous variety of psycho-
pathological alterations summarized under the term “schizophre-
nia” do not represent a disease entity, but rather a hypothetical
construct that was created many decades ago by leading authorities
in the field and is now defined by international classifications com-
mittees, which have inclusion criteria that are changed from issue
to issue. However, this crossover can lead to misdiagnosis in psy-
chiatric practice. As an example, one investigation found that more
than 30 % of patients who actually had bipolar disorder were ini-
tially diagnosed as having schizophrenia [9]. Another study chal-
lenged the basis of the current classifications systems by pointing
out that there are no current methods to validate the basic con-
cepts which are independent of the same concepts [10]. In any
event, psychiatrists do not always use these classification systems
for making a diagnosis. In many cases, diagnosis may be made

based on experience and personal views in a more heuristic man-
ner. Again, this is not ideal as it can result in errors based on mis-
conceptions, biases or selective memories.
Aside from these issues, the DSM and ICD-10 classification
systems work based on the framework that mental disorders such
as schizophrenia are distinct diseases with common etiologies
which can be defined by criteria based on signs and symptoms. In
reality, it is often not the case that specific symptoms are linked to
defined diseases. For example, individuals with traumatic disorders,
infectious diseases, metabolic conditions or even those under the
influence of certain substances can present with symptoms that
occur in schizophrenia [11, 12]. In addition, it is not uncommon
for a diagnosis to change over time. A long-term study found sig-
nificant changes in diagnosis from major depressive disorder to
bipolar disorder and schizophrenia [13] and another found that
only half of the patients stayed on their initial diagnosis [14].
2.1 The Importance The concordance rate for a diagnosis of schizophrenia in identical
of Early Diagnosis twins ranges from 10 to 70 % [15–17]. Although this provides evi-
dence that there can be a genetic predisposition for schizophrenia,
it also indicates that an individual will not necessarily develop
schizophrenia even when a potential genetic effect is present. In
fact environmental and other nongenetic factors are also impor-
tant. Factors which could precipitate schizophrenia include preg-
nancy or delivery complications, such as infections, hypoxia or
malnutrition [18, 19], as well as nonbiological factors, including
social stressors such as experiencing a natural disaster, loss of a fam-
ily member, or the chronic experiences of an unbearable environ-
ment such as an intolerable work situation, a dysfunctional family
life, or an abusive relationship [20]. On a positive note, the pres-
ence of an environmental component also suggests that disease
prevention or minimization might be possible if the responsible
factors can be identified and avoided.
It is not difficult to imagine that certain environmental factors
such as poor nutrition, social stress, and physical trauma can affect
a person’s physiological state. Several research groups have now
shown that metabolic abnormalities such as insulin resistance occur
in 20–50 % of schizophrenia patients at their first clinical presenta-
tion [21–23]. Furthermore, multiple research groups have found
alterations in circulating inflammatory and immune response
markers in first onset schizophrenia patients [24, 25]. Two studies
have now shown that such changes can occur months to years
before full clinical manifestation of schizophrenia symptoms, sug-
gesting that perturbations in these molecular pathways may play a
role in the disease etiology [26, 27]. This also gives hope for iden-
tifying individuals at risk of developing the disease at the earliest
possible phase, as described above This is important as numerous

reports have now described importance of early intervention
therapeutics for individuals at high risk of developing schizophre-
nia [28–30]. Any delay in diagnosis can have detrimental effects
on the lives of the patients, such as the patient experiencing a full
blown psychosis leading to other problems including substance
abuse, alienation from family and friends, increased accidents, and
the potential of self-harm [31, 32]. There is the problem of misdi-
agnosis which can lead to inappropriate treatments, which can
either be ineffective or even harmful to the patient. In addition,
misdiagnosis followed by inappropriate treatment can have a num-
ber of socioeconomic consequences, such as inflated medical costs,
work absence, and harmful effects on family and relationships [33].
2.2 The Development The European health authorities have lent support to the develop-
of Biomarker Assays ment and implementation of biomarkers through agencies such as
for Diagnosis the Innovative Medicines Initiative [34, 35]. This began as a part-
of Schizophrenia nership between the European Commission and the major phar-
maceutical companies with the overall objective of promoting
more efficient discovery and development of better medicines. A
key objective is the discovery of translational biomarkers which, in
this case, means incorporating them into drug discovery pipelines
for use in human clinical studies. The European Commission con-
tributed one billion Euros to this project and this has been matched
in kind by contributions from the participating companies.
Diagnostic biomarker tests in the USA are regulated by the
Clinical Laboratory Improved Amendments (CLIA) agency [36].
These imposed regulatory standards govern any tests that are
performed in a clinical setting on human samples for the purpose
of diagnosis, disease prevention, treatment or assessment of health.
Commercially available tests marketed under CLIA are categorized
by the FDA depending on the potential risks for health. The devel-
opment of diagnostic biomarker tests for all diseases including psy-
chiatric disorders requires repeated demonstrations of precise
performance characteristics including scores such as sensitivity and
especially specificity, given the symptomatic and molecular overlap
among all psychiatric disorders. This is an absolute requirement
since biomarker measurements can be affected by many factors
including biological, ethnicity, gender, environmental, sample col-
lection, and analytical variables. For example, development of mul-
tiplexed immunoassays requires the testing and validation of each
component immunoassay as well as the combination of assays used
in each multiplex to maximize repeatability, precision and accuracy.
This includes selection and immobilization of capture ligands on
microbeads, calibration steps, testing for reagent–antibody com-
patibility, and ensuring each individual assay has sufficient dynamic
range and the required limits of detection [37].
www.Ebook777.com
Another criterion of biomarker tests that is often overlooked is

that they must be in a format that is high throughput, accurate and
user friendly to allow use by clinicians, hospital staff and scientists.
Along these lines, we suggest that the discovery and implementa-
tion platforms should be different to maximize development of
tests with the highest performance. Thus, although they are pow-
erful discovery tools, mass spectrometry and two-dimensional gel
electrophoresis techniques may be too cumbersome in their larger
formats and may require a high level of expertise to be considered
as realistic options for clinical use. In contrast, an automated plat-
form based on multiplexed immunoassay is a more likely candidate
as a clinically friendly platform as it has already shown some prom-
ise in this area. However, even this would be too slow in its discov-
ery format.
3 Biomarkers Identified for Schizophrenia
Although genomic studies are able to identify genes conferring

susceptibility to a particular disease, the functional abnormalities of
most disorders are reflected ultimately in the proteome and metab-
olome. This is because proteins and metabolites represent the
molecular phenotype of a disease in parallel with the clinical mani-
festation. Recent years have seen the increasing use of proteomics
as a tool for the discovery of biomarkers for diagnosis, monitoring
disease progression, treatment response and for the identification
of novel therapeutic targets. It is also important to remember that
analysis of central nervous system (CNS) disorders is difficult as the
brain is not readily accessible for invasive diagnostic purposes.
Thus, sources such as serum and plasma have been undergoing
increasing scrutiny as they have a higher utility in the clinic.
3.1 Inflammation A multiplex immunoassay profiling study which used cytokine

Biomarkers arrays identified increased levels of interleukin (IL)-1β in cerebro-
in Schizophrenia spinal fluid from first onset schizophrenia patients, suggesting that
the inflammation response may be perturbed in the brains of some
patients [38]. This is consistent with other studies which demon-
strated that brain development can be disturbed by changes in the
balance of pro-inflammatory and anti-inflammatory cytokines [39,
40]. In addition, altered inflammation has been linked to changes
in the glutamate system, the main excitatory neurotransmitter in
the brain. Transcriptomic and proteomic profiling studies of
post mortem brains from schizophrenia patients have identified
increased levels of inflammation-related gene products in oligo-
dendrocytes and endothelial cells in comparison to non-psychiatric
control subjects [41, 42]. However, it is possible that this is a con-
founding factor of prolonged drug treatment or an unhealthy life-
style, as often occurs in the chronic or latter stages of individuals
suffering from this disorder (Fig. 2) [43].
Hypothalamus
Pituitary
Cytokines
Prolactin Cytokines
Interleukins
GH Interleukins
ACTH Clotting
Cascade
Cortisol Thymus Transport

proteins
Insulin
Adrenals
Spleen
Clotting
Testosterone Cascade Cytokines
Pancreas Transport Interleukins

Progesterone Insulin proteins
Gonads
Bone marrow
Fig. 2 Peripheral and central signaling molecules affected in schizophrenia with a focus on inflammation
(white) and hormonal/metabolic (yellow) pathways. The dashed arrows indicate connections via the blood-
stream. ACTH = adrenocorticotrophic hormone; GH = growth hormone. Note that components of the interleu-
kins, cytokines, transport proteins, and clotting cascade are not listed individually for presentation reasons.
See text for more detail
The finding of circulating changes in molecules such as inflam-

matory factors in psychiatric disorders like schizophrenia is what
makes biomarker testing for psychiatric disorders feasible [44, 45].
In addition, these factors may be informative as either trait or state
biomarkers. A meta analysis of circulating inflammation-related
changes in schizophrenia patients showed that cytokines including
IL-12, soluble IL-2 receptor, interferon-γ, and tumor necrosis
factor-α may be useful as trait biomarkers, giving a stable indication
that the disease is present [44]. In contrast, cytokines such as
IL-1β, IL-6, and transforming growth factor-β may represent state
biomarkers, which means that these may be used as readouts for

acute changes in the disease. In addition, there have been many
reports on the discovery of blood-based biomarker signatures
comprised of a large number of inflammation-related proteins
including some components of the clotting cascade and transport
proteins in first onset schizophrenia patients [46, 47].
Inflammation in the periphery can affect brain function
through effects on the hypothalamic–pituitary–adrenal (HPA) axis
(see below) [48, 49]. Activation of inflammatory pathways stimu-
lates release of corticotrophin releasing factor from the hypo-
thalamic region of the brain and this initiates a cycle causing
adrenocorticotrophic hormone (ACTH) to be released from the
pituitary, which in turn drives cortisol release from the adrenal cor-
tex [50]. Along with several other effects in the body, cortisol also
exerts a negative feedback control on the HPA axis by binding to
specific receptors in the brain and pituitary [51]. The link to psy-
chiatric disorders comes from the fact that the HPA cycle also alters
neurotransmitter systems throughout the brain, which are involved
in regulation of mood and behavior. Given this link, it is not
surprising that some investigators have tested the use of anti-
inflammatory drugs such as aspirin or cyclooxygenase-2 inhibitors
in combination with traditional antipsychotics as a novel treatment
approach to relieve some symptoms of schizophrenia, with some
success [52–55]. However, these findings require validation in fur-
ther studies involving larger cohorts.
3.2 Neuroendocrine- Several studies have now shown effects on a number of hormonal
Related Biomarkers systems related to metabolic homeostasis in schizophrenia. A num-
ber of studies over the past decade have shown that impaired fast-
ing glucose tolerance, high insulin levels and insulin resistance
occurs in both first onset [21, 22] and chronic schizophrenia
patients [55, 57], as can occur in type 2 diabetes patients. One
study showed the presence of hepatic insulin resistance in schizo-
phrenia patients using a hyperinsulinemic clamp [58]. In terms of
biomarkers, two studies found that first onset schizophrenia
patients had increased levels of circulating insulinrelated peptides
and high levels of chromogranin A, pancreatic polypeptide, prolac-
tin, progesterone and cortisol, along with lower levels of growth
hormone, in comparison to controls [23, 59], This indicated
altered secretion from several neuroendocrine glands including
pancreatic β cells, pancreatic PP cells, the anterior pituitary, the sex
organs and adrenal glands (Fig. 2). This could have important
implications since chronically high insulin levels can have disrup-
tive effects on brain function such as inducing increased brain
inflammation, aberrant phosphorylation of filamentous structural
proteins and increased deposition of amyloid plaques [60–62].
High insulin levels have also been found to lead to altered function
of neurotransmitter pathways [63] and perturb synaptic plasticity
in brain regions such as the hippocampus [64]. The increased

cortisol secretion is indicative of an activation of the HPA axis, as
described above, which has been identified as a risk factor for
schizophrenia in adolescents [65]. Another study showed gender-
specific changes in the sex hormones estradiol and testosterone in
schizophrenia patients, suggesting effects on the gonadal tissues
[66]. More recent studies found decreased serum levels of
thyroxine, triiodothyronine, and thyroid-stimulating hormone in
schizophrenia patients [67], which may be tied in with the metab-
olism-related hormone changes described above. Another factor to
consider is that many hormones are influenced by circadian rhythms
and it is likely that some of those described above are co-regulated
as part of an oscillatory feedforward–feedback mechanism between
pancreatic islet cells, the pituitary and other components of the
diffuse neuroendocrine system. For example, high insulin secretion
has been associated with increased prolactin levels [68] and dis-
rupted pulsatile release of growth hormone [69].
The repeated finding that hyperinsulinemia occurs in some
first onset schizophrenia patients suggests that drugs which alle-
viate insulin resistance may offer a novel treatment approach.
Furthermore, chronically treated patients can also exhibit high
insulin levels since antipsychotic drugs can induce metabolic side
effects such as insulin resistance and weight gain. Interestingly, the
weight gain appears to be linked to antipsychotic therapeutic effi-
cacy. One investigation showed that changes in body weight, blood
glucose, and leptin levels were associated with improvement in
both positive and negative symptoms of schizophrenia [70]. How-
ever these effects may not be an absolute requirement for improve-
ment as studies that used the insulinsensitizing agents metformin
and rosiglitazone to treat the antipsychotic-induced insulin resis-
tance did so without disrupting the psycho-therapeutic benefits [71].
Therefore, the relationship between metabolism and psychiatric
symptoms requires further scrutiny. It is possible that insulin-sen-
sitizing agents may have a direct effect on alleviating some symp-
toms, such as the cognitive deficits. In support of this possibility,
one study found that patients with mild Alzheimer’s disease who
were given pioglitazone showed improvements in cognition along
with increased regional cerebral blood flow [72].
Drugs which target other hormone systems have also been
tested as a novel means of treating schizophrenia symptoms. Dehy-
droepiandrosterone (DHEA), an adrenal steroid-like compound,
has been tested as a potential addon therapeutic with antipsychotics
and this led to improvements in depression and anxiety symptoms
in some schizophrenia patients [73]. Furthermore, treatment with
the selective estrogen receptor modulator raloxifene resulted in
reduced negative symptoms in postmenopausal females with
schizophrenia compared to controls [74].
3.3 Biomarkers Biomarker tests that can be used for better classification of
for Prediction schizophrenia patients opens up the possibility of better treatment
of Treatment Response options. For example, biomarkers that can be used to predict
response of schizophrenia patients to treatment would be an
important step forward for the well-being of the patients and it will
assist the prescribing physicians, as well as pharmaceutical compa-
nies conducting clinical trials. Genetic studies have shown that
polymorphisms in the histamine 2 receptor (HRH2) gene can be
used to predict response to clozapine treatment in 76 % of schizo-
phrenia cases [75]. Other studies have shown that variants in genes
for dopamine receptors, serotonin receptors and enzymes involved
in drug metabolism or neurotransmitter turnover can have influ-
ence of patient response to treatment including the propensity
to develop certain side effects [76]. Another way of predicting
response is through the use of physiometric measurements such as
waist circumference, adiposity, body mass index (BMI), which
have already been used to predict the development of side effects
such as metabolic syndrome or other insulin resistance with good
sensitivity and specificity [77, 78]. As for blood-based proteomic
biomarkers, one study showed that schizophrenia patients with
higher levels of serum prolactin have a better outcome following 5
years of antipsychotic treatment [79]. Two multiplex immunoassay
serum profiling studies found that the levels of insulin were predic-
tive of improvement in negative symptoms [80] and those of spe-
cific apolipoproteins, growth factors, hormones and interleukins
could be used to predict weight gain [81] in first-onset schizophre-
nia patients after 6 weeks of antipsychotic treatment (Table 1).
Another study showed that the levels of fatty acid binding protein
could be used to predict response to olanzapine [82]. It should be
kept in mind that these three investigations involved study of first
or recent onset patients and biomarker profiles may be different for
more chronic patients. Further studies aimed at retesting these
prototype biomarker panels may lead to development of validated
molecular tests that can be used to identify those patients who are
more likely to respond to particular antipsychotic medications as
well as those who are likely to benefit from add-on compound that
target either the inflammatory or metabolic symptoms. This could
also lead to the opportunity for clinicians to take actions such as
patient assessment, counseling, or even readjusting treatments in
accordance with measured biomarker readouts.
4 Point-of-Care Methods for Use in Schizophrenia
For decades psychiatrists and have acted on the assumption that

psychiatric disorders such as schizophrenia are caused by defects in
brain. However, developments over recent years have resulted in a
new concept involving the whole body in the precipitation and
www.Ebook777.com
Table 1
Significant associations between molecular levels at baseline and changes in either (A)
psychopathology scores (positive and negative syndrome scale—PANSS) or (B) body mass indices
(BMI) after 6 weeks of antipsychotic treatment. R = Spearman correlation coefficient. NS = not
significant. ANCOVA = analysis of covariance [80, 81]
Positive symptoms Negative symptoms
Protein P-value R P-value R

Insulin NS – 0.005 −0.37
Protein ANCOVA R
Apolipoprotein CIII 0.019 −0.33
Apolipoprotein H 0.005 −0.33
Epidermal growth factor 0.025 −0.28
Follicle-stimulating hormone 0.043 −0.28
Interleukin 18 0.015 0.24
Interleukin 25 0.024 −0.26
Interleukin 6 receptor 0.031 −0.30
Matrix metalloproteinase 1 0.011 −0.24
Placental growth factor 0.016 −0.24
Thyroid-stimulating hormone 0.026 −0.23
progression of these conditions. This is because the brain is holistically

integrated in most fundamental biological functions of the body
and therefore the functioning of this organ can be monitored by
examining changes in the molecular composition of the blood.
This is the basis for the use of blood serum or plasma in the study
of psychiatric diseases [46, 47, 83]. This is useful since blood can
be taken from living patients at different stages of the disease or
throughout a treatment course. In the foreseeable future, it is likely
that increased biomarker testing by clinicians will lead to more
extensive “bio-” signatures in individuals that reflect the physio-
logical status occurring in health or disease. Blood serum and
plasma samples contain many molecules such as hormones, growth
factors and cytokines which can only be detected using methods
that are highly sensitive. One of the best methods to achieve this is
the sandwich format of immunoassay [84, 85] and this is the basis
for the multiplex immunoassay platform described above.
4.1 Credit Card- Multiplex immunoassay biomarker tests have now been available
Sized Devices for more than a decade on medium-sized laboratory equipment
and Mobile Phone and with typical turnaround times of around 1 week from the
Apps for User Friendly sample preparation stages to the final results analysis. More recently,
Rapid Testing multiplex methods have been developed using microfluidic
approaches to yield a devices which are approximately the size of a
credit card [86]. This offers the possibility of inexpensive analysis
using either electrochemical or optical read-outs. Most impor-
tantly, this approach is user friendly as no expertise is required for
operation. The protocol involves application of a blood drop to the
card followed by insertion of the card into a book-sized reader/
analyzer and then a diagnosis score can be read out in less than
15 min. The major benefit of this approach is the rapid turnover
time and this will help to minimize waiting times for lab test results,
which can often take several days or even weeks using standard
methods. Furthermore, these devices can connect to a computer
for transmission of data to a smartphone device. Large consumer
market companies like Apple and Google are now showing interest
in the diagnostic market and are exploiting the potential of linking
diagnostic test results with an app driven by smart software. This
would allow testing using real-time, multiplexed sensors, linked
with artificial intelligence through mobile communication systems.
This is of particular relevance to mental disorders such as schizo-
phrenia, since it is generally a long term disease that requires con-
stant monitoring and treatment. A recent review of trials involving
medical care interventions facilitated by smartphones showed that
patient outcomes were improved more than 60 % of the time [87].
Recently, multiplex immunoassay based tests have been developed
on a handheld smartphone-based colorimetric reader using a
3D-printed optomechanical interface [88]. To date, this approach
has been tested successfully in a clinical microbiology laboratory
using mumps, measles and herpes simplex I and II virus immuno-
globulin tests. It is not hard to imagine that similar tests for other
diseases such as psychiatric disorders will be available in the not so
distant future.
5 Conclusions
This chapter describes recent advances using biomarker tests which

can be used for improved diagnosis and classification of individuals
with schizophrenia. The ultimate goal is to provide more informed
treatment options for improved patient outcomes. The use of the
multiplex biomarker approach provides a way of unraveling the
convoluted array of molecular pathways affected in this disorder

and potentially facilitate identification of disease subtypes which
require different treatment approaches. For example, many patients
show distinct patterns of blood-based molecules which suggest the
presence of perturbed inflammation- or metabolism-related path-
ways as described in this chapter. Thus, improved classification of
such patients based on biomarker profiling would enable selection
of better treatment options such as the potential of including
add-on therapeutics targeting these pathways. Finally, the use of
multiplex tests on handheld devices capable of distinguishing schi-
zophrenia patients who are most likely to respond to specific psy-
chiatric medications would be an important breakthrough for
point-of-care applications. This could help to improve the lives of
individuals suffering from this debilitating disorder and have ben-
eficial effects on society as well as significant cost savings for the
healthcare services in general.
Acknowledgments
D.M.S. and the Laboratory of Neuroproteomics, UNICAMP are

funded by FAPESP (São Paulo Research Foundation) grant num-
ber 13/08711-3.
References
1. van Os J, Kapur S (2009) Schizophrenia. Global Burden of Disease Study 2010. Lancet
Lancet 374:635–645 382:1575–1586
2. American Psychiatric Association (2013) Diag- 7. Liu Y, Dalal K, Stollenwerk B (2013) The asso-
nostic and Statistical Manual of Mental ciation between health system development
Disorders, Fifth Edition (DSM-5). American and the burden of cardiovascular disease: an
Psychiatric Publishing, Arlington, VA. ISBN analysis of WHO country profiles. PLoS One
10: 0890425558 8:e61718
3. World Health Organization (1 Jan. 1992) 8. Tsuang MT (1975) Heterogeneity of schizo-
ICD-10: The ICD-10 classification of mental phrenia. Biol Psychiatry 10:465–474
and behavioural disorders: clinical descriptions 9. Gonzalez-Pinto A, Gutierrez M, Mosquera F,
and diagnostic guidelines. World Health Ballesteros J, Lopez P, Ezcurra J et al. (1998)
Organisation. ISBN-10: 9241544228 First episode in bipolar disorder: misdiagnosis
4. Penttilä M, Jääskeläinen E, Hirvonen N, and psychotic symptoms. J Affect Disord
Isohanni M, Miettunen J (2014) Duration of 50:41–44
untreated psychosis as predictor of long-term 10. Follette WC, Houts AC (1996) Models of
outcome in schizophrenia: systematic review scientific progress and the role of theory in
and meta-analysis. Br J Psychiatry 205:88–94 taxonomy development: a case study of
5. Jablensky A (2000) Epidemiology of schizo- the DSM. J Consult Clin Psychol 64:
phrenia: the global burden of disease and dis- 1120–1132
ability. Eur Arch Psychiatry Clin Neurosci 11. Yolken RH, Dickerson FB, Fuller TE (2009)
250:274–285 Toxoplasma and schizophrenia. Parasite
6. Whiteford HA, Degenhardt L, Rehm J, Baxter Immunol 31:706–715
AJ, Ferrari AJ, Erskine HE et al (2013) Global 12. Lovatt A, Mason O, Brett C, Peters E (2010)
burden of disease attributable to mental and Psychotic-like experiences, appraisals, and
substance use disorders: findings from the trauma. J Nerv Ment Dis 198:813–819
13. Clayton PJ, Guze SB, Cloninger CR, Martin persons experiencing high-risk symptoms: pre-
RL (1992) Unipolar depression: diagnostic liminary results
inconsistency and its implications. J Affect from the NAPLS project. Schizophr Bull 41:
Disord 26:111–116 419–428
14. Bromet EJ, Kotov R, Fochtmann LJ, Carlson 27. Chan MK, Krebs MO, Cox D, Guest PC,
GA, Tanenberg-Karant M, Ruggero C et al. Yolken RH, Rahmoune H et al. (2015)
(2011) Diagnostic shifts during the decade fol- Development of a blood-based molecular bio-
lowing first admission for psychosis. Am marker test for identification of schizophrenia
J Psychiatry 168:1186–1194 before disease onset. Transl Psychiatry 5, e601.
15. Torrey EF (1992) Are we overestimating the doi:10.1038/tp.2015.91
genetic contribution to schizophrenia? Schi- 28. Agius M, Shah S, Ramkisson R, Murphy S,
zophr Bull 18:159–170 Zaman R (2007) Three year outcomes of an
16. McGue M (1992) When assessing twin con- early intervention for psychosis service as com-
cordance, use the probandwise not the pairwise pared with treatment as usual for first psychotic
rate. Schizophr Bull 18:171–176 episodes in a standard community mental
17. Tsuang M (2000) Schizophrenia: genes and health team. Preliminary results. Psychiatr
environment. Biol Psychiatry 47:210–220 Danub 19:10–19
18. Dauncey MJ, Bicknell RJ (1999) Nutrition and 29. Salokangas RK, McGlashan TH (2008) Early
neurodevelopment: mechanisms of develop- detection and intervention of psychosis. A
mental dysfunction and disease in later life. review. Nord J Psychiatry 62:92–105
Nutr Res Rev 12:231–253 30. Yap HL (2010) Early psychosis intervention.
19. Schlotz W, Phillips DI (2009) Fetal origins of Singapore Med J 51:689–693
mental health: evidence and mechanisms. Brain 31. Thomas P (2004) The many forms of bipolar
Behav Immun 23:905–916 disorder: a modern look at an old illness.
20. Koenig JI, Kirkpatrick B, Lee P (2002) Gluco- J Affect Disord 79(Suppl 1):S3–S8
corticoid hormones and early brain development 32. Hirschfeld RM (2001) Bipolar spectrum disor-
in schizophrenia. Neuropsychopharmacology der: improving its recognition and diagnosis.
27:309–318 J Clin Psychiatry 62(Suppl 14):5–9
21. Ryan MC, Collins P, Thakore JH (2003) 33. Post RM (2005) The impact of bipolar depres-
Impaired fasting glucose tolerance in first- sion. J Clin Psychiatry 66(Suppl 5):5–10
episode, drug-naive patients with schizophre- 34. Kamel N, Compton C, Middelveld R,
nia. Am J Psychiatry 160:284–489 Higenbottam T, Dahlén SE (2008) The
22. Spelman LM, Walsh PI, Sharifi N, Collins P, Innovative Medicines Initiative (IMI): a new
Thakore JH (2007) Impaired glucose toler- opportunity for scientific collaboration bet-
ance in first-episode drug-naive patients with ween academia and industry at the European
schizophrenia. Diabet Med 24:481–445 level. Eur Respir J 31:924–926
23. Guest PC, Wang L, Harris LW, Burling K, 35. Hunter AJ (2008) The Innovative Medicines
Levin Y, Ernst A et al. (2010) Increased levels Initiative: a pre-competitive initiative to
of circulating insulin-related peptides in first- enhance the biomedical science base of
onset, antipsychotic naive schizophrenia Europe to expedite the development of new
patients. Mol Psychiatry 15:118–119 medicines for patients. Drug Discov Today
24. Szulc A, Galińska B, Konarzewska B, Gudel- 13:371–373
Trochimowicz I, Popławska R (2001) Immu- 36. http://www.fda.gov/MedicalDevices/Device
nological marker activity in first episode RegulationandGuidance/IVDRegulator y
schizophrenic patients. Pol Merkur Lekarski Assistance/ucm124105.ht
10:450–452 37. Ellington AA, Kullo IJ, Bailey KR, Klee GG
25. Van Venrooij JA, Fluitman SB, Lijmer JG, (2010) Antibody-based protein multiplex
Kavelaars A, Heijnen CJ, Westenberg HG et al. platforms: technical and operational challenges.
(2012) Impaired neuroendocrine and immune Clin Chem 56:186–193
response to acute stress in medication-naive 38. Söderlund J, Schröder J, Nordin C, Samuelsson
patients with a first episode of psychosis. M, Walther-Jallow L, Karlsson H et al. (2009)
Schizophr Bull 38:272–279 Activation of brain interleukin-1beta in schizo-
26. Perkins DO, Jeffries CD, Addington J, Bearden phrenia. Mol Psychiatry 14:1069–1071
CE, Cadenhead KS, Cannon TD et al. (2014) 39. Merrill JE (1992) Tumor necrosis factor alpha,
Towards a psychosis risk blood diagnostic for interleukin 1 and related cytokines in brain
development: normal and pathological. Dev 52. Müller N, Riedel M, Schwarz MJ (2004)
Neurosci 14:1–10 Psychotropic effects of COX-2 inhibitors – a
40. Meyer U, Feldon J, Yee BK (2009) A review of possible new approach for the treatment of
the fetal brain cytokine imbalance hypothesis of psychiatric disorders. Pharmacopsychiatry 37:
schizophrenia. Schizophr Bull 35:959–972 266–269
41. Saetre P, Emilsson L, Axelsson E, Kreuger J, 53. Akhondzadeh S, Tabatabaee M, Amini H,
Lindholm E, Jazin E (2007) Inflammation- Ahmadi Abhari SA, Abbasi SH, Behnam B
related genes up-regulated in schizophrenia (2007) Celecoxib as adjunctive therapy in
brains. BMC Psychiatry 7:46 schizophrenia: a doubleblind, randomized and
42. Schmitt A, Leonardi-Essmann F, Durrenberger placebo-controlled trial. Schizophr Res 90:
PF, Parlapani E, Schneider-Axmann T, 179–185
Spanagel R et al. (2011) Regulation of immun- 54. Müller N, Krause D, Dehning S, Musil R,
emodulatory genes in left superior temporal Schennach-Wolff R, Obermeier M et al. (2010)
cortex of schizophrenia patients: a genome- Celecoxib treatment in an early stage of schizo-
wide microarray study. World J Biol Psychiatry phrenia: results of a randomized, double-blind,
12:201–215 placebocontrolled trial of celecoxib augmenta-
43. Montejo AL (2010) The need for routine tion of amisulpride treatment. Schizophr Res
physical health care in schizophrenia. Eur 121:118–124
Psychiatry 25(Suppl 2):S3–S5 55. Laan W, Grobbee DE, Selten JP, Heijnen CJ,
44. Pedrini M, Massuda R, Fries GR, de Kahn RS, Burger H (2010) Adjuvant aspirin
BittencourtPasquali MA, Schnorr CE, Moreira therapy reduces symptoms of schizophrenia
JC, et al. (2012) Similarities in serum oxidative spectrum disorders: results from a randomized,
stress markers and inflammatory cytokines in double-blind, placebo-controlled trial. J Clin
patients with overt schizophrenia at early and late Psychiatry 71:520–527
stages of chronicity. J Psychiatr Res 46:819–824 56. Arranz B, Rosel P, Ramirez N, Duenas R,
45. Miller BJ, Buckley P, Seabolt W, Mellor A, Fernandez P, Sanchez JM et al (2004) Insulin
Kirkpatrick B (2011) Meta-analysis of cytokine resistance and increased leptin concentrations
alterations in schizophrenia: clinical status in noncompliant schizophrenia patients but
and antipsychotic effects. Biol Psychiatry 70: not in antipsychotic-naive first episode schizo-
663–671 phrenia patients. J Clin Psychiatry 65:
1335–1342
46. Schwarz E, Izmailov R, Spain M, Barnes A,
Mapes JP, Guest PC et al. (2010) Validation of 57. Cohn TA, Remington G, Zipursky RB, Azad
a blood-based laboratory test to aid in the con- A, Connolly P, Wolever TM (2006) Insulin
firmation of a diagnosis of schizophrenia. resistance and adiponectin levels in drug-free
Biomark Insights 5:39–47 patients with schizophrenia: a preliminary
report. Can J Psychiatry 51:382–386
47. Schwarz E, Guest PC, Rahmoune H, Harris
LW, Wang L, Leweke FM et al. (2012) 58. van Nimwegen LJ, Storosum JG, Blumer RM,
Identification of a biological signature for Allick G, Venema HW, de Haan L et al (2008)
schizophrenia in serum. Mol Psychiatry 17: Hepatic insulin resistance in antipsychotic
494–502 naive schizophrenic patients: stable isotope
studies of glucose metabolism. J Clin Endo-
48. Späth-Schwalbe E, Born J, Schrezenmeier H, crinol Metab 93:572–577
Bornstein SR, Stromeyer P, Drechsler S et al.
(1994) Interleukin-6 stimulates the 59. Guest PC, Schwarz E, Krishnamurthy D, Harris
hypothalamus-pituitary-adrenocortical axis in LW, Leweke FM, Rothermundt M (2011)
man. J Clin Endocrinol Metab 79:1212–1214 Altered levels of circulating insulin and other neu-
roendocrine hormones associated with the onset
49. Straub RH, Buttgereit F, Cutolo M (2011) of schizophrenia. Psychoneuroendocrinology
Alterations of the hypothalamic-pituitary- 2011(36):1092–1096
adrenal axis in systemic immune diseases – a
role for misguided energy regulation. Clin Exp 60. Taguchi A, Wartschow LM, White MF (2007)
Rheumatol 29:S23–S31 Brain IRS2 signaling coordinates life span and
nutrient homeostasis. Science 317:369–372
50. Bremner JD (2006) Traumatic stress: effects
on the brain. Dialogues Clin Neurosci 8: 61. Convit A (2005) Links between cognitive
445–461 impairment in insulin resistance: an explanatory
model. Neurobiol Aging 26(Suppl 1):31–35
51. Spijker AT, van Rossum EF (2012) Gluco-
corticoid sensitivity in mood disorders. 62. Craft S (2007) Insulin resistance and
Neuroendocrinology 95:179–186 Alzheimer’s disease pathogenesis: potential
www.Ebook777.com
mechanisms and implications for treatment. placebo controlled trial. Schizophr Res 79:
Curr Alzheimer Res 4:147–152 251–256
63. Bello NT, Hajnal A (2006) Alterations in 74. Usall J, Huerta-Ramos E, Iniesta R, Cobo J,
blood glucose levels under hyperinsulinemia Araya S, Roca M et al (2011) Raloxifene as an
affect accumbens dopamine. Physiol Behav adjunctive treatment for postmenopausal
2006(88):138–145 women with schizophrenia: a double-blind,
64. O’Malley D, Shanley LJ, Harvey J (2003) randomized, placebo-controlled trial. J Clin
Insulin inhibits rat hippocampal neurones via Psychiatry 72:1552–1557
activation of ATP-sensitive K+ and large con- 75. Arranz MJ, Munro J, Birkett J, Bolonna A,
ductance Ca2+-activated K+ channels. Mancama D, Sodhi M et al (2000) Pharma-
Neuropharmacology 44:855–863 cogenetic prediction of clozapine response.
65. Corcoran CM, Smith C, McLaughlin D, Lancet 355:1615–1616
Auther A, Malaspina D, Cornblatt B (2012) 76. Zhang JP, Malhotra AK (2011) Pharma-
HPA axis function and symptoms in adoles- cogenetics and antipsychotics: therapeutic effi-
cents at clinical high risk for schizophrenia. cacy and side effects prediction. Expert Opin
Schizophr Res 135:170–174 Drug Metab Toxicol 7:9–37
66. Gorobets LN, Matrosova MI (2010) Specialties 77. Gebhardt S, Haberhausen M, Heinzel-
of prolactin secretion and peripheral reproduc- Gutenbrunner M, Gebhardt N, Remschmidt
tive sex hormones in patients with of first epi- H, Krieg JC et al (2009) Antipsychotic-induced
sode of schizophrenia. Zh Nevrol Psikhiatr Im body weight gain: predictors and a systematic
S S Korsakova 110:17–22 categorization of the long-term weight course.
67. Akiibinu MO, Ogundahunsi OA, Ogunyemi J Psychiatr Res 43:620–626. doi:10.1016/j.
EO (2012) Inter-relationship of plasma mark- jpsychires.2008.11.001, Epub 2008 Dec 24
ers of oxidative stress and thyroid hormones in 78. Lau SL, Muir C, Assur Y, Beach R, Tran B,
schizophrenics. BMC Res Notes 5:169 Bartrop R et al (2016) Predicting weight gain
68. Ben-Jonathan N, Hugo ER, Brandebourg TD, in patients treated with clozapine: the role of
LaPensee CR (2006) Focus on prolactin as a sex, body mass index, and smoking. J Clin
metabolic hormone. Trends Endocrinol Metab Psychopharmacol 36:120–124
17:110–116 79. Shrivastava A, Johnston M, Bureau Y, Shah N
69. Tannenbaum GS, Martin JB, Colle E (2012) Baseline serum prolactin in drug naive,
(1976) Ultradian growth hormone rhythm in first-episode schizophrenia and outcome at five
the rat: effects of feeding, hyperglycemia, and years: is it a predictive factor? Innov Clin
insulin-induced hypoglycemia. Endocrinology Neurosci 2012(9):17–21
99:720–727 80. Schwarz E, Guest PC, Steiner J, Bogerts B, Bahn
70. Meltzer HY, Perry E, Jayathilake K (2003) S (2012) Identification of blood based molecular
Clozapine-induced weight gain predicts signatures for prediction of response and relapse in
improvement in psychopathology. Schizophr schizophrenia patients. Transl Psychiatry 2, e82
Res 59:19–27 81. Schwarz E, Steiner J, Guest PC, Bogerts B,
71. Bahtiyar G, Weiss K, Sacerdote AS (2007) Bahn S (2015) Investigation of molecular
Novel endocrine disrupter effects of classic and serum profiles associated with predisposition to
atypical antipsychotic agents and divalproex: antipsychotic-induced weight gain. World
induction of adrenal hyperandrogenism, J Biol Psychiatry 16:22–30
reversible with metformin or rosiglitazone. 82. Tomasik J, Schwarz E, Lago SG, Rothermundt
Endocr Pract 13:601–608 M, Leweke FM, van Beveren NJ et al (2016)
72. Sato T, Hanyu H, Hirao K, Kanetaka H, Pretreatment levels of the fatty acid handling
Sakurai H, Iwamoto T (2011) Efficacy of proteins H-FABP and CD36 predict response
PPARgamma agonist pioglitazone in mild to olanzapine in recent-onset schizophrenia
Alzheimer disease. Neurobiol Aging 32: patients. Brain Behav Immun 52:178–186
1626–1633 83. Domenici E, Willé DR, Tozzi F, Prokopenko I,
73. Nachshoni T, Ebert T, Abramovitch Y, Assael- Miller S, McKeown A et al (2010) Plasma pro-
Amir M, Kotler M, Maayan R et al (2005) tein biomarkers for depression and schizophre-
Improvement of extrapyramidal symptoms fol- nia by multi analyte profiling of case-control
lowing dehydroepiandrosterone (DHEA) collections. PLoS One 5(2):e9166
administration in antipsychotic treated schizo- 84. Salmon SE, Mackey G, Fudenberg HH (1969)
phrenia patients: a randomized, double-blind “Sandwich” solid phase radioimmunoassay
for the quantitative deterimination of human of-care multiparameter analysis. Lab Chip 12:
immunoglobulins. J Immunol 103:129–137 464–473
85. Salmon SE, Smith BA (1970) Sandwich 87. Krishna S, Boren SA, Balas EA (2009) Health-
solid phase radioimmunoassays for the char- care via cell phones: a systematic review.
acterization of human immunoglobulins Telemed J E Health 15:231–240
synthesized in vitro. J Immunol 104: 88. Berg B, Cortazar B, Tseng D, Ozkan H, Feng
665–672 S, Wei Q et al (2015) Cellphone-based hand-
86. Schumacher S, Nestler J, Otto T, Wegener M, held micro-plate reader for point-of-care test-
Ehrentreich-Förster E, Michel D et al (2012) ing of enzyme-linked immunosorbent assays.
Highly-integrated lab-on-chip system for point- ACS Nano 9:7857–7866
Chapter 3
Multiplex Biomarker Approaches in Type 2 Diabetes

Mellitus Research
Susan E. Ozanne, Hassan Rahmoune, and Paul C. Guest
Abstract
Type 2 diabetes mellitus is a multifactorial condition resulting in high fasting blood glucose levels.
Although its diagnosis is straightforward, there is not one set of biomarkers or drug targets that can be
used for classification or personalized treatment of individuals who suffer from this condition. Instead, the
application of multiplex methods incorporating a systems biology approach is essential in order to increase
our understanding of this disease. This chapter reviews the state of the art in biomarker studies of human
type 2 diabetes from a proteomic and metabolomic perspective. Our main focus was on biomarkers for
disease prediction as these could lead to early intervention strategies for the best possible patient
outcomes.
Key words Type 2 diabetes mellitus, Biomarkers, Drug targets, Genomics, Proteomics, Metabolomics
1 Introduction
Type 2 diabetes mellitus is a complex metabolic disorder resulting

from a combination of insulin resistance and insulin deficiency.
This type of diabetes accounts for 90 % of the cases worldwide and
can be distinguished from type 1 diabetes by the fact that it is not
autoimmune in origin. After a marked increase over the last three
decades, the prevalence of this disorder has now reached epidemic
proportions, with almost 400 million adult cases throughout the
world [1]. One reason that has been proposed to account for this
increase is a possible epidemiologic transition away from commu-
nicable diseases as being the major cause of early death. However,
it is also likely that this is due to a transition towards increasingly
unhealthy dietary habits and lower levels of physical activity [2, 3].
Therefore, there is an urgent need for the development of early
intervention strategies to address the public health, economic,
scientific and ethical calls to the individual, as well as the societal
and healthcare burdens associated with this disease.
37
38 Susan E. Ozanne et al.
Non-modifiable risk factors Modifiable risk factors
Sub-optimal early environment
Insulin resistance
Family history of diabetes

Sedentary lifestyle
High cholesterol
Ethnic background
Risk factors of Obesity
type 2 diabetes
Advanced age Excessive fatty food
consumption
Excessive simple sugar

Presence of genetic risk consumption
factors
High blood pressure
Fig. 1 Schematic diagram showing the impact of various risk factors for development of type 2 diabetes. The
factors on the right and left are listed as non-modifiable and modifiable, respectively but it should be noted that
these are generalizations
To date, type 2 diabetes has been difficult to predict and there

is no cure. However, abnormalities in glucose metabolism includ-
ing states of moderate hyperglycaemia are known to occur well
before disease onset [4, 5]. Therefore, availability of a reliable bio-
marker test for identification prior to development of hyperglycae-
mia would enable high risk individuals to adopt certain nutritional
and physical activity lifestyle changes to delay or even halt progres-
sion of the disease [6–8]. In turn, this would help to minimize
its impact and improve the long term health outcome of the
individual.
Known risk factors for type 2 diabetes include advanced age,
high body mass index (BMI), poor diet, low physical activity, unfa-
vorable intrauterine environment and certain adverse genetic asso-
ciations [1, 7, 9–12] (Fig. 1). Molecular studies have also pointed
at insulin resistance and knock-on effects in liver, skeletal muscle
and adipose tissues as potential predictive markers of type 2 diabe-
tes development [4, 5, 13–15]. Genome Wide Association Studies
(GWAS) have identified several genetic risk allele [14]. Many of
these are thought to be involved in regulation of pancreatic β-cell
function, although the mechanisms by which they do so are poorly
defined.
Multiplex Biomarkers in Type 2 Diabetes 39
To improve patient outcomes, prediction of diabetes should be

achieved at the earliest possible stage of the disease. Here we review
the various multiplex proteomic and metabolomic studies which
have been performed in an effort to identify predictive biomarkers
for this disorder. These two approaches were chosen since pro-
teomic and metabolomic markers can be detected and routinely
monitored in the blood, making these more amenable for clinical
investigations. Nevertheless, the results of other biomarker-related
approaches will be presented to demonstrate consistency of the
biological findings.
2 Epigenetic Biomarkers for Prediction of Type 2 Diabetes
Most genetic studies of type 2 diabetes have not attempted to

account for information on interaction with potential environmen-
tal factors such as nutrition, body mass, activity levels or lifestyle.
However, is likely that the genetic variants associated identified
through GWAS act in combination with physiological perturba-
tions caused by environmental factors such poor nutrition or high
body mass indices to increase the risk of diabetes. The interaction
between the environment has been termed epigenetics and explains
why one genotype can give rise to multiple phenotypes. The main
epigenetic modifications that mediate this interaction between the
genotype and the environment are DNA methylation (on cytosine
residues), histone acetylation/methylation and alterations in levels
of small noncoding RNAs such as miRNAs.
Several studies have now found significant differences in site-
specific DNA methylation patterns between tissues from type 2
diabetics and glucose tolerant individuals as well as during aging in
humans [16–21]. Although in many cases the analysis has been of
a clinically accessible tissue such as peripheral blood mononuclear
cells (PBMCs) some have been carried out in more metabolically
relevant tissues such as pancreatic islets, adipose tissue liver and
skeletal muscle [22–26]. There is much debate as to how reflective
DNA methylation changes in PMBCs are of the same loci in tissues
involved in glucose homeostasis such as liver, muscle, adipose tis-
sue and pancreatic islets. However regardless of whether the
changes are indicative of the underlying biology they could still
represent useful biomarkers [6, 27]. One study using PBMCs from
more 800 individuals who were not diabetic identified an associa-
tion between differential methylation in a cholesterol transport
gene (ABCG1) and fasting insulin levels [28]. Another investiga-
tion found methylation differences in the nuclear factor-kB path-
way gene (MALTI) and the G-protein receptor 6 gene (GPR61) in
PBMCs from twins who were discordant for type 2 diabetes [29]
and another identified DNA hypo-methylation of specific loci in
young individuals who later developed diabetes [30]. In addition,
www.Ebook777.com
a study identified methylation differences in genes such as PPARG

and IRS1 in adipose tissue from individuals with type 2 diabetes
compared to nondiabetic controls [25]. Both of these genes
encode proteins that are involved in insulin signaling.
Other studies have identified small noncoding RNAs (e.g.,
microRNAs) as another possible source of biomarkers for type 2
diabetes [31–33]. Guay and colleagues reviewed the potential of
using circulating microRNA profiles as a means of monitoring spe-
cific aspects of health and they identified patterns which appeared
to be predictive of long-term complications in patients with type 1
and type 2 diabetes [34].
3 Proteomic and Metabolomic Technologies in Type 2 Diabetes Research
Biomarkers are physical characteristics or molecules that can be

monitored and used as an indication of physiological parameters
such as health, disease or drug response. For practical purposes,
biomarkers should be easily accessible due to the potential need of
routine sampling. This section will give a brief description of pro-
teomic and metabolomic profiling in blood samples (for more in
depth descriptions, see other chapters in this volume). It should be
kept in mind that changes in physiological state can be dynamic
and thus methods that are capable of coping with this are essential.
Most researchers now study proteins and metabolites, considering
that these molecules actually carry out or respond to most pro-
cesses in cells, tissues or organisms. In addition, most drugs in use
today target proteins such as enzymes or receptors and induce
metabolic change, evidenced as the synthesis or degradation of
both proteins and metabolites. With this in mind, most diabetes
researchers have employed methods such as multi-protein arrays,
two dimensional gel electrophoresis (2DGE), mass spectrometry
or nuclear magnetic resonance (NMR) profiling platforms.
3.1 Multiplex Blood serum and plasma contain many important bioactive or reg-
Immunoassay ulatory molecules, such as hormones, growth factors and cytokines
which are present only at low concentrations. Therefore, detection
of these by high sensitivity platforms is essential. This can be
achieved by multiplex immunoassay as described by Fulton and
coworkers in 1997 [35]. The assay consists of multiple specific
antibodies each covalently linked to spectrally distinct microbeads.
When the sample is added the target molecules will bind to the
appropriate beads and fluorescently labeled detection antibodies
are then added which bind to the target molecules in a sandwich-
like format. Finally, the beads are passed though a reader in a
narrow steam for analysis by two lasers for identification and quan-
tification of the targeted analyte.
3.2 Two-Dimensional In this method, protein samples are applied to a strip gel with an
Gel Electrophoresis immobilized-pH gradient and the resident proteins are separated
(2DE) according to their isoelectric points (pI) by isoelectric focusing.
Next the strips are applied onto a sodium dodecyl sulfate (SDS)–
polyacrylamide gel and electrophoresed to separate the proteins
according to their molecular weights (MW). The resulting protein
spot constellations can be visualized with post-electrophoretic
stains such as Coomassie Blue or Sypro Ruby and quantitated
using appropriate imaging software. Alternatively, multiple protein
samples can be pre-labeled with spectrally distinct fluorescent dyes
and then electrophoresed on the same gels to allow direct compari-
son of the different proteomes [36]. This latter technique known
as 2D-DIGE (two-dimensional difference gel electrophoresis)
helps by eliminating the need for gel to gel comparisons. The reso-
lution power of 2DE and 2D-DIGE are remarkable for proteome
investigations but limitations include the poor detection of very
acidic and very basic proteins as well as low abundant proteins
[37]. Also, there is a need to use other techniques to identify the
separated proteins using a technique such as mass spectrometry.
3.3 Mass In proteomics, mass spectrometry was initially used to identify pro-
Spectrometry teins previously separated by 2DE, employing peptide mass finger-
prints [38]. Given 2DE limitations and the fact that mass
spectrometers equipped with ESI sources could be connected
online to liquid chromatography systems, the concept of shotgun
proteomics or shotgun mass spectrometry has emerged [39]. This
has revolutionized the field towards the end of the Human Genome
Project, due to its sensitivity and high throughput power for pro-
teomic biomarker identification [39]. The term “shotgun” comes
from the point that the target proteins can be cleaved with enzymes
to generate smaller peptides, which are the actual analytes in this
approach (see other chapters in this book for more detailed infor-
mation). Cleavage is performed as most intact proteins are too
large and complex at the structural level to enable direct analysis.
After cleavage, the peptides are separated according to their phys-
iochemical properties by liquid chromatography and they enter the
mass spectrometer by electrospray ionization. This ionization stage
allows the peptides to be accelerated by magnets in the mass spec-
trometer towards a detector at a speed that is inversely propor-
tional to their mass/charge ratio. Quantitation occurs at the
detector essentially via the number of impacts of each given pep-
tide ion. At the same time, the peptide sequence can be deter-
mined by bombardment with a streaming gas to allow fragmentation
into smaller pieces. Determining the masses of these fragments can
then be used to derive the amino acid sequences that make up the
peptides, which could then be used to search a protein database
to obtain the identification of the parent protein. This method
can also be used for metabolomic analysis, with some modifications.
In this case, there is no need for enzymatic cleavage of the mole-

cules as these are already of a manageable size. Thus, the sample
can be infused directly into the mass spectrometer, the quantity
measured as above, and the identity of the metabolite can be deter-
mined through comparison with known standards.
1
3.4 1H-Nuclear H-NMR spectroscopy is used for metabolite and small molecule
Magnetic Resonance analyses. One major advantage of this method is the simplicity and
(NMR) Spectroscopy reproducibility of the sample preparation stage. However, it is less
sensitive than mass spectrometry approaches for metabolite detection
[40, 41]. The technique gives structural information about the target
molecules and so it is ideal for identification purposes. This works by
monitoring the behavior of protons on molecules in a strong mag-
netic field. The nuclei of the protons line up with the magnetic field in
a similar manner as a compass needle aligns with the Earth’s magnetic
field. The NMR procedure is initiated through application of radio
pulses to the sample and this stimulates the nuclei to rotate around the
axis of the magnetic field with a frequency related to the physiochemi-
cal environment of the surrounding atoms within the molecule. NMR
can be used to monitor relative changes in the levels of small mole-
cules such as amino acids, sugars and lipids.
4 Proteomic Biomarkers
4.1 Biomarkers There is no question about whether beta cell dysfunction leading
for Prediction of Type to impaired insulin production and insulin resistance are estab-
2 Diabetes lished risk factors for type 2 diabetes. However, since insulin is
known to be involved in regulation of metabolism and/or growth
of virtually all cells in the body, there are likely to be many more
associated proteins which can be detected in serum or plasma that
are biomarkers of type 2 diabetes. An early study used surface
enhanced laser desorption/ionization time-of-flight (SELD-TOF)
mass spectrometry for identification of potential serum biomarkers
associated with type 2 diabetes [42]. This led to detection of four
proteins which showed significant quantitative differences between
the diabetic and control groups and these were identified by
peptide mass fingerprinting as albumin, apolipoprotein C3, trans-
ferrin, and transthyretin (Table 1). Another investigation found
decreased levels of apolipoprotein A1 and increased levels of apoli-
poprotein E, leptin and C-reactive protein in patients with type 2
diabetes compared to controls, using a combination of 2DGE
and matrix-assisted laser desorption/ionization-time of flight
(MALDI-TOF) mass spectrometry [43]. Furthermore they vali-
dated these changes using immunoassays. Another 2DGE study
analyzed levels of proteins in plasma samples from patients with
poorly controlled diabetes and nondiabetic controls [44]. This
resulted in identification of increased levels of fibrinogen and hap-
Table 1
Potential serum/plasma proteomic biomarkers for risk of type 2 diabetes and associated
complications
Biomarker Function Citation

Type 2 diabetes
α1-antitrypsin Inflammation [41]
Albumin Molecular transport [39]
Albumin-cysteinylated Molecular transport [42]
Albumin-glycated Molecular transport [42]
Apolipoprotein A1 Molecular transport [40, 41]
Apolipoprotein A1-glycated Molecular transport [42]
Apolipoprotein A1-oxidized Molecular transport [42]
Apolipoprotein C3 Molecular transport [39]
Apolipoprotein E Molecular transport [40]
C-reactive peptide Inflammation [40]
Hemoglobin-glycated Molecular transport [42]
Hemoglobin-nitrosylated Molecular transport [42]
Leptin Hormone signaling [40]
Transferrin Molecular transport [39]
Transthyretin Molecular transport [39, 41]
Vitamin D binding protein Molecular transport [41]
Diabetes-associated cardiovascular disease
Interleukin 6 Inflammation [51]
Plasminogen activator inhibitor-1 Coagulation [51]
Tumor necrosis factor-α Inflammation [51]
von Willebrand factor Coagulation [51]
Diabetes-associated nephropathy
Albumin Molecular transport [15, 53]
C-reactive protein Inflammation [60]
E-selectin Inflammation [60]
Intercellular cell adhesion molecule 1 Inflammation [58, 59]
Interleukin 6 Inflammation [58, 59]
Interleukin 18 Inflammation [54–58]
(continued)
Table 1
(continued)

Tissue plasminogen activator Coagulation [60]
Vascular cell adhesion molecule 1 Inflammation [58, 59]
von Willebrand factor Coagulation [58–60]
Diabetes-associated retinopathy
α1-antitrypsin Inflammation [62]
α2-macroglobulin Inflammation [62]
Afamin Molecular transport [62]
Antithrombin-III Coagulation [62]
Apolipoprotein A1 Molecular transport [62]
Apolipoprotein A4 Molecular transport [64]
Apolipoprotein E Molecular transport [62]
CD40 ligand Inflammation [63]
Clusterin Inflammation [64]
Complement C1/C3 Inflammation 64)
Complement C7 Inflammation [64]
Fibrinogen Coagulation [62]
Gelsolin Coagulation [62]
Haptoglobin Molecular transport [62]
Inter-α-trypsin inhibitor H2 Inflammation [64]
Interleukins 6, 7, 9, 13, 15, 17, S2Rα Inflammation [63]
Kininogen-1 Coagulation [62]
Leucine-rich alpha-2-glycoprotein Inflammation [62]
Monocyte chemoattractant protein-1 Inflammation [63]
Protein arginine N-methyltransferase 5 Growth signaling [62]
Serum amyloid P Molecular transport [62]
Tumor necrosis factor-β Inflammation [63]
Vitamin D-binding protein Molecular transport [62]
Vitronectin Inflammation [62]
toglobin along with decreased levels of α1-antitrypsin, apolipopro-

tein A1, transthyretin, and vitamin D binding protein in plasma
from the diabetic individuals. A more recent study described
the development of a new on-chip liquid chromatography-
nanoelectrospray ionization mass spectrometry blood-based gly-
caemia monitoring assay which afforded simultaneous measurement
of glucose and glycated forms of hemoglobin (HbA1c), albumin
and apolipoprotein A1 [45]. This study incorporated assays for
cysteinylated albumin, S-nitrosylated hemoglobin, and methionine-
oxidized apolipoprotein A1 for assessing oxidative stress and car-
diovascular risk [45]. The results showed that this assay was capable
of distinguishing type 2 diabetes patients from healthy controls
although further testing using additional cohorts in multiple clini-
cal settings is required to confirm the findings.
4.2 Biomarkers Diabetes is the fifth leading cause of death worldwide being associ-
for Assessing the Risk ated with more than 5 % of all deaths. This is because diabetes is a
of Diabetes- known risk factor for development of several other disorders,
Associated including cardiovascular disease, stroke and peripheral vascular dis-
Cardiovascular ease [46]. Approximately two-thirds of the deaths attributed to
Disease type 2 diabetes result from comorbid cardiovascular events such as
coronary artery disease [47]. Thus, the life expectancy for persons
with type 2 diabetes is around 5–10 years lower than for non-
diabetic individuals [48]. However, the underlying mechanisms
linking diabetes with cardiovascular conditions have not been
completely elucidated. Considering this, there is a critical need for
further studies of molecular risk factors and the development of
biomarker assays that can be used to predict, characterize and
monitor this disorder along with its associated comorbidities and
outcomes. Of particular note, several studies over the past few
years have reported that treatment of type 2 diabetes patients using
glucose lowering agents alone does not lead to a reduction in
severity of a heart disease-related comorbidity [49–52]. This finding
suggests that therapeutic measures in addition to glycemic control
are needed. Furthermore, the Food and Drug Administration has
now issued a new guidance to drug companies concerning restric-
tions on the development of novel antidiabetic pharmaceuticals
[53]. In essence, this states that new drugs in development must be
evaluated for any cardiovascular disease risk. Therefore, there is also
a clinical need for the development of new biomarker tests which
can be used to predict the propensity towards this risk as well as the
physiological effects of novel antidiabetic compounds.
Recent proteomic studies have suggested that there is an inter-
relation between the metabolic disruptions in type 2 diabetes and
inflammation, which leads to a disease state through an increased
coagulation response and consequential detrimental effects on
endothelial cells and the vasculature (Table 1) [15]. The increased
expression of inflammatory cytokines and adhesion molecules can
amplify the inflammatory responses, leading to an aggravation of

diabetic vascular complications. In turn, this may initiate a process
of atherosclerosis and the formation of arterial thrombi. This
appears to result in the changes in circulating pro-coagulant and
inflammatory biomarkers including von Willebrand factor, inter-
leukin-6, tumor necrosis factor-α, and plasminogen activator
inhibitor-1. In addition several of the serum and plasma proteins
detected as potential biomarkers for of type 2 diabetes have also
been implicated in separate studies of coronary heart disease. This
includes the proteins albumin and C-reactive protein, as well as
many others associated with coagulation or inflammation [54].
Such molecules may serve as potential biomarkers for predicting
the risk of cardiovascular and renal perturbations in diabetic
patients and for the monitoring of patient responses to therapeu-
tics targeting these widespread complications.
4.3 Biomarkers Diabetic nephropathy is the leading cause of chronic renal failure
for Prediction in Western countries. It has been associated with early death and
of Diabetic can also negatively affect patient quality of life, as well as being a
Nephropathy significant burden on the healthcare systems [55]. One of the first
clinical signs of such microvascular damage in diabetes is an increase
in the levels of albumin [56, 57]. This has led to screening of albu-
min levels in patients with diabetes to identify those at risk of
microvascular complications such as nephropathy [53]. Recent
longitudinal studies have also been carried out and these have led
to the identification of proteomic and metabolomic biomarkers
which can predict the onset or progression of nephropathy in
patients with type 2 diabetes. The most robust markers for prediction
of onset of neuropathy identified to date are plasma asymmetric
dimethylarginine, serum interleukin 18 and urinary ceruloplasmin,
immunoglobulin G and transferrin (Tables 1 and 2) [58–62]. For
predicting neuropathy progression, plasma levels of intercellular
cell adhesion molecule 1, interleukin 6, plasma asymmetric dimeth-
ylarginine, vascular cell adhesion molecule 1, and von Willebrand
factor have been consistently identified as the most robust bio-
markers [62, 63]. For prediction of both the onset and progression
of neuropathy, the most accurate biomarkers were plasma C-reactive
protein, E-selectin, tissue-type plasminogen activator, triglycer-
ides, and von Willebrand factor [64]. This indicated that the lists
of biomarkers for predicting onset and progression were distinct.
Another study carried out multiplex immunoassay analyses of the
levels of 27 cytokines in urine from patients with type 2 diabetes
and micro- or normo-albuminuria [65]. They found that the levels
of eotaxin, granulocyte colony-stimulating factor, interferon-γ-
inducible protein 10, interleukin 8, monocyte chemoattractant
protein-1, RANTES and tumor necrosis factor-α were increased
significantly in micro-albuminuric patients compared to levels in
normo-albuminuric diabetic patients or controls. On the other
Table 2
Potential plasma/serum metabolomic biomarkers for risk of type 2 diabetes and associated
complications

Type 2 diabetes
a -hydroxybutyric acid Lipid-related [70, 71]
Acetone Lipid-related [70]
Acetoacetate Lipid-related [70]
Acyl-alkyl-phosphatidylcholine Lipid-related [66]
b -hydroxybutyrate Lipid-related [70]
Diacylphosphatidylcholine C32:1 Lipid-related [66]
Glucose Sugar [65]
Glycine Amino acid [66, 69]
Isoleucine Branched chain amino acid [65–67]
Leucine Branched chain amino acid [65–67]
Linoleoyl-glycerophosphocholine Lipid-related [71]
lysophosphatidylcholine C18:2 Lipid-related [65, 66, 69]
Phenylalanine Aromatic amino acid [65, 66]
Sphingomyelin C16:1 Lipid-related [66]
Tyrosine Aromatic amino acid [65, 66]
Valine Branched chain amino acid [65–67]
Diabetes-associated cardiovascular disease
No distinct markers NA [74]
Diabetes-associated nephropathy
a -hydroxybutyric acid Lipid-related [76, 77]
Asymmetric dimethylarginine Oxidation/reduction [54–59]
Isoleucine Branched chain amino acid [76, 77]
Leucine Branched chain amino acid [76, 77]
Phenylalanine Aromatic amino acid [76, 77]
Triglycerides Lipid-related [60]
Tyrosine Aromatic amino acid [76, 77]
Valine Branched chain amino acid [76, 77]
Diabetes-associated retinopathy
Ascorbic acid Oxidation/reduction [78]
Galactitol Sugar [78]
hand, the levels of granulocyte-macrophage colony-stimulating

factor, macrophage inflammatory protein-1α and macrophage
inflammatory protein-1β levels were elevated in micro-albuminuric
patients in comparison to controls. Furthermore, the levels of
interferon-γ-inducible protein 10 and monocyte chemoattractant
protein-1 were significantly correlated with the urinary albumin
excretion and estimated glomerular filtration rates, and the levels
of eotaxin, granulocyte-macrophage colony-stimulating factor,
interferon-γ-inducible protein 10, monocyte chemoattractant pro-
tein-1, and RANTES were correlated with glycated hemoglobin
(HbA1c). These findings suggest that monitoring of cytokine lev-
els in urine may be helpful for early diagnosis and treatment of
patients with diabetic nephropathy.
4.4 Biomarkers A proteomics study using a two-dimensional difference gel electro-

for Determining Risk phoresis (2D-DIGE) approach found 28 plasma protein differ-
of Diabetic ences between diabetic patients with and without retinopathy and
Retinopathy these proteins were identified by matrix-assisted laser desorption/
ionization-time of flight (MALDI-TOF) mass spectrometry [66].
The proteins which were altered in the patients with retinopathy
were mainly involved in the inflammation response and the coagu-
lation cascade (Table 1). This included identification of multiple
proteins that had been reported previously as diabetic retino-
pathy markers, including α1-antitrypsin, α2-macroglobulin,
antithrombin-III, apolipoprotein A1, apolipoprotein E, complement
C1/C3, fibrinogen, haptoglobin, kininogen-1, serum amyloid
P-component, and vitronectin. This demonstrated the consistency
of these findings with previous studies. In addition, this study also
identified potential novel diabetic retinopathy biomarkers such as
afamin, gelsolin, leucine-rich alpha-2-glycoprotein, protein argi-
nine N-methyltransferase 5 and vitamin D-binding protein. These
studies are still in an early phase and awaiting validation testing
using additional cohorts but could in the future be useful clinically.
However, the finding that some of these had already been iden-
tified by previous investigations lends some support to the possi-
bility that these proteins and their associated pathways may
be involved in the onset and progression of diabetes-associated
retinopathy.
One study assessed the levels of plasma cytokines in 59 diabetic
patients and 19 nondiabetic controls using a multiplex immunoas-
say approach [67]. They found that the levels of monocyte che-
moattractant protein-1, tumor necrosis factor-β, soluble CD40
ligand, and the interleukins 6, 7, 9, 13, 15, 17 and soluble 2Rα
were increased significantly in diabetic patients compared to the
controls. The authors also found that the plasma levels of tumor
necrosis factor-α plasma were significantly higher in patients with
diabetic retinopathy compared to patients without retinopathy. In
contrast, the levels of Flt-3 ligand and interleukins 1Ra, 3, 5, and
12 (p40) were present at lower levels in the diabetes patients com-

pared to the control group. Jin and colleagues performed a com-
prehensive proteomic analysis using a selective reaction monitoring
mass spectrometry approach to discover biomarkers for diabetic
retinopathy [68]. The authors targeted 96 vitreous humor proteins
which were also expressed in plasma and found that a 4-plex panel
comprised of apolipoprotein A4, complement C7, clusterin, and
inter-alpha-trypsin inhibitor H2 could be used to distinguish dia-
betes patients with early stages of retinopathy from controls. A
drawback of many of these studies is that they include samples that
are collected from people who already have diabetes. In these situ-
ations it is not clear if the changes identified are simply a conse-
quence of the disease. To identify biomarkers of future disease risk
longitudinal prospective cohort studies are require so that samples
are available from an individual both prior to and after develop-
ment of type 2 diabetes.
5 Metabolomic Biomarkers
5.1 Metabolite Metabolomic studies have identified blood-based biomarkers

Biomarkers which may have predictive value in type 2 diabetes (Table 2). For
for Prediction of Type example, increased levels of branched-chain amino acids (leucine,
2 Diabetes isoleucine, and valine) have been identified by several different
research groups and in some cases these changes were found to
occur more than 10 years before clinical diagnosis of the disease
[69–71]. It is possible that changes in these amino acids are
involved in the pathogenesis of type 2 diabetes although similar
changes have also been associated with insulin resistance which can
occur in a variety of disorders [72]. Metabolomic studies have also
identified elevated levels of circulating aromatic amino acids such
as phenylalanine and tyrosine several years ahead of actual diagno-
sis [69, 70]. Only one amino acid, glycine, was found to be reduced
before the onset of type 2 diabetes and this may be a result of
increased gluconeogenesis [70, 73].
A large number of metabolomic profiling studies have also
found increased levels of a -hydroxybutyric acid up to 10 years
before clinical presentation of the disease [74, 75]. This is thought
to occur due to perturbed redox pathways resulting in increased
oxidative stress and lipid oxidation [75, 76]. Several studies have
also found serum or plasma changes in b -hydroxybutyrate, ace-
tone and acetoacetate prior to overt manifestation of type 2 diabe-
tes, and these changes are also likely to be associated with increased
lipid oxidation [74]. Alterations in other circulating lipids have
also been identified in metabolomic profiling studies, including
elevated levels diacylphosphatidylcholine C32:1 [70] and decreased
levels of sphingomyelin C16:1 [70], acyl-alkyl-phosphatidylcholine
[70], lysophosphatidylcholine C18:2 [69, 70, 73], and linoleoyl-
www.Ebook777.com
glycerophosphocholine [71]. The latter molecule may be linked to

diabetes as its normal function is to increase glucose-stimulated
insulin secretion from pancreatic β-cells [75].
One of the most common means to assess predictive power of
potential biomarker signatures is the application of receiver opera-
ting characteristic-area under the curve (ROC-AUC) statistics.
Using this approach, one study found that a metabolomic signa-
ture could predict the development of diabetes with a similar
ROC-AUC score as found using the Diabetes Risk Score, which is
based on lifestyle, diet and anthropometric factors [77]. However,
by combining a biomarker signature composed of fasting glucose,
HbA1c, and metabolites with the Diabetes Risk Score, the predic-
tive power could be increased by approximately 6 % to a value of
0.912 (1.0 = perfection) [69]. Another study found a ROC-AUC
of 0.82 using a metabolomic signature for prediction of type 2
diabetes compared to a slightly lower score of 0.79 using only fast-
ing glucose levels [74]. Although these improvements in predictive
power are only marginal, they support the need for identification
of further biomarker-based signatures associated with disease risk
and highlight the potential of combining molecular, dietary, life-
style, and anthropomorphic data into predictive algorithms.
5.2 Metabolite One study attempted to identify metabolomic signatures associ-

Biomarkers ated with cardiovascular risk factors such as high blood pressure,
of Diabetes-Related liver complications and coronary heart disease in patients with type
Cardiovascular 2 diabetes [78]. Although the researchers were successful and
Disease found distinct metabolomic signatures between patients with these
different comorbidities, no conclusions could be made given the
absence of a control group with diabetes and no comorbidities.
However, a number of metabolomic studies of risk factors for car-
diovascular disorders have been carried out and these have identified
several markers which also related to diabetes risk, such as branched
chain and aromatic amino acids [79, 80]. It is clear for the scarcity
of research in this particular aspect of type 2 diabetes that consider-
able further work is required. This is especially important in this
particular case considering the devastating effects of heart disease
throughout the world at the individual, societal, and economic
levels.
5.3 Metabolite Another study screened urine metabolites of diabetic patients with
Biomarkers and without chronic renal disease and this led to identification of
for Prediction 13 molecules that were present at significantly lower levels in those
of Diabetes-Related with the kidney disorder [81]. Most of these molecules were
Renal Disease involved in mitochondrial function consistent with the possibility
that impaired mitochondrial metabolism was an underlying factor
in this comorbidity. In addition, a prospective metabolomic study
identified increased levels of uremic solutes and acylcarnitines prior
to end-stage kidney disease in type 2 diabetes patients, compared
to matched controls with no renal dysfunction [82]. The same

study also found decreased levels of branched chain and aromatic
amino acids and the derivative a -hydroxybutyric acid, suggestive
of increased mitochondrial amino acid b -oxidation. These find-
ings were consistent with another metabolomic study in which
plasma from diabetes patients with and without diabetic nephropa-
thy was compared (Table 2) [83].
5.4 Metabolite A metabolomic profiling study of vitreous humor carried out on

Biomarkers diabetic retinopathy patients found reduced levels of galactitol and
for Prediction ascorbic acid (Table 2) [84]. However, the findings may have been
of Diabetes-Related limited by the fact that the control group used in this study was
Retinopathy made up of patients with macular degeneration, rather than dia-
betic patients without retinopathy. Nevertheless, further studies
using definitive disease and control cohorts should be carried out
since the production of a predictive biomarker test and application
to individual risk assessment has the potential of initiating early
interventions which could improve patient outcomes.
6 Conclusions and Future Directions
Novel research in the fields of proteomics and metabolomics has

opened up new opportunities of identifying biomarkers for risk
and progression of type 2 diabetes mellitus. Through profiling of
blood-based biomarkers, new information about these markers
and how they interact with age or lifestyle-related risk factors have
now been identified. Further insights into these associations will
lead to improvements in existing predictive biomarkers of type 2
diabetes and associated comorbidities and should therefore help to
prevent or delay onset in individuals at high risk. Most of this
research has revealed a tight link with disturbances in inflammatory
response and redox potential in the disease itself as well as in the
common comorbidities. However, considerable further work is
required combining the use of the omic platforms in multicentre
and longitudinal studies to provide specific biomarker panels that
can distinguish the risk potential for each of these possibilities.
Furthermore, translation of these biomarkers onto rapid and user-
friendly platforms for point-of-care testing is essential to optimize
potential clinical applications. This may include the lab-on-a-chip
and mobile phone applications as described in the last few chapters
of this book.
Future challenges include targeted research that will bring us
closer to a better understanding of the link between molecular bio-
markers and the link with physiometric or lifestyle-related factors at
the level of the individual patient. This is important as biomarker
testing has already established that algorithms that combine such
different factors can lead to tests with greater performance, com-
www.Ebook777.com
pared to those using a single factor alone. Therefore, future

research on biomarkers and type 2 diabetes has the potential to
produce an abundance of specific biomarkers that might aid in the
individually targeted prevention of the disease and the associated
comorbidities. This is critical considering that this condition is so
devastating to individual health as well as to society and the health-
care services throughout the world.
Acknowledgments
SEO is funded the UK Medical Research Council

(MC_UU_12012/4).
References
1. http://www.diabetesatlas.org/ 10-year follow-up of diabetes incidence and
2. Steyn NP, Mann J, Bennett PH, Temple N, weight loss in the diabetes prevention program
Zimmet P, Tuomilehto J et al (2004) Diet, outcomes study. Lancet 374:1677–1686
nutrition and the prevention of type 2 diabetes. 9. Hales CN, Barker DJ (1992) Type 2 (non-
Public Health Nutr 7:147–165 insulin-dependent) diabetes mellitus: the thrifty
3. Mattei J, Malik V, Wedick NM, Hu FB, phenotype hypothesis. Diabetologia 35:595–601
Spiegelman D, Willett WC et al (2015) 10. Nielsen JH, Haase TN, Jaksch C, Nalla A,
Reducing the global burden of type 2 diabetes Søstrup B, Nalla AA et al (2014) Impact of
by improving the quality of staple foods: the fetal and neonatal environment on beta cell
global nutrition and epidemiologic transition function and development of diabetes. Acta
initiative. Global Health 11:23 Obstet Gynecol Scand 93:1109–1122
4. Perseghin G, Ghosh S, Gerow K, Shulman GI 11. Vaag A, Brons C, Gillberg L, Hansen NS,
(1997) Metabolic defects in lean non Diabetic HjortL AGP et al (2014) Genetic, nongenetic
offspring of NIDDM parents: a cross-sectional and epigenetic risk determinants indevelop-
study. Diabetes 46:1001–1009 mental programming of type 2 diabetes. Acta
5. Henriksen JE, Levin K, Thye-Rønn P, Alford Obstet Gynecol Scand 93:1099–1108
F, Hother-Nielsen O, Holst JJ et al. (2000) 12. Coope A, Torsoni AS, Velloso L (2016) Mech-
Glucose-mediated glucose disposal in insulin- anisms in endocrinology: metabolic and inflam-
resistant normoglycemic relatives of type 2 dia- matory pathways on the pathogenesis of type 2
betic patients. Diabetes 49:1209–1218 diabetes. Eur J Endocrinol 174:R175–R187
6. Tuomilehto J, Lindstrom J, Eriksson JG, Valle 13. DeFronzo RA (1992) Pathogenesis of type2
TT, Hamalainen H, Ilanne-Parikka P (2001) (non-insulin dependent) diabetes mellitus: a
Prevention of type 2 diabetes mellitus by balanced overview. Diabetologia 35:389–397
changes in life style among subjects with 14. Dimas AS, Lagou V, Barker A, Knowles JW,
impaired glucose tolerance. N Engl J Med Mägi R, Hivert MF et al (2014) Impact of type
344:1343–1350 2 diabetes susceptibility variants on quantita-
7. Knowler WC, Barrett-Connor E, Fowler SE, tive glycemic traits reveals mechanistic hetero-
Hamman RF, Lachin JM, Walker EA et al geneity. Diabetes 63:2158–2171
(2002) Reduction in the incidence of type 2 15. Domingueti CP, Dusse LM, Carvalho MD, de
diabetes with lifestyle intervention or metfor- Sousa LP, Gomes KB, Fernandes AP (2016)
min. N Engl J Med 346:393–403 Diabetes mellitus: the linkage between oxida-
8. Diabetes Prevention Program Research Group, tive stress, inflammation, hypercoagulability
Knowler WC, Fowler SE, Hamman RF, and vascular complications. J Diabetes
Christophi CA, Hoffman HJ et al (2009) Complications 30:738–745
16. Ling C, Poulsen P, Simonsson S, Ronn T, cose, insulin, and HOMA-IR in the genetics of
Holmkvist J, Almgren P et al (2007) Genetic lipid lowering drugs and diet network study.
and epigenetic factors are associated with Diabetes 63:801–807
expression of respiratory chain component 29. Yuan W, Xia Y, Bell CG, Yet I, Ferreira T, Ward
NDUFB6 in human skeletal muscle. J Clin KJ et al (2014) An integrated epigenomic anal-
Invest 117:3427–3435 ysis for type 2 diabetes susceptibility loci in
17. Ronn T, Poulsen P, Hansson O, Holmkvist J, monozygotic twins. Nat Commun 5:5719
Almgren P, Nilsson P et al (2008) Age influ- 30. Toperoff G, Aran D, Kark JD, Rosenberg M,
ences DNA methylation and gene expression Dubnikov T, Nissan B et al (2012) Genome-
of COX7A1 in human skeletal muscle. wide survey reveals predisposing diabetes type
Diabetologia 51:1159–1168 2-related DNA methylation variations in
18. Heyn H, Li N, Ferreira HJ, Moran S, Pisano human peripheral blood. Hum Mol Genet
DG, Gomez A et al (2012) Distinct DNA 21:371–383
methylomes of newborns and centenarians. 31. Ferland-McCollough D, Ozanne SE, Siddle K,
Proc Natl Acad Sci USA 109:10522–10527 Willis AE, Bushell M (2010) The involvement
19. Horvath S, Zhang Y, Langfelder P, Kahn RS, of microRNAs in Type 2 diabetes. Biochem
Boks MP, van Eijk K et al (2012) Aging effects Soc Trans 38:1565–1567
on DNA methylation modules in human brain 32. Esguerra JL, Mollet IG, Salunkhe VA, Wendt
and blood tissue. Genome Biol 13:R97.10 A, Eliasson L (2014) Regulation of pancreatic
20. Hannum G, Guinney J, Zhao L, Zhang L, beta cell stimulus-secretion coupling by
Hughes G, Sadda S et al (2013) Genome-wide microRNAs. Genes (Basel) 5:1018–1031
methylation profiles reveal quantitative views 33. Esguerra JL, Eliasson L (2014) Functional
of human aging rates. Mol Cell 49:359–367 implications of long non-coding RNAs in the
21. Dayeh T, Volkov P, Salo S, Hall E, Nilsson E, pancreatic islets of Langerhans. Front Genet
Olsson AH et al (2014) Genome-wide DNA 5:209
methylation analysis of human pancreatic islets 34. Guay C, Regazzi R (2013) Circulating microR-
from type 2 diabetic and non-diabetic donors NAs as novel biomarkers for diabetes mellitus.
identifies candidate genes that influence insulin Nat Rev Endocrinol 9:513–521
secretion. PLoS Genet 10:e1004160 35. Fulton RJ, McDade RL, Smith PL, Kienker LJ,
22. Small KS, Hedman AK, Grundberg E, Nica Kettman JR Jr (1997) Advanced multiplexed
AC, Thorleifsson G, Kong A et al (2007) analysis with the FlowMetrix system. Clin
Identification of an imprinted master trans reg- Chem 43:1749–1756
ulator at the KLF14 locus related to multiple 36. Knowles MR, Cervino S, Skynner HA, Hunt
metabolic phenotypes. Nat Genet 43:561–564 SP, de Felipe C, Salim K et al (2003) Multiplex
23. Johansson A, Enroth S, Gyllensten U (2013) proteomic analysis by two-dimensional differ-
Continuous aging of the human DNA methy- ential in-gel electrophoresis. Proteomics 3:
lome throughout the human lifespan. PLoS 1162–1171
One 8:e67378 37. Oliveira BM, Coorssen JR, Martins-de-Souza
24. Steegenga WT, Boekschoten MV, Lute C, D (2014) 2DE: the phoenix of proteomics.
Hooiveld GJ, de Groot PJ, Morris TJ et al J Proteomics 104:140–150
(2014) Genome-wide age-related changes in 38. Larsen MR, Roepstorff P (2000) Mass spectro-
DNA methylation and gene expression in metric identification of proteins and character-
human PBMCs. Age (Dordr) 36:9648 ization of their post-translational modifications
25. Horvath S (2013) DNA methylation age of in proteome analysis. Fresenius J Anal Chem
human tissues and cell types. Genome Biol 366:677–690
14:R115 39. Link AJ, Eng J, Schieltz DM, Carmack E, Mize
26. Voight BF et al, MAGIC investigators, GIANT GJ, Morris DR et al (1999) Direct analysis of
Consortium (2010) Twelve type 2 diabetes protein complexes using mass spectrometry.
susceptibility loci identified through large-scale Nat Biotechnol 17:676–682
association analysis. Nat Genet 42: 579–589 40. Griffin JL (2003) Metabonomics: NMR
27. Ozanne SE (2015) Epigenetics and metabolism spectroscopy and pattern recognition analysis
in 2014: Metabolic programming--knowns, of body fluids and tissues for characterisation of
unknowns and possibilities. Nat Rev Endocrinol xenobiotic toxicity and disease diagnosis. Curr
11:67–68 Opin Chem Biol 7:648–654
28. Hidalgo B, Irvin MR, Sha J, Zhi D, Aslibekyan 41. Beckonert O, Keun HC, Ebbels TM, Bundy J,
S, Absher D et al (2014) Epigenome-wide Holmes E, Lindon JC et al (2007) Metabolic
association study of fasting measures of glu- profiling, metabolomic and metabonomic
procedures for NMR spectroscopy of urine, nistration, Center for Drug Evaluation and
plasma, serum and tissue extracts. Nat Protoc Research (CDER); 2008
2:2692–2703 54. Donahue MP, Rose K, Hochstrasser D,
42. Sundsten T, Eberhardson M, Göransson M, Vonderscher J, Grass P, Chibout SD et al
Bergsten P (2006) The use of proteomics in (2006) Discovery of proteins related to coro-
identifying differentially expressed serum pro- nary artery disease using industrial-scale
teins in humans with type 2 diabetes. Proteome proteomics analysis of pooled plasma. Am
Sci 4:22 Heart J 152:478–485
43. Riaz S, Alam SS, Akhtar MW (2010) Proteomic 55. Gordois A, Scuffham P, Shearer A, Oglesby A
identification of human serum biomarkers in (2004) The health care costs of diabetic
diabetes mellitus type 2. J Pharm Biomed Anal nephropathy in the United States and the
51:1103–1107 United Kingdom. J Diabetes Complications
44. Bhonsle HS, Korwar AM, Chougale AD, Kote 18:18–26
SS, Dhande NL, Shelgikar KM et al (2013) 56. Mogensen CE, Neldam S, Tikkanen I, Oren S,
Proteomic study reveals downregulation of Viskoper R, Watts RW et al (2000) Randomised
apolipoprotein A1 in plasma of poorly con- controlled trial of dual blockade of renin–
trolled diabetes: a pilot study. Mol Med Rep angiotensin system in patients with hyperten-
7:495–498 sion, microalbuminuria, and non-insulin
45. Mao P, Wang D (2014) Top-down proteomics dependent diabetes: the candesartan and lisin-
of a drop of blood for diabetes monitoring. oprilmicroalbuminuria (CALM) study. Br Med
J Proteome Res 13:1560–1569 J 321:1440–1444
46. Erion DM, Park HJ, Lee HY (2016) The role 57. Ninomiya T, Perkovic V, de Galan BE, Zoungas
of lipids in the pathogenesis and treatment of S, Pillai A, Jardine M et al (2009) Albuminuria
type 2 diabetes and associated co-morbidities. and kidney function independently predict car-
BMB Rep 49(3):139–148 diovascular and renal outcomes in diabetes.
47. Coccheri S (2007) Approaches to prevention J Am Soc Nephrol 20:1813–1821
of cardiovascular complications and events in 58. van der Velde M, Halbesma N, de Charro FT,
diabetes mellitus. Drugs 67:997–1026 Bakker SJ, de Zeeuw D, de Jong PE et al
48. Tuomilehto J, Lindstrom J, Qiao Q (2005) (2009) Screening for albuminuria identifies
Strategies for the prevention of type 2 diabetes individuals at increased renal risk. J Am Soc
and cardiovascular disease. Eur Heart J Suppl Nephrol 20:852–862
7:D18–D22 59. Kazumi T, Hozumi T, Ishida Y, Ikeda Y, Kishi
49. Brunner EJ, Shipley MJ, Witte DR, Fuller JH, K, Hayakawa M et al (1999) Increased urinary
Marmot MG (2006) Relation between blood transferrin excretion predicts microalbumin-
glucose and coronary mortality over 33 years uria in patients with type 2 diabetes. Diabetes
in the Whitehall Study. Diabetes Care 29: Care 22:1176–1180
26–31 60. Narita T, Hosoba M, Kakei M, Ito S (2006)
50. Nissen SE, Wolski K (2007) Effect of rosigli- Increased urinary excretions of immunoglobu-
tazone on the risk of myocardial infarction and lin G, ceruloplasmin, and transferrin predict
death from cardiovascular causes. N Engl development of microalbuminuria in patients
J Med 356:2457–2471 with type 2 diabetes. Diabetes Care 29:
51. Action to Control Cardiovascular Risk in 142–144
Diabetes Study Group, Gerstein HC, Miller 61. Araki S, Haneda M, Koya D, Sugimoto T,
ME, Byington RP, Goff DC Jr, Bigger JT et al. Isshiki K, Chin-Kanasaki M et al (2007)
(2008) Effects of intensive glucose lowering in Predictive impact of elevated serum level of
type 2 diabetes. N Engl J Med IL-18 for early renal dysfunction in type 2 dia-
358:2545–2559 betes: an observational follow-up study.
52. Duckworth W, Abraira C, Moritz T, Reda D, Diabetologia 50:867–873
Emanuele N, Reaven PD et al (2009) Glucose 62. Hanai K, Babazono T, Nyumura I, Toya K,
control and vascular complications in veterans Tanaka N, Tanaka M et al (2009) Asymmetric
with type 2 diabetes. N Engl J Med 360: dimethylarginine is closely associated with the
129–139 development and progression of nephropathy
53. FDA Guidance for Industry: Diabetes in patients with type 2 diabetes. Nephrol Dial
Mellitus — Evaluating Cardiovascular Risk in Transplant 24:1884–1888
New Antidiabetic Therapies to Treat Type 2 63. Persson F, Rossing P, Hovind P, Stehouwer
Diabetes. U.S. Department of Health and CD, Schalkwijk CG, Tarnow L et al (2008)
Human Services, Food and Drug Admi- Endothelial dysfunction and inflammation pre-
dict development of diabetic nephropathy in
the Irbesartan in patients with type 2 diabetes litus and its pathophysiology. PLoS One
and microalbuminuria (IRMA 2) study. Scand 9:e85082
J Clin Lab Invest 68:731–738 75. Ferrannini E, Natali A, Camastra S, Nannipieri
64. Stehouwer CD, Gall MA, Twisk JW, Knudsen M, Mari A, Adam KP et al (2013) Early
E, Emeis JJ, Parving HH (2002) Increased uri- metabolic markers of the development of
nary albumin excretion, endothelial dysfunc- dysglycemia and type 2 diabetes and their
tion, and chronic low-grade inflammation in physiological significance. Diabetes 62:
type 2 diabetes: progressive, interrelated, and 1730–1737
independently associated with risk of death. 76. Gall WE, Beebe K, Lawton KA, Adam KP,
Diabetes 51:1157–1165 Mitchell MW, Nakhle PJ et al (2010) Alpha-
65. Liu J, Zhao Z, Willcox MD, Xu B, Shi B hydroxybutyrate is an early biomarker of insulin
(2010) Multiplex bead analysis of urinary resistance and glucose intolerance in a nondia-
cytokines of type 2 diabetic patients with betic population. PLoS One 5:e10883
normo- and microalbuminuria. J Immunoassay 77. Schulze MB, Hoffmann K, Boeing H, Linseisen
Immunochem 31:279–289 J, Rohrmann S, Möhlig M et al (2007) An
66. Lu CH, Lin ST, Chou HC, Lee YR, Chan HL accurate risk score based on anthropometric,
(2012) Proteomic analysis of retinopathy- dietary, and lifestyle factors to predict the
related plasma biomarkers in diabetic patients. development of type 2 diabetes. Diabetes Care
Arch Biochem Biophys 529:146–156 30:510–515
67. Hang H, Yuan S, Yang Q, Yuan D, Liu Q 78. Wu T, Xie G, Ni Y, Liu T, Yang M, Wei H et al
(2014) Multiplex bead array assay of plasma (2015) Serum metabolite signatures of type 2
cytokines in type 2 diabetes mellitus with dia- diabetes mellitus complications. J Proteome
betic retinopathy. Mol Vis 20:1137–1145 Res 14:447–456
68. Jin J, Min H, Kim SJ, Oh S, Kim K, Yu HG 79. Shah SH, Sun JL, Stevens RD, Bain JR,
et al (2016) Development of diagnostic bio- Muehlbauer MJ, Pieper KS et al (2012)
markers for detecting diabetic retinopathy at Baseline metabolomic profiles predict cardio-
early stages using quantitative proteomics. vascular events in patients at risk for coronary
J Diabetes Res 2016:6571976. doi:10.1155/ artery disease. Am Heart J 163:844–850
2016/6571976, Epub 2015 Nov 9 80. Würtz P, Havulinna AS, Soininen P, Tynkkynen
69. Walford GA, Porneala BC, Dauriz M, Vassy JL, T, Prieto-Merino D, Tillin T et al (2015)
Cheng S, Rhee EP et al. (2014) Metabolite Metabolite profiling and cardiovascular event
traits and genetic risk provide complementary risk: a prospective study of 3 population-based
information for the prediction of future type 2 cohorts. Circulation 131:774–785
diabetes. Diabetes Care 37:2508–2514 81. Sharma K, Karl B, Mathew AV, Gangoiti JA,
70. Floegel A, Stefan N, Yu Z, Mühlenbruch K, Wassel CL, Saito R et al (2013) Metabolomics
Drogan D, Joost HG et al (2013) Identification reveals signature of mitochondrial dysfunction
of serum metabolites associated with risk of in diabetic kidney disease. J Am Soc Nephrol
type 2 diabetes using a targeted metabolomic 24:1901–1912
approach. Diabetes 62:639–648 82. Niewczas MA, Sirich TL, Mathew AV, Skupien
71. Wang TJ, Larson MG, Vasan RS, Cheng S, J, Mohney RP, Warram JH et al (2014) Uremic
Rhee EP, McCabe E et al (2011) Metabolite solutes and risk of end-stage renal disease in
profiles and the risk of developing diabetes. type 2 diabetes: metabolomic study. Kidney Int
Nat Med 17:448–453 85:1214–1224
72. Batch BC, Hyland K, Svetkey LP (2014) 83. Huang M, Liang Q, Li P, Xia J, Wang Y, Hu P
Branch chain amino acids: biomarkers of health et al (2013) Biomarkers for early diagnosis of
and disease. Curr Opin Clin Nutr Metab Care type 2 diabetic nephropathy: a study based on
17:86–89 an integrated biomarker system. Mol Biosyst
73. Wang-Sattler R, Yu Z, Herder C, Messias AC, 9:2134–2141
Floegel A, He Y et al (2012) Novel biomarkers 84. Barba I, Garcia-Ramírez M, Hernández C,
for pre-diabetes identified by metabolomics. Alonso MA, Masmiquel L, García-Dorado D
Mol Syst Biol 8:615 et al (2010) Metabolic fingerprints of proli-
74. Padberg I, Peter E, González-Maldonado S, ferative diabetic retinopathy: an 1H-NMR-
Witt H, Mueller M, Weis T et al (2014) A new based metabonomic approach using vitreous
metabolomic signature in type-2 diabetes mel- humor. Invest Ophthalmol Vis Sci 51:
4416–4421
Chapter 4
LC-MSE, Multiplex MS/MS, Ion Mobility, and Label-Free

Quantitation in Clinical Proteomics
Gustavo Henrique Martins Ferreira Souza, Paul C. Guest,
Abstract
Proteomic tools can only be implemented in clinical settings if high-throughput, automated, sensitive, and
accurate methods are developed. This has driven researchers to the edge of mass spectrometry (MS)-based
proteomics capacity. Here we provide an overview of recent achievements in mass spectrometric technolo-
gies and instruments. This includes development of high and ultra definition-MSE (HDMSE and UDMSE)
through implementation of ion mobility (IM) MS towards sensitive and accurate label-free proteomics
using ultra performance liquid chromatography (UPLC). Label free UPLC-HDMSE is less expensive than
labeled-based quantitative proteomics and has no limits regarding the number of samples that can be ana-
lyzed and compared, which is an important requirement for supporting clinical applications.
Key words Proteomics, Label-free, MSE, DIA, HDMSE, UDMSE, Data independent acquisition
1 Introduction
Proteomics has emerged as a promising field for biomarker identi-

fication in the post-genomic era. By the end of the 1990s, the
combination of liquid chromatography and electrospray ionization
mass spectrometry (ESI-MS) became the main tool for proteomic
characterization. Technological advances in this area and the num-
ber of potential proteomic biomarkers have grown substantially in
recent years, revealing a broader understanding of diseases.
Considering the heterogeneity among proteomic datasets originat-
ing from these studies, scientists must now focus their efforts on
validating this information, and establish the relevance of these
proteins through applications in clinical practice [1, 2]. In order to
achieve this, proteomic-based studies must first demonstrate
improvements in design and standardized protocols to produce
uniform and reliable results. Researchers should pay attention to
pivotal details when defining patient cohorts, such as clinical stage,
inclusion/exclusion criteria, sample quality, quality of controls,
57
58 Gustavo Henrique Martins Ferreira Souza et al.
sample preservation/storage methods, protocol complexity, repro-

ducibility, and adequacy of statistical analyses. This is a daunting
task considering as most clinical studies are complex and can some-
times require large numbers of samples/participants as well as
extended data processing periods.
Although proteomics is in an early stage of exponential growth,
the field currently offers extensive capabilities for elucidating func-
tional protein interactions and for discovering novel biomarkers for
therapeutics. In addition, proteomics can be used to determine the
functional status of a disease and subsequently help to select treat-
ment options. This approach has been termed pharmacopro-
teomics [3]. Despite sophisticated proteomic study designs using
bottom-up shotgun proteomics, subsequent research has chal-
lenged the reliability or reproducibility of the results. These prob-
lems impact all of the “omics” fields and not just one technology
[4]. This chapter describes some of the latest developments in the
field of label-free MS approaches which may help to improve the
discovery and validation of multiplex biomarkers for use in labora-
tory, preclinical and clinical studies.
2 Systems Biology
The use of quantitative high-throughput MS-based proteomics

and comprehensive systems biology approaches has generated
complex datasets which have the possibility to assist researchers in
discovering the meaning of disease-related molecular changes,
especially when combined with information generated by other
omic approaches. In this context, clinical proteomics has the
potential to enable early disease discovery through the develop-
ment of multiplex assays that can aid clinical decisions. However,
the term “biomarker” can be applied only to those that have been
validated and approved following the rules of regulatory agencies.
The discovery, validation and translation of biomarkers depend on
pairing of the correct proteomic approach with a reliable study
design. The throughput achieved in protein MS enables the dis-
covery of new biomarkers in virtually all types of biological sam-
ples. Recent MS-based approaches have allowed the identification
of new biomarkers not previously described with high sensitivity
and confidence. Furthermore, the development of tools for in
silico analysis of datasets allows for the interconnection of the iden-
tified proteins to increase our understanding their interactions in a
systems biology manner. The completion of this cycle could lead to
the identification of a biomarker with real clinical significance and
one that discriminates between disease and healthy controls with
higher confidence. Major diseases are not caused by the action of
single genes, but rather by alterations in the functioning of a com-
plex web of networks and pathways. Thus, measurement at the
Label-Free Mass Spectrometry 59
level of proteome and metabolome presents new opportunities for

both data integration and the potential translation of findings to
medical practice. Although these strategies are useful in identifying
disease-associated pathways or biomarkers, bench scientists and cli-
nicians recognize the need to translate these methods and knowl-
edge into clinical practice. However, the interdisciplinary nature of
these studies and the lack of robust tools have impeded progress.
Nevertheless, recent examples show that reliability of the data can
be achieved and efforts are underway to improve translation of
multiplex biomarkers into clinical practice.
3 Recent Advances in MS-Based Proteomic Analysis Applied to Biomarker

Discovery
Proteomics has evolved to focus on functionality of the huge data-

sets acquired through a variety of analytical technologies. The
quality of procedures, selectivity and specificity has recently
received increased attention. The potential of protein analysis using
complex samples such as chromosome-based, cell and biological
fluid-based proteomics is only just now coming to fruition. Many
researchers claim to have identified proteins but the lack of analyti-
cal quality and statistics is high. For example, the cell is a complex
structure and working with cellular proteins remains challenging.
Proteins are more difficult to understand than DNA or
RNA. Proteins sometimes act alone, in one-to-one interactions or
in large groups. If we want to understand cell behavior, we first
need to understand proteins, taking them into consideration as an
entire system. Consequently, the protein analysis problem is com-
plex and the vast combination of the common and uncommon
amino acids increases this complexity. This includes the fact that
proteins can undergo a large range of enzymic and non-enzymic
interactions, not all tissues express the proteins at the same time,
and the same protein can be expressed at different levels in differ-
ent tissues. In addition, not all proteins are expressed at the same
concentration in the cell and the concentration of a given protein
can change over time. A mass spectrometer is ideally suited to
acquire sufficient data to cover the maximum dynamic range as
needed. It has been noted that an increased dynamic range in the
analytical instrument increases the chances of success due to access
to proteins of lower abundance, especially in the label-free quanti-
tation approaches. MS as a non-target approach relies on methods
such as data dependent acquisition (DDA), data-independent
acquisition (DIA), multiplexed MSE / HDMSE acquisitions, or
selected reaction monitoring (SRM). These approaches assist in
increasing the analytical capability of MS-base approaches and
therefore allow more thorough analyses of samples in biomarker
research.
4 Shotgun Proteomics: Complex Samples Data Acquisition and Molecular Ion

Density
For complex samples, data-dependent acquisitions (DDA) are

largely incapable of accommodating the chimeric and stochastic
nature of the data and peptide mass sufficiency distribution (Fig. 1).
Peptides with an overlapping mass-to-charge ratio (m/z) and simi-
lar retention times make assigning more than one peptide per MS/
MS spectrum far more difficult to interpret, even with application
of complex algorithms [5]. However, it has been found that DDA
and DIA product ion spectra have high quality similarity when one
peptide is isolated at the selection window.
A significant source of error in proteomic experiments arises
from the algorithmic interpretation of product ion spectra derived
from the chimeric and composite MS spectra. In the case of com-
plex mixture experiments, most search engines do not acknowl-
edge the fact that a typical DDA product ion spectrum is most
likely to arise from co-fragmented peptides. Approximately two-
thirds of all precursor ion detections in a complex protein digest
mixture are at least 2.5 orders of magnitude lower in intensity than
the most abundant ions. Consequently, the incidence of overlap-
ping isotopic clusters of similar m/z and intensity is significant.
The specificity of DDA acquisitions is challenged under such con-
ditions, especially when the search engine peptide score is based
primarily on the intensity of the matched product ions relative to
that of the unmatched ions. The ion density together with dynamic
range and molecular weight distributions of proteins in a complex
mixture, allow many analytes to be packaged in a small analytical
Fig. 1 Mass sufficiency distribution—1 Da wide. The bar plot illustrates the ions
of m/z 537–538 assuming a mass resolving power of 25,000 FWHM at mass
500. There are 50 bins, each bin is 20 millidaltons wide and the colors indicate
charge-state. Over 45 independent ions on m/z 537.28 +/−20 mDa within that
1 min of gradient elution are represented. For a given charge-state of an accu-
rate mass measurement, an orthogonal approach such as ion mobility separa-
tion could help to increase selectivity
www.Ebook777.com
space, which includes both chromatography and mass analysis.

Since these are unevenly distributed, areas of high analyte density
can also suffer from issues such as the dynamic range of the sample
being greater than the dynamic range of MS. Besides this, the data
density can be followed by constricted m/z spaces were two or
more peptides are fragmented at once (this is called the chimeracy
effect). According to Michalski et al., up to 50 % of total peptides
are chimeric and this problem is not solved by speed or sensitivity.
The DDA method plays a key role making analyses irreproducible
due the stochastic nature of DDA MS/MS [6]. In contrast, DIA
methods are theoretically limited only by instrument peak capacity
[7]. Acquiring DDA data more quickly or with higher sensitivity
insignificantly reduces these sources of analytical mistakes. An
increase in speed and sensitivity without a concurrent increase in
specificity will generally produce compromised information by
generating supplementary low abundance mixed mass spectra.
Thus, modern acquisition techniques such as a multiplex high-
resolution format MSE and high-definition HDMSE (drift time-
conformation and charge differences) are recommended for
proteomics and complex samples [8].
5 Ion Mobility: Conformation-Based Separations in Gas-Phase
Known as “ion chromatography” and “plasma chromatography,”

ion-mobility spectrometry began as far back as 1930, when Tyndall
and Powell published studies showing that different conformations
could be discriminated in the gas phase due the inference of DC
and specific differential apertures capable of accelerating ions into a
low-pressure cell [9, 10]. In 1932, Bradbury and Neilson described
the first ion-gate that resulted in development of the ion trap prior
to IM-MS separation [11, 12]. In 1967, the first commercial mass
spectrometer was produced which was capable of producing an out-
standing 1 Torr (1.33 mbar) on a drift cell device with a maximum
pressure and this could be coupled to a time of flight (TOF) ana-
lyzer. In 1991, Von-Helden and collaborators described an ion-
mobility mass spectrometer with drift-cell capable of handling up to
4 Torr and carried out the first experimental measurements of cross
section values from different conformations and molecular species
[13]. Next, Valentine and collaborators described development of
the first ion-mobility MS with a dynamic accumulating ion trap
upstream to increase duty-cycle. In 2001, an important article came
out [14] describing separation of peptides from different charge
states that dramatically expanded the applications for clinical and
medical fields by making it applicable for both proteomic and
metabolomic analyses. The first commercially available instrument
was released in 2006 which had additional benefits such as improved
accuracy and specificity due the discrimination of s/n ratios (Fig. 2).
a 100
0
b 100
0 m/z
730 735 740 745 750 755 760 765 770 775 780 785 790 795 800 805 810
Fig. 2 Improvements of signal-to-noise ratio (s/n). The MS spectra demonstrate the separation at gas phase
from a peptide. (a) Standard MS spectrum. (b) Drift-time stripped Ion-Mobility based MS spectrum. Both data-
sets were acquired with a 100 attomoles of ß-lactoglobulin nanoESI(+)
As mentioned above, the introduction of Synapt High Definition

mass spectrometer in 2006 provided a new way to visualize mass
spectrometric data and calculating molecular cross sections by com-
bining the techniques of ultra performance liquid chromatography,
ion mobility spectrometry with MS (Fig. 3) [15]. By definition, ion-
mobility spectrometry is capable of separating gas-phase ionic spe-
cies as they drift through a gas under positive pressure and under the
influence of an electric field. The rate of the drift is proportional and
dependent on the mobility through the gas. Mobility is driven based
on mass, charge, conformation, gas pressure and polarizability, elec-
trical field such as wave velocity of the drift cell (m/s) and height
(V). Thus with the correct parameters, it is possible to discriminate
different charges and isobaric species due to the specificity of differ-
ent cross sections. For protein and peptide measurements, measur-
ing the mobility of molecular ions can yield structural information,
since small and compact ions drift quicker than large extended ones.
Some classical limitations of ion-mobility separation in standard
drift-tubes is low duty cycle, ions lost at the gate, short separation
times such as 10–20 ms, diffusive losses that cause the ion cloud to
be drawn out during separation and losses on the walls and the exit
aperture before reaching the detector. Also, one of the main require-
ments was the need for a high voltage potential gradient across a
long drift tube for efficient separation. Current ion-mobility devices
are more effective as these use radiofrequencies (RF) applied over
Fig. 3 Schematic of an instrument with ion mobility separation. The scheme demonstrates the drift time pusher
extraction pulses and molecular ions being discriminated (red, green and yellow) by size, shape and charge.
Arrival times of ions are recorded by synchronization of the oa-TOF acquisitions with gated release of ions from
the Trap T-wave to the IM-MS. After the gate pulse, the subsequent 200 orthogonal acceleration pushes of the
TOF analyzer are recorded giving on overall mobility recording time of: 200× tpp where tpp is the pusher fre-
quency. Following the next gated release of ions a further 200 mass spectra are acquired and added to the
corresponding spectra from the previous acquisition. This process is repeated for a predefined period and
subsequently 200 spectra are saved and the next summation period begins. RP Rotary Pump, TP Turbo-
molecular pump, oa-TOF: orthogonal acceleration time-of-flight
the driving field extension at the electrodes, giving radial confine-

ment and greater sensitivity by allowing low and high impact ion
currents to transit across the electric field more efficiently.
6 Dynamic Range, Resolution, and Collision Energy
Advances in ion-mobility separation, conferring significantly

increased peak capacity and peptide identification has shown that
resolution is dependent only on m/z but also on collision cross
sections and multiplex DIA acquisitions such as high and ultra
definition-MSE (HDMSE and UDMSE) [16–20]. More specifically
the UDMSE approach allows all of the benefits of HD with a more
effective collision energy control at the transfer cell. In HDMSE,
the ions are separated based on mobility in the IM-MS cell via a
traveling voltage wave and the precursors are subjected to frag-
mentation in the transfer cell normally through application of
www.Ebook777.com
ramped collision energies (eV) [21]. On the other hand in UDMSE,

drift time-specific collision energies are used to fragment precursor
ions prior to TOF-analysis. Ion mobility roughly correlates with
m/z for all charge states. Thus, it is possible to assign quasi m/z-
specific collision energies to the different drift time bins, increasing
the intensity and MS/MS performance for peptide sequencing
correlation.
Ion mobility approaches such as traveling wave ion mobility MS
(TWIM-MS) are usually located inside the mass spectrometer [22].
The IM-MS located after the quadrupole and before a TOF ana-
lyzer increases selectivity since discrimination between charged
molecules can be performed even for those that elute with the same
retention times as well those with the same m/z, given different
cross sections relative to conformation. For example, peptides from
complex samples with the same constitutive amino acids but differ-
ent amino acid positions will be orthogonally separated and dis-
criminated into two or more different signals through drift time
(dt). Until recently, separated ions were typically detected using a
device called a multichannel plate (MCP). Briefly, this type of detec-
tor corresponds to two V-shaped aligned plates behaving as a fast-
electron multiplier, such that a single ion event can result in 1 × 107
electrons being produced over 4–5 ns. This event is followed by a
time-to-digital converter (TDC) that is a positive chemical ioniza-
tion (PCI)-based acquisition device. A high-impact ion flux caused
by multi-charged ions such as those from peptides, strikes the front
MCP, causing an electron shower within the plate that is then mul-
tiplied across a pair of plates. The voltage strike created on the
anode is recorded by the TDC as an ion arrival event and this infor-
mation is combined with TOF data, buffered and passed to the host
embedded PC acquisition system (EPCAS), for processing and
data-event recording [23]. This ion flux momentum over a single
MS spectra-detection amplitude range corresponds to at least 3 logs
of dynamic range (log10max = 3). New detectors such as the ETP-
MagneTOF and the hybrid analog-to-digital converter (ADC)
combined with ion mobility dramatically enhances the duty-cycle
and peak capacity, contributing to the accuracy of measuring ion
arrival times in a mass spectrometer with a wider dynamic range
(log10max = 4–6). With the unbiased DIA acquisition with alternating
low and high energy, the detection dynamic range over each sample
can also be calculated. All of those features have now been com-
bined into one instrument.
7 Quantitative Proteomics and Peak Capacity
Absolute quantitation can assist in understanding proteomic data

[24]. For an LC-MS system, 1D and 2D nano chromatography
and can be employed [25]. In 1D chromatography, a column with
an inner diameter of 75 μm can usually be loaded with up to 500 ng

of total protein and this can be increased in accordance with the
number of fractions. This helps to avoid saturation of the column
and concomitantly increases identification of peptides due to
enhanced peak-to-peak chromatographic resolution. The number
of peptides identified in a protein should be proportional to the
intact protein’s molecular weight and the amount of that particular
protein in the column, as described previously [26]. In order to
address peak capacity and avoid charging effects, stacked-ion ring
traveling-wave (T-wave) [22] devices have been used recently to
increase the ion optics capability, thus enabling the amount of ions
that the mass spectrometer can handle. In the example of “step
wave” (SW) devices, the conjoined ion optics transmission T-wave
helps to increase ion sensitivity, focusing and robustness. The SW
function is based on stacked ring ion guides, which are designed to
maximize ion transmission from the source to the mass analyzer. It
also allows for active removal of neutral contaminants, enhancing
the overall signal to noise ratio. The patented design enables effi-
cient capture of the diffuse ion cloud entering the first stage, which
is then focused into the upper ion guide for transfer to the mass
analyzer. It actively extracts the ion beam into the upper stage and
passes the high gas flow, excess solvent and neutral species to the
exhaust. This protects the critical upper ion guide and the subse-
quent aperture from the direct line of sight of contaminants, ensur-
ing that the methods remain robust for longer, even with complex
matrices as in the case of omics samples.
In order to facilitate biomarker identification, MS discovery
and validation platforms must be developed in parallel. The ion
optics must be equal or superior to those in SRM analysis and this
can be achieved with t-wave devices (such as the StepWaveTM tech-
nology mentioned above). This has many applications besides ion
mobility and ion optics and can been used as collision cells in trap
collision-induced dissociation (CID) or electron transfer dissocia-
tion (ETD), as well as in transfer cells in CID that can be turned
on individually or in groups [27]. As previously mentioned, hybrid
ADC detectors dramatically enhance peak capacity and detection
with a dynamic range of up to 4–6 logs. This target can be achieved
in a straightforward manner for analysis of complex samples using
high-resolution ion mobility instruments with a continuous ion
current, low energy and high energy DIA multiplex MS/MS.
8 Data Quality Control in Label-Free Proteomics
Figures of merit (FOM) are based on how consistent a proteomics

experiment can be graded in terms of consistency and robustness
by the end of the study. At the peptide level, the following points
may be considered:
● Percentage of missed cleavages.

● Percentage of peptides that have fragmented in the source.
● Mass accuracy (normality of distribution across 0 ppm).
● Mass error distribution across m/z range.
● Percentage of detected peptides at a maximum of 5 ppm.
The following can also be considered at the protein level
● Average peptide per protein.
● The calculated false positive/discovery rate (FPR/FDR).
● Total number of proteins identified.
● Dynamic range of the analysis.
● Protein chosen to normalize data.
● Coefficient of variation of all proteins and the chosen protein
(spectra count or ion-accounting).
These considerations can assist in understanding the quality
and integrity of MS data and allow choices to be made for further
validation [6]. The FOM can also be enhanced using the increased
power of identification algorithms, such as those based on ion
accounting [28], and the inclusion of higher numbers of samples.
Due to the complexity of the samples and low component concen-
trations, the data can sometimes be incomplete or can include
interferences, and the assignment of data to peptides or clusters is
often uncertain. Thus, reproducibility of the data is important
when searching for a potential biomarker. The identification-based
algorithms must rely not only on the mass accuracy of the precur-
sors and fragment ions but also on other previously described
physicochemical parameters. After a biomarker is chosen, a new
method is developed and the process advances to the target analy-
sis stage. Here, the focus is on measuring a target peptide and a
related protein biomarker across several replicates and conditions.
Recent instrument developments and LC-MS-based proteomics
techniques have considerably improved the speed of analysis, depth
of protein coverage and amount of information that can be
obtained from complex biological sample mixtures. Despite these
developments, variations in identification and quantification
remain a concern, and alternative and complementary methods
such as SRM are required. The value and use of a data-independent
fragment ion repository has therefore been explored. In line with
this, the required sensitivity and selectivity for the purpose of pro-
tein identification and quantification have been defined.
Conceptually, the following requirements can be considered as
guidelines for validation, hypothesis-driven studies or SRM method
development [29]:
● A minimum of three peptides and three product ions should

be identified for each protein (fragment selection is based on
the highest replication rate and the smallest signature product
ion variation across experiments and samples).
● Similar precursor and product ions should be selected from the
second and third proteins that are consistently present with the
protein of interest (together, these proteins outline a fragment
ion signature).
● Unknown samples can be subsequently mapped against the
fragmentation database signature to validate the presence of
the target protein.
Various statistical and computational tools and methods are
currently being tested using the fragment ion repository to facili-
tate the processes outlined above and to increase mathematical
accounting information in the relational database [30]. Currently,
the database can only be populated with qualitative results obtained
through DIA experiments, specifically LC-MSE and/or LC-HDMSE
approaches. DDA experiments generally do not yield precursor
and product ion intensity measurements across the complete chro-
matographic peak or MS and MS/MS intensity recording for the
same amount of time using the same gain. As spectral libraries and
fragment ion repositories are more widely used in proteomics, the
process will improve [31]. In addition to the identification of frag-
ments, peptides, and proteins, data-independent fragment ion
repositories have great potential in regard to the quantification of
protein abundances, stoichiometry, and the reliable quantification
of posttranslational modifications [32].
9 Applications of Label-Free Shotgun MS in Biomarker Studies
9.1 Analysis In accordance with the need to identify biomarkers in clinically

of Biofluids accessible biosamples, label-free shotgun MS profiling analyses
have been applied routinely in the analysis of media which are
obtainable from subjects with minimal invasiveness, such as urine,
cerebrospinal fluid, sweat, serum, and plasma. This has been par-
ticularly useful in the study of cancer, considering the great need
for such biomarkers to support clinical studies and the discovery of
new drugs. One label-free profiling study investigated urine from
prostate cancer patients in an attempt to indentify urinary bio-
markers [33]. Twenty potential biomarkers in were identified in
the prostate cancer patients with fold differences of more than two
standard deviations observed for 17 proteins using an intensity-
based absolute quantification approach. In a similar manner,
Beretov et al. performed a comparative proteomic analysis using
ion count relative quantification label free liquid chromatography
MS analysis and identified 59 urinary proteins that were present at

significantly different levels in breast cancer compared to control
females [34]. Some of these markers were associated with pre- or
early-invasive phases and others were linked with metastatic stages
of the cancer. Therefore, further work in this area could lead to
identification of serum biomarkers associated with staging of breast
cancer. In another study, Fan et al. carried out a high-performance
liquid chromatography label-free MS profiling analysis of serum
from colorectal cancer patients and healthy controls and found 69
proteins that were linked to the cancer [35]. Two of these proteins
(macrophage mannose receptor 1 and S100A9) were validated as
being increased by western blot analysis and could be used to dis-
tinguish colorectal cancer patients from controls with high accu-
racy through immunoassay analyses.
Studies of other diseases have also been performed such as
comparative label-free liquid chromatography-tandem MS analysis
of cerebrospinal fluid from patients with sporadic amyotrophic lat-
eral sclerosis (ALS) [36]. This study identified several proteins
which were significantly altered in ALS patients compared to
patients with other neurological disorders and controls. These pro-
teins were linked with diverse biological pathways such as inflam-
mation, neuronal activity, and extracellular matrix regulation. The
authors also used the identified profile to create a support vector
machine classifier which could distinguish the ALS from the non-
ALS individuals with 83 % sensitivity and 100 % specificity in an
independent test set, and they validated four of the classifier pro-
teins in an independent test set. Another report described a label-
free MS profiling approach to provide information about the
normal composition of human sweat, as a means of investigating
this chemical barrier against pathogens on the surface of skin [37].
These researchers identified 95 proteins and 20 of these were
novel. The most abundant proteins in this medium were dermci-
din, apolipoprotein D, clusterin, prolactin-inducible protein, and
serum albumin, which comprised more than 90 % of secreted sweat
proteins and these were validated using other multiplex biomarker
profiling approaches. Another label-free MS investigation of serum
showed that alpha-1-antitrypsin was the most discriminative pro-
tein found in patients with tuberculosis compared to controls [38].
In addition, Zhang et al. applied a similar shotgun profiling analy-
sis of plasma which led to identification of three glycated albumin
peptides which could be used for early diagnosis of type 2 diabetes
mellitus [39].
Although psychiatric and neurological disorders have tradi-
tionally been considered to be diseases of the central nervous sys-
tem, serum and plasma can be used to study such conditions. This
is possible since the bloodstream contains a large number of mol-
ecules which are involved in linking peripheral and central func-
tions such as mood, behavior, and memory. For example, one
group analyzed serum from depressed patients and control sub-

jects using a label-free liquid chromatography-tandem MS profil-
ing method and identified ten proteins that were present at
significantly different levels in the depressed patients [40]. The dif-
ferences in three of these proteins were validated by immunoassay
and these were ceruloplasmin, inter-alpha-trypsin inhibitor heavy
chain H4 and complement component 1qC, which were all
increased in the depressed patients. Another investigation analyzed
blood plasma samples of patients with mild and severe forms of
cognitive impairment using a label-free shotgun proteomics
approach [41]. This led to selection of a multivariate model which
demonstrated an accuracy of 79 % in predicting progressive cogni-
tive impairment. Within the model, sex-specific protein biomarkers
were also identified which included alpha-2-macrogloblin as cor-
relating with progression to more severe forms of impairment in
females only. Another study found sex-specific serum biomarker
profiles in patients with Asperger syndrome using a label-free shot-
gun proteomics approach [42]. Twelve proteins were altered in
Asperger syndrome females and one protein was altered in Asperger
syndrome males. These results indicate that the search for bio-
markers or novel drug targets in autism-related disorders may
require stratification into male and female subgroups, which could
lead to the production of new targeted approaches to treatment.
9.2 Analysis Label-free liquid chromatography MS profiling has also been

of Tissues applied in the study of tissues in several distinct medical areas. As
with biofluids, one of the largest areas of use in the analysis of tis-
sue samples has been in the field of cancer. One investigation used
a label-free MS profiling approach to study renal carcinoma tissue
in comparison to normal tissue from the same patients [43]. The
authors found changes in proteins indicative of mitochondrial dys-
function and cell death, along with perturbations in metabolism
and acetylation. In a similar approach, Dai and coworkers analyzed
proteins from matched pairs of human gastric cancer and adjacent
tissues [44]. They found 146 proteins which were altered by more
than twofold in the cancerous tissue. The changes in four nucleic
acid binding proteins (heterogeneous nuclear ribonucleoprotein
(hnRNP)A2B1, hnRNPD, hnRNPL, and Y-box binding protein
1) were validated by other multiplex biomarker techniques such as
quantitative polymerase chain reaction and immunoassay. Other
researchers have used the label-free method to study normal bio-
logical processes. In an investigation of the effects of aging, Theron
and coworkers carried out a shotgun proteomic analysis of muscle
tissue from mature and older females and found that 35 proteins
were linked to aging in muscle and most of these showed decreased
levels with age [45]. This included proteins involved in energy
metabolism and myofilament and cytoskeleton functions.
One area of widespread use of the label-free shotgun pro-

teomic profiling approach has been in the analysis of tissues associ-
ated with the central nervous system. This has mainly involved the
study of psychiatric and neurological disorders. Martins-de Souza
and coworkers presented the first characterization of the human
occipital lobe (primary visual cortex) and cerebellum proteomes
using a label-free shotgun MS profiling approach [46]. They iden-
tified proteins which have been associated previously with condi-
tions such as neurological disorders, progressive motor neuropathy,
Parkinson’s disease, and schizophrenia. Therefore, these proteins
may serve as biomarkers in the study of neurological processes and
identify potential novel drug targets for therapeutic treatment
approaches. Two label-free shotgun MS studies of postmortem
pituitary tissues resulted in identification of candidate biomarkers
for schizophrenia [47], bipolar disorder and major depressive dis-
order [48]. Schizophrenia patients had altered levels of pro-
adrenocorticotropic hormone, arginine vasopressin precursor,
agouti-related protein, growth hormone, prolactin, and secretago-
gin. By comparison, bipolar disorder patients had significantly
increased levels of pro-opiomelanocortin (POMC) and galanin,
and major depressive disorder patients had significantly decreased
levels of the prohormone-converting enzyme carboxypeptidease E
and decreased activity of prolyl-oligopeptidase convertase. Given
that the pituitary directly releases a variety of hormones and other
bioactive molecules into the circulation, many of the proteins iden-
tified in the above study could serve as focal points in the search for
peripheral biomarkers in clinical or drug treatment studies of psy-
chiatric patients. Also in the field of psychiatric studies, Broek and
colleagues carried out a label-free shotgun proteomic analysis of
postmortem brain samples and found opposite directional changes
in proteins which have a role in synaptic connectivity and myelina-
tion in the prefrontal cortex and cerebellum [49]. This is consis-
tent with existing theories of the occurrence of altered structural
and/or functional connectivity in the prefrontal cortex and cere-
bellum in autism patients. As with all biomarker-related studies,
the findings of the above studies can only be considered as prelimi-
nary and require validation using larger independent cohorts of
patient and control samples in follow up investigations.
10 Conclusions and Perspectives
In conclusion, repositories are a valuable addition to the require-

ments of system biology, not only allowing quantitative analysis of
low-abundant proteins but also delivering reliable quantitative data
when proteins are analyzed across multiple samples in multiple
laboratories. When requiring more than one peptide for identifica-
tion, the label-free approach gives superior information particu-
www.Ebook777.com
larly when coverage is taken into account. Both the label and
label-free approaches can provide relative quantification datasets
although the latter has advantages in terms of sample require-
ments, sample preparation and instrumental time. Due to signifi-
cant sample complexity and wide concentration ranges, confidence
in qualitative protein identifications and accuracy of the quantita-
tive assessment of such proteins is imperative. A multiplexed acqui-
sition approach such as label-free MSE can facilitate more protein
identifications by offering greater selectivity and specificity while
also providing the desired quantitative reproducibility for further
validation and SRM/MRM acquisitions. Spectra databanks can
increase the data integrity for quantitative proteomics discovery
and reduce cost, sample size, instrument time, and sample prepara-
tion. This will increase confidence in identification and quantifica-
tion, the average number per protein identified and reproducibility.
Also will be most important for clinical applications and the use of
these approaches in several areas of medicine has increased mark-
edly over the last decade.
Acknowledgments
Prof. Daniel Martins-de-Souza is supported by the São Paulo

Research Foundation (FAPESP) grants 13/08711-3 and
14/10068-4 and by the Brazilian National Council for Scientific
and Technological Development (CNPq) grant 460289/2014-4.
References
1. Lee JM, Kohn EC (2010) Proteomics as a 6. Geromanos SJ, Hughes C, Golick D, Ciavarini
guiding tool for more effective personalized S, Gorenstein MV, Richardson K et al (2011)
therapy. Ann Oncol 21:205–210 Simulating and validating proteomics data and
2. Tchabo NE, Liel MS, Kohn EC (2005) search results. Proteomics 11:1189–1211
Applying proteomics in clinical trials: assessing 7. Geromanos SJ, Vissers JP, Silva JC, Dorschel
the potential and practical limitations in ovar- CA, Li GZ, Gorenstein MV et al (2009) The
ian cancer. Am J Pharmacogenomics detection, correlation, and comparison of pep-
5:141–148 tide precursor and product ions from data
3. Wulfkuhle JD, Edmiston KH, Liotta LA, independent LC-MS with data dependent
Petricoin EF 3rd (2006) Technology insight: LC-MS/MS. Proteomics 9:1683–1695
pharmacoproteomics for cancer—promises of 8. Chakraborty AB, Berger SJ, Gebler JC (2007)
patient-tailored medicine using protein micro- Use of an integrated MS-multiplexed MS/MS
arrays. Nat Clin Pract Oncol 3:256–268 data acquisition strategy for high-coverage
4. Patel VJ, Thalassinos K, Slade SE, Connolly peptide mapping studies. Rapid Commun
JB, Crombie A, Murrell JC et al (2009) A Mass Spectrom 21:730–744
comparison of labeling and label-free mass 9. Tyndall AM, Starr LH, Powell CF (1928) The
spectrometry-based proteomics approaches. mobility of ions in air. Part IV. Investigations
J Proteome Res 8:3752–3759 by two new methods. Proc R Soc Lond A
5. Michalski A, Cox J, Mann M (2011) More 121:172–184
than 100,000 detectable peptides species elute 10. Tyndall AM, Grindley GC, Sheppard PA
in single shotgun proteomics runs but the (1928) The mobility of ions in air. Part V. The
majority is inaccessible to data-dependent transformation of positive Ions at short ages.
LC-MS/MS. J Proteome Res 10:1785–1793 Proc R Soc Lond A 121:185–194
11. Bradbury NE (1932) Photoelectric currents in dem mass spectrometry on a novel quadrupole/
gases between parallel plates as a function of orthogonal-acceleration time-of-flight mass
the potential difference. Phys Rev 40:980 spectrometer. Rapid Commun Mass Spectrom
12. Bradbury NE, Nielsen RA (1936) Absolute 10:889–896
values of the electron mobility in hydrogen. 24. Yu YQ, Gilar M, Lee PJ, Bouvier ES, Gebler JC
Phys Rev 49:388–393 (2003) Enzyme-friendly, mass spectrometry-
13. von Helden G, Hsu M-T, Kemper PR, Bowers compatible surfactant for ion-solution enzy-
MT (1991) Structures of carbon cluster ions matic digestion of proteins. Anal Chem
from 3 to 60 atoms: linears to rings to fuller- 75:6023–6028
enes. J Chem Phys 95:3835–3837 25. Gilar M, Olivova P, Daly AE, Gebler JC (2005)
14. Valentine SJ, Kulchania M, Srebalus Barnes Two-dimensional separation of peptides using
CA, Clemmer DE (2001) Multidimensional RP-RP-HPLC system with different pH in first
separations of complex peptide mixtures: a and second separation dimensions. J Sep Sci
combined high-performance liquid chroma- 28:1694–1703
tography/ion mobility/time-of-flight mass 26. Silva JC, Gorenstein MV, Li GZ, Vissers JP,
spectrometry approach. Int J Mass Spectrom Geromanos SJ (2006) Absolute quantification
212:97–109 of proteins by LCMSE: a virtue of parallel MS
15. Ruotolo BT, Giles K, Campuzano I, Sandercock acquisition. Mol Cell Proteomics 5:144–156
AM, Bateman RH, Robinson CV (2005) 27. Williams JP, Brown JM, Campuzano I, Sadler
Evidence for macromolecular protein rings in the PJ (2010) Identifying drug metallation sites on
absence of bulk water. Science 310:1658–1661 peptides using electron transfer dissociation
16. Giles K, Williams JP, Campuzano I (2011) (ETD), collision induced dissociation (CID)
Enhancements in travelling wave ion mobility and ion mobility-mass spectrometry (IM-MS).
resolution. Rapid Commun Mass Spectrom Chem Commun 46:5458–5460
25:1559–1566 28. Li GZ, Vissers JP, Silva JC, Golick D, Gorenstein
17. Ruotolo BT, Benesch JL, Sandercock AM, MV, Geromanos SJ (2009) Database searching
Hyung SJ, Robinson CV (2008) Ion mobility- and accounting of multiplexed precursor and
mass spectrometry analysis of large protein product ion spectra from the data independent
complexes. Nat Protoc 3:1139–1152 analysis of simple and complex peptide mix-
18. Lalli PM, Corilo YE, de Sa GF, Daroda RJ, de tures. Proteomics 9:1696–1719
Souza V, Souza GHMF et al (2011) Intrinsic 29. Turtoi A, Mazzucchelli GD, De Pauw E
mobility of gaseous cationic and anionic aggre- (2010) Isotope coded protein label quantifica-
gates of ionic liquids. ChemPhysChem tion of serum proteins--comparison with the
12:1444–1447 label-free LC-MS and validation using the
19. Lalli PM, Corilo YE, Fasciotti M, Riccio MF, MRM approach. Talanta 80:1487–1495
de Sa GF, Daroda RJ et al (2013) Baseline res- 30. Lucas JE, Thompson JW, Dubois LG,
olution of isomers by traveling wave ion mobil- McCarthy J, Tillmann H, Thompson A et al
ity mass spectrometry: investigating the effects (2012) Metaprotein expression modeling for
of polarizable drift gases and ionic charge dis- label-free quantitative proteomics. BMC
tribution. J Mass Spectrom 48:989–997 Bioinformatics 13:74
20. Distler U, Kuharev J, Navarro P, Levin Y, 31. Benjamin AM, Thompson JW, Soderblom EJ,
Schild H, Tenzer S (2014) Drift time-specific Geromanos SJ, Henao R, Kraus VB et al
collision energies enable deep-coverage data- (2013) A flexible statistical model for align-
independent acquisition proteomics. Nat ment of label-free proteomics data-
Methods 11:167–170 incorporating ion mobility and product ion
21. Distler U, Kuharev J, Navarro P, Tenzer S information. BMC Bioinformatics 14:364
(2016) Label-free quantification in ion 32. Thalassinos K, Vissers JP, Tenzer S, Levin Y,
mobility-enhanced data-independent acquisi- Thompson JW, Daniel D et al (2012) Design
tion proteomics. Nat Protoc 11:795–812 and application of a data-independent precur-
22. Giles K, Pringle SD, Worthington KR, Little sor and product ion repository. J Am Soc Mass
D, Wildgoose JL, Bateman RH (2004) Spectrom 23:1808–1820
Applications of a travelling wave-based ratio- 33. Adeola HA, Soares NC, Paccez JD, Kaestner
frequency-only stacked ring ion guide. Rapid L, Blackburn JM, Zerbini LF (2015) Discovery
Commun Mass Spectrom 18:2401–2414 of novel candidate urinary protein biomarkers
23. Morris HR, Paxton T, Dell A, Langhorne J, for prostate cancer in a multiethnic cohort of
Berg M, Bordoli RS et al (1996) High sensitiv- South African patients via label-free mass spec-
ity collisionally-activated decomposition tan- trometry. Proteomics Clin Appl 9:597–609
www.Ebook777.com
34. Beretov J, Wasinger VC, Millar EK, Schwartz 42. Steeb H, Ramsey JM, Guest PC, Stocki P,
P, Graham PH, Li Y (2015) Proteomic analysis Cooper JD, Rahmoune H et al (2014) Serum
of urine to identify breast cancer biomarker proteomic analysis identifies sex-specific differ-
candidates using a label-free LC-MS/MS ences in lipid metabolism and inflammation
approach. PLoS One 10:e0141876 profiles in adults diagnosed with Asperger syn-
35. Fan NJ, Chen HM, Song W, Zhang ZY, drome. Mol Autism 5:4
Zhang MD, Feng LY, Gao CF (2016) 43. Zhao Z, Wu F, Ding S, Sun L, Liu Z, Ding K, Lu
Macrophage mannose receptor 1 and S100A9 J (2015) Label-free quantitative proteomic analy-
were identified as serum diagnostic biomark- sis reveals potential biomarkers and pathways in
ers for colorectal cancer through a label-free renal cell carcinoma. Tumour Biol 36:939–951
quantitative proteomic analysis. Cancer 44. Dai P, Wang Q, Wang W, Jing R, Wang W,
Biomark 16:235–243 Wang F et al (2016) Unraveling molecular dif-
36. Collins MA, An J, Hood BL, Conrads TP, ferences of gastric cancer by label-free quanti-
Bowser RP (2015) Label-free LC-MS/MS tative proteomics analysis. Int J Mol Sci
proteomic analysis of cerebrospinal fluid identi- 17(1):E69. doi:10.3390/ijms17010069, pii
fies protein/pathway alterations and candidate 45. Théron L, Gueugneau M, Coudy C, Viala D,
biomarkers for amyotrophic lateral sclerosis. Bijlsma A, Butler-Browne G et al (2014)
J Proteome Res 14:4486–4501 Label-free quantitative protein profiling of vas-
37. Csősz É, Emri G, Kalló G, Tsaprailis G, Tőzsér tuslateralis muscle during human aging. Mol
J (2015) Highly abundant defense proteins in Cell Proteomics 13:283–294
human sweat as revealed by targeted pro- 46. Martins-de-Souza D, Guest PC, Guest FL,
teomics and label-free quantification mass Bauder C, Rahmoune H, Pietsch S et al (2012)
spectrometry. J Eur Acad Dermatol Venereol Characterization of the human primary visual
29:2024–2031 cortex and cerebellum proteomes using shot-
38. Song SH, Han M, Choi YS, Dan KS, Yang gun mass spectrometry-data-independent anal-
MG, Song J (2014) Proteomic profiling of yses. Proteomics 12:500–504
serum from patients with tuberculosis. Ann 47. Krishnamurthy D, Harris LW, Levin Y,
Lab Med 34:345–353 Koutroukides TA, Rahmoune H, Pietsch S
39. Zhang M, Xu W, Deng Y (2013) A new strat- et al (2013) Metabolic, hormonal and stress-
egy for early diagnosis of type 2 diabetes by related molecular changes in post-mortem
standard-free, label-free LC-MS/MS quantifi- pituitary glands from schizophrenia subjects.
cation of glycated peptides. Diabetes World J Biol Psychiatry 14:478–489
62:3936–3942 48. Stelzhammer V, Alsaif M, Chan MK, Rahmoune
40. Lee J, Joo EJ, Lim HJ, Park JM, Lee KY, Park H, Steeb H, Guest PC et al (2014) Distinct pro-
A et al (2015) Proteomic analysis of serum teomic profiles in post-mortem pituitary glands
from patients with major depressive disorder to from bipolar disorder and major depressive dis-
compare their depressive and remission sta- order patients. J Psychiatr Res 60:40–48
tuses. Psychiatry Investig 1:249–259 49. Broek JA, Guest PC, Rahmoune H, Bahn S
41. Yang H, Lyutvinskiy Y, Herukka SK, Soininen (2014) Proteomic analysis of post mortem
H, Rutishauser D, Zubarev RA (2014) brain tissue from autism patients: evidence for
Prognostic polypeptide blood plasma biomark- opposite changes in prefrontal cortex and cer-
ers of Alzheimer’s disease progression. ebellum in synaptic connectivity-related pro-
J Alzheimers Dis 40:659–666 teins. Mol Autism 5:41
Chapter 5
Phenotyping Multiple Subsets of Immune Cells In Situ

in FFPE Tissue Sections: An Overview of Methodologies
James R. Mansfield
Abstract
The recent clinical success of new cancer immunotherapy agents and methods is driving the need to under-
stand the role of immune cells in solid tissues, especially tumors. Immune cell phenotyping via flow cytom-
etry, while a cornerstone of immunology, is not spatially resolved and cannot analyze immune cell subsets
in situ in clinical biopsy sections or to determine their interrelationships. To address this problem, a num-
ber of methodologies have been developed in attempts to phenotype immune and other cells in images
acquired from tissue sections and to assess their organization in the tumor and its microenvironment. This
chapter review the staining and multiplex image analysis methods that have been developed for phenotyp-
ing immune and other cells in formalin-fixed, paraffin-embedded (FFPE) tissue sections.
Key words Immunotherapy, Cancer, Tumor, Immune cells, Multiplex imaging
1 Introduction
The growth in cancer immunotherapy and immunology as treat-

ment methodology and paradigm for cancer research, respectively,
is unprecedented. From its humble beginnings over a century ago
through the work of William B. Coley to an increasing rate of US
Food and Drug Association (FDA) approvals for cancer immuno-
therapies across several cancer types, cancer immunotherapy has
taken its place as one of the five pillars of cancer therapy, along with
surgery, radiotherapy, chemotherapy, and targeted therapies. It was
chosen as the Breakthrough of the Year by Science Magazine in
2013 and continues to play a central role at major cancer confer-
ences such as the American Association of Cancer Researchers
(AACR) or the American Society for Clinical Oncology (ASCO),
with many of the advances being treatment of otherwise untreat-
able solid tumors. Many excellent reviews on the subject have been
written [1–4] and immune-oncology and the tumor microenviron-
ment are now included as a part of the hallmarks of cancer [1].
75
76 James R. Mansfield
Given the new importance of the role of the immune system in

solid-tumor oncology, and its continued importance in many other
diseases, it is vital to be able to understand the specific role each
different type of immune cells plays in situ in the tumor and its
microenvironment, something that is still practically challenging.
There have been decades of research on the different roles of each
type of immune cell [1], and many thousands of immune cell types
have been described. Flow cytometry has been a key technology
used in immunology and has been employed for decades as one of
the primary tools used to differentiate the various immune cells
[2]. The technique relies on the multiplex fluorescence labeling of
cells followed by passing them one at a time through a flow cham-
ber where the fluorescent signal of each marker is measured [3].
The determination of the phenotype of each cell is made through
a process of “gating,” or separating the various cell types based on
the fluorescence intensity of each marker or marker combination.
The markers used in flow cytometry are typically cell membrane-
bound antigens and it is the differential expression of these cell
surface markers that defines the phenotype of the cell [4].
However, flow cytometry cannot give any information about
the distribution of the various cells in solid tissues. In order to be
analyzed by flow cytometry, a solid tissue must first be disaggre-
gated and then the cells passed through the flow cytometer. In the
field of solid-tumor oncology, it is critical to be able to visualize the
phenotypic distributions of the various immune and other cells and
to investigate how these relate to the tumor and its microenviron-
ment. This requires being able to obtain the same kind of pheno-
typic information that is already well understood from flow
cytometry of individual cells but to do so on cells still in situ. It is
also important to be able to obtain an image of the cells in the
context of the tissue architecture so that the spatial relationships of
those cells can be viewed and evaluated in some manner.
Traditional pathology methods for the visualization of cellular
phenotypes in situ utilize a one-marker-at-a-time approach in thin for-
malin-fixed, paraffin-embedded (FFPE) sections, typically using
3,3′-diaminobenzidine (DAB) to stain for a single protein marker and
hematoxylin as a counterstain to display cell nuclei and some tissue
architecture. This is a useful approach as long as a single protein is
enough to accomplish the objectives. However, for cancer immunol-
ogy and immunotherapy research this is typically insufficient as the
phenotypes in question are those from flow cytometric analyses and
can require knowing the expression of several, if not dozens, of pro-
teins simultaneously in a single cell. A good example of this is the flow
cytometric definition of what constitutes a regulatory T cell (CD3+,
CD4+, CD25 high, and FOXP3+), something beyond a one-marker-
at-a-time approach. It is this requirement to be able to measure the
level of multiple proteins in a specific cell, while at the same time seeing
the cell in situ in the context of the other cells and tissues in a sample,
which is driving the development of multiplexed imaging methods.
Multiplex Phenotyping of Immune Cells 77
This chapter covers the range of technologies that have been

developed to try to determine complex phenotypes of cells in situ in
FFPE tissue sections. A wide range of technologies have been applied
to this problem, which will be grouped into two major categories:
(1) simultaneous acquisition and (2) sequential acquisition of imag-
ery. In simultaneous acquisition, all of the markers are present in the
sample at one time and the images or data are collected from the
sample in one pass. Since there are a number of technologies which
are simultaneous, this section will be subdivided into staining, imag-
ing and image analysis subsections. For sequential acquisition, mark-
ers are present in small groups (typically 2–4), imagery is acquired
and the sample cleared or wiped of the first group of markers to
prepare for the next group. The final images or data require register-
ing the resulting set into one larger dataset.
2 Simultaneous Acquisition Methods
In simultaneous acquisition methods, all of the markers on which

the per-cell analyses will be based are applied onto the tissue section
prior to imaging and then an appropriate imaging method acquires
data from all markers in a single step. The multimarker staining is
typically done via some form of immunohistochemistry, using anti-
bodies which bind to specific antigens in the tissue section, followed
by the application of some form of marker (chromogenic, fluores-
cent or metal isotopes) which will then be imaged using an appro-
priate technology. Choosing the most appropriate staining method
depends on the number of markers one wishes to image.
For a small number of markers (three or fewer), chromogenic
methods are typically easiest to implement and best matched to the
types of samples used most often in pathology, although fluores-
cence can also be used. For a medium number of markers [4–12],
fluorescence is typically easiest to implement, although there are
examples of chromogenic staining being used for up to 6 markers.
Chromogenically and fluorescently labeled samples can each be
imaged using commercially available optical imaging systems. For
large numbers of markers (6 or more), there have been a number
of publications describing the use of using metal isotopes to label
the sample and then some form of mass spectrometry imaging to
acquire the data. Mass spectrometry imaging can theoretically
assess more than 100 markers simultaneously, although the practi-
cal limit for staining has typically been around 40.
In addition to the markers required for each antigen one wishes
to label, a method of staining the nucleus is also needed since
downstream image analysis methods all rely on using nuclei to find
cells. For chromogenic methods, nuclear counterstaining is typi-
cally carried out in a blue color using hematoxylin, although there
are other colors/reagents available (methyl green, nuclear fast red,
etc.). For fluorescence methods, 4′,6-diamidino-2-phenylindole
www.Ebook777.com
(DAPI) is typically used as the nuclear counterstain, although

other colors/regents are also available (Hoechst, Draq5, etc.). For
mass spectrometry imaging, one typically needs to use an antibody
for a nuclear protein to delineate the nucleus. It should be noted
that using an antibody to demarcate the nucleus would be a prob-
lem for chromogenic or fluorescent methods as only a limited
number of markers can be applied. However, in cases where it is
possible to use up to 40 markers, as is the case for mass spectrom-
etry imaging, using one of these analytes for nuclear counterstain-
ing is feasible.
Figure 1 shows a schematic of sequential methods. The top
panel shows the overview of the staining, imaging, image analysis
and summary data analyses that are typically performed for this
Multiplex Imaging Image Analysis Summary Data

Staining • Whole-slide • Tissue • Extracting cellular
• Fluorescence scanning Segmentation data from different
• Brightfield • Multispectral • Cellular tissue
• Wide-field versus Segmentation compartments
• Mass spectrometry
confocal • Cellular • Intra-cellular
Phenotyping distance and spatial
metrics
Tissue Cellular Cellular

Segmentation Segmentation Phenotyping
• Hand-drawn • Nuclear • “Flow-like” gating
regions segmentation • Multivariate gating
• Algorithm-based • Cyoplasmic and
• Developed a priori membrane
• User-trained segmentation
Image Analysis
Fig. 1 The top panel shows a schematic of the workflow for a simultaneously labeled sample, involving stain-
ing, imaging, image analysis, and summary data analysis. The image analysis portion is further broken out into
the lower panel, showing tissue segmentation, cellular segmentation, and cellular phenotyping
kind of assay. The lower panel shows a further break down of the
image analysis step into tissue segmentation, cellular segmentation
and cellular phenotyping.
3 Staining and Imaging: Chromogenic
The simplest approach to visualize or quantitate specific cell types

in situ is to use standard chromogenic methods. The most com-
mon method is to use a DAB chromogen as substrate for an immu-
noenzyme reaction. This deposits a brown, alcohol- and
water-insoluble precipitate at the site of enzymatic activity. A pri-
mary antibody for the protein of interest is placed onto the slide
and then a secondary antibody that is selective for the species or
type of the primary antibody is applied. The secondary antibody is
typically conjugate to an enzyme such as horseradish peroxidase
(although others are used), such that the immunoreaction marks
the distribution of the protein with the chromogen. By choosing
antibodies for proteins that are specific for a particular cell type,
one can delineate specific cell phenotypes, such as CD4 for helper
T cells, CD8 for cytotoxic T cells and cytokeratin for epithelial
cells. A blue hematoxylin counterstain is commonly used to high-
light the nuclei in the sample and to enable visualization of the
tissue architecture. The vast majority of immunohistochemistry in
pathology is done using DAB as a marker and hematoxylin as a
counterstain, considering the utility of this approach. It is easily
assessed visually, either through the oculars of a microscope, or by
using a color brightfield digital slide scanner and viewing the result-
ing image on a computer screen. For samples which have been
digitized into an image, it is possible to perform analysis to quan-
titate the protein distribution, either through counting positive
cells or determining the amount of chromogen present in each cell
or subcellular structure.
Despite its ease of use and ubiquity, DAB has some well-known
limitations. Although it is possible to measure the intensity of the
DAB staining and quantitate protein expression on a per-cell basis,
the dynamic range of expression levels that can be measured using
DAB is less than that achieved with fluorescence [5]. DAB is also
problematic for tissues that have natural brown pigments, such as
skin or other epithelial tissues that contain melanin. In those cases,
a red chromogenic substrate can be used. However, the main limi-
tation of DAB is that it can only mark the distribution of a single
protein, which can make differentiating between complex cell types
difficult or impossible. To assess the distributions of more than one
protein in a tissue, a commonly used method is to stain serial sec-
tions of a specimen using different antibodies and then interpolat-
ing between the sections to get an estimate of the inter-distribution
of the proteins. This works well on a gross morphology level.
However, flow cytometry and other techniques have given research-

ers a wide range of cellular phenotypes that are important, particu-
larly in immunology, and understanding these phenotypes requires
using more than one marker per individual cell. By taking serial
sections, one has to rely on a particular cell being split between the
two sections, with one portion in one section and another portion
in the sequential section, and then finding these cells and making
the connection between the halves. This is a difficult task and only
works for two markers at a time. Many immune cells are smaller
than the thickness of a tissue section and are not split across sec-
tions. To assess these cells, the markers need to be deposited on the
same microscope slide and in the same cell using a multiplex chro-
mogenic approach.
There are many methods of performing multiplex chromo-
genic immunohistochemistry, dating back to the 1970s. It is
beyond the scope of this chapter to describe all of these, but there
are a number of excellent review articles [6–8]. Two of the more
common and useful methods will be described here.
3.1 Double This was so named by the original author because it is the simplest
for Dummies means of performing double staining and can be done using any
pair of different species antibodies (e.g., one mouse and one rabbit
antibody) [9]. The two primary antibodies are put on the sample
together, either in a cocktail or individually. The markers are visual-
ized using two separate processes for chromogen deposition, first
with a secondary antibody for one species (e.g., anti-mouse) and
one color, and then using a secondary antibody for the other spe-
cies (e.g., anti-rabbit) and a different color chromogen. This has
the advantage of not requiring any extra heat steps to strip previ-
ously used primary and secondary antibodies. However, it does
require having primary antibodies of different species, which can
be difficult for some marker combinations for which monoclonal
rabbit antibodies may be the only ones obtainable.
3.2 Sequential This is probably the most difficult of the methods, but it is the most
Staining Using generalizable and can be used for up to six different antibodies of
Antibody Stripping the same species. A simple version is a universal type of sequential
double alkaline phosphatase staining. This includes visualization in
red and blue, and any cross-reaction between the two staining
sequences is abolished by a heat-induced epitope retrieval step. This
removes both primary and secondary antibodies but does not gen-
erally affect the quality of subsequent steps [10, 11]. This can be
extended to use of a larger number of colors, with three immuno-
markers being practical and 4–6 being maximal [12–14].
Of course there are many other methods for multiplex immu-
nohistochemistry. For a more detailed discussion, the reader is
encouraged to review the work of the late Chris van der Loos, who
was widely acknowledged to be the leader in this field [15, 16].
4 Imaging Considerations in Brightfield
If all that is required is a photograph of a sample, then any color

camera on a microscope can be used to take an image. Care should
be taken for white balancing of the image and for flat-fielding
(removing the observed intensity differences in the white light dis-
tribution across the sample), but these are optional as long as the
image quality is good. However, if one wishes to perform some
kind of analysis on the image, more care must be taken to provide
quantitative data. Color balancing and image flat-fielding are stan-
dard on all whole-slide scanning systems used in digital pathology,
so the use of those kinds of imaging systems is preferable to
employing regular microscopes and cameras.
4.1 Optical Density The immune-enzyme methods used in chromogenic staining

(OD) Conversion deposit an amount of dye that is proportionate to the amount of
antigen present in the sample [11]. According to the Beer–Lambert
law, the amount of color measured in an image should translate to
the amount of antigen present in each pixel of an image [17, 18].
To be able to apply the Beer–Lambert relationship, the light inten-
sities must first be converted to absorbance (A) or optical density
(OD) units, after which the absorbance of the sample will be pro-
portional to the amount of antigen present [19]. This can only be
a relative quantitation unless there is a reference standard or series
of standards that can be used to calibrate the absorbance, which
has proved elusive in FFPE sections. Despite this limitation, rela-
tive quantitation of antigen intensity within an image, sample or
between samples can be useful.
4.2 Color The quantitative imaging of a chromogenic sample stained singly

Versus Multispectral with either DAB or Fast Red, for example, and then counter-
Imaging stained with hematoxylin, requires only a color imaging system. In
general, the quantitative separation of co-localized colors requires
at least one more wavelength than the number of colors being
imaged [20]. Therefore, separating brown and blue using a three-
wavelength RGB system is possible and all standard digital pathol-
ogy systems are capable of acquiring images which can be used for
this. However, if there are more than two co-localized colors,
more wavelengths are required which necessitates using a multi-
spectral approach [21, 22]. Although quantitative color separa-
tion can be useful for the visualization of multiple markers in a
single sample, the most important aspect of color separation is for
the quantitation of color intensity. If two or more colors are co-
localized, it is difficult to measure the absorbance of any of the
colors unless they are separated. For a two-color system, this can
be done with a red–green–blue (RGB) camera, but more than two
co-localized colors requires multispectral imaging and unmixing
www.Ebook777.com
or some other quantitative color separation method [23]. There

are a number of algorithms that attempt to do a blind separation
of multiple colors in multispectral imaging [24] and these can be
useful for data exploration. However, because these algorithms
attempt to decompose or separate each individual multispectral
image, they are prone to errors when unanticipated components
are present in an image or when one or more of the colors is not
present. For intensity quantitation, it is more reliable to use an
unmixing, or separation, method that utilizes a basis set, or spec-
tral library, that has been developed in advance to match the sam-
ples being quantitated. A predeveloped spectral library will ensure
that all samples are quantitated identically and intensity measure-
ments can then be compared between images. Although theoreti-
cally determined spectra can be used, this is not optimal. There are
both subtle and not-so-subtle differences in the spectral proper-
ties of various commercial chromogens, depending on the sample
and the specific staining method used, which necessitates the use
of sample-specific spectral libraries, usually developed from singly
stained control samples [24]. In addition, DAB can display differ-
ent spectra depending on many factors, which make it less than
optimal for spectral imaging and unmixing [20].
The second problem with multispectral imaging systems is
that they are generally slow compared with regular whole-slide
scanning systems. At the time of writing, there are no commer-
cial multispectral whole-slide scanning systems on the market.
However, there are a few systems which combine the ability to
do multispectral imaging on a per-field-of-view basis with whole-
slide scanning [13, 25]. These so-called “survey and drill” sys-
tems use a regular color imaging system to generate a
low-magnification overview of the sample and then return to
acquire higher resolution multispectral images from a subset of
fields of view. This typically yields 10–50 fields of multispectral
data, compared to the sometimes thousands of fields obtained
with a traditional whole-slide scanner. However, a pre-commer-
cial system has been developed which can scan an entire micro-
scope multispectrally in a reasonable timeframe [26]. This will
help with the expansion of these methods into more routine use
in pathology research.
Despite all of the potential pitfalls, it is possible to obtain quan-
titative separations of many chromogens in a single sample. The
biggest challenges are typically with optimization and validation.
There are simple methods available for two markers and it can be
fairly straightforward to develop a three-marker (four-color) assay.
However, it can be difficult and time-consuming to develop stain-
ing assays for more than four colors. For more than three markers,
a fluorescence approach is more common and generally considered
to be easier.
5 Staining and Imaging: Fluorescence
For fluorescence imaging, there are a large number of options for

labeling multiple fluorophores in FFPE tissue sections. Each of
these has their pros and cons, depending on the level of multiplex-
ing required and the abundance of the protein being labeled.
5.1 Direct Labeling One of the simplest methods for multiplexed fluorescence staining
is to use antibodies which have already been conjugated to fluoro-
phores of different colors. Other than issues with spectral overlap
between channels during imaging, there is no theoretical limit to
the number of markers which can be used in this manner. However,
there are some practical limitations to this. First of all, one must
have antibodies which have been directly conjugated to the fluoro-
phores. There are a few antibody–fluorophore conjugates which
are available commercially, although the vast majority of antibodies
one would wish to use are not available in pre-conjugated form.
This means that the conjugation must be done in the lab, using
one of the commercial conjugation kits or by developing the chem-
istry independently. In addition, since this method results in only
one or a few fluorophores for each antibody bound to the sample,
there may not be enough fluorescence to detect a signal, especially
against an autofluorescent sample background.
5.2 Primary– One means of avoiding a conjugation step to directly label an anti-
Secondary body with a fluorophore is to use primary–secondary antibody
Conjugation Methods methodologies. For this, a primary antibody of a specific species is
introduced into the sample and then a secondary antibody directed
against that species and labeled with a fluorophore is used. The
secondary antibody binds to the primary antibodies in the sample,
labeling them with a fluorophore. This methodology can be
extended to multiple fluorophores by using primary antibodies of
different species, and then using secondary antibodies (with differ-
ent colored fluorophores) specific for those same species. This
relies on having a separate species of primary antibody for each
marker to be labeled. It can be challenging to find appropriate rab-
bit, mouse, rat, goat, chicken, camel combinations for the exact
markers one needs. In addition, this method can still suffer from
sensitivity problems, at it results in only a single antibody–
fluorophore-labeling event for each primary antibody present.
5.3 Quantum Dots Quantum dots (QDs), small nanoscale semiconductor devices with
useful fluorescence properties, have been used in an effort to
address both the issue of needing multiple nonoverlapping fluoro-
phores and the sensitivity problems of direct and primary–second-
ary methods. By controlling the size of the nanoparticle, the Stokes
shift of QDs can be “tuned” to have them emit at narrow and
www.Ebook777.com
specific wavelengths using a single excitation wavelength, enabling

the use of many of these in a single sample without significant cross
talk. In addition, QDs are typically brighter than the organic small
molecule fluorophores normally used in microscopy. QDs have
been used in FFPE tissue sections to achieve multiplexing of up to
six colors in a single sample [27, 28]. However, there are still some
limitations to this technology. The QDs are large, often 20–30 nm
in diameter (once coatings have been put around them to prevent
their oxidization and make them soluble in water and biologically
compatible), and this can prevent their penetration into the speci-
men, limiting their access to only the outermost portion of the
tissue. QDs can be used both for direct labeling and for primary–
secondary staining methods. Although individual QD molecules
can have narrow fluorescence spectra (2–3 nm), the commercial
production is based on separating QDs of similar size into lots. The
end result is that a range of sizes is separated into a single lot, giv-
ing a distribution of individually narrow fluorescent peaks. This
distribution can be 10–30 nm in width, which still limits the num-
ber of QDs which can be multiplexed in a single sample [29].
Despite their potential advantages, QDs do not seem to work for
all markers and are not a single solution to multiplexed fluores-
cence imaging. However, they can be used in combination with
other methods and labels, which can make them useful in some
circumstances. Nonetheless, QDs have been used in a large num-
ber of studies describing the use of multiplexed QD immunohisto-
chemistry in pathology [28, 30, 31].
5.4 Immuno-enzyme In general, the problem of “one fluorophore per antibody” can
Amplification limit the utility of direct and primary–secondary methods. Abundant
proteins are not too much of a problem but for proteins which have
a very low copy number in a cell, regular conjugation methods are
insufficient. These sensitivity issues can largely be overcome using
an enzyme-mediated amplification method involving HRP analo-
gous in some ways to that used for chromogenic staining [32].
Tyramide signal amplification (TSA) involves the same workflow as
the chromogenic amplification method: application of a primary
antibody, conjugation with a species-specific secondary antibody
which contains HRP and then binding of multiple fluorophores to
the sample using an HRP-catalyzed reaction [33]. The primary use
of TSA has been for signal amplification. However, as a part of the
tyramide reaction, the fluorophores are covalently bound to the
sample, unlike direct and regular primary–secondary methods, in
which the fluorophore is only weakly bound through non-covalent
mechanisms. As a result, tyramide-reaction-bound fluorophores
will remain in place on the sample during a boiling/heating proce-
dure that will strip the primary and secondary antibodies. This will
leave the sample prepared for a second round of TSA primary–sec-
ondary-amplification process using a primary antibody of any
species, including the same species as the other primaries being
used. The heat step, typically through boiling the sample in a buf-
fer, is similar to those used for chromogenic multiplexing and simi-
lar to the heat-induced epitope retrieval step often performed on
FFPE tissues. This species-independent multiplexing, combined
with the signal amplification benefits of TSA, makes it useful for
multiplexed immunofluorescence [34]. The primary drawback of
TSA multiplexing is that it is a sequential method, requiring the
repetition of the entire staining cycle for each color/marker being
used. For a 5- or 6-color staining protocol, this can take 18–24 h of
processing time. In addition, sometimes the heating step is per-
formed using a microwave, which can hinder automation of the
process.
5.5 Autostaining Although not an immunohistochemistry method per se, autostain-

ers are an important part of in situ multiplexing. To be able to
utilize a multiplex immunohistochemical approach in pathology
and medical research, it is important to be able to fully automate
the staining of the sample since a clinical study can involve hun-
dreds or thousands of samples, each with one or more tissue sec-
tions to be analyzed. Having to manually stain these sections,
particularly with a sequential method like TSA, can be onerous at
best and practically impossible in the worst-case scenario.
Fortunately, many of these methods can be automated in commer-
cial automated slide staining systems [35]. In addition, there are
now commercially available solutions for multiplexed TSA staining
for up to five markers, with future expansions planned.
5.6 Combining IHC Phenotype visualization and quantitation via immunohistochemis-

and ISH try is valuable. However, its utility can be increased when com-
bined with in situ hybridization for mRNA or DNA, giving a read
out on the genotype, phenotype and the mechanism by which one
is translated into the other. Although relatively new, there have
already been a number of publications showing the utility of this
kind of multiplexing [36–38].
6 Imaging Considerations in Fluorescence
Most conventional microscopes use a fluorescence imaging

scheme that involves having one emission wavelength for each
excitation wavelength. This is typically achieved by using a filter
cube with an excitation filter, a dichroic mirror and an emission
filter. The excitation filter selects which excitation wavelength
range will be extracted from a broad-spectrum excitation light
source. The excitation light is shone onto the dichroic mirror,
which is chosen to reflect the excitation light and transmit the
emission light. The dichroic mirror reflects the excitation light
through the objective lens onto the sample. The fluorescence
emission thus generated is collected by the objective lens and
transmitted to the dichroic mirror, which passes the light to the

emission filter. The emission filter is chosen to absorb light in the
range of the excitation wavelengths (to prevent those from reach-
ing the camera) and to pass a typically narrow range of emission
wavelengths on to the camera. The excitation, dichroic mirror,
and emission filters are carefully chosen to match the absorbance
and emission properties of the fluorophore of interest for that fil-
ter cube. This type of fluorescence imaging utilizes a monochrome
camera to take an image of the sample with each filter cube. The
resulting gray-scale images, often referred to as “channels,” are
then typically pseudo-colored into any color of choice. In this
way, a multilayer pseudo-color image is created with one layer per
filter cube (and per fluorophore).
In conventional fluorescence microscopy, methods have been
worked out for imaging four colors simultaneously with little prob-
lem in non-fixed samples such as tissues or cells, and there have
been a high number of publications using those methods. Unlike
cell samples, for FFPE tissue sections there is the problem of tissue
autofluorescence, which can mask or completely obscure signals of
interest and can interfere with quantitation of even those signals
that can be seen above the autofluorescence. In addition, the desire
to multiplex more than four colors at one time can make cross-talk
between channels harder to manage. To address this, there are a
number of cross-talk correction algorithms based on linear unmix-
ing that can be used to ensure that the fluorophores are viewed or
quantitated in the correct channel [39, 40].
6.1 Color In pathology, fluorescently labeled tissue sections are most often
Versus Multispectral imaged using a commercial whole-slide scanning system, which
Imaging can rapidly digitize the entire tissue section. The resulting whole-
slide can be used for viewing or quantitation. If the fluorophores in
the sample are bright enough to be easily seen above the intrinsic
tissue autofluorescence then visualization is easy. If the fluoro-
phores are weak relative to the autofluorescence, then slide visual-
ization can be difficult or impossible with a regular whole-slide
scanner. Fluorescence images from a regular microscope or a
whole-slide scanner can also be used for intensity quantitation of a
pixel or region of the sample. Again, if the fluorescence signals in
the sample are strong relative to the intrinsic autofluorescence,
then quantitation can be straight forward. However, even small
amounts of tissue autofluorescence in a given channel can cause
problems. A 5 % contribution of autofluorescence is probably not
very important. However, given the strong autofluorescence in
formalin-fixed tissue, many samples can contain 20 %, 50 % or even
90 % autofluorescence in a given channel, especially in the green-
yellow emission range. These interfering signals can make signal
level quantitation impossible on a regular microscope or whole-
slide scanner.
www.Ebook777.com
For situations where autofluorescence is a problem, either for

simple visualization or for intensity quantitation, a multispectral
imaging approach is required [41, 42]. By acquiring more spectral
information about the sample, it is possible to unmix or separate
the autofluorescence from the fluorophores of interest, increasing
the contrast and legibility of a slide and improving intensity quan-
titation. The same unmixing algorithms that can correct for auto-
fluorescence can be used to correct cross-talk between channels.
The use of multispectral imaging to remove autofluorescence has
been applied over many years in a number of fields, including small
animal imaging and in microscopy [41]. It has proved especially
useful in pathology, with hundreds of publications using this
approach over the last 10 years. This has been documented in a
recent review article [43].
Although excellent for fluorescence imaging of FFPE tissue
sections, there are two major drawbacks with multispectral imag-
ing approaches. Firstly, obtaining correct results with linear unmix-
ing requires having the correct spectral “library,” or spectral basis
set. The preferred method for developing a spectral library is to use
samples that have been labeled with only one fluorophore as a
“spectral control.” In the case of autofluorescent samples, how-
ever, this can be difficult as there are typically no pixels in an image
that contain the fluorophore of interest and are also free of auto-
fluorescence. To overcome this, an unstained sample is used to
obtain a spectrum of autofluorescence and then that is used to
isolate purified signatures from the control sample spectra [42].
There is a debate over how many extra wavelengths need to be
acquired in a multispectral approach for autofluorescence removal.
Theoretically, a minimum of N + 1 wavelengths for N fluorophores
is required. However, in practice more wavelengths are required to
sufficiently measure the spectral properties of each fluorophore and
to quantitatively separate them from each other and from autofluo-
rescence. However, it is not yet clear just how many are required
for a quantitative separation in the case of FFPE autofluorescence
and the kinds of fluorophores used in pathology.
6.2 Confocal Both confocal and widefield fluorescence systems have been widely
and Wide-Field used for multiplexed imaging in the life sciences. However, the
Imaging confocal approach has been typically focused on cellular imaging,
while widefield fluorescence has been used for pathology samples,
although there are some notable and interesting exceptions [44,
45]. Confocal systems are designed to eliminate interference from
out-of-focus light in a sample through the use of pinholes to reject
light from regions above and below the focal plane. This means
that the majority of light detected comes from a specified voxel in
the sample, with localization in all three dimensions. Most com-
mercial confocal systems have the ability to scan in all three dimen-
sions and can therefore build up a three-dimensional map of the
sample [46]. For thick samples (>10 μm) or when full three-
dimensional information is required, this is clearly an advantage.
However, when imaging with a 20× objective, which is the de facto
standard in pathology, most if not all of a 5-μm section is in focus
in a non-confocal image even for high numerical aperture objec-
tives, which means that there is little advantage to using a confocal
system. However for 40× or 60× objectives, there can be a signifi-
cant amount of out-of-focus light that will cause blur in an image.
In addition, it can be advantageous to be able to do three-
dimensional co-localization analyses. In those situations, confocal
imaging can be useful. In particular, three-dimensional resolution
of small spots can be assisted by confocal imaging. DNA or RNA
in situ hybridization (FISH or CISH) spots can be small and at
20× it can be difficult to resolve individual foci, particularly when
they are aligned vertically in the sample. This problem can be exac-
erbated when imaging 5-μm sections because of the way sectioning
can leave only portions of individual cells in the sample. Confocal
approaches can enable the use of thicker sections containing cells
which can be analyzed in their entirety, albeit at the cost of taking
more time to acquire imagery.
7 Highly Multiplexed Imaging
Traditional immunohistochemistry and imaging, as described

above, is an excellent means of combining pathology and tissue
architecture with the visualization and quantitation of protein dis-
tributions. However, even the most ambitious of methods will be
limited in the number of markers that can be assessed, with some-
where between 6 and 12 being a practical limit. However, there are
good reasons for wanting to use more than 6–12 markers.
Fluorescence flow cytometry can handle up to 19 or 20 markers
and newer mass spectrometry-based flow cytometric methods have
been demonstrated with 40 markers, up to a limit of potentially
around 100 [47–50]. These higher multiplexing methods have
been shown to be useful, particularly for cancer immunological
applications. Mass spectroscopy imaging has been applied to FFPE
tissue sections for a number of years and is approaching a more
main-stream acceptance, as instrumentation becomes less complex
and more affordable [51, 52]. There have been two good recent
reviews of mass spectroscopy imaging [53, 54].
Another new technology involves using UV light to release
previously bound antibodies and their bar-code fluorescent tags
from spatially resolved regions of an FFPE section. The released
markers are then collected and analyzed in a separate analysis
instrument, thus giving a read out of up to 800 markers simultane-
ously from a small region of the sample. The analysis technology
(nCounter™) has been in use for a number of years, and has
numerous applications in pathology for protein, DNA and RNA

quantitation in any combination [55–57]. The new version of that
technology can analyze the same markers as before, but provides
the information from smaller regions of an FFPE section. Although
new, this technology has the potential to deliver spatially resolved
phenotypic and genotypic information from regions as small as
single cells.
8 Sequential Acquisition Methods
Another approach to obtaining complex phenotypic information

from cells in situ is to acquire images from a sample containing a
smaller number of markers (typically 2–4) and then remove these
markers and re-image the same field of view with a new set of
markers, repeating as many cycles as necessary. One such study
showed the detection of 61 protein epitopes in a single FFPE
section using pair-wise immunofluorescence [58]. This technology
has been commercialized under the name MultiOmyx™ and sold
as a service [59]. This kind of technology provides for a highly
multiplexed assay, with up to 100 markers acquired simultaneously.
There are other methods for achieving high multiplicity in single
cells using cyclic immunofluorescence, in which two- or four-color
staining alternates with inactivation of fluorophores to progres-
sively build a multichannel image. The use of standard reagents
and instrumentation may be suitable for high-throughput assays
and screening applications [60, 61].
Given the sequential nature of these processes, it can be slow
to acquire all of the imagery needed for a single field of view from
a single sample. In addition, care needs to be taken with image
registration to ensure that the data from each pixel in each of the
sequential images is coming from the exact same spot on the sam-
ple. Despite these limitations, a number of publications have come
out using this approach [62–65].
9 Image Analysis Methods
Multiplex staining and image acquisition methods are just two

parts of the story. The more colors that are put into a single slide,
the more complicated it becomes to be able to visually assess the
slide, particularly if there is some co-localization of the markers.
Even with 4–6 markers plus a counterstain, it can be difficult to
make sense of the images by eye. However, once the very highly
multiplexed methods come into more mainstream use, it will be
virtually impossible to visualize these markers all at once using any
kind of standard histopathology display method. Therefore, some
kind of image analysis and redisplay is required. In addition, the
primary driver of adding so many markers to FFPE samples is to

better understand the phenotype of each cell, something that is not
obvious from simply viewing markers. Again, this highlights the
importance of image analysis in multiplexed imaging.
Image analysis in pathology can be divided broadly into three
general steps: (1) tissue segmentation and quantitation; (2) cellular
segmentation and quantitation; and (3) image/slide summary and
meta-analysis. The goal in tissue segmentation is to use some kind
of tissue morphology assessment algorithm to subdivide the image
into morphologically relevant regions that can then be analyzed
individually for area and as a part of the per-cell summary informa-
tion. Cellular segmentation covers a range of analyses applied to
individual cells in an image. It is often a multistep process, with a
variety of algorithms employed, typically starting with the finding
of each cell’s nucleus based on the nuclear counterstain. Cellular
cytoplasm and membrane regions are then found based on the
nuclear location and shape. Once each cellular region has been
defined, the intensity of each marker can be extracted. This subcel-
lular per-marker intensity information can then be translated into a
multimarker phenotype for each cell. Once each cell’s phenotype is
known, summary data can be compiled for each field of view and
for each microscope slide, counting the number of each cell type in
each morphologic region of the sample. For example, this approach
can be used to determine how many CD4+/CD25+/FOXP3+
cells are in a tumor, tumor margin, and/or stroma on a slide.
Tissue, or morphologic, segmentation is often the most dif-
ficult step. The morphology of clinical FFPE sections is complex
at best as this varies from sample to sample and tumor type to
tumor type, and can be equivocal even for experienced patholo-
gists [66, 67]. One approach is to have an expert manually out-
line the morphologic regions of interest in each image. This can
work if there are only a few images that need analyzing but is
evidently impractical for larger data sets with thousands of images.
Developing an automated algorithm that will assess an image and
segment it into morphologic regions is difficult, but there are
successful examples.
There are two general approaches. One is parametric: an algo-
rithm is developed using human-chosen image criteria (nuclear
size, shape, distance metrics, colors, etc.) to segment unknown
images [68, 69]. Another method is to employ user-drawn training
regions as examples of each morphological class and then to use a
computer program to determine an effective, typically black-box
algorithm for distinguishing example regions. This is sometimes
termed “machine-learning” or “deep learning” and has been
applied in a range of image analysis problems, including pathology
[70, 71]. A number of groups and commercial software packages
have incorporated one or both of these schemes, which can be
applied to a range of digital pathology image and sample types
[72–77].
Cellular segmentation and analysis is typically the next step. In

this, the location and shape of each cell is determined using a cell
segmentation algorithm, typically utilizing the nuclear counterstain
to denote the nuclear location and then finding the rest of the cel-
lular subcompartments using a variety of other methods. Cellular
segmentation is not new and has been applied to a large variety of
cell types and situations [78]. However, cells in situ in FFPE tissue
sections present a more difficult challenge than live or fixed cells on
a glass slide or in a well. A 5 μm section of FFPE tissue cuts through
cells, leaving partial cells in the section and the possibility of spatial
overlap between cells. This makes nuclear segmentation difficult,
requiring special algorithms for FFPE tissue compared to other
samples [79, 80]. In addition, because of the three-dimensional
aspects of a section and the partial cells found there, performing
cytoplasmic and membrane segmentation on FFPE sections is also
more difficult. Many nuclear segmentation algorithms make the
assumption that “one nucleus equals one cell,” which can make
segmentation of multinucleated cells difficult. Cytoplasmic seg-
mentation tends to be done by a watershed expansion starting from
each cell’s nucleus, which requires an active membrane signal to
determine correct cellular boundaries. Membrane segmentation
typically requires having a good membrane stain from which to
work, a strategy that works well when there is a clear membrane
pattern, but which can be problematic for cells with no obvious
membrane staining. In spite of these complications, there are many
commercial and user-developed software packages and algorithms
available which do an effective job of performing cellular segmenta-
tion on brightfield and fluorescence images from FFPE sections.
Once the cellular and subcellular boundaries and compart-
ments have been decided, then it is a relatively simple matter to
extract the fluorescence or absorbance data from each specific cell
and subcellular (e.g., nuclear, cytoplasmic and membrane) region.
This gives a table of data in which each row represents the data for
a specific cell, which is analogous to a table of data obtained from
flow cytometry. For example, a sample with 5-plex staining and a
nuclear counterstain, and subdivision of each cell into nuclear,
cytoplasmic and membrane compartments, this would result in 20
variables (nuclear intensities for all five fluors, cytoplasmic intensity
for all five fluors, the membrane intensity for all five fluors and the
total intensity for all five fluors). In addition, it is often possible to
extract other summary data about the fluors in each subcompart-
ment (minimum, maximum, standard deviation, and sometimes
texture-based readings), and each cell typically comes with a range
of shape/size variables (area, long axis, short axis, circularity, etc.),
as well as texture variables about the staining. This results in a very
long list of variables for each cell in the image.
This table of per-cell data forms the basis of cellular pheno-
typing. The simplest approach for analyzing this kind of table is
to use a flow-cytometry sequential gating approach. Most cell
www.Ebook777.com
segmentation data can be exported and read into flow-cytome-

try commercial software for analysis [81, 82]. However, direct
application of the sequential gating methods used so effectively
in cytometry to cellular data derived from cells in situ in FFPE
sections may not be satisfactory [83]. In addition to the three-
dimensional problems of overlapping cells, there is a large cel-
lular contact problem. In a flow cytometer, the cells are analyzed
one at a time and the levels of each marker are measured for that
cell without interference from other cells. However in tissue
sections, cells can touch each other and, in the case of smaller
immune cells, this can involve many cells. The nature of mem-
brane segmentation of these cells in situ means that they share
the membrane with neighboring cells, which means that if a
CD4+/CD8− cell shares 25 % of its membrane region with a
touching CD4-/CD8+ cell, the nice, neat dual cluster scatter
plot seen for CD4/CD8 differentiation will not exist for these
cells in tissue. The two neat clusters will spread into one another,
depending on the sharing of membranes and spatial overlap.
One approach to dealing with this tissue cytometry problem
was developed by Ron Germain and coworkers at the National
Institute of Health and has been applied to a number of immunol-
ogy applications. These include the role of immune homeostasis by
co-localized effector and regulatory T cells [84] and assessment of
dendritic cell subpopulations [45]. Another approach is to have a
user-guided training of an algorithm based on examples the user
selects for each multimarker phenotype of interest. The computer
then generates an algorithm which creates the best multivariate
gating scheme to separate the cells into phenotypes based on the
long list of per-cell variables in the data table. One implementation
of this machine learning for cellular phenotyping involved using a
multinomial logistic regression algorithm to find the multivariate
gating scheme in a 6-plex assay to divide cells into categories of
tumor cells, cytotoxic T cells, helper T cells, regulatory T cells, and
macrophages, producing cell phenotype maps retaining spatial
arrangements. This kind of multivariate gating scheme requires
users to select a minimum number of examples of the cells exem-
plifying the phenotypes of interest to train the algorithm. This is
likely to be the most promising approach to this kind of highly
multiplexed imaging and analysis problem.
Once each cell has been located and phenotyped, with its
marker intensities extracted on a subcellular basis, there are a range
of options for analysis. A simple visual picture of the distributions
of the various cell types in the tumor and tumor microenvironment
can be useful. However, visual assessment of the inter-distributions
of multiple cells types is difficult. For this reason a number of sum-
mary data analyses have been developed to begin to address this
data analysis problem. These kinds of summary analyses can be
done on a per-field (per-image) basis or across all of the images
from a section, or even across a cohort of patients in a study. A

simple metric is to simply count the number of each phenotypical
cell in each morphological region. These summary metrics can be
augmented by a range of spatial mapping algorithms. Typically
these can be used to explore whether or not the distances between
the cells of a specific phenotype are related to a clinical/experimen-
tal parameter such as survival, dose or age. These can be simple
spatial distance metrics such as a Euclidean or Mahalanobis dis-
tances or more complicated metrics involving the relationships
between multiple cellular phenotypes and tissue morphological
regions. There are also hypothesis-free methods which can be used
to determine what spatial patterns of cells correlate with clinical or
experimental parameters such as survival [85, 86]. This all contrib-
utes to being able to map the spatial heterogeneity in the tumor
microenvironment, something which is understood to be of
increasing importance [87].
10 Literature Highlights
Two groups at the National Institute of Health have developed

multiplexed imaging and per-cell analysis technologies utilizing
confocal microscopes [88]. Sometimes termed histo-cytometry,
this method has been used to stain and image up to ten fluorescent
markers, and then open source software is used to perform per-cell
analysis [45]. These have been primarily used in neuroscience and
immune-oncology applications, from investigating how helper B
cells can promote cytotoxic T cell survival and proliferation inde-
pendently of antigen presentation through CD27/CD70 interac-
tions [89], to how immune homeostasis is enforced by co-localized
effector and regulatory T cells [84] to investigations of the spatio-
temporal basis of immunity in lymphoid tissue [90].
Multispectral imaging has been used for over a decade as the
basis for multiplexed imaging in pathology, with hundreds of
papers published. Full coverage of the uses of this technology for
the multiplexed imaging of clinical biopsy samples is outside of the
scope of this chapter but there has been a recent literature review
which addresses this [43].
One recent example of multiplexed imaging and per-cell anal-
ysis is the investigation of tumor infiltrating plasma cells and their
association with tertiary lymphoid structures, cytolytic T cell
responses and prognosis in ovarian cancer [91]. Another is a study
into tumor-derived lipocalin-2 and whether or not it promotes
breast cancer metastases. To this end, co-localized PyMT- and
Ki-67-double positive tumor cells were detected and quantified
[92]. In immune-oncology, the T cell landscape in primary mela-
noma was investigated and found to predict the survival of patients
with metastatic disease after their treatment with dendritic cell
vaccines [93]. In another immune-oncology paper, the authors

discuss how CD169 can identify an activated CD8+ T cell subset
in lymph nodes which can predict a favorable prognosis in colorec-
tal cancer [94]. This same set of technologies has been applied to
the studying the effects of transforming growth factor-β to limit
secretion of lumican by activated stellate cells within primary pan-
creatic adenocarcinoma tumors [95].
Recent work from the group of Bernard Fox has shown a meth-
odology that can be consistently applied to investigate FFPE tissues
by multiplexing up to six markers. This helped to enumerate the
complicated phenotypes of immune cell subsets and allowed spatial
distribution analyses. They validated the use of multiplex immuno-
histochemistry for detection of CD3+CD4+ and CD3+/CD8+ T
cell subsets in murine spleen and tumors [96] and applied the same
methodology to analyzing tumor-infiltrating lymphocytes in mela-
noma [97]. The institute led by James Allison at MD Anderson has
been among the leaders in immune-oncology research for many
years. In a recent publication, they analyzed immune signatures in
longitudinal tumor samples for insight into biomarkers of response
and mechanisms of resistance to immune checkpoint blockade. This
work demonstrated that adaptive immune signatures in tumor
biopsies can be predictive of response to checkpoint blockade treat-
ment. This study is also of high interest because it combined multi-
plexed protein analysis, per-cell quantitation and phenotyping with
highly multiplexed non-imaging analyses of homogenized tissues
[98]. These combinations of traditional pathology imaging and
analysis with newer multiplexed quantitative pathology methods
along with multiplexed analyses of homogenized tissues are likely to
become increasingly useful as tumor–immune interactions are
investigated further.
11 Conclusions
The field of multiplexed imaging and per-cell phenotyping of cells

in situ in FFPE sections has undergone a rapid growth thanks to
the interest in immune–tumor interactions, fuelled by the success
of new cancer immunotherapies. There are a range of technolo-
gies that have been developed, including simple two- or three-
plex chromogenic staining with a visual assessment all the way to
highly multiplexed methods which can analyze hundreds of mark-
ers at a time from individual cells. All of these techniques require
some kind of marker labeling strategy, an imaging strategy and an
analysis strategy, each with its own pros and cons. Up to six-plex
per-cell phenotyping using immunofluorescence labeling, multi-
spectral imaging, and morphologic and per-cell analysis software
has become attainable, using a few different commercial staining
methods and image analysis packages. Beyond that, there are a
few highly multiplexed methods that are just becoming available.

Although these can provide potentially up to 800-plex analyses,
the acquisition time is slow and the scope of an experiment using
these approaches may be limited. One area that has not seen much
development so far is in the combination of genotyping and phe-
notyping of cells—combining ISH for DNA and RNA, and immu-
nohistochemistry for protein staining along with imaging and
cellular analyses. These hold the potential to expand our under-
standing of tumor–immune interactions even further. However,
further developments are necessary due to the spot-counting
nature of many ISH analysis problems and the three-dimensional
problems of cells in the context of the tumor microenvironment.
Addressing these will require the development of three-dimen-
sional imaging methods for thicker (10–30 μm) samples.
References
1. Medzhitov R, Shevach EM, Trinchieri G, 10. van der Loos CM, Teeling P (2008) A gener-
Mellor AL, Munn DH, Gordon S et al (2011) ally applicable sequential alkaline phosphatase
Highlights of 10 years of immunology in immunohistochemical double staining.
nature reviews immunology. Nat Rev Immunol J Histotechnol 31:119–127. doi:10.1179/
11:693–702 his.2008.31.3.119
2. Mandy FF, Bergeron M, Minkus T (1997) 11. Osman TA, Øijordsbakken G, Costea DE,
Evolution of leukocyte immunophenotyping as Johannessen AC (2013) Successful triple
influenced by the HIV/AIDS pandemic: a immunoenzymatic method employing primary
short history of the development of gating antibodies from same species and same immu-
strategies for CD4^+ T-cell enumeration. noglobulin subclass. Eur J Histochem 57:e22
Cytometry 30:157–165 12. Blenman KR, Lee PP (2014) Quantitative
3. Loken MR (1990) Immunofluorescence and spatial image analysis of tumor and drain-
Techniques. In: Flow cytometry and sorting. ing lymph nodes using immunohistochemis-
2nd edn. Wiley-Blackwell; ISBN-13: 978– try and high-resolution multispectral imaging
0471562351, pp 341–353 to predict metastasis. Methods Mol Biol
4. Zola H, Swart B, Banham A, Barry S, Beare A, 1102:601–621
Bensussan A et al (2007) CD molecules 2006- 13. Setiadi AF, Ray NC, Kohrt HE, Kapelner A,
-human cell differentiation molecules. Carcamo-Cavazos V, Levic EB et al (2010)
J Immunol Methods 318:1–5 Quantitative, architectural analysis of immune
5. Rimm DL (2006) What brown cannot do for cell subsets in tumor-draining lymph nodes
you. Nat Biotechnol 24:914–916 from breast cancer patients and healthy lymph
6. Taylor CR, Levenson RM (2006) Quantification nodes. PLoS One 5:e12420
of immunohistochemistry—issues concerning 14. Oguejiofor K, Hall J, Slater C, Betts G, Hall G,
methods, utility and semiquantitative assess- Slevin N (2015) Stromal infiltration of CD8 T
ment II. Histopathology 49:411–424 cells is associated with improved clinical out-
7. Potts SJ, Johnson TD, Voelker FA, Lange H, come in HPV-positive oropharyngeal squa-
Young GD (2011) Multiplexed measurement of mous carcinoma. Br J Cancer 113:886–893
proteins in tissue in a clinical environment. Appl 15. van der Loos CM (2010) Chromogens in mul-
Immunohistochem Mol Morphol 19:494–498 tiple immunohistochemical staining used for
8. Dixon AR, Bathany C, Tsuei M, White J, visual assessment and spectral imaging: the col-
Barald KF, Takayama S (2015) Recent devel- orful future. J Histotechnol 33:31–40.
opments in multiplexing techniques for immu- doi:10.1179/his.2010.33.1.31
nohistochemistry. Expert Rev Mol Diagn 16. van der Loos CM (2008) Multiple immunoen-
15:1171–1186 zyme staining: methods and visualizations for
9. van der Loos CM, Naruko T, Becker AE the observation with spectral imaging.
(1996) The use of enhanced polymer one-step J Histochem Cytochem 56:313–328
staining reagents for immunoenzyme double- 17. Farkas DL, Ballou B, Du C, Fisher GW, Lau C,
labelling. Histochem J 28:709–714 Levenson RM (1997, September). Optical
image acquisition, analysis and processing for dots versus organic dyes as fluorescent labels.
biomedical applications. In: Image analysis and Nat Methods 5:763–775
processing. Session 11: biomedical applications 30. Byers RJ, Hitchman ER (2011) Quantum dots
image analysis and processing. Volume 1311 of brighten biological imaging. Prog Histochem
the series lecture notes in computer science, Cytochem 45:201–237
pp 663–671 31. Tholouli E, MacDermott S, Hoyland J, Yin JL,
18. Mansfield JR, Sowa MG, Scarth GB, Somorjai Byers R (2012) Quantitative multiplex quan-
RL, Mantsch HH (1997) Fuzzy C-means clus- tum dot in-situ hybridisation based gene
tering and principal component analysis of time expression profiling in tissue microarrays iden-
series from near-infrared imaging of forearm tifies prognostic genes in acute myeloid leukae-
ischemia. Comput Med Imaging Graph mia. Biochem Biophys Res Commun
21:299–308 425:333–339
19. Mansfield JR, Hoyt C, Levenson RM (2008) 32. Chao J, DeBiasio R, Zhu Z, Giuliano KA,
Visualization of microscopy‐based spectral Schmidt BF (1996) Immunofluorescence sig-
imaging data from multi‐label yissue sections. nal amplification by the enzyme-catalyzed
Curr Protoc Mol Biol Chapter 14:Unit 14.19. deposition of a fluorescent reporter substrate
doi:10.1002/0471142727.mb1419s84 (CARD). Cytometry 23:48–53
20. Neher R, Neher E (2004) Optimizing imaging 33. Warford A, Akbar H, Riberio D (2014)
parameters for the separation of multiple labels Antigen retrieval, blocking, detection and visu-
in a fluorescence image. J Microsc 213:46–62 alisation systems in immunohistochemistry: a
21. Rothmann C, Bar-Am I, Malik Z (1998) review and practical evaluation of tyramide and
Spectral imaging for quantitative histology and rolling circle amplification systems. Methods
cytogenetics. Histol Histopathol 13:921–926 70:28–33
22. Macville MV, Van der Laak JA, Speel EJ, Katzir 34. Wang G, Achim CL, Hamilton RL, Wiley CA,
N, Garini Y, Soenksen D (2001) Spectral imag- Soontornniyomkij V (1999) Tyramide signal
ing of multi-color chromogenic dyes in patho- amplification method in multiple-label immu-
logical specimens. Anal Cell Pathol 22:133–142 nofluorescence confocal microscopy. Methods
23. Wemmert C, Kruger JM, Forestier G, 18:459–464
Sternberger L, Feuerhake F, Gançarski P 35. Yarilin D, Xu K, Turkekul M, Fan N, Romin Y,
(2013). Stain unmixing in brightfield multi- Fijisawa S (2015) Machine-based method for
plexed immunohistochemistry. In: Image pro- multiplex in situ molecular characterization of
cessing (ICIP), 2013 20th IEEE International tissues by immunofluorescence detection. Sci
Conference. IEEE pp. 1125–1129. Doi: Rep 5:9534. doi:10.1038/srep09534
http://dx.doi.org/10.1016/j. 36. Stempel AJ, Morgans CW, Stout JT,
compmedimag.2015.04.001 Appukuttan B (2014) Simultaneous visualiza-
24. Choi S, Cichocki A, Park H-M, Lee SY (2005) tion and cell-specific confirmation of RNA and
Blind source separation and independent com- protein in the mouse retina. Mol Vis
ponent analysis: a review. Neural Inform 20:1366–1373
Process - Lett Rev 6(1):1–57 37. Kunju LP, Carskadon S, Siddiqui J, Tomlins
25. Ghaznavi F, Evans A, Madabhushi A, Feldman SA, Chinnaiyan AM, Palanisamy N (2014)
M (2013) Digital imaging in pathology: whole- Novel RNA hybridization method for the in
slide imaging and beyond. Annu Rev Pathol situ detection of ETV1, ETV4, and ETV5
8:331–359 gene fusions in prostatecancer. Appl
26. Lindner M, Shotan Z, Yuval G (2016) Rapid Immunohistochem Mol Morphol 22:e32–e40
microscopy measurement of very large spectral 38. Li P, Janczewski WA, Yackle K, Kam K,
images. Opt Express 24:9511–9527 Pagliardini S, Krasnow MA et al (2016) The
27. Vu TQ, Lam WY, Hatch EW, Lidke DS (2015) peptidergic control circuit for sighing. Nature
Quantum dots for quantitative imaging: from 530:293–297
single molecules to tissue. Cell Tissue Res 39. Zimmermann T, Rietdorf J, Pepperkok R
360:71–86 (2003) Spectral imaging and its applications in
28. Zhou L, Yan J, Tong L, Han X, Wu X, Guo P live cell microscopy. FEBS Lett 546:87–92
(2016) Quantum dot-based immunohisto- 40. Zimmermann T (2005) Quantum dot-based
chemistry for pathological applications. Cancer immunohistochemistry for pathological appli-
Transl Med 2:21–28. doi:10.4103/2395- cations. Adv Biochem Engin Biotechnol
3977.177562 95:245–265
29. Resch-Genger U, Grabolle M, Cavaliere- 41. Dickinson ME, Bearman G, Tille S, Lansford
Jaricot S, Nitschke R, Nann T (2008) Quantum R, Fraser SE (2001) Multi-spectral imaging
and linear unmixing add a whole new dimen- 55. Ullal AV, Peterson V, Agasti SS, Tuang S, Juric
sion to laser scanning fluorescence microscopy. D, Castro CM et al (2014) Cancer cell profil-
Biotechniques 31:1272–1279 ing by barcoding allows multiplexed protein
42. Mansfield JR, Gossage KW, Hoyt CC, analysis in fine-needle aspirates. Sci Transl Med
Levenson RM (2005) Autofluorescence 6(219):219ra9
removal, multiplexing, and automated analysis 56. Reis PP, Waldron L, Goswami RS, Xu W, Xuan Y,
methods for in-vivo fluorescence imaging. Perez-Ordonez B et al (2011) mRNA transcript
J Biomed Opt 10:41207 quantification in archival samples using multi-
43. Mansfield JR (2014) Multispectral imaging: a plexed, color-coded probes. BMC Biotechnol
review of its technical aspects and applications in 11:46. doi:10.1186/1472-6750-11-46
anatomic pathology. Vet Pathol 51:185–210 57. Kolbert CP, Feddersen RM, Rakhshan F, Grill
44. Gerner MY, Torabi-Parizi P, Germain RN DE, Simon G, Middha S et al (2013) Multi-
(2015) Strategically localized dendritic cells platform analysis of microRNA expression
promote rapid T cell responses to lymph-borne measurements in RNA from fresh frozen and
particulate antigens. Immunity 42:172–185 FFPE tissues. PLoS One 8:e52517
45. Gerner MY, Kastenmuller W, Ifrim I, Kabat J, 58. Gerdes MJ, Sevinsky CJ, Sood A, Adak S, Bello
Germain RN (2012) Histo-cytometry: a MO, Bordwell A et al (2013) Highly multi-
method for highly multiplex quantitative tissue plexed single-cell analysis of formalin-fixed,
imaging analysis applied to dendritic cell subset paraffin-embedded cancer tissue. Proc Natl
microanatomy in lymph nodes. Immunity Acad Sci U S A 110:11982–11987
37:364–376 59. Hollman-Hewgley D, Lazare M, Bordwell A,
46. Sheppard CJR, Gan X, Gu M, Roy M, Pawley Zebadua E, Tripathi P, Ross AS et al (2014) A
JB (2006) In: Pawley JB (ed) Handbook of single slide multiplex assay for the evaluation of
biological confocal microscopy, 3rd edn. classical Hodgkin lymphoma. Am J Surg Pathol
Springer, New York, pp 442–451. ISBN ISBN- 38:1193–1202
13: 978–0387259215 60. Lin JR, Fallahi-Sichani M, Sorger PK (2015)
47. Giesen C, Wang HA, Schapiro D, Zivanovic N, Highly multiplexed imaging of single cells
Jacobs A, Hattendorf B et al (2014) Highly using a high-throughput cyclic immunofluo-
multiplexed imaging of tumor tissues with sub- rescence method. Nat Commun 6:8390.
cellular resolution by mass cytometry. Nat doi:10.1038/ncomms9390
Methods 11:417–422 61. Clarke GM, Zubovits JT, Shaikh KA, Wang D,
48. Tanner SD, Baranov VI, Ornatsky OI, Bandura Dinn SR, Corwin AD et al (2014) A novel,
DR, George TC (2013) An introduction to automated technology for multiplex biomarker
mass cytometry: fundamentals and applications. imaging and application to breast cancer.
Cancer Immunol Immunother 62:955–965 Histopathology 64:242–255
49. Schwamborn K, Caprioli RM (2010) Molecular 62. Sood A, Miller AM, Brogi E, Sui Y, Armenia J,
imaging by mass spectrometry—looking McDonough E, Pang Z et al (2016)
beyond classical histology. Nat Rev Cancer Multiplexed immunofluorescence delineates
10:639–646 proteomic cancer cell states associated with
50. Spitzer MH, Nolan GP (2016) Mass cytometry: metabolism. JCI Insight 1(6):e87030
single cells, many features. Cell 165:780–791 63. Adams A, Bordwell A, Bouman D, Fisher D,
51. Angelo M, Bendall SC, Finck R, Hale MB, Hollman D, Lazare M et al (2013)
Hitzman C, Borowsky AD (2014) Multiplexed MultiOmyxTM: a novel chemistry and visualiza-
ion beam imaging of human breast tumors. tion tools to aid resolution of discrepant
Nat Med 20:436–442 Hodgkin lymphoma cases-a single slide multi-
52. Schüffler PJ, Schapiro D, Giesen C, Wang HA, plex assay for the evaluation of classical
Bodenmiller B, Buhmann JM (2015) Hodgkin lymphoma. Blood 122:2994
Automatic single cell segmentation on highly 64. Gerdes MJ, Sood A, Sevinsky C, Pris AD,
multiplexed tissue images. Cytometry A Zavodszky MI, Ginty F (2014) Emerging
87:936–942 understanding of multiscale tumor heteroge-
53. Norris JL, Caprioli RM (2013) Analysis of tis- neity. Front Oncol 4:366. doi:10.3389/
sue specimens by matrix-assisted laser desorp- fonc.2014.00366
tion/ionization imaging mass spectrometry in 65. Uhlik MT, Liu J, Falcon BL, Iyer S, Stewart J,
biological and clinical research. Chem Rev Celikkaya H et al (2016) Stromal-based signa-
113:2309–2342 tures for the classification of gastric cancer.
54. Bodenmiller B (2016) Multiplexed epitope- Cancer Res 76:2573–2586
based tissue imaging for discovery and health- 66. Rizzardi AE, Johnson AT, Vogel RI,
care applications. Cell Syst 2:225–238 Pambuccian SE, Henriksen J, Skubitz AP et al
(2012) Quantitative comparison of immuno- magazine, 29(5), pp. 140–145. September

histochemical staining measured by digital 2012
image analysis versus pathologist visual scoring. 79. Irshad H, Veillard A, Roux L, Racoceanu D
Diagn Pathol 7:1596–1597 (2014) Methods for nuclei detection, segmen-
67. Mengel M, von Wasielewski R, Wiese B, tation, and classification in digital histopathol-
Rüdiger T, Müller-Hermelink HK, Kreipe H ogy: a review-current status and future
(2002) Inter‐laboratory and inter‐observer potential. IEEE Rev Biomed Eng 7:97–114
reproducibility of immunohistochemical assess- 80. Xing F, Yang L (2016) Robust nucleus/cell
ment of the Ki‐67 labelling index in a large detection and segmentation in digital pathol-
multi‐centre trial. J Pathol 198:292–299 ogy and microscopy images: a comprehensive
68. Beck AH, Sangoi AR, Leung S, Marinelli RJ, review. IEEE Rev Biomed Eng Jan 6. [Epub
Nielsen TO, van de Vijver MJ et al (2011) ahead of print]
Systematic analysis of breast cancer morphol- 81. Lugli E, Roederer M, Cossarizza A (2010)
ogy uncovers stromal features associated with Data analysis in flow cytometry: the future just
survival. Sci Transl Med 3:108ra113 started. Cytometry A 77:705–713
69. Baatz M, Zimmermann J, Blackmore CG 82. Aghaeepour N, Finak G, FlowCAP
(2009) Automated analysis and detailed quan- Consortium, DREAM Consortium, Hoos H,
tification of biomedical images using definiens Mosmann TR et al (2013) Critical assessment
cognition network technology. Comb Chem of automated flow cytometry data analysis
High Throughput Screen 12:908–916 techniques. Nat Methods 10:228–238
70. Fernandez DC, Bhargava R, Hewitt SM, Levin 83. Hedley DW (1989) Flow cytometry using par-
IW (2005) Infrared spectroscopic imaging for affin‐embedded tissue: five years on. Cytometry
histopathologic recognition. Nat Biotechnol 10:229–241
23:469–474 84. Liu Z, Gerner MY, Van Panhuys N, Levine
71. Levenson R (2008) Putting the “more” back AG, Rudensky AY, Germain RN (2015)
in morphology: spectral imaging and image Immune homeostasis enforced by co-local-
analysis in the service of pathology. Arch Pathol ized effector and regulatory T cells. Nature
Lab Med 132:748–757 528:225–230
72. Feldman MD (2008) Beyond morphology: 85. Rose CJ, Naidoo K, Clay V, Linton K, Radford
whole slide imaging, computer-aided detec- JA, Byers RJ (2013) A statistical framework for
tion, and other techniques. Arch Pathol Lab analyzing hypothesized interactions between
Med 132:758–763 cells imaged using multispectral microscopy
73. Veta M, Pluim JP, van Diest PJ, Viergever MA and multiple immunohistochemical markers.
(2014) Breast cancer histopathology image J Pathol Inform 4(Suppl):S4.
analysis: a review. IEEE Trans Biomed Eng doi:10.4103/2153-3539.109856
61:1400–1411 86. Nelson LS, Mansfield JR, Lloyd R, Oguejiofor
74. Pantanowitz L, Valenstein PN, Evans AJ, K, Salih Z, Menasce LP, Linton KM et al
Kaplan KJ, Pfeifer JD, Wilbur DC et al (2011) (2015) Automated prognostic pattern detec-
Review of the current state of whole slide imag- tion shows favourable diffuse pattern of
ing in pathology. J Pathol Inform 2:36. FOXP3(+) tregs in follicular lymphoma. Br
doi:10.4103/2153-3539.83746 J Cancer 113:1197–1205
75. Isse K, Lesniak A, Grama K, Roysam B, 87. Heindl A, Nawaz S, Yuan Y (2015) Mapping
Minervini MI, Demetris AJ (2012) Digital spatial heterogeneity in the tumor microenvi-
transplantation pathology: combining whole ronment: a new era for digital pathology. Lab
slide imaging, multiplex staining and automated Invest 95:377–384
image analysis. Am J Transplant 12:27–37 88. Germain RN (2005) Imaging dynamic interac-
76. Bueno G, Fernández-Carrobles MM, Deniz O, tions in immune responses. Semin Immunol
García-Rojo M (2016) New trends of emerg- 17:385–386
ing technologies in digital pathology. 89. Deola S, Panelli MC, Maric D, Selleri S,
Pathobiology 83:61–69 Dmitrieva NI, Voss CY et al (2008) Helper B
77. Akbar S, Jordan LB, Purdie CA, Thompson cells promote cytotoxic T cell survival and pro-
AM, McKenna SJ (2015) Comparing liferation independently of antigen presenta-
computer-generated and pathologist-generated tion through CD27/CD70 interactions.
tumour segmentations for immunohistochemi- J Immunol 180:1362–1372
cal scoring of breast tissue microarrays. Br 90. Qi H, Kastenmüller W, Germain RN (2014)
J Cancer 113:1075–1080 Spatiotemporal basis of innate and adaptive
78. Meijering E (2012) Cell segmentation: 50 immunity in secondary lymphoid tissue. Annu
years down the road. IEEE signal processing Rev Cell Dev Biol 30:141–167
91. Kroeger DR, Milne K, Nelson BH (2016) stellate cells within primary pancreatic adeno-
Tumor infiltrating plasma cells are associated carcinoma tumors. Clin Cancer Res, Apr 28.
with tertiary lymphoid structures, cytolytic T pii: clincanres.2780.2015. [Epub ahead of
cell responses, and superior prognosis in ovar- print]
ian cancer. Clin Cancer Res 22:3005–3015 96. Feng Z, Jensen SM, Messenheimer DJ, Farhad
92. Ören B, Urosevic J, Mertens C, Mora J, Guiu M, Neuberger M, Bifulco CB et al (2016)
M, Gomis RR et al (2016) Tumour stroma- Multispectral imaging of T and B cells in
derived lipocalin-2 promotes breast cancer murine spleen and tumor. J Immunol
metastasis. J Pathol 239:274–285 196:3943–3950
93. Vasaturo A, Halilovic A, Bol KF, Verweij DI, 97. Feng Z, Puri S, Moudgil T, Wood W, Hoyt
Blokx WA, Punt CJ et al (2016) T cell land- CC, Wang C et al (2015) Multispectral imag-
scape in a primary melanoma predicts the sur- ing of formalin-fixed tissue predicts ability to
vival of patients with metastatic disease after generate tumor-infiltrating lymphocytes from
their treatment with dendritic cell vaccines. melanoma. J Immunother Cancer 3:47.
Cancer Res 76:3496–3506 doi:10.1186/s40425-015-0091-z
94. Zhang J, Xu J, Zhang RX, Zhang Y, Ou QJ, Li 98. Chen PL, Roh W, Reuben A, Cooper ZA,
JQ et al (2016). CD169 identifies an activated Spencer CN, Prieto PA et al (2016) Analysis of
CD8+ T cell subset in regional lymph nodes immune signatures in longitudinal tumor sam-
that predicts favorable prognosis in colorectal ples yields insight into biomarkers of response
cancer patients. OncoImmunology (in press) and mechanisms of resistance to immune
95. Kang YA, Roife D, Lee Y, Lv H, Suzuki R, checkpoint blockade. Cancer Discov 6:827–
Ling J et al (2016) Transforming growth 837, Jun 14. pii: CD-15-1545. [Epub ahead of
factor-β limits secretion of lumican by activated print]
Part II
Statistical Considerations
Chapter 6
Identification and Clinical Translation of Biomarker

Signatures: Statistical Considerations
Emanuel Schwarz
Abstract
Powerful machine learning tools exist to extract biological patterns for diagnosis or prediction from high-
dimensional datasets. Simultaneous advances in high-throughput profiling technologies have led to a rapid
acceleration of biomarker discovery investigations across all areas of medicine. However, the translation of
biomarker signatures into clinically useful tools has thus far been difficult. In this chapter, several important
considerations are discussed that influence such translation in the context of classifier design. These include
aspects of variable selection that go beyond classification accuracy, as well as effects of variability on assay
stability and sample size. The consideration of such factors may lead to an adaptation of biomarker discov-
ery approaches, aimed at an optimal balance of performance and clinical translatability.
Key words Biomarker discovery, Machine learning, Variable selection, Classification, Molecular
diagnostics
1 Introduction
Technological advancements in high-throughput technology have

tremendously accelerated the search for biological patterns that
have clinical utility for diagnosis and prediction. Among these are
multiplexed assays that facilitate simultaneous measurement of
analytes in small sample volumes, with high-throughput and low
variability often comparable to the single-plex gold standard meth-
odology. The developments in these approaches have been paral-
leled by a tremendous increase in application of multivariate
methods to identify biological signatures within the generated
high-dimensional datasets, a process that has been accelerated by
the availability of complex algorithms in standard software pack-
ages. These facilitate the extraction of complex biological patterns
from high-dimensional data that can already be transferred effi-
ciently into dedicated multiplexed measurement systems. To aid in
this process, computational pipelines have been developed that
support translation from study design and initial biomarker
103
104 Emanuel Schwarz
screening to clinically applicable multiplexed tests [1]. However,

such transferability is platform dependent and results from high-
throughput profiling within a research setting may not be easily
transferred into clinically usable assay systems [2]. Despite these
technological advances, the translation of biomarker candidates
into clinical tests has been slow. For example in the cancer field,
many biomarker candidates that showed promise during initial
stages of the development process did not turn out to be clinically
useful, and this was frequently not realized until the later stages of
the process [3]. While there are numerous aspects that affect the
discovery and clinical translation of biomarker signatures, includ-
ing clinical, methodological, and regulatory challenges [4–8], this
chapter will focus on statistical considerations regarding the opti-
mal identification of biomarker sets.
The computational identification of biological signatures typi-
cally falls within the realm of supervised machine learning, where a
given part of labeled data is used for “training,” to select an “opti-
mal” combination of measurements for a predefined classification
or prediction task. The algorithm is subsequently tested in a part of
the data not used during training and the accuracy of the predic-
tion in this test data is used as an estimation of how the classifier
will perform in future, independent datasets. In practice, training
and testing is usually performed by splitting datasets into a training
and a test set, by cross-validation or similar techniques. In cross-
validation, the data is split randomly into a given number of junks
and each of these junks is used as a test set until all subjects have
been classified once. Estimates of classifier accuracy are then deter-
mined across the entire cohort used during cross-validation. The
identification of biomarker signatures using machine learning
methodology is influenced by a significant number of choices, such
as algorithm selection, data transformation, or parameter specifica-
tion. As these aspects have been extensively reviewed elsewhere
[9–11], this chapter focuses on areas that are less frequently con-
sidered during identification of machine learning classifiers but
which may be of significant importance for translating signatures
efficiently into clinically usable tests.
2 Considerations for Identifying Optimal Predictive Signatures
The application of machine learning methods to identify biological

signatures with potential clinical relevance is common in current
research and reaches across all areas of medicine [12]. There are a
large variety of different algorithms that are used for the identifica-
tion of such biomarker signatures. Although benchmarking studies
have compared the performance of these algorithms using diverse
datasets [13], the choice for or against a given algorithm is not
trivial as it frequently depends on the data under investigation
Statistical Considerations for Biomarker Translation 105
[14]. For high-dimensional data, assessment of the dependency

between data dimensionality and accuracy has shown that some
algorithms give higher performance than others [15]. Nevertheless,
for nearly all algorithms, there is an advantage in applying so-called
variable selection, the identification of an optimal subset of all mea-
sured variables [16]. The selection of a smaller number of variables
is frequently beneficial for performance and generalizability of
algorithms. This is due to the fact that algorithms are affected less
by sources of variability compared to those incorporating a large
number of predictors. Therefore, subject-specific fluctuations in
the measured analyte levels or random effects from variables that
do not show substantial association with the outcome are less likely
to impact on classification or prediction. On the other hand, for
accurate classification of complex illnesses, or other multifactorial
clinical outcomes, such as treatment response, it is beneficial to
include larger numbers of predictors that each contributes a small
piece of independent information. As a consequence, to find an
optimal balance between accuracy and generalizability, biomarker
panels identified in the literature frequently feature between 5 and
80 different predictors.
In these approaches, the defining criterion for selecting a given
set of predictors is classification accuracy. However, in high-
dimensional data it is frequently observed that there is not a single
optimal solution to the variable selection problem as different com-
binations of variables may lead to comparable performance esti-
mates. For this reason, there is an opportunity to utilize other
metrics during the variable selection process that capture important
properties of a given predictor set beyond the classification accu-
racy. One of these metrics is the stability of the variable selection
under sampling variability [17], an issue that is particularly promi-
nent for the usual scenario of variable numbers largely exceeding
the number of investigated subjects [18]. The general idea of stabil-
ity assessment is the repeated sampling of training sets from a given
dataset and to explore how consistently a given variable gets selected
as an important predictor. Numerous methods exist to quantify
variable selection stability and these are reviewed in [17] and [19].
Importantly, different variable selection methods differ regarding
the stability and accuracy of the selected variables [20]. Among the
reasons for this difference is their differential treatment of corre-
lated variables with some methods explicitly trying to avoid redun-
dancy. This may in turn lead to a loss of stability as, for example,
many genes may encode functionally related proteins [20].
Another potentially relevant property of predictor sets is cost-
effectiveness. In particular for multiplexing, addition of a given
analyte may be associated with a disproportionally higher cost
compared to other predictors. Therefore, it may be desirable to
include an estimation of cost-effectiveness that quantifies the bal-
ance between the cost of utilizing a given predictor and its impact
106 Emanuel Schwarz
on classification performance in the variable selection process [21].

Such considerations are also relevant for multimodal classifiers that
aim to extract predictive patterns across multiple technological
platforms, including proteomics, RNA sequencing, genome-wide
association data, or neuroimaging. For efficient clinical translation
of classifiers, an estimation of cost-effectiveness during variable
selection is almost mandatory. In addition to including such met-
rics during variable selection, other approaches have been devel-
oped that may show utility for such applications. Examples include
the integration of machine learning with constraint programming,
where optimal variable combinations are found within the limits of
prespecified constraints [22, 23]. Another interesting approach is
the utilization of more complex, multimodal data during training
of a classifier, but not during testing. “Learning with privileged
information” facilitates an adaptation of the parameters learned on
such less complex classifiers through the additional data available
during training [24–26]. This would allow the utilization of a
wider spectrum of data during classifier design, while profiting
from improved accuracy, higher cost-effectiveness, and simpler
logistics during classifier application. Another related approach is
that of multi-task learning, which is based on the assumption that
a single classifier cannot learn a given task well from a complex
dataset [27, 28]. Instead, multiple classifiers are learned on differ-
ent tasks that may or may not be related [29]. During this process,
information is exchanged between the classifiers. This leads to an
inductive adaptation of the original classifier based on the addi-
tionally learned classifiers and has previously been applied for mul-
timodal machine learning in the imaging genetics field [30]. An
interesting adaptation of this method called robust multi-task
learning explores the actual dependency between tasks and may
outperform standard multi-task learning [31, 32].
Another interesting consideration is cross-platform transfer-
ability. For clinical translation of biomarker tests, signatures that
have initially been identified by comprehensive high-throughput
platforms, such as DNA microarrays, need to be transferred to
more targeted platforms with lower variability like quantitative
real-time PCR. Another example where platform migration has
been performed successfully is the development of a dedicated
51-plex multiplexed immunoassay panel for aiding the diagnostic
process of schizophrenia [33]. In this context, metrics such as the
log-ratio discrepancy have been devised for prediction of how suc-
cessfully a given model can be migrated [34]. The log-ratio dis-
crepancy quantifies the absolute difference in log-ratio as measured
between platforms, averaged over all possible pairs of samples and
has a minimum value of zero when there is perfect agreement
between platforms. However, more research is needed to identify
algorithms that can accurately predict platform migration issues
using results from a given omic platform, prior to availability of
data on the same subjects from the migration target.
3 Measurement Error, Sample Size, and Generalizability
The main reasons for performing high-throughput experiments such

as those with multiplexed assays include the increase in efficiency, the
reduction in required sample volume, and, potentially, a reduction of
measurement variability. One aspect that is frequently overlooked is
that the joint measurement of multiple analytes gives rise to corre-
lated measurement errors, but the effect on subsequent inference of
outcome associations is typically not considered. Pollack et al per-
formed an interesting investigation of such effects through modeling
of correlated measurement error, correlations between predictors,
and measurement error as a function of predictor level under varying
levels of overall errors in measurement [35]. This study showed that
while bias may be substantial under different settings of correlation
and measurement error, outcome associations were mostly biased
toward the null. In addition, bias was nearly absent when the true
underlying association had an odds ratio of 1. Therefore, correlated
measurement error is not likely to lead to false-positive, but rather to
false-negative findings [35]. Another interesting conclusion was that
measurements that are subject to a lower limit of detection were less
affected by bias when values below these limits were replaced by sta-
tistically valid substitution values. The transformation of data values
regarding effects of measurement reliability and accuracy is an impor-
tant aspect that should be considered during classifier design.
The variability of measurements that includes both the biological
inter-subject variability as well as measurement-related variability also
has a crucial impact on the sample size required for particular classifi-
cation problems. In practice, machine learning approaches are most
frequently performed post-hoc on an already existing database. Here,
sample size is typically determined by practical considerations (i.e.,
the patient recruitment rate within a fixed time-window) or more
formal considerations, such as sample-size calculation for univariate
analysis. However, such considerations rarely include estimations of
required sample size for accurate prediction based on multivariate
methods. This is statistically an arguably more complex task com-
pared to univariate sample size estimation. Statistical learning theory
provides bounds on the generalization performance of a given classi-
fier that depends on the sample size and these bounds can in turn be
used to estimate sample sizes for a given target performance [36].
However, such bounds on classification performance are
known to be conservative and frequently higher generalizability is
observed than predicted by the model. As a consequence, the
required sample number would likely be lower to achieve a given
performance than predicted by the model. For example, the model
predicts an approximation toward the optimal performance pro-
portional to the inverse of the sample number. In practice, such
approximation can, however, be observed to be exponential with
the sample number, at least in some cases [37].
108 Emanuel Schwarz
The required sample number clearly depends on factors such as

the complexity of the classification task, the underlying variability in
the investigated cohort, and platform measurement specific effects
[38]. Therefore, strategies that simulate the required sample size
based on pilot data or that perform an adaptive sample size determi-
nation may be of higher practical utility compared to theoretical
bounds. For example, Shao and coworkers determined the mini-
mum required sample size for microarray prediction of outcomes
with various complexities using an adaptive procedure. A signal-to-
noise-ratio-based metric that is used to quantify the observed differ-
ence between classes was used as a criterion to stop or continue
sample collection to obtain an optimally sized training sample [39].
Similarly, Dobbin et al showed that the identification of good classi-
fiers from high-dimensional microarray data depends on standard-
ized fold change, class prevalence, and number of genes or features
on the arrays [40]. From this they proposed a method for ex post
facto assessment of whether the size of a training set used to develop
a classifier was adequate. Hwang et al used power analysis to esti-
mate the minimum sample size required to build a discriminant
analysis algorithm on transcriptomics data with statistical reliability
[41]. De Valpine applied Monte Carlo simulation to model required
sample sizes for classification and concluded that many existing stud-
ies will identify suboptimal classification performance, leading to
poor validation results, due to limitations in sample size [42]. More
recently, Beleites et al used learning curves to estimate the impact of
sample size on generalization performance. It was found that the
typical sample sizes of up to 25 subjects per class may lead to accept-
able classification performance but learning curves can easily be
masked by the large uncertainty due to the small sample numbers. In
fact, they estimated that 75-100 samples will be required to achieve
good classification performance and such samples can subsequently
be used for more refined sample size estimation [43]. The simula-
tion of suitable sample sizes should not only consider platform-spe-
cific methodological aspects or the nature of the classification task
but also the expected heterogeneity of the effect across the target
population. For example, a frequent observation is that biomarker
signatures vary substantially across clinical sites, even if strict stan-
dard operating procedures are followed for sample collection and
data acquisition. In particular for complex algorithms that integrate
a large number of predictors that show a heterogeneous response to
site-to-site variability, normalization of such effects may be difficult
and larger training samples will only be useful to the extent that they
reflect the heterogeneity in the target population. Variable selection
strategies that reduce the complexity of algorithms can help to
reduce this problem. In particular, they allow identification of vari-
able subsets that are reproducibly altered across clinical sites, a mea-
sure of variable stability in the context of population heterogeneity.
4 Longitudinal Stability of Classifiers
This section explores a variability-related problem that impacts on

classifier performance. High-throughput technologies for identifi-
cation of biomarker signatures have a non-hypothesis-driven nature
and are typically applied during an early stage of clinical tool devel-
opment. However, high-throughput profiling platforms are for
measurement variability, cost, and logistics reasons less suitable for
clinical translation of biomarker signatures. These issues may be
overcome through migration of the biomarker panel to dedicated
measurement platforms, such as multiplex systems targeting the
analytes used within the respective classifiers. This is also important
from a regulatory perspective, since clinical translation of a given
biomarker signature requires fixed algorithms that are specified on
one occasion with little possibility for recalibration. This requires
high measurement accuracy and substantial stability over extended
periods of time. While variability cannot be avoided for many rea-
sons, such as changes in reagent lots, little is known about its effects
on the performance of multivariate classifiers. For example, an
important question is whether or not the algorithm performance is
more sensitive to future measurement noise when they incorporate
a higher number of variables to achieve the same given perfor-
mance. Such error accumulation would have an important impact
already on classifier design and implementation of dedicated mea-
surement platforms, since performance optimization would not
only relate to current performance on a given dataset but also to
future performance under methodological noise.
To explore this, we performed a simulation for training classi-
fiers of different complexity and evaluated the impact of noise dur-
ing testing. The simulation was performed as follows:
1. A training dataset was created that contained 100 “patients”
and 100 “controls.”
2. For each group a number (n) of uncorrelated variables (with
n = 5 through n = 100) were drawn randomly from a normal
distribution with a standard deviation of 1.
3. The mean of the distribution was 0 for controls and dynami-
cally adjusted for patients.
4. The adjustment of the patient mean was performed such that
the classification accuracy in the training data was 80 %, aver-
aged over many repetitions of the procedure.
5. Classification was performed using linear discriminant analysis
and accuracy was assessed using cross-validation (the dynamic
adjustment of the patient mean, which corresponds in this case
to the effect size, is necessary since classifiers with higher num-
ber of variables can achieve the same classification accuracy at
lower effect sizes across its predictors.
110 Emanuel Schwarz
6. A test dataset was simulated using the same parameters as iden-

tified for the training data.
7. To each observation in the test data noise from a normal distri-
bution with a mean of 0 and variable standard deviation was
added.
8. The classifier trained on the training data was then applied to
the test data and classification performance assessed using the
area under the receiver operating characteristic curve (AUC).
Figure 1a shows that the decrease in classification perfor-
mance with higher levels of added noise does not depend on
the number of predictors in the classifier. In this example, noise
was additive, meaning that the introduced amount of noise did
not depend on the magnitude or concentration of the variable
it was added to. In practice, variability effects are frequently
multiplicative, affecting variables with higher magnitude more
compared to those at lower magnitudes. Another characteristic
that is often found in real-life datasets is that distributions of
variables are heavy-tailed and primarily observations within the
tail differentiate patients from control subjects. For this reason,
the same simulation described above was performed with the
following modifications.
9. During creation of the training data, the absolute value of the
normal random variables was used, sampling for both groups
from a distribution with a mean of 0, although in this case the
standard deviation was dynamically adjusted in patients, lead-
ing to increasing portions of the patient group that differed
from the controls.
a) b) c)
1.0
4
1.0
3
0.9
0.9
2
Variable 2
1
AUC
AUC
0.8
0.8
0
0.7
0.7
-1
Patients
-2
Controls
0.6
0.6
0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 -2 0 2 4
Noise Noise Variable 1
Fig. 1 Simulation of longitudinal measurement variability effects performance of classifiers with differing num-
bers of predictors. The figure shows the dependency of performance (AUC) on noise that was additive (panel
a; noise is taken from a normal distribution with mean 0 and standard deviation is plotted along the x axis) and
multiplicative (b; noise is taken from a normal distribution with mean 1 and standard deviation is plotted along
the x axis). (c) Example of a decision boundary for two-dimensional normal distributions associated with aver-
age cross-validation AUC of 0.80. In panels (a) and (b), darker gray colors refer to decision rules with higher
numbers of predictors
10. Noise was sampled from a normal distribution with a mean of

1 and varying standard deviation and the absolute values of
such noise were multiplied with the simulated test data.
Figure 1b shows that the decrease in classification perfor-
mance with increasing amounts of noise did not depend on the
numbers of predictors within the classifier, as described under
the original conditions. These results suggest that noise
encountered after classifier design will not affect complex clas-
sifiers more than those with fewer predictors. In practice, the
situation may be different due to several factors. First, more
complex rules frequently integrate a large number of predic-
tors that are substantially more difficult to measure and, there-
fore, affected by disproportionally higher amounts of
longitudinal measurement variability. The expected measure-
ment variability and variance of a given analyte in the popula-
tion may already be a useful metric during variable selection for
designing the classifier. Secondly, the example presented here
is based on simulation of simple, linear decision boundaries. A
two-dimensional example of this is shown in Fig. 1c. Here, the
effect size in each variable was adapted such that linear dis-
criminant analysis yielded an average AUC of 0.80. With non-
linear decision boundaries, effects of noise may be stronger, in
particular when more advanced algorithms, such as random
Forests or neural networks, need to be used. For this reason, it
may be advantageous to perform classification with simpler
algorithms, such as linear support vector machines, even if vari-
ables are selected using more complex methods, such as ran-
dom Forest variable importance measures. Such hybrid
machine learning approaches aim to combine advanced vari-
able selection with improved generalizability and may perform
better than either of the methods individually [44].
5 Conclusions
The application of machine learning algorithms for identification

of biological signatures with clinical utility is widespread and spans
across most high-throughput profiling technologies. Freely avail-
able software packages facilitate intuitive use of such methods mak-
ing them accessible to a broad community of scientists without the
requirement for in-depth knowledge of the algorithms’ inner
workings. However, the increasing popularity of these approaches
and the potential clinical relevance of biomarker signatures also
contribute to the problem that machine learning is applied on
studies that were not designed for such analyses, which are fre-
quently too small and lead to irreproducible findings. Here some
considerations were discussed that may have utility for classifier
112 Emanuel Schwarz
development throughout the different phases of biomarker discov-

ery and clinical translation. These comprise approaches to estimate
sample sizes appropriate for multivariate classification and potential
bounds on test performance. They also include properties of pre-
dictors that go beyond their impact on classification accuracy such
as stability, cost-effectiveness, or the possibility for platform migra-
tion. While these properties are frequently overlooked in favor of
optimizing classification accuracy, they play a crucial role when
translating a biomarker signature into an assay system for clinical
use. Finally, a simulation was performed to estimate the effect of
longitudinal measurement variation on a defined classifier. These
results suggest that there is no accumulation of noise that would
affect algorithms with higher numbers of predictors more than
those with fewer. This finding is generally positive for biomarker
discovery in complex illnesses, where large numbers of individually
weak predictors need to be combined to achieve clinically relevant
predictive performance. In practice, it will still be desirable to
reduce the complexity of decision rules as far as possible and the
considerations discussed here may aid in driving this selection pro-
cess toward an optimal balance of performance and clinical
translatability.
Acknowledgments
This study was supported by the DFG Emmy-Noether-Program

SCHW 1768/1-1.
References
1. Cohen Freue GV, Meredith A, Smith D, 5. Jin G, Zhou X, Wang H, Wong STC (2010)
Bergman A, Sasaki M, Lam KK et al (2013) The challenges in blood proteomic biomarker
Computational biomarker pipeline from dis- discovery. In: Pham T (ed) Comput Biol.
covery to clinical implementation: plasma pro- Springer, New York, pp 273–299
teomic biomarkers for cardiac transplantation. 6. Rifai N, Gillette MA, Carr SA (2006) Protein
PLoS Comput Biol 9:e1002963 biomarker discovery and validation: the long
2. Zhang Z, Chan DW (2010) The road from and uncertain path to clinical utility. Nat
discovery to clinical diagnostics: lessons learned Biotechnol 24:971–983
from the first FDA-cleared in vitro diagnostic 7. Füzéry AK, Levin J, Chan MM, Chan DW
multivariate index assay of proteomic biomark- (2013) Translation of proteomic biomarkers
ers. Cancer Epidemiol Biomarkers Prev into FDA approved cancer diagnostics: issues
19:2995–2999 and challenges. Clin Proteomics 10:13
3. Alymani NA, Smith MD, Williams DJ, Petty 8. Goodsaid F, Mattes WB (2013) Thepath from
RD (2010) Predictive biomarkers for person- biomarker discovery to regulatory qualifica-
alised anti-cancer drug use: discovery to clinical tion. 1 edn., Academic Press. Accessed 16 July
implementation. Eur J Cancer 46:869–879 2013. ISBN: 0123914965
4. Deyati A, Younesi E, Hofmann-Apitius M, 9. Kotsiantis SB (2007) Supervisedmachine
Novac N (2013) Challenges and opportunities learning: a review of classification techniques.
for oncology biomarker discovery. Drug Informatica 31:249–268. doi:10.1115/
Discov Today 18:614–624 1.1559160
10. Hastie T, Tibshirani R, Friedman J (2009) The 24. Lapin M, Hein M, Schiele B (2014) Learning
elements of statistical learning. Elements using privileged information: SV M+ and
1:337–387. doi:10.1007/b94608 weighted SVM. Neural Netw 53:95–108
11. Saeys Y, Inza I, Larrañaga P (2007) A review of 25. Pechyony D, Vapnik V (2010) On the theory
feature selection techniques in bioinformatics. of learning with privileged information. Nips
Bioinformatics 23:2507–2517 pp 1894–1902
12. Kononenko I (2015) Machine learning for 26. Vapnik V, Vashist A (2009) A new learning
medical diagnosis: history, state of the art and paradigm: learning using privileged informa-
perspective. Artif Intell Med 23:89–109. tion. Neural Netw 22:544–557
doi:10.1016/S0933-3657(01)00077-X 27. Chapelle O, Shivaswamy P, Vadrevu S,
13. Caruana R, Niculescu-Mizil A (2006) An Weinberger K, Zhang Y, Tseng B (2011) Boosted
empirical comparison of supervised learning multi-task learning. Mach Learn 85:149–173
algorithms. Proceedings of 23rd international 28. Evgeniou T, Pontil M (2004) Regularized
conference machine learning. pp 161–168. multi--task learning. Proceedings of 10th ACM
doi: 10.1145/1143844.1143865 SIGKDD pp 109–117
14. Guo Y, Graber A, McBurney RN, 29. Romera-Paredes B, Argyriou A, Pontil M,
Balasubramanian R (2010) Sample size and Berthouze N (2012) Exploiting unrelated
statistical power considerations in high- tasks in multi-task learning. Proceedings of
dimensionality data settings: a comparative 15th international conference of artificial intel-
study of classification algorithms. BMC ligence statistics, vol 22, pp 951–959
Bioinformatics 11:447 30. Wang H, Nie F, Huang H, Risacher SL, Saykin
15. Caruana R, Karampatziakis N, Yessenalina A AJ, Shen L et al (2012) Identifying disease sen-
(2008) An empirical evaluation of supervised sitive and quantitative trait-relevant biomarkers
learning in high dimensions. Proceedings of from multidimensional heterogeneous imaging
25th international conference machine learning. genetics data via sparse multimodal multitask
pp 96–103. doi: 10.1145/1390156.1390169 learning. Bioinformatics 28:i127–i136
16. Guyon I, Elisseeff A (2003) An introduction to 31. Gong P, Ye J, Zhang C (2012) Robust multi-
variable and feature selection. J Mach Learn task feature learning. KDD 2012:895–903
Res 3:1157–1182 32. Ishibuchi H, Nojima Y (2007) Analysis of inter-
17. He Z, Yu W (2010) Stable feature selection for pretability-accuracy tradeoff of fuzzy systems by
biomarker discovery. Comput Biol Chem multiobjective fuzzy genetics-based machine
34:215–225 learning. Int J Approx Reason 44:4–31
18. Loscalzo S, Yu L, Ding C (2009) Consensus 33. Schwarz E, Izmailov R, Spain M, Barnes A,
group stable feature selection. Proceedings of Mapes JP, Guest PC et al (2010) Validation of
15th ACM SIGKDD international conference a blood-based laboratory test to aid in the con-
on knowledge discovery and data mining. firmation of a diagnosis of schizophrenia.
pp 567–576 Biomark Insights 5:39–47
19. Awada W, Dittman D, Wald R, Napolitano 34. Gyorffy B, Molnar B, Lage H, Szallasi Z,
A, Khoshgoftaar TM(2012) A review of the Eklund AC (2009) Evaluation of microarray
stability of feature selection techniques for preprocessing algorithms based on concor-
bioinformatics data. In: Proceedings of 2012 dance with RT-PCR in clinical samples. PLoS
IEEE 13th international conference informa- One 4(5):e5645
tion reuse and integration IRI 2012. 35. Pollack AZ, Perkins NJ, Mumford SL, Ye A,
pp 356–363 Schisterman EF (2013) Correlated biomarker
20. Haury AC, Gestraud P, Vert JP (2011) The measurement error: an important threat to
influence of feature selection methods on accu- inference in environmental epidemiology. Am
racy, stability and interpretability of molecular J Epidemiol 177:84–92
signatures. PLoS One 6(12), e28210 36. Shawe-Taylor J, Anthony M, Biggs NL (1993)
21. Braun DC, Reynolds JD (2012) Cost-effective Bounding sample size with the Vapnik-
variable selection in habitat surveys. Methods Chervonenkis dimension. Discret Appl Math
Ecol Evol 3:388–396 42:65–73
22. Guns T, Nijssen S, De Raedt L (2011) Itemset 37. Cohn D, Tesauro G (1991) Howtight are the
mining: a constraint programming perspective. Vapnik-Chervonenkisbounds? Neural Comput
Artif Intell 175:1951–1983 4:249–269
23. Talbi EG (2013) Combining metaheuristics 38. Dobbin K, Simon R (2005) Sample size deter-
with mathematical programming, constraint mination in microarray experiments for class
programming and machine learning. 4OR Q comparison and prognostic classification.
J Oper Res 11:101–150 Biostatistics 6:27–38
114 Emanuel Schwarz
39. Shao L, Fan X, Cheng N, Wu L, Cheng Y 42. De Valpine P, Bitter HM, Brown MPS, Heller
(2013) Determination of minimum training J (2009) A simulation-approximation
sample size for microarray-based cancer out- approach to sample size planning for high-
come prediction-an empirical assessment. dimensional classification studies. Biostatistics
PLoS One 8:e68579 10:424–435
40. Dobbin KK, Zhao Y, Simon RM (2008) How 43. Beleites C, Neugebauer U, Bocklitz T et al
large a training set is needed to develop a clas- (2013) Sample size planning for classification
sifier for microarray data? Clin Cancer Res models. Anal Chim Acta 760:25–33
14:108–114 44. Menze BH, Kelm BM, Masuch R, Himmelreich
41. Hwang D, Schmitt WA, Stephanopoulos G, U, Bachert P, Petrich W et al (2009) A compari-
Stephanopoulos G (2002) Determination of son of random forest and its Gini importance
minimum sample size and discriminatory with standard chemometric methods for the fea-
expression patterns in microarray data. ture selection and classification of spectral data.
Bioinformatics 18:1184–1193 BMC Bioinformatics 10:213
Chapter 7
Opportunities and Challenges of Multiplex Assays:

A Machine Learning Perspective
Junfang Chen and Emanuel Schwarz
Abstract
Multiplex assays that allow the simultaneous measurement of multiple analytes in small sample quantities
have developed into a widely used technology. Their implementation spans across multiple assay systems
and can provide readouts of similar quality as the respective single-plex measures, albeit at far higher
throughput. Multiplex assay systems are therefore an important element for biomarker discovery and
development strategies but analysis of the derived data can face substantial challenges that may limit the
possibility of identifying meaningful biological markers. This chapter gives an overview of opportunities
and challenges of multiplexed biomarker analysis, in particular from the perspective of machine learning
aimed at identification of predictive biological signatures.
Key words Biomarker discovery, Machine learning, Confounding, Bias, Multiplex
1 Introduction: Opportunities of Multiplexed Assay Systems
Due to substantial advancements of high-throughput omics

technologies, the acquisition of high-dimensional biological
datasets has become routine practice. Genetic association,
expression, or methylation testing can be performed at the level
of the genome- and proteomics often measures in excess of 1000
variables in a given sample [1–3]. The availability of such high-
dimensional datasets affords tremendous possibilities of per-
forming a non-hypothesis-driven search for biological patters
that predict a given clinical outcome. This strategy particularly
applies to complex clinical outcomes in which individual predic-
tors may have low effect sizes and a combination of numerous
predictors is needed to achieve clinically useful accuracy.
However, high-throughput measurement of molecular concen-
trations using proteomic or transcriptomic techniques can be
affected by substantial measurement variability and therefore
such data has its greatest use during the initial stages of the bio-
marker development process. As the development of accurate
115
116 Junfang Chen and Emanuel Schwarz
classifiers is dependent on the longitudinal stability of measured

concentrations, a central aim is the reduction of measurement
noise. This is an important requirement for validation of biologi-
cal patterns as otherwise a subtle but predictive biological signa-
ture may be drowned out by experimental variability. In this
scenario, multiplex assays have substantial utility as they can
facilitate the transition from a global biomarker screening tool to
a dedicated measurement platform with low variability.
Some previous studies have found the analytical performance
of multiplex assays to be comparable to that of single-plex assays
[4, 5]. However, such comparability of analytical performance
may depend on the measured analytes and the experimental pro-
cedures [6]. In addition to reducing measurement noise, multi-
plexed measurement of biological analytes may be of help to
improve throughput. For example, some liquid chromatogra-
phy-based mass spectrometry methods may not be able to gener-
ate datasets of a size sufficient for the identification of accurate
biological signatures. Meaningful application of machine learn-
ing tools to high-dimensional biological datasets requires large
sample numbers, in particular when effects of individual predic-
tors are small. For example, Ein-Dor et al. have shown that
thousands of samples are needed for robust prediction of out-
come in cancer [7]. Similarly, Kim and coworkers showed that
independently generated gene signatures predictive of breast
cancer using 600 samples per experiment show an overlap of
only 16.5 % [8]. While such sample sizes may already exceed the
technological possibilities of some omics assay systems, the
required sample size is far higher in practice. This is due to the
fact that for development of algorithms that generalize well to
new samples, the data used for algorithm training needs to reflect
the properties of the target population. In practice, this means
that the spectrum of physiological inter-individual variability
needs to be reflected in both the data and the differences between
factors such as ethnicities or clinical sites, an effect that can be
substantial even if standard operating procedures are in place
[9]. In addition, diagnostic classifiers are rarely useful when
purely focusing on case-control differentiation and therefore,
training data should incorporate measurements from relevant
differential diagnoses. Under these circumstances, multiplex
immunoassays offer the possibility of translating promising
omics results into an assay system with which high-quality, large-
scale data can be acquired for more extensive assessment of pre-
dictive algorithm performance. Such assessment can help to
overcome small sample effects in training data that are known to
frequently lead to false-positive findings [10].
Opportunities and Challenges of Multiplex Assays 117
2 Recent Examples of Multiplex Assays for Disease Diagnosis
An example of multiplexed assay-based biomarker development is

the development of a test for schizophrenia [11]. Here, a multiplex
panel of 51 analytes was created that combined promising features
from proteomics and immunoassay measurements into a single
measurement system. This system was then used to acquire protein
concentration data on over 800 subjects for algorithm training.
Similarly, Surinova et al used a phased mass-spectrometry approach
to identify a biomarker signature for colorectal cancer. Candidates
derived from an initial profiling of about 300 secreted and cell sur-
face candidate glycoproteins were translated into an 88-plex tar-
geted SRM assay to measure the concentrations of these in large
populations [12]. This was ultimately transferred into a five-protein
signature with high accuracy.
One of the primary strengths of multiplexed assay systems is
the possibility to customize the set of measurement targets to suit
the investigated biological system ideally. While the simultane-
ously measured analytes may cover the most predictive set as
described above, it may also try to capture important elements of
a given biological pathway or known molecular targets of a given
illness. For example, mass spectrometry-based multiplex selected
reaction monitoring (SRM) has been used to identify dysregula-
tion of glycolysis-associated enzymes in schizophrenia [13].
Hembrough and colleagues reported a significant improvement
in tumor tissue analysis due to the use of multiplexed SRM that
enables the accurate measurement of the expression levels of a
panel of oncological protein targets [14]. Similarly, Xie et al
designed a novel multiplex assay that quantifies autoantibodies
against a large collection of clinically vital tumor-associated anti-
gens, combined with a classical cancer biomarker [15]. More
recently, Arjomandi et al proposed a novel algorithm to improve
the performance of an ovarian cancer detection test using a mul-
tiplex approach [16]. In this study, measurements of autoanti-
bodies to p53 in sera of patients against selected confirmatory
epitopes were acquired using a multiplexed-based immunoassay.
Such a focus on a selected set of target analytes facilitates biologi-
cally meaningful stratification of biological measures for multi-
variate or functional downstream analysis. One example of this is
biologically stratified association analyses (also known as “path-
way-wide association study,” PWAS), which can aid in differenti-
ating illness relevant variants from background [17, 18]. A related
analysis focusing on genetic co-expression networks (“integrative
network-based association study,” INAS) has been used to iden-
tify illness-associated genes from tissue-to-tissue co-expression
networks [17, 19]. Similarly, meaningful biological stratification
of measured analytes may help machine learning approaches to
identify biological signatures and remove noise from measure-

ments that are not likely to be illness related. This particularly
applies for interactions between measured analytes, which are dif-
ficult to extract from noisy, high-dimensional datasets. The inte-
gration of interactions in machine learning models such as
support vector machines or random forests can aid to improve
performance [20–22].
These studies show that the application of multiplexed immu-
noassays for biomarker development goes beyond large-scale vali-
dation of candidate markers identified through omics techniques.
In fact, Stahl-Zeng and coworkers have shown that multiplexed
selected reaction monitoring of N-glycosites can be used to detect
plasma proteins at concentrations in the ng/mL or sub-ng/mL
range and quantify their concentrations accurately over five orders
of magnitude [23]. Methodological innovations, such as data
acquisition paradigms to simultaneously quantify and confirm the
identity of the targeted peptides can further expand the multiplex
capacity of SRM approaches [24].
3 Challenges: Biological and Technical Confounders
Multiplexed measurement of biological analytes gives rise to sev-

eral challenges some of which may affect particularly multivariate
downstream analysis. One of the most prominent factors is cross-
reactivity and analytes that interact with other antibodies or irrel-
evant interfering factors that may have to be excluded from analysis
[25]. For example, serum rheumatoid factor and other heterophilic
antibodies can cause significant interference with antibody-based
multiplex immunoassays through binding to detection reagents
[26, 27]. Similar cross-reactivity effects are known for non-
immunoassay-based measurement platforms, such as SRM. In the
case of the latter, ion suppression can impact on the formation of
the analyte ions because of sample-matrix interactions [28, 29].
Another potential challenge affecting multiplex assays is that
assay performance may vary between different laboratories or plat-
forms and may therefore impede multi-site biomarker develop-
ment efforts. Zhang et al. investigated the impact of variability
between three different laboratories and multiplex systems on the
clinical interpretation of pneumococcal serology assay measure-
ments in 57 antibody-deficient patients [30]. Despite substantial
variability of quantitative antibody levels, a decision concordance
of up to 80 % was determined across laboratories. This suggests
that the integration of multiplex data using multivariate algorithms
may overcome the problems of variable individual predictors.
However, this study also highlights the need for novel paradigms
to control multiplex data quality across reference laboratories to
help reduce experimental variation [30, 31].
While technical variation is typically lower than biological dif-

ferences, some sources of such variability do not only hinder iden-
tification of predictive signatures but they can lead to discovery of
artefactual illness fingerprints [32, 33]. Frequently, multivariate
and machine learning analyses are performed using multiplex data
that were not acquired for such analyses and therefore study design
may not be suitable to exclude some confounding factors. This
particularly relates to batch effects that are virtually unavoidable
for high-throughput measurements [34] and which can be sub-
stantially amplified through aggregation of their effects across
measured analytes. Such effects include changes in personnel, dif-
ferent laboratories, different run dates, and inconsistent experi-
mental design between assays or laboratories. If such effects are
associated with the outcome of interest they may lead to substan-
tial bias and application of statistical remedies may not be sufficient
to remove their effects on the data [34].
Clarke and colleagues systematically investigated technical con-
founding factors in multiplex immunoassay data [33]. Mixed-effects
modeling was used to eliminate the effects of both the technical and
biological sources of variation followed by the analytical evaluation
of normalized multiplex data. Similarly, Browne et al. applied a sta-
tistical batch correction method to reduce variance between mea-
surement plates for multiplex cytokine assays in serum and saliva
[35]. While such methods reduce the impact of batch effects on
downstream analysis, residual variation in strongly confounded data
may still impact particularly on sensitive machine learning algo-
rithms. Soneson et al. have used different machine learning algo-
rithms to explore the influence of batch effect severity on classification
performance [36]. When high levels of the confounding factor
existed, the performance estimates obtained from cross-validation
were found to be highly biased. Parker and coworkers developed a
novel method of batch effect removal for multivariate prediction
that they called frozen surrogate variable analysis [37]. This approach
uses a training dataset to correct batch effects at the individual sub-
ject level. As batch effects may remain undetected especially in
smaller cohorts, multiplex assays and careful study design may help
to identify less biased biological signatures that show better general-
izability to new samples. Also, some multiplex systems have been
optimized for improved longitudinal measurement stability, thereby
reducing problems caused by batch-to-batch assay variation [38].
Finally, an interesting challenge of multiplexed assays is that a
given factor of variability might simultaneously affect the different
measured analytes and therefore lead to correlated measurement
error. Pollack et al. explored the impact of such effects on
downstream inference on outcome associations and found that
outcome associations were mostly biased toward the null [39].
This effect is therefore more likely to lead to false-negative rather
than false-positive outcome associations.
4 Conclusions
The combination of experimental and computational methods is

indispensable for the identification of novel biomarkers and the
exploration of biological mechanism of complex illnesses. Due to
their high throughput, typically low technical variation, and the
possibility to efficiently acquire data on biological analytes that are
difficult to measure otherwise, multiplex assays can be an impor-
tant element of biomarker development strategies [40]. At the
same time, their application faces several challenges that may par-
ticularly impact on sensitive downstream machine learning analy-
sis. These may reduce the ability to identify biological illness
signatures or lead to artefactual findings when sources of variation
are associated with the clinical outcome of interest. While statistical
methods exist to remove such effects in multiplex data, residual
confounding factors may remain and require the generation of
independent data for validation of biomarker signatures. Finally,
sharing of data, protocols, and experimental designs as well as
strategies to ensure consistent data quality across laboratories
would aid in validation efforts of promising biomarker signatures.
Acknowledgments
This study was supported by the DFG Emmy-Noether-Program

SCHW 1768/1-1.
References
1. Gutstein HB, Morris JS, Annangudi SP, 6. Tighe PJ, Ryder RR, Todd I, Fairclough LC
Sweedler JV (2008) Microproteomics: analysis (2015) ELISA in the multiplex era: potentials
of protein diversity in small samples. Mass and pitfalls. Proteomics Clin Appl 9:406–422
Spectrom Rev 27:316–330 7. Ein-Dor L, Zuk O, Domany E (2006) Thousands
2. Bibikova M, Barnes B, Tsan C, Ho V, Klotzle of samples are needed to generate a robust gene
B, Le JM et al (2011) High density DNA list for predicting outcome in cancer. Proc Natl
methylation array with single CpG site resolu- Acad Sci U S A 103:5923–5928
tion. Genomics 98:288–295 8. Kim SY (2009) Effects of sample size on
3. van’t Veer LJ, Dai H, van de Vijver MJ, He robustness and prediction accuracy of a prog-
YD, Hart AA, Mao M et al (2002) Gene nostic gene signature. BMC Bioinformatics
expression profiling predicts clinical outcome 10:147
of breast cancer. Nature 415:530–536 9. Ellington AA, Kullo IJ, Bailey KR, Klee GG
4. Elshal MF, McCoy JP (2006) Multiplex bead (2009) Measurement and quality control issues
array assays: performance evaluation and com- in multiplex protein assays: a case study. Clin
parison of sensitivity to ELISA. Methods Chem 55:1092–1099
38:317–323 10. Mollenhauer B Parnetti L, Rektorova I,
5. Dupont NC, Wang K, Wadhwa PD, Culhane Kramberger MG, Pikkarainen M, Schulz-
JF, Nelson EL (2005) Validation and compari- Schaeffer WJ et al (2015) Biological confounders
son of luminex multiplex cytokine analysis kits for the values of cerebrospinal fluid proteins in
with ELISA: determinations of a panel of nine Parkinson's disease and related disorders.
cytokines in clinical sample culture superna- J Neurochem. doi: 10.1111/jnc.13390. [Epub
tants. J Reprod Immunol 66:175–191 ahead of print]
11. Schwarz E, Izmailov R, Spain M, Barnes A, tivity detection of plasma proteins by multiple
Mapes JP, Guest PC et al (2010) Validation of reaction monitoring of N-glycosites. Mol Cell
a blood-based laboratory test to aid in the con- Proteomics 6:1809–1817
firmation of a diagnosis of schizophrenia. 24. Kiyonami R, Schoen A, Prakash A, Zabrouskov
Biomark Insights 12:39–47 V, Peterman S, Picotti P et al (2011) Increased
12. Surinova S, Choi M, Tao S, Schüffler PJ, Chang selectivity, analytical precision, and throughput
CY, Clough T et al (2015) Prediction of colorec- in targeted proteomics. Mol Cell Proteomics
tal cancer diagnosis based on circulating plasma 10:M110–M002931. doi:10.1074/mcp.
proteins. EMBO Mol Med 7:1166–1178 M110.002931.EpubJul27
13. Martins-de-Souza D, Alsaif M, Ernst A, Harris 25. Skogstrand K, Thorsen P, Nørgaard-Pedersen
LW, Aerts N, Lenaerts I et al (2012) The appli- B, Schendel DE, Sørensen LC, Hougaard DM
cation of selective reaction monitoring con- (2005) Simultaneous measurement of 25
firms dysregulation of glycolysis in a preclinical inflammatory markers and neurotrophins in
model of schizophrenia. BMC Res Notes neonatal dried blood spots by immunoassay
5:146 with xMAP technology. Clin Chem
14. Hembrough T, Thyparambil S, Liao WL, 51:1854–1866
Darfler MM, Abdo J, Bengali KM et al (2013) 26. Todd DJ, Knowlton N, Amato M, Frank MB,
Application of selected reaction monitoring for Schur PH, Izmailova ES et al (2011) Erroneous
multiplex quantification of clinically validated augmentation of multiplex assay measurements
biomarkers in formalin-fixed, paraffin-embedded in patients with rheumatoid arthritis due to
tumor tissue. J Mol Diagn 15:454–465 heterophilic binding by serum rheumatoid fac-
15. Xie C, Kim HJ, Haw JG, Kalbasi A, Gardner tor. Arthritis Rheum 63:894–903
BK, Li G et al (2011) A novel multiplex assay 27. Churchman SM, Geiler J, Parmar R, Horner
combining autoantibodies plus PSA has EA, Church LD, Emery P et al (2012)
potential implications for classification of Multiplexing immunoassays for cytokine detec-
prostate cancer from non-malignant cases. tion in the serum of patients with rheumatoid
J Transl Med 9:43 arthritis: lack of sensitivity and interference by
16. Arjomandi A, Delanoy ML, Walker RP, Binder rheumatoid factor. Clin Exp Rheumatol
SR (2015) A novel algorithm to improve speci- 30:534–542
ficity in ovarian cancer detection. Clin Ovarian 28. Annesley TM (2003) Ion suppression in mass
Other Gynecol Cancer DOI: http://dx.doi. spectrometry. Clin Chem 49:1041–1044
org/10.1016/j.cogc.2015.05.002 29. Grebe SK, Singh RJ (2011) LC-MS/MS in the
17. Califano A, Butte AJ, Friend S, Ideker T, clinical laboratory-where to from here. Clin
Schadt E (2012) Leveraging models of cell Biochem Rev 32:5–31
regulation and GWAS data in integrative 30. Zhang X, Simmerman K, Yen-Lieberman B,
network-based association studies. Nat Genet Daly TM (2013) Impact of analytical variability
44:841–847 on clinical interpretation of multiplex pneumo-
18. Wang K, Li M, Hakonarson H (2010) Analysing coccal serology assays. Clin Vaccine Immunol
biological pathways in genome-wide association 20:957–961
studies. Nat Rev Genet 11:843–854 31. Hill HR, Pickering JW (2013) Reference labo-
19. Dobrin R, Zhu J, Molony C, Argman C, ratory agreement on multianalyte pneumococ-
Parrish ML, Carlson S et al (2009) Multi-tissue cal antibody results: an absolute must! Clin
coexpression networks reveal unexpected sub- Vaccine Immunol 20:955–956
networks associated with disease. Genome Biol 32. Blankley RT, Fisher C, Westwood M, North
10:R55 R, Baker PN, Walker MJ et al (2013) A label-
20. Zhao Z, Liu H (2007) Searching for interacting free selected reaction monitoring workflow
features. Published in: IJCAI’07 Proceedings of identifies a subset of pregnancy specific glyco-
the 20th international joint conference on artif- proteins as potential predictive markers of
ical intelligence. pp 1156–1161 early-onset pre-eclampsia. Mol Cell
21. Cortes C, Vapnik V (1995) Support-vector Proteomics 12:3148–3159
networks. Mach Learn 20:273–297 33. Clarke DC, Morris MK, Lauffenburger DA
22. Ho TK (1995) Random decision forests. In: (2010) Normalization and statistical analysis of
Document analysis and recognition, proceed- multiplexed bead-based immunoassay data
ings of the third international conference on using mixed-effects modeling. Mol Cell
IEEE, vol 1, pp 278–282 Proteomics 12:245–262
23. Stahl-Zeng J, Lange V, Ossola R, Eckhardt K, 34. Leek JT, Scharpf RB, Bravo HC, Simcha D,
Krek W, Aebersold R et al (2007) High sensi- Langmead B, Johnson WE et al (2010)
Tackling the widespread and critical impact of 38. Cham GK, Kurtis J, Lusingu J, Theander TG,
batch effects in high-throughput data. Nat Rev Jensen AT, Turner L (2008) A semi-automated
Genet 11:733–739 multiplex high-throughput assay for measuring
35. Browne RW, Kantarci A, LaMonte MJ, IgG antibodies against Plasmodium falciparum
Andrews CA, Hovey KM, Falkner KL et al erythrocyte membrane protein 1 (PfEMP1)
(2013) Performance of multiplex cytokine domains in small volumes of plasma. Malar
assays in serum and saliva among community- J 7:108
dwelling postmenopausal women. PLoS One 39. Pollack AZ, Perkins NJ, Mumford SL, Ye A,
8, e59498 Schisterman EF (2013) Correlated biomarker
36. Soneson C, Gerster S, Delorenzi M (2014) measurement error: an important threat to
Batch effect confounding leads to strong bias inference in environmental epidemiology. Am
in performance estimates obtained by cross- J Epidemiol 177:84–92
validation. PLoS One 9, e100335 40. Mani A, Ravindran R, Mannepalli S, Vang D,
37. Parker HS, Corrada Bravo H, Leek JT (2014) Luciw PA, Hogarth M et al (2015) Data min-
Removing batch effects for prediction probing strategies to improve multiplex microbead
lems with frozen surrogate variable analysis. immunoassay tolerance in a mouse model of
Peer J 2, e561 infectious diseases. PLoS One 10:e0116262
Part III
Protocols
Chapter 8
Multiplex Analyses Using Real-Time Quantitative PCR

Abstract
Quantitative polymerase chain reaction (qPCR) is a routinely used method for the detection and quantita-
tion of gene expression in real time. Multiplex qPCR requires the use of probe-based assays, in which each
probe is labeled with a unique fluorescent dye, resulting in different observed colors for each assay. The
signal from each dye is used to quantitate the amount of each target separately in the same tube or well.
The availability to multiplex therefore allows the measurement of the expression levels of several targets or
genes of interest quickly. Here, we describe a method using the SensiFAST and SensiFAST One-Step
probe kits which allows simultaneous real-time quantitation of up to 5 amplicons.
Key words qPCR, Fluorescent dyes, Taq polymerase, Quantitation, mRNA, cDNA, Amplicon,
Multiplex analysis
1 Introduction
Polymerase chain reaction (PCR) method was a revolutionary

innovation by Kary Mullis in the 1980s [1, 2]. Since this time, it
has seen widespread use in biomedical research since it can detect
and quantify small amounts of specific nucleic acid sequences. For
example, small levels of messenger RNA (mRNA) can be quanti-
fied through the combination of reverse transcription (RT) to yield
complementary DNA (cDNA) and PCR amplification to produce
exponentially higher levels of these cDNA strands [3] (Fig. 1). In
addition to increased levels of the amplified products (amplicons),
the reliability and reproducibility of measurements between differ-
ent laboratories are essential, especially if the method is to be per-
formed in a clinical setting. This is critical for patient outcomes as
well as for reducing healthcare costs since approximately one third
of medical care budgets result from measurements and tests associ-
ated with diagnosis [4].
Quantitative PCR (qPCR) is a later development of the method
that allows users to monitor the progress of a PCR reaction in real
time [5]. In brief, the method uses a DNA-based sequence-specific
125
126
Primer
5’ 3’ 5’ 3’ 5’ 3’
+
Repeat
5’ cDNA 3’
Denaturation Annealing Elongation
20-40 cycles
3’ 5’
Repeat
+
3’ 5’ 3’ 5’ 3’ 5’
Nucleotides
Fig. 1 Schematic diagram of PCR

Quantitative PCR 127
probe with a fluorescent reporter molecule at one end and a mol-

ecule that quenches this fluorescence at the other. The proximity
of the reporter to the quench molecule prevents the detection of
fluorescence and cleavage of the probe by the 5′ to 3′ exonuclease
activity of Taq polymerase results in unquenched emission of fluo-
rescence. Thus, the increase in the cDNA amplicon targeted by the
reporter probe during each PCR cycle leads to a proportional
increase in fluorescence due to cleavage of the probe and release of
the reporter (Fig. 2). The available fluorescent reporter molecules
include dyes that bind to double-stranded DNA such as SYBR®
Green (Thermo Fisher Scientific; Waltham, MA, USA) or sequence
specific probes like Molecular Beacons (Newark, NJ, USA),
Scorpions (DxS Ltd), or TaqMan® Probes (Roche Molecular
Diagnostics; Basel, Switzerland). As with standard PCR, qPCR is
normally performed using a thermal cycler, which can rapidly heat
and cool samples to allow the melting, annealing, and extension
phases of replication. However in the case of qPCR, the thermocy-
cler should also have the ability to illuminate each sample with
specific wavelengths of light for the detection of the fluorescence
emitted following excitation of the probe.
PCR normally consists of a series of temperature changes that
are repeated approximately 30 times. Each cycle consists of two or
three steps. In the three step cycling approach, the first step is car-
ried out at approximately 95 °C, which allows separation of the
double-stranded nucleic acid chains (denaturation). The second
phase is performed at around 55 °C to allow binding of the prim-
ers to the DNA/cDNA template (annealing). Finally, the third
step is carried out at 72 °C to facilitate polymerization using DNA
polymerase (elongation). In the two step cycling method, the
Reporter Quencher
5’ 3’
5’ 3’
Probe Primer Nucleotides
cDNA
3’ 5’ Denaturation
3’ 5’
Taq polymerase cleaves

reporter yielding unquenched Annealing
fluorescent signal
5’ 3’ 5’ 3’
Repeat
~2 x 1020-40 Elongation
reporter 20-40 cycles
molecules
Repeat
3’ 5’ 3’ 5’
Fig. 2 Schematic diagram of real-time qPCR procedure

128 Steve F.C. Hawkins and Paul C. Guest
annealing and elongation steps are combined at the temperature

annealing temperature. In qPCR, it should be noted that 40 cycles
are performed and that the temperatures and associated times used
in each cycle depend on a variety of factors, such as the polymerase
used, the concentration of deoxyribonucleotides (dNTPs), and the
optimum binding temperature of the primers.
In general, two basic methods are used in qPCR and these are
based on ether relative quantification and absolute quantification
[6]. Relative quantification is based on comparisons with standard
DNA/cDNAs within the sample for measurement of ratiometric
differences. The absolute quantitation approach can yield the pre-
cise number of resulting amplicons by comparison with DNA stan-
dards using a calibration curve. This requires that PCR of the
DNA/cDNA in the sample and the standard have the same ampli-
fication efficiency. In addition to widespread use in research stud-
ies, qPCR has already been applied in many studies for the discovery
of biomarkers for applications in clinical studies such as evaluating
the status of certain cancers or for monitoring disease progression
or treatment response [7–9]. Here, we describe the use of the
SensiFAST™ Probe (Bioline; London, UK) that uses a unique buf-
fer chemistry to enable fast and reproducible multiplex qPCR
determinations. This property makes this an ideal approach for
routine clinical use.
2 Materials (See Note 1)
1. 400 nM oligonucleotide primers (see Notes 2 and 3).

2. 100 nm probes (see Notes 4 and 5).
3. Templates: approximately 1 mg genomic DNA or 100 ng
cDNA or 1 × 10−6–1.0 μg total RNA or 0.01 pg mRNA (see
Note 6).
4. 1× SensiFAST Probe Mix, containing hot-start DNA poly-
merase, dNTPs, stabilizers, and enhancers.
5. Reverse transcriptase (see Note 7).
6. RNase inhibitor (see Note 7).
7. qPCR thermocycler (see Note 8).
3 Methods (See Note 9)
1. Isolate DNA or RNA as required using standard methods.

2. Select amplicons of interest (see Note 10).
3. For DNA and cDNA templates, prepare a PCR master mix based
on a standard 20 μL final reaction volume containing the prim-
ers, probes, template, and probe mix (see Notes 6 and 11).
0.44
0.40
0.36
0.32
0.28
Fluorescence
0.24
0.20
0.16
0.12
0.08
0.04
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45
Cycle
Fig. 3 Five replicates were run using a conventional TaqMan primer/probe set under fast cycling conditions
(3 min 95 °C followed by 45 cycles at 95 °C for 10 s and 60 °C for 10 s). Singleplex reactions (blue line) and
quadruplex reaction (red line) for the γ-actin and JOE dye are indistinguishable. However, slightly lower fluo-
rescence intensity is often seen for the multiplex reactions, as reagents are consumed more quickly
4. For total RNA and mRNA templates, prepare a PCR master

mix based on a standard 20 μL final reaction volume contain-
ing the 1:100 reverse transcriptase, 1:50 RiboSafe RNase
Inhibitor, primers, probes, template, and probe mix (see Notes
6 and 11).
5. Suggested thermal cycling conditions for DNA and cDNA: 1
cycle at 95 °C for 2–5 min for polymerase activation, then 40
cycles at 95 °C for 10 s for denaturation and 60 °C for 20–50 s
for annealing/extension (see Notes 12 and 13) (Fig. 3).
6. Suggested thermal cycling conditions for RNA: 1 cycle at
45 °C for 10 min for reverse transcription, 1 cycle at 95 °C for
2 min for polymerase activation, then 40 cycles at 95 °C for 5 s
for denaturation and 60 °C for 20 s for annealing/extension
(see Notes 13 and 14).
7. Data analysis (see Note 15).
4 Notes
1. These guidelines refer to the design and setup of TaqMan

probe-based PCR. Please refer to the relevant literature when
using other probe types. The specific amplification, yield, and
overall efficiency of any qPCR can be critically affected by the
sequence and concentration of the probes and primers, and
amplicon length.
2. Use primer-design software, such as Primer3 (http://frodo.

wi.mit.edu/primer3/) or visual OMP™ (http://dnasoftware.
com/). Primers should have a melting temperature (Tm) of
approximately 60 °C; the Tm of the probe should be approxi-
mately 10 °C higher than that of the primers. Tm can be deter-
mined using software; however, a good approximation is
(2 × number of As and Ts) + (4 × number of Gs and Cs).
Importantly, a 40–60 % GC content is recommended for all
primers and to avoid long stretches of any one base. There is a
range of about 6–8 °C over which the PCR will work well. The
closer you are to the top of this range, the more specificity you
will have. For fast reaction kits such as the SensiFAST we also
recommend adding a further 5 °C as they have a higher salt
concentration.
3. A final primer concentration of 400 nM is suitable for most
probe-based reactions; however to determine the optimal con-
centration we recommend titrating in the range 200–
1000 nM. The forward and reverse primers concentration
should be equimolar. We recommend aliquoting the primers
to avoid repeated freeze/thaw of the primary source, as this
will effect PCR efficiency and sensitivity. Aliquots should be
used for up to six freeze/thaw cycles.
4. A probe concentration of 100 nM is recommended for multi-
plexing since higher concentrations can result in cross-channel
fluorescence. Probe sequence should be designed as above.
5. For each probe, consider the spectral properties of the dyes in
terms of intensity of fluorescence and spectral overlap with
other dyes within the reaction. It is important to determine the
dyes for which your qPCR instrument has been calibrated or is
capable of detecting once calibrated. The manufacturer can
provide instrument excitation and detectable emission wave-
lengths (Table 1). Some of the older instruments require the
use of a passive reference such as ROX or fluorescein, to nor-
malize expression levels between wells of a 96 or 384-well
plate. If normalization is required with these instruments, this
will reduce the selection of fluorescent dyes that can be used.
6. It is important that the DNA template is suitable in terms of
purity and concentration. The template must be devoid of any
contaminating PCR inhibitors (e.g., EDTA). The recom-
mended amount of template for PCR is dependent upon the
type of DNA used. For genomic DNA, use up to 1 mg extracted
DNA using a kit such as the Bioline ISOLATE II Genomic
DNA Kit or a phenol/chloroform-based method such as
Bioline TriSure (ensure that samples are washed thoroughly as
even small amounts of phenol are inhibitory to PCR). For
cDNA, highly pure RNA is recommended and to perform a
two-step RT-PCR. It is also important to use a pre-optimized
Table 1
Common fluorophores and quenchers used for qPCR probes
Fluorophore Absorption (nm) Emission (nm) Suggested compatible quencher

FAM 495 517 TAMRA, BHQ-1, Dabcyl
JOE 520 548 TAMRA, BHQ-1, Dabcyl
VIC 528 546 TAMRA, BHQ-1, Dabcyl
HEX 537 553 TAMRA, BHQ-1, Dabcyl
NED 546 575 TAMRA, BHQ-1, Dabcyl
TAMRA 550 576 BHQ-2
Cy3 550 570 BHQ-2
ROX 581 607 BHQ-2
Cy5 650 667 BHQ-2/BHQ-3
mix such as the SensiFAST cDNA Synthesis Kit for reverse

transcription. The optimal amount of cDNA to use in a single
PCR is dependent upon the copy number of the target gene.
We suggest using 100 ng cDNA per reaction; however, it may
be necessary to vary this amount. For RNA, it is important that
the template is intact and devoid of DNA or contaminating
inhibitors of both reverse transcription and PCR. The recom-
mended amount of template for one-step real-time RT-PCR is
dependent upon the type of RNA used. For total RNA, we
recommend using 1 × 10−6 pg to 1 μg and for mRNA 0.01 pg
per 20 μL reaction.
7. For use with RNA templates. It is important to use and RNAase
inhibitor such as the RiboSafe RNase inhibitor (Bioline)
although others can be used.
8. Many instruments can be used here such as the 7500 FAST,
7900HT FAST, ViiA7™, and StepOne™ from Applied
Biosystems (Waltham, Massachusetts, USA), the Mx4000™
from Stratagene/Agilent (Santa Clara, California, USA), the
iCycler™ and MyiQ5™ from Bio-Rad (Hercules, California,
USA), the LightCycler® from Roche (Basel, Switzerland), the
RotorGene™ from Qiagen (Hilden, Germany), and the MIC
from Bio Molecular Systems (Upper Coomera, Queensland,
Australia). Although other instruments can be used but it is
important to check with the manufacturer for compatibility, as
described above.
9. qPCR is extremely sensitive and so to help prevent any carry-
over DNA contamination, separate areas for reaction setup,
PCR amplification and any post-PCR gel analysis should be
maintained. It is essential that any tubes containing amplified
PCR product are not opened in the PCR setup area. As with all
types of PCR, follow the three-room rule. One of the biggest
causes of contamination and background is from using the
same pipettes for extraction, PCR setup, and post-run analysis.
Even if aerosol resistant tips are used all the time, this is not a
good idea. Instead, you should have a dedicated set of pipettes
for each stage. In addition to pipettes, you should have a dif-
ferent location, either hoods with UV lamps (or preferably a
completely different room) for extractions, PCR setup, and
any post PCR analysis. In addition, it is important to detect the
presence of contaminating DNA that may affect the reliability
of the data by including a no-template control reaction, replac-
ing the template with PCR-grade water.
10. For multiplex analyses, the length of the amplicons should be
similar and between 50 and 150 bp for optimal PCR efficiency.
This is important since each amplicon competes for the same
reagents in the probe mix (dNTPs and polymerase).
11. Always mix the reagents well before use. This may sound obvi-
ous but this is a very sensitive system and the reagents contain
dyes, dNTPs, and enzymes that may have settled while sitting
in the freezer or refrigerator.
12. The conditions are suitable for the SensiFAST Probe Kit tar-
geting amplicons up to 200 bp. For the polymerase activation
step, 2 min are required for cDNA and 5 min for genomic
DNA. For all other steps, the temperatures may vary depend-
ing on the primer sequences and up to 50 cycles may be
required in multiplex experiments.
13. When testing a mix, template, or primers, it is important to
amplify from a tenfold template dilution series. Loss of detec-
tion at low template concentrations is the only direct measure-
ment of sensitivity and can also indicate the presence of
inhibitors. If inhibition is observed, either the DNA needs to
be used at lower concentrations or it requires re-purification.
Ideally, samples should cross the threshold (Ct) between cycles
20–30. Therefore, individual reactions should be optimized
prior to multiplexing, with efficiencies as close to 100 % as
possible.
14. The conditions are suitable for the SensiFAST Probe One-Step
Kit, targetting amplicons of up to 200 bp. However, they can
be varied to suit different machine-specific protocols. The
reverse transcription reaction time can be extended up to
20 min and/or the temperature can be increased up to
48 °C. For the annealing/extension stage, temperatures may
vary depending on primer sequences and up to 50 s may be
necessary for multiplexing with more than two probes.
15. Optimal analysis settings, such as baseline and threshold val-

ues, for each primer/probe set are a prerequisite for accurate
quantification data. It is therefore important to analyze the
data for each channel separately as the qPCR instrument
default settings may not provide accurate results. It is recom-
mended to keep the multiplex reactions after amplification so
that if there is any doubt in the results, the PCR products can
be checked on an agarose gel.
References
1. Mullis KB, Faloona FA (1987) Specific syn- 7. Deng Q, Yang H, Lin Y, Qiu Y, Gu X, He P
thesis of DNA in vitro via a polymerase-cata- et al (2014) Prognostic value of ERCC1 mRNA
lyzed chain reaction. Methods Enzymol expression in non-small cell lung cancer, breast
155:335–350 cancer, and gastric cancer in patients from
2. Mullis KB (1990) The unusual origin of the Southern China. Int J Clin Exp Pathol
polymerase chain reaction. Sci Am 262:56–61 7:8312–8321
3. Higuchi R, Fockler C, Dollinger G, Watson R 8. Böttcher R, Henderson DJ, Dulla K, van Strijp
(1993) Kinetic PCR analysis: real-time monitor- D, Waanders LF, Tevz G et al (2015) Human
ing of DNA amplification reactions. phosphodiesterase 4D7 (PDE4D7) expression
Biotechnology 11:1026–1030 is increased in TMPRSS2-ERG-positive primary
prostate cancer and independently adds to a
4. h t t p : / / w w w. m a n a g e d c a r e m a g . c o m /
reduced risk of post-surgical disease progres-
archives/2009/5/managing-cost-diagnosis
sion. Br J Cancer 113:1502–1511
5. Valasek MA, Repa JJ (2005) The power of real-
9. Bahnassy A, Mohanad M, Ismail MF, Shaarawy S,
time PCR. Adv Physiol Educ 29:151–159
El-Bastawisy A, Zekri AR (2015) Molecular bio-
6. Wong ML, Medrano JF (2005) Real-time markers for prediction of response to treatment
PCR for mRNA quantitation. Biotechniques and survival in triple negative breast cancer patients
39:75–85 from Egypt. Exp Mol Pathol 99:303–311
Chapter 9
Multiplex Analysis Using cDNA Transcriptomic Profiling

Abstract
DNA microarrays contain microscopic DNA spots attached to a solid surface. Each spot contains picomolar
levels of a specific DNA probe sequence and hybridization to the corresponding gene products can be
detected and quantitated through the use of fluorescently labeled target DNA. In this format, DNA micro-
arrays can be used to measure the expression level of thousands of genes in a single experiment. Here, we
present a method to detect the mRNA transcriptional changes in neuronal precursor cells following dif-
ferentiation using high density cDNA microarrays.
Key words Fluorescent dyes, Taq polymerase, Quantitation, mRNA, cDNA, Multiplex analysis
1 Introduction
The DNA microarray technique was developed originally in a rudi-

mentary format in 1975 [1] but emerged more or less in its present
form beginning in the 1990s [2, 3]. This approach gave scientists
the capability of simultaneously profiling the expressed genes, or
messenger RNAs (mRNAs), within a cell, tissue, or organism for
the first time. As with most molecular biology detection methods,
the DNA microarray approach works by exploiting the quality of
nucleic acid sequences to hybridize with other nucleic acids
and form complementary sequences. Microarrays consist of a pre-
designed set of synthetic nucleic acid probes immobilized and
arrayed in a grid-like pattern on a solid surface [4]. Microarrays
were first made by immobilizing probes onto filter paper although
this progressed to attaching the probes to solid surfaces, such as
glass or silicon chips [5–7]. Currently, most approaches use a robot
to print nucleic acid probes onto a chemical matrix or they employ
a photoactivated chemistry and masking technique to synthesize
probes directly on the surface of choice in predesignated locations.
In a typical protocol, reverse transcription (RT) of extracted mRNA
is carried out in the presence of the four nucleotides, one of which
is labeled with a fluorescent dye, to generate labeled complementary
135
Fig. 1 Schematic diagram of the cDNA microarray technique
DNA (cDNA) that can then be hybridized with the probe array
(Fig. 1).
In comparative microarray studies, one sample usually serves as
the control (e.g., a sample from a healthy person) and the other as
an experimental sample (e.g., a sample from an individual with a
specific disease). This begins with isolation of mRNA from the two
samples followed by RT using a nucleotide mixture labeled with
dyes that fluoresce at different wavelengths. This is done so that
the two samples can be hybridized and eventually distinguished on
the same microarray [8]. The hybridization step involves addition
of the labeled cDNA mixture onto the microarray where this will
bind to the attached cDNA probes, followed by a series of washes
to remove the nonspecific cDNAs. Following this, the fluorescent
labels on captured cDNA strands are excited by a laser and these
release light at a specific wavelength and intensity. The strength of
the released light depends upon the amount of target sample bind-
ing to the probes present on the corresponding spot in the micro-
array. This correlates to the amount of each specific mRNA in the
sample. Therefore, calculating the ratio of the two dyes in a com-
parative DNA array experiment yields the relative levels of the same
mRNA in the associated samples (Fig. 2).
Microarray technology has high multiplexing capacity with the
capability of simultaneously profiling thousands of mRNA tran-
scripts. It has already been used in numerous studies covering such
applications as single nucleotide polymorphism identification,
studies of cell fate and disease mechanisms [5, 9, 10]. Here, we
cDNA Microarray Analysis 137
Fig. 2 Schematic diagram of a comparative microarray study. In the example shown, the transcript levels in a
control and diseased cell are compared. If a particular mRNA is present at a higher level due to the disease,
the spot on that part of the microarray will appear to have a higher amount of red coloration due to binding of
a greater amount of red-labeled cDNA. In contrast, higher levels of an mRNA in the control will appear to be
green on the corresponding position on the microarray and those with approximately equal levels in the dis-
ease and control will appear as yellow (an equal mixture of red and green)
present the transcriptional changes in adult mouse subventricular

zone progenitor cells induced by differentiation [11] using the
Agilent high density cDNA microarrays.
2 Materials (see Note 1)
1. DNAase- and RNase-free water (see Note 2).

2. 20 μg total RNA samples (see Note 3).
3. 6 μM oligo20 dT primers.
4. 20 nM cyanine 3-dCTP (see Note 4).
5. 20 nM cyanine 5-dCTP (see Note 4).
Fig. 3 Schematic of Agilent high density microarray slide
6. First strand synthesis master mix: 1× first strand reaction buffer,

10 mM dithiothreitol, 400 U Moloney murine leukemia virus
reverse transcriptase, 40 U RiboSafe RNase inhibitor, 40 nM
ATP, 40 nM TTP, 40 nM GTP and 20 nM CTP (see Note 5).
7. cDNA wash buffer PE (Qiagen; Manchester, UK).
8. Elution buffer: 10 mM Tris–HCl, pH 8.5.
9. QIAquick spin column (Qiagen).
10. Deposition hybridization buffer (see Note 6).
11. cDNA microarray on 7.62 × 2.54 × 0.10 cm glass slides and
appropriate slide cover slips (Agilent Technologies; Santa Clara,
CA, USA) (Fig. 3) (see Note 7).
12. Microarray hybridization chamber (see Note 8).
13. Wash solution 1: 0.5× saline sodium citrate (SSC) and 0.01 %
sodium dodecyl sulphate (SDS).
14. Wash solution 2: 0.06× SSC.
3 Methods
1. Combine RNA and oligo dT primer in a sterile nuclease-free

tube in 25 mL total volume and incubate at 70 °C for 10 min
and then place on ice for 5 min (see Note 9).
2. Add cyanine 3-dCTP to the RNA sample from undifferenti-
ated cells and add cyanine 5-dCTP to the RNA sample from
differentiated cells (see Note 10).
3. Add the master mix to each sample and incubate at 42 °C for
60 min.
4. Heat at 70 °C for 10 min and then place on ice for 5 min

(see Note 11).
5. Centrifuge briefly to ensure all liquids are at the bottom of
the tube and add 20 U RNase 1A, mix by pipetting gently
up and down, and incubate at room temperature for 30 min
(see Note 12).
6. Combine the cyanine 3- and cyanine 5-cDNA reactions for
each microarray hybridization experiment into one tube to
give a final volume of 100 μL (see Note 13).
7. Centrifuge at 10,000 × g for 1 min using a QIAquick column
(see Note 14).
8. Discard flow-through and wash the column with 400 μL of
wash buffer, centrifuge for 60 s, and repeat (see Note 15).
9. Elute the cyanine-labeled cDNA by adding 30 μL of elution
buffer, leaving this for 1 min, and centrifuging for 60 s, collect-
ing the eluate into a nuclease-free collection tube.
10. Repeat wash and centrifugation, and combine the two eluates
from each sample into a single collection tube.
11. Add 7.5 μL of each cyanine-labeled sample to a fresh tube con-
taining 1× deposition hybridization buffer and nuclease-free
water to give a final volume of 25 μL.
12. Mix well, incubate at 98 °C for 2 min, and leave the solution
at room temperature until use (see Note 16).
13. Place the microarray slide so that the barcode side is facing
down and add the hybridization mixture onto the center of
each microarray section (Fig. 3) and place a glass cover slip
over each one in use (see Note 17).
14. Place the slide into the hybridization chamber, which contains
nuclease-free water, close the chamber, and incubate at 65 °C
overnight or for approximately 16 h (see Note 18).
15. The next day, remove the slides from the chamber and remove
cover slips by gently dipping in wash solution 1.
16. Place the slide in a rack in a container containing wash solution
1 on a magnetic stirrer and stir for 5 min.
17. Transfer to the same setup in a container containing wash solu-
tion 2 and stir for 2 min.
18. Transfer the slides to plastic containers in a swinging bucket
rotor and centrifuge at 400 × g for 2 min (see Note 19).
19. Scan the slides using an appropriate imager and analyze the
spot volumes in the channels for cyanine 3 and cyanine 5 to
quantitate ratios of expression levels of specific transcripts in
the two samples under comparison (Table 1) (see Note 20).
Table 1
Top differentially (Diff) regulated genes in neural progenitor cells (NPCs) following a 24 h
differentiation protocol. The values are given as the levels in Diff/NPC cells
Symbol Name Ratio

ALDH1A3 Aldehyde dehydrogenase family 1, subfamily A3 10.44
CHGB Chromogranin B 9.43
PADI2 Peptidyl arginine deiminase, type II 7.47
TRP53INP1 Transformation related protein 53 inducible nuclear protein 1 7.39
POU3F1 POU domain, class 3, transcription factor 1 6.97
GFAP Glial fibrillary acidic protein 6.83
NDRG2 N-myc downstream regulated gene 2 6.54
DUSP4 Dual specificity phosphatase 4 6.35
EDNRB Endothelin receptor type B 6.28
AGT Angiotensinogen 6.15
IER3 Immediate-early response 3 5.89
APLN Apelin 5.81
SPRY1 Sprouty homologue 1 (Drosophila) 4.96
SPP1 Secreted phosphoprotein 1 4.84
C1QR1 Complement component 1, subcomponent q1 4.71
4 Notes
1. The protocol below is specific for Agilent cDNA microarrays

on 7.62 × 2.54 × 0.10 cm glass slides containing amplified
Human UniGene 1, Human Drug Target, and Human
Foundation Series clone sets [12, 13]. Each microarray con-
tains approximately 22,500 cDNAs and each spot corresponds
to a different clone. For other microarrays the protocol may be
slightly different.
2. Care should be taken throughout due to the ease of losing
DNA and RNA to unwanted nuclease activity.
3. We used RNA extracted from neuronal precursor cells under
proliferation (culture in the presence of 20 ng⁄mL fibroblast
growth factor 2 and 20 ng⁄mL epidermal growth factor) and
differentiating (culture in the absence of growth factors) con-
ditions. The RNA can be extracted using standard protocols.
4. Minimize exposure to light as both the cyanine dyes are
photolabile.
5. This step is for use with total RNA templates and conversion to
cDNA. It is important to use and RNase inhibitor such as the
RiboSafe RNase inhibitor (Bioline) although others can be
used.
6. The buffer contains lithium chloride and lithium lauryl sul-
phate, so the usual laboratory care should be taken.
7. Each slide contains two microarrays so two 2.4 × 3.0 cm cover
slips are required.
8. Many chambers can be used but these must be compatible
with slide/cover slip hybridization and watertight.
9. This step denatures the template and primers.
10. It is recommended to carry out reciprocal labeling reactions to
account for any preferential labeling of either of the cyanine
dyes. In this case one comparison would be Cy3-undifferentiated
versus Cy5-differentiated as shown and another could be Cy5-
undifferentiated versus Cy3-differentiated.
11. This step inactivates the enzyme.
12. This step degrades the RNA.
13. This allows hybridization of the two different samples to the
same microprobes on the cDNA array.
14. In this step, the cyanine-labeled DNA is bound to the column.
Here, we use the Qiagen quick spin columns but other col-
umns from other manufacturers would also work well.
15. We use the Qiagen buffer PE but other wash buffers would
also work here (these buffers usually contain salts, ethanol, and
other reagents).
16. This is to denature the DNA.
17. Be careful not to touch the surface of the microarray as it is
easily damaged. The presence of two sections on each slide
allows the reciprocal labeling strategy to be easily applied (as
described above).
18. The chamber used must be watertight when sealed and
the addition of water to the chamber helps to prevent
evaporation.
19. This step is for drying the slides and must be performed quickly
to avoid leaving a residue on the slides which could interfere
with scanning.
20. There are many scanners available on the market for this
purpose, although these should be capable of exciting and
detecting the hybridized cyanine 3- (550 nm and 570 nm,
respectively) and cyanine 5- (650 nm and 670 nm, respec-
tively) labeled cDNAs.
References
1. Grunstein M, Hogness DS (1975) Colony analysis using oligonucleotide arrays produced
hybridization: a method for the isolation of by maskless photolithography. Genome Res
cloned DNAs that contain a specific gene. Proc 12:1749–1755
Natl Acad Sci U S A 72:3961–3965 8. Shalon D, Smith SJ, Brown PO (1996) A DNA
2. Craig AG, Nizetic D, Hoheisel JD, Zehetner G, microarray system for analyzing complex DNA
Lehrach H (1990) Ordering of cosmid clones samples using two-color fluorescent probe
covering the herpes simplex virus type I (HSV-I) hybridization. Genome Res 6:639–645
genome: a test case for fingerprinting by hybridi- 9. Hacia JG, Fan JB, Ryder O, Jin L, Edgemon
sation. Nucleic Acids Res 18:2653–2660 K, Ghandour G et al. (1999) Determination of
3. Lennon GG, Lehrach H (1991) Hybridization ancestral alleles for human single-nucleotide
analyses of arrayed cDNA libraries. Trends polymorphisms using high-density oligonucle-
Genet 7:314–317 otide arrays. Nat Genet 22:164–167
4. Leung YF, Cavalieri D (2003) Fundamentals of 10. de Souto MC, Costa IG, de Araujo DS,
cDNA microarray data analysis. Trends Genet Ludermir TB, Schliep A (2008) Clustering
19:649–659 cancer gene expression data: a comparative
5. Bumgarner R (2013) Overview of DNA micro- study. BMC Bioinformatics 9:497
arrays: types, applications, and their future. 11. Bonnert TP, Bilsland JG, Guest PC, Heavens
Curr Protoc Mol Biol Chapter 22:Unit 22.1. R, McLaren D, Dale C (2006) Molecular char-
doi:10.1002/0471142727.mb2201s101 acterization of adult mouse subventricular
6. Goldmann T, Gonzalez JS (2000) DNA- zone progenitor cells during the onset of dif-
printing: utilization of a standard inkjet printer ferentiation. Eur J Neurosci 24:661–675
for the transfer of nucleic acids to solid 12. http://www.ncbi.nlm.nih.gov/geo/query/
supports. J Biochem Biophys Methods 42: acc.cgi?acc = GPL10
105–110 13. http://www.ncbi.nlm.nih.gov/geo/query/
7. Nuwaysir EF, Huang W, Albert TJ, Singh J, acc.cgi?acc = GPL875
Nuwaysir K, Pitas A et al (2002) Gene expression
Chapter 10
Multiplex Single Nucleotide Polymorphism Analyses

Abstract
Quantitative polymerase chain reaction (qPCR) is a routinely used method for detection and quantitation
of gene expression in real time. This is achieved through the incorporation and measurement of fluorescent
reporter probes in the amplified cDNA strands, since the fluorescent signals increase as the reaction pro-
gresses. The availability of multiple probes that fluoresce at different wavelengths allows for multiplexing
as this gives rise to amplicons with unique fluorescent signatures. Here, we describe a method using the
SensiFAST and SensiFAST One-Step probe kits which allows simultaneous real-time quantitation of up to
5 amplicons.
Key words Single nucleotide polymorphism, Allele, PCR, Fluorescent dyes, Taq polymerase,
Quantitation, Multiplex analysis
1 Introduction
A single nucleotide polymorphism (SNP) is a variation in a single

nucleotide at a specific position in the genome [1]. As an example, the
base cytosine (C) may be present at a specific position in the human
genome although this is replaced by an adenosine (A) base may in a
small percentage of individuals. Many SNPs are benign although some
may underlie differences across individuals in their susceptibility to
specific diseases [2], such as Alzheimer disease [3], sickle-cell anaemia
[4], certain metabolic disorders [5], various types of cancer [6], and
heart disease [7]. For example, a C/T substitution at amino acid posi-
tion 130 of the APOE gene is associated with a higher risk for
Alzheimer’s disease [8].
Several platforms are now in use for detection of SNPs, the
most efficient way to link a SNP with phenotypes is the so-called
genome-wide association (GWA) study, in which hundreds of
thousands or even millions of polymorphisms are scanned per sam-
ple using DNA microarrays. For determination of new SNPs, next
generation sequencing (NGS) is used; however for high-throughput
screening of individual SNPs (often linked to diseases) qPCR is the
143
method of choice. The TaqMan® SNP genotyping approach uses

the 5′ nuclease activity of Taq polymerase to produce a fluorescent
signal during the polymerase chain reaction (PCR) stage of the
analysis [9]. The assay normally uses two probes that are identical
except for the sequence at the site of the allele in question. One
probe is complementary to the native sequence whereas the other
targets the mutant allele. Each probe also contains a distinct 5′
fluorescent reporter dye and a 3′ quencher dye. As with standard
qPCR (see Chapter 11), fluorescence of the reporter dye is sup-
pressed in intact probes due to the proximity of the quencher. In
the annealing stage of PCR, the probes hybridize to the site in
question. During the extension stage, the reporter and quencher
are released by the 5′ nuclease activity of Taq polymerase, resulting
in fluorescence of the reporter dye. However, cleavage of the probe
occurs only in the case of perfectly matched probes as only these
will be recognized by the polymerase (Fig. 1). At the end of the
PCR experiment, the fluorescent signal for the two reporter dyes is
measured. The ratio of the signals will be indicative for the geno-
type of the sample. For increased accuracy, the DNA sample should
be analyzed with two probes. One of these should contain the
native sequence and the other should contain the mutant allele.
Here, we describe the use of the SensiFAST Genotyping Kit
for high throughput of SNPS (see Note 1). This assay is used to
detect a C/T transition substitution on gene SLCO2B1 (a solute
carrier organic anion transporter family member 2B1). The
sequence around the substitution is given below:
Testing for C/A polymorphism Testing for native allele

Correct identification using probe with T/A modification Correct identification using probe containing native G/C sequence
Primer Primer
T G
A C
Primer Primer
T G
A C
Primer Primer
A C
Probe cleaved = signal Probe cleaved = signal
No identification using probe containing native G/C sequence No identification using probe containing T/A sequence
Primer Primer
G T
A C
Primer Primer
G T
A C
Primer G Primer T
A C
Probe displaced = no signal Probe displaced = no signal
Fig. 1 Schematic diagram showing how the levels of a mutant (C/A) and native (T/A) allele can be determined
by PCR analysis of the same DNA sample using two probes specific for each of these sequences
Multiplex Single Nucleotide Polymorphism Analyses 145
A G G AT G C C A G G G TA G AT TA A C C C G G [C / T ]
GAGGCTGAAGTCTAAATAACTGGAA
1. 0.9 μM forward primer (ThermoFisher Scientific;

Loughborough, Leicestershire, UK; assay: M__25345745_10;
catalogue number: 4351384) (see Note 2).
2. 0.9 μM reverse primer (ThermoFisher Scientific; assay: cata-
logue number: 4351384).
3. 0.2 μM allele 1 probe (ThermoFisher Scientific; catalogue
number: 4351384) (see Note 3).
4. 0.2 μM allele 2 probe (ThermoFisher Scientific; catalogue
number: 4351384).
5. 20 ng cDNA template (see Note 4).
6. 1× SensiFAST Genotyping Mix, containing hot-start DNA
polymerase, dNTPs, stabilizers, and enhancers.
7. qPCR thermocycler (see Note 5).
1. Prepare a PCR master mix based on a standard 20 μL final

reaction volume containing the primers, probes, template, and
Genotyping mix (see Notes 4 and 7).
2. Suggested thermal cycling conditions for DNA: 1 cycle at
95 °C for 2–5 min for polymerase activation, then 40 cycles at
95 °C for 10 s for denaturation and 60 °C for 20–50 s for
annealing/extension (see Notes 8 and 9).
3. Data analysis (Fig. 2) (see Note 10).
4 Notes
1. The SensiFAST Genotyping Kit is compatible with many dual-

label probe assays. The overall yield and efficiency relates to the
sequence and concentration of all reagents including template,
primers, and probes, as well as the targeted amplicon length.
Please refer to the appropriate literature during the design and
set-up stages. The following information relates to the design
and setup of TaqMan probe-based qPCR.
2. The length, sequence, and concentration of the primers and
probes are critical for specific amplification. Here, we have
used primers and probes from the indicated ThermoFisher kit,
www.Ebook777.com
12
11
10
Fluorescence (553-580 nm)

9
8
7
6
5
4
3
2
1
1 2 3 4 5 6 7 8
Fluorescence (465-510 nm)
Fig. 2 DNA samples from 96 individuals were genotyped using a custom TaqMan
drug metabolism genotyping assay. The SensiFAST Genotyping Kit was used on
the Roche LightCycler® 480 (autocaller confidence level 95 %). Cycling condi-
tions were: 95 °C for 3 min, then 35 cycles at 95 °C for 10 s and 60 °C for 30 s.
The results show that the SensiFAST Genotyping Kit was able generate distinct
clusters across the 96 individuals
targeting the specific mutation site within the SLCO2B1gene.

For user-defined targets that are not commercially available,
we recommend the use of primer-design software, such as
Primer3 (http://frodo.wi.mit.edu/primer3/) or visual
OMP™ (http://dnasoftware.com/). For the best results
primers and probes should have melting temperatures of
approximately 60 °C and the targeted amplicon length should
be 80–200 base pairs. For the SLCO2B1 gene, the amplicon
length is approximately 150 base pairs (SNP ID: rs4226825,
chromosome 7: 99658318 on NCBI Build 37, polymorphism:
C/T transition substitution). Forward and reverse primers
should be equimolar and titration in the range of 0.2–1 μM
should be used to find the ideal concentration.
3. For user-defined targets that do not have commercially avail-
able reagents, we recommend the use of probe design soft-
ware, as indicated above. A final concentration of 0.2 μM for
each probe is sufficient for most reactions although these
should be at least twofold lower than the primer concentra-
tions. As for the primers, the concentration of both probes
should be equimolar and determined by titration as above.
Each probe should contain dyes that are similar in intensity of
fluorescence but with a distinct spectral separation. Here, we
have used probes from the indicated ThermoFisher kit, target-
Multiplex Single Nucleotide Polymorphism Analyses 147
ing the specific mutation site within the SLCO2B1gene. The

probes are labeled with VIC® (528-546 nm) and
6-carboxyfluorescein (FAM; 495–517 nm), which are spec-
trally distinct and can be quenched with tetramethylrhodamine
(TAMRA).
4. The SensiFAST Genotyping Kit can be used with sample
lysates or purified genomic DNA. It is important that the DNA
template does not contain any inhibitors, such as EDTA.
5. Many instruments can be used here but it is important to check
with the manufacturer for compatibility, as described above. In
this study, we used the Roche LightCycler® 480.
6. Due to the sensitivity of qPCR, clean room conditions should
be used to avoid any contamination of the samples.
7. As reagents can settle or partition during storage, always mix
well when setting up reactions. However, do not use sonica-
tion methods for this due to the likely outcome of shearing the
DNA template.
8. Positive and controls of known genotype should be included
to ensure distinct genotype calling by the instrument. A con-
trol that does not contain any template should also be used to
detect any contamination.
9. Cycling conditions can be varied to suit machine-specific pro-
tocols. Initially, running 30 cycles with a 30 s extension is rec-
ommended and then adding cycles in increments of five, as
required. Longer extension times may be required for ampli-
cons larger than 200 bp. For low concentrations of template
(<1 ng). up to 45 cycles may be necessary. However, it is not
recommended to exceed a total of 45 cycles for optimal call-
ing. Here are the conditions that we used for this experiment:
95 °C for 3 min, then 35 cycles at 95 °C for 10 s and 60 °C for
30 s.
10. Most qPCR machines will create scatter plots and use algo-
rithms to assign genotypes based on reporter probe signals at
the end of amplification steps. Failure of this autocalling is usu-
ally due to problems associated with outliers that skew the
clusters. Removing the outliers and reanalyzing the data should
allow the program to adjust the scaling so that distinct clusters
can be seen.
References
1. Ligtenberg MJ, Gennissen AM, Vos HL, 2. Bacolod MD, Schemmann GS, Giardina SF, Paty
Hilkens J (1991) A single nucleotide polymor- P, Notterman DA, Barany F (2009) Emerging
phism in an exon dictates allele dependent paradigms in cancer genetics: some important find-
differential splicing of episialin mRNA. Nucleic ings from high-density single nucleotide polymor-
Acids Res 19:297–301 phism array studies. Cancer Res 69:723–727
3. Li L, Yin Z, Liu J, Li G, Wang Y, Yan J, Zhou H of three gene variants association with risk of
(2013) CYP46A1 T/C polymorphism associ- prostate cancer: an update. Urol J 12:
ated with the APOE ε4 allele increases the risk of 2138–2147
Alzheimer’s disease. J Neurol 260:1701–1708 7. Zhang MM, Xie X, Ma YT, Zheng YY, Yang
4. Vicari P, Adegoke SA, Mazzotti DR, Cançado YN, Li XM, Fu ZY, Liu F, Chen BD (2015)
RD, Nogutti MA, Figueiredo MS (2014) Association of COX-2 -765G > C genetic
Interleukin-1β and interleukin-6 gene polymor- polymorphism with coronary artery disease: a
phisms are associated with manifestations of meta-analysis. Int J Clin Exp Med 8:7412–
sickle cell anemia. Blood Cells Mol Dis 7418
54:244–249 8. Namboori PK, Vineeth KV, Rohith V, Hassan I,
5. Ghalandari H, Hosseini-Esfahani F, Mirmiran P Sekhar L, Sekhar A, Nidheesh M (2011) The
(2015) The association of polymorphisms in ApoE gene of Alzheimer’s disease (AD). Funct
leptin/leptin receptor genes and ghrelin/ghre- Integr Genomics 11:519–522
lin receptor genes with overweight/obesity and 9. Glaab WE, Skopek TR (1999) A novel assay for
the related metabolic disturbances: a review. Int allelic discrimination that combines the fluoro-
J Endocrinol Metab 13:e19073 genic 5′ nuclease polymerase chain reaction
6. Chen Y, Zhong H, Gao JG, Tang JE, Wang R (TaqMan) and mismatch amplification mutation
(2015) A systematic review and meta-analysis assay. Mutat Res 430:1–12
Chapter 11
Pulsed SILAC as a Approach for miRNA Targets

Identification in Cell Culture
Daniella E. Duque-Guimarães, Juliana de Almeida-Faria,
Thomas Prates Ong, and Susan E. Ozanne
Abstract
Pulsed stable isotope labeling by amino acids in cell culture (pSILAC) comprises a variation of the classical
SILAC proteomic methodology that enables the identification of short-term proteomic responses such as
those elicited by micro RNAs (miRNAs). Here, we describe a detailed pSILAC protocol for global identi-
fication and quantification of protein translation alterations induced by a miRNA using 3T3-L1 pre-
adipocytes as a model system.
Key words Pulsed stable isotope labeling by amino acids, Cell culture, MicroRNAs, Targets,
Proteomics
1 Introduction
MicroRNAs (miRNAs) have been shown to play a key role in the

control of gene expression, modulating the expression of many
proteins in a wide range of biological processes [1]. These small
noncoding RNA molecules can individually regulate multiple tar-
get genes by binding to target sites found within the 3′ untrans-
lated region of the targeted mRNA, resulting in posttranscriptional
gene silencing of gene networks. The posttranscriptional effect
results from degradation of the targeted mRNA or repression of its
translation [2, 3]. Most of the targets contain perfect complemen-
tary sites in their 3′-untranslated regions to the seed sequence of
the miRNA. This conserved seed sequence of typically six to eight
nucleotides is often used as the main feature for miRNA target site
prediction [4]. Several computational algorithms such as
TargetScan [5] and Diana-MicroT [6] have been developed to
predict targets based on a combination of different parameters
including these sequences. However, additional factors can impede
these bioinformatic approaches, resulting in false positives and thus
149
150 Daniella E. Duque-Guimarães et al.
limiting their usefulness [7]. Experimental methods are therefore

required for robust target predication. Approaches used to date
include in vitro luciferase reporter assays for individual miRNAs
and targets or high-throughput experimental methods such as
next-generation sequencing following immunoprecipitation [8].
However, these have their limitations and the methodology is still
challenging. There is no clear consensus that a functional miRNA
target will be always identified.
Stable Isotope Labeling by Amino acids in Cell culture
(SILAC) is a proteomic approach that has been used to identify
miRNA targets and investigate the consequences on protein levels
[9, 10]. SILAC is a quantitative method based on whole proteome
metabolic labeling with stable isotope-labeled amino acids in cell
culture [11]. This method is based on the introduction of a mass
difference between two proteomes, which creates two versions of
every peptide (e.g.,12C versus13C) that can be distinguished in mass
spectrometry (MS)-based analysis [11, 12]. The classical version of
SILAC is based on growing cells in media with natural “light” or
“heavy” amino acids for several days or for approximately five cell
doublings. This allows virtually complete incorporation of the
heavy amino acids into cellular proteomes. SILAC-based
approaches have also been used to identify miRNA targets in many
cell systems including HeLa cells [9], as well as in SW480 colon
cancer [13] and MiaPaCa-2 pancreatic cancer [14] cells.
In recent years, an increasing number of studies have adopted
a variation of the classical SILAC method termed pulsed SILAC
(pSILAC) to investigate miRNA proteomic effects [10, 15, 16].
The pSILAC method is based on growing two populations of cells
(e.g., miRNA-treated and control cells) that are pulse-labeled with
two different heavy SILAC amino acids [15]. Both populations of
cells are grown initially in standard media containing normal light
amino acids [16]. After miRNA transfection, the culture medium
from control and miRNA-transfected cells is replaced with SILAC
medium containing either medium-heavy or heavy amino acids,
respectively. After the pulse labeling step, all newly synthesized
proteins will incorporate medium-heavy or heavy amino acids,
while proteins that were synthesized before the labeling step will
contain only light amino acids and are not considered for protein
quantification [15, 16].
In this chapter, we describe a detailed pSILAC protocol for
global identification and quantification of protein translational
alterations induced by miRNA using 3T3-L1 pre-adipocytes as a
model-system (Fig. 1). Cells are grown in two cell culture dishes
for 24 h and then transfected with either the miRNA mimic or a
transfection control. After 8 h, the normal growth media from
both cell populations is replaced with media containing either
heavy lysine (Lys8) or medium-heavy lysine (Lys4) for the miRNA
mimic transfected and miRNA control cells, respectively. After
MicroRNA Profiling using Pulsed SILAC 151
Transfection with Lys0 Lys0 Lys4

miRNA negative control Lys0 Mass spectrometry
Lys0 Lys0
Lys4
Lys0
Lys0
Lys4 Lys4
Lys4
3T3-L1 cells
Intensity
3T3-L1 cells
24-hour growth
in regular DMEM SILAC DMEM Lys8
with Lys0 with Lys4
8 hours 24 hours Combine SDS PAGE
Change to Pulsed SILAC proteomes Tryptic m/z
SILAC DMEM labelling digestion
3T3-L1 cells 3T3-L1 cells
24-hour growth
in regular DMEM SILAC DMEM Lys0
Lys0 Lys8 / Lys4 < 1
with Lys0 with Lys8
Lys8
Lys0
Lys0
Lys8
Lys0 Potential
miRNA target
Transfection with Lys0
miRNA mimic
Fig. 1 pSILAC experimental design. Cells are cultivated in regular DMEM media containing light lysine (Lys 0)
in two cell culture dishes for 24 h and then transfected with the miRNA mimic of interest or the negative con-
trol. After 8 h, normal growth medium from both cell populations is replaced with SILAC media containing
either heavy lysine [13C615N2 -l-lysine (Lys8); red font] and medium-heavy lysine [2H4-l-lysine (Lys4); blue font]
for the miRNA mimic cells and miRNA negative control cells, respectively. After 24 h, protein extracts from both
cell populations are combined and subjected to SDS-PAGE separation, digested with trypsin and submitted to
high performance liquid chromatography tandem mass spectrometry (HPLC/MS/MS). The ratio of heavy and
medium-heavy peptides reflects differences in the translation of the corresponding protein induced by the
miRNA mimic and the light peptides are ignored. A ratio of heavy (Lys8)/medium-heavy (lys4) peptides smaller
than 1 indicates a potential miRNA target
24 h, protein extracts from both cell populations are combined and

subjected to sodium dodecyl sulphate polyacrylamide gel electro-
phoresis (SDS-PAGE) to separate the proteins. The protein bands
are then excised from the gels, enzymatically digested with trypsin
and submitted to high performance liquid chromatography tan-
dem mass spectrometry (HPLC/MS/MS) analysis. The ratio of
heavy and medium-heavy peptides reflects differences in the trans-
lation of the corresponding protein induced by the miRNA, while
the light peptides are ignored.
2.1 pSILAC Reagents 1. L-arginine-HCL (Arg0) in phosphate buffered saline (PBS)

(see Note 2).
2. 4, 4, 5, 5-D4 L-Lysine-2HCl (Lys4) in PBS
(see Note 2).
3. 13C615N2 l-Lysine-2HCl (Lys8) in PBS (see Note 2).
4. Dulbecco’s Modified Eagle Medium (DMEM), containing

10 % heat-inactivated calf serum, antibiotic-free (see Note 3).
5. SILAC DMEM media, lacking lysine and arginine.
6. Dialyzed fetal bovine serum (FBS) (see Note 4).
7. Medium-heavy Lys4 SILAC DMEM: 0.08 mg/mL L-arginine,
0.15 mg/mL Lys4, and 10 % FBS in 500 mL SILAC DMEM
(0.22 μm filter-sterilized).
8. Heavy Lys8 SILAC DMEM: 0.08 mg/mL L-arginine,
0.15 mg/mL Lys8, and 10 % FBS in 500 mL SILAC DMEM
(0.22 μm filter-sterilized).
2.2 Cell Culture 1. Undifferentiated 3T3-L1 pre-adipocytes (see Note 5).

and miRNA Mimic 2. Sterile 10 cm diameter cell culture dishes, 15 mL-capacity
Transfection Falcon tubes, and 1.5 mL-capacity microcentrifuge tubes.
3. Tissue culture trypsin solution for dissociation of cells.
4. Opti-MEM® medium and Lipofectamine®RNAiMAX reagent
(Invitrogen).
5. 10 μM mirVana™ miRNA126 mimic and 10 μM mirVana™
miRNA negative control #1 in nuclease-free water
(ThermoFisher Scientific) stored at or below 20 °C (see Note
6).
6. BLOCK-iT™ Alexa Fluor® Red Fluorescent Oligo (Invitrogen)
(see Note 7).
7. Epifluorescence inverted microscope.
8. RIPA buffer: 25 mM Tris–HCl (pH 7.6), 150 mM NaCl, 1 %
NP-40, 1 % sodium deoxycholate, 0.1 % SDS.
9. 100× Halt™ protease and phosphatase single-use inhibitor
cocktail (ThermoFisher Scientific).
10. Bicinchoninic acid (BCA) reagent, 0.125–1.0 mg/mL bovine
serum albumin protein standards.
11. TRI® reagent (Sigma-Aldrich).
12. Direct-zol™ RNA MiniPrep (Zymo Research).
13. TaqMan® microRNA reverse transcription kit (ThermoFischer
Scientific).
2.3 SDS-PAGE 1. 4× Novex Nu PAGE LDS sample buffer (Thermo Fisher

Scientific).
2. Proteomics grade SDS-PAGE gel solutions or precast gels
(see Note 8).
3. Stain solution: 1 mg/mL CoomassieBlue R250, 40 % metha-
nol, 10 % acetic acid (filter sterilized).
4. Destain solution: 20 % methanol, 10 % acetic acid (filter
sterilized).
2.4 In-Gel Trypsin 1. 100 mM ammonium bicarbonate.

Digestion 2. 100 mM ammonium bicarbonate, 50 % acetonitrile.
3. 20 mM ammonium bicarbonate.
4. Reducing solution: 100 mM ammonium bicarbonate, 10 mM
dithiothreitol (see Note 9).
5. Alkylating solution: 100 mM ammonium bicarbonate, 20 mM
iodoacetamide (see Note 9).
6. Trypsin digestion solution: 12.5 μg/mL trypsin (mass spectrom-
etry grade), 20 mM ammonium bicarbonate (see Note 10).
7. 1 % formic acid.
2.5 HPLC, MS 1. MS solvent: 0.1 % trifluoroacetic acid, 3 % acetonitrile.

Analysis, and Data 2. Solvent A: 0.1 % formic acid.
Processing
3. Solvent B: 100 % acetonitrile.
4. NanoAcquity uPLC with a 100 Å, 5 μm, 180 μm × 20 mm C18
Symmetry trap and 130 Å, 1.7 μm, 75 μm × 250 mm C18
BEH analytical columns (Waters Corporation).
5. Accurate mass/high resolution MS instrument such as the
LTQ-OrbiTrap XL (ThermoFisher Scientific).
6. Maxquant software (version 1.5.3.30 or higher; www.max-
quant.org).
2.6 Analysis 1. Several software options: Microsoft Office Excel (https://

of pSILAC Data www.office.com/), R (https://www.r-project.org) in con-
junction with Bioconductor (http://www.bioconductor.org),
Spotfire (http://spotfire.tibco.com/), Matlab (http://uk.
mathworks.com/products/matlab/), and Graphpad Prism
(http://www.graphpad.com/scientific-software/prism/).
3 Methods
3.1 Cell Culture 1. Seed 0.4 × 106 3T3-L1 cells in two dishes in 10 mL antibiotic-
and miRNA Mimic free DMEM and culture for 24 h at 37 °C under 5 % CO2(see
Transfection Note 11).
2. Add 50 μL of miRNA mimic stock solution to 500 μL Opti-
MEM medium in a sterile 1.5 mL tube.
3. Add 43 μL of Lipofectamine RNAiMAX to 500 μL Opti-MEM
medium in a sterile 1.5 mL tube.
4. Add diluted miRNA mimic solution to diluted Lipofectamine
RNAiMAX solution, gently mix and incubate for 5 min at
room temperature.
5. Add the miRNA mimic-lipid complex to 60–80 % confluent
cells, gently mix, and incubate for 8 h at 37 °C under 5 % CO2.
6. Add 50 μL of miRNA control stock solution to 500 μL Opti-

MEM medium in a sterile 1.5 mL tube.
7. Add 43 μL of Lipofectamine RNAiMAX to 500 μL Opti-MEM
medium in a sterile 1.5 mL tube.
8. Add diluted miRNA control solution to diluted Lipofectamine
RNAiMAX solution, gently mix, and incubate for 5 min at
room temperature.
9. Add the miRNA control-lipid complex to 60–80 % confluent
cells, gently mix, and incubate for 8 h at 37 °C under 5 % CO2.
10. Remove the regular medium from the dish containing miRNA
mimic-transfected cells, replace with 10 mL pre-warmed
(37 °C) heavy Lys8 SILAC medium, and incubate for 24 h at
37 °C under 5 % CO2.
11. Remove the regular medium from the dish containing miRNA
control-transfected cells, replace with 10 mL pre-warmed
(37 °C) medium-heavy Lys4 SILAC medium, and incubate for
24 h at 37 °C under 5 % CO2 (see Notes 12 and 13).
3.2 Protein and RNA 1. After 24 h, add trypsin solution to remove cells from the plates
Extraction and wash by centrifugation in ice-cold PBS at 750 × g, for
5 min at 4 °C, in a 15 mL Falcon tube.
2. Suspend the pellets gently in PBS and repeat the centrifugation
wash step.
3. Resuspend cells in 10 mL ice-cold PBS, transfer 5 mL homo-
geneous aliquots to two different Falcon tubes designated for
protein or RNA extraction, and pellet the cells by centrifuga-
tion at 750 × g for 5 min at 4 °C.
4. For protein extraction, add 250 μL cold RIPA buffer containing
protease and phosphatase inhibitors to one aliquot of cells, swirl
on ice for 5 min, centrifuge at 13,000 × g to remove the cell debris,
and transfer the supernatants to fresh tubes (see Note 14).
5. Mix 50 μg proteins extracted from cells transfected with
miRNA mimic with 50 μg of proteins extracted from cells
transfected with control.
6. For total RNA extraction, add 500 μL TRI reagent to the cell
aliquot and extract mRNAs and small RNAs using the Direct-
zol RNA kit according to the manufacturer’s instructions.
7. Prepare cDNA using the reverse transcription kit and carry out
polymerase chain reactions following the manufacturer’s pro-
tocol (see Note 15).
3.3 SDS-PAGE 1. Add the necessary volume of 4× sample buffer to the miRNA
and Treatment of Gel mimic/control mixture.
Pieces 2. Electrophorese the samples.
3. Stain the gel for approximate 2 h, incubate in destain solution

overnight on a rocking table, and take a photograph of the
protein bands.
4. Slice the gel lanes into the desired number of horizontal pieces
and then chop these further to produce 1 mm3 (approximate)
pieces using a new scalpel blade for each slice.
5. The gel pieces from each slice can now be used immediately for
the next step or stored at −80 °C.
6. Wash the gel pieces from each slice with 1 mL water for 15 min,
centrifuge at 750 × g for 5 min, and discard the supernatant.
7. Add 300 μL of acetonitrile to the gel pieces, wash for 15 min,
centrifuge as above, and discard the supernatant.
8. Wash the pieces with 300 μL of 100 mM ammonium bicar-
bonate for 15 min, centrifuge as above, and discard the
supernatant.
bonate, 50 % acetonitrile for 15 min, centrifuge as above, and
discard the supernatant (see Note 16).
10. If the gel pieces are still blue, repeat steps 8 and 9.
11. Wash the pieces as above with 100 μL of acetonitrile for 5 min
(which will cause them to shrink and turn white), centrifuge as
above, and discard the supernatant.
12. Dry the pieces in a sterile hood for 10–15 min.
13. Add 50 μL of reducing solution and incubate for 1 h at 56 °C,
centrifuge as above, and discard the supernatant.
14. Add 50 μL alkylation solution, incubate for 30 min at room
temperature, centrifuge as above, and discard the
supernatant.
15. Wash the pieces twice with 300 μL of 100 mM ammonium
bicarbonate for 15 min, centrifuge as above, and discard the
supernatant.
bonate, 50 % acetonitrile for 15 min, centrifuge as above, and
discard the supernatant.
17. Add 100 μL acetonitrile, incubate for 5 min, centrifuge as
above, and discard the supernatant.
18. Dry the gel pieces in a sterile hood for 10–15 min.
3.4 In-Gel Trypsin 1. Add 30 μL digestion solution to the gel pieces from each slice
Digestion and let stand for 30 min at room temperature (see Note 17).
2. Add a small volume of 20 mM ammonium bicarbonate (with-
out trypsin) so that this solution just covers the gel pieces and
incubate at 30 °C overnight (at least 16 h) (see Note 18).
www.Ebook777.com
3. Add an equal volume of acetonitrile and incubate at 30 °C for

30 min.
4. Centrifuge as above and transfer the supernatant containing
peptide digests to a fresh tube.
5. Add 50 μL 1 % formic acid to the gel pieces and incubate at
room temperature for 20 min.
6. Centrifuge as above and add the supernatant containing pep-
tide digests to the same tube as above.
7. Repeat steps 3–6.
8. Add 50 μL of acetonitrile to the gel pieces and incubate for
10 min (they should shrink and turn white again).
9. Centrifuge as above and add the supernatant to the same tube
as above giving a final volume of 200–400 μL containing the
peptide digest corresponding to each gel slice.
10. Dry samples in a vacuum centrifuge at 60 °C (see Note 19).
11. Suspend the peptide pellets in 25–50 μL of 1 % formic acid,
agitate gently for 30 min at room temperature, and store the
samples at −80 °C if required.
3.5 HPLC, MS 1. Thaw the samples and centrifuge at 13,000 × g for 10 min.
Analysis, and Data 2. Take aliquots from the supernatant gently and transfer to an
Processing MS tube.
3. Dry the tryptic peptides almost to completion in a centrifugal
vacuum concentrator and suspend in 10 μL MS solvent.
4. Analyze peptides using solvents A and B over the desired linear
gradient using liquid chromatography tandem mass
spectrometry.
5. Raw MS files can be processed in parallel using MaxQuant and
data can be searched against the International Protein Index
mouse database using MASCOT Daemon and/or Andromeda
softwares (see Note 20).
4 Notes
1. Make up all solutions and dilutions with ultra-pure water unless

another reagent is required. It is important to work clean and
sterile and all solutions must be made fresh before use.
2. These are the stock solutions and can be stored at
−20 °C. Different companies (ThermoFisher, Cambridge
Isotope Laboratories, Life Technologies, among others)
provide SILAC kits, usually containing only one heavy amino
acid (e.g., Lys6). Although they are suited for classical
SILAC experiments, they normally lack a second heavy
amino acid that is needed for pSILAC experiments. Thus, in

our laboratory we acquire the Lys4 (medium-heavy) and
Lys8 (heavy)reagents for the pSILAC experiments sepa-
rately. This introduces a 4 Da mass shift in the heavy Lys-
labeled peptide as compared to the medium-heavy
Lys-labeled peptide, which allows identification and quanti-
fication of peptides in MS analysis.
3. Because antibiotics can interfere with the miRNA mimic trans-
fection based on the liposome reagent, these should not be
added to the regular DMEM media in which cells are grown
initially.
4. Dialyzed fetal bovine serum should be used in pSILAC experi-
ments to avoid incorporation of unlabeled Lys0 that is nor-
mally present in this medium. Importantly, some cell lines may
display altered growth patterns in SILAC media because dia-
lyzed serum lacks growth factors. Therefore, when optimizing
pSILAC experiments, it is recommended to verify if the cells
under study are affected by the pSILAC media conditions.
5. The cell line should be chosen according to their suitability for
the miRNA(s) of interest. To overexpress a specific miRNA, it
is preferable to choose a cell line that has low endogenous
expression of the same miRNA. In contrast, to inhibit a miRNA
of interest, choose a cell line with high expression of that
miRNA.
6. It is important to optimize the correct dose and incubation
time for the mimic and/or inhibitor chosen.
7. BLOCK-iT™ Alexa Fluor® Red Fluorescent Oligo is a double
stranded RNA oligonucleotide that can provide a good indica-
tion of transfection efficiency (as indicated by a strong fluores-
cence) [15].
8. Precast gels are the best option but it is possible to use home-
made gels if desired. In this case, prepare “clean” stocks of all
gel solutions and store as recommended.
9. Approximately 200 μL will be needed per gel slice. Prepare this
solution fresh on the day of use.
10. Prepare 40 μL of this solution for each gel slice fresh on the
day of use.
11. One dish is for miRNA mimic transfection and the other is for
miRNA negative control transfection. Transfection should
be performed when cells reach 60–80 % confluence after a
24 h growth period. Thus, the starting amount of cells
should be defined for each cell line in use. For undifferenti-
ated 3T3-L1 pre-adipocytes, we use 0.4 × 106 cells in a
10 mL dish and this yields approximately 70 % confluence at
the time of transfection.
12. In our laboratory we run at least four independent biological

replicates.
13. For each biological replicate, a transfection efficiency control
with BLOCK-iTAlexa Fluor Red Fluorescent Oligocan can be
run in parallel. Efficiency can be visually monitored using an
epifluorescence microscope 8 h after transfection and 24 h
after pulse labeling with SILAC media.
14. For each biological sample, it is important to save at least
60 μg of protein for validation studies (e.g., by Western blot
analysis). Based on the final pSILAC analysis, it is possible
to select the proteins showing the most significant changes
in abundance and validate these using commercially avail-
able antibodies.
15. Performing reverse transcription polymerase chain reaction
analysis of the miRNA of the interest is a good way to check
success of the transfection. It is also important to save RNA for
further mRNA expression studies of dysregulated proteins
since targets are usually regulated at both the mRNA and pro-
tein levels by miRNAs. In this case, use High-Capacity cDNA
reverse transcription kit.
16. The gel pieces should shrink during this stage.
17. In this step, the gel pieces should be restored to their original
sizes.
18. Check after 30 min to determine if there is enough solution
covering the gel pieces and, if not, add more. Make a note
of how much buffer was added in total to each set of gel
pieces.
19. This could take 2–3 h.
20. Based on the final list of dysregulated proteins, it is possible to
select potential direct targets. Check if the 3′ un-translated
region of the gene matches the miRNA seed sequence for the
miRNA of interest. While Western blot analyses can be per-
formed for general target validation, luciferase assays can be
used to validate direct targets.
Acknowledgments
DDG was the recipient of a FAPESP Post-Doctoral Fellowship

(BEPE-PD 2014/20380-5) and is funded by the BHF
(PG/14/20/30769. JdAF was the recipient of a FAPESP Doctoral
Fellowship (BEPE-DR 2014/17012-4),TPO was a recipient of a
Visiting Scientist CAPES Science Without Borders Fellowship
(BEX 11766-13-1). SEO is funded by the UK Medical Research
Council (MC_UU_12012/4).
References
1. Bartel DP (2004) MicroRNAs: genomics, bio- 10. Kaller M, Oeljeklaus S, Warscheid B,
genesis, mechanism, and function. Cell Hermeking H (2014) Identification of
116:281–297 microRNA targets by pulsed SILAC. Methods
2. Flynt AS, Lai EC (2008) Biological principles of Mol Biol 1188:327–349
microRNA-mediated regulation: shared themes 11. Ong SE, Mann M (2006) A practical recipe for
amid diversity. Nat Rev Genet 9:831–842 stable isotope labeling by amino acids in cell
3. Pasquinelli AE (2012) MicroRNAs and their culture (SILAC). Nat Protoc 1:2650–2660
targets: recognition, regulation and an emerg- 12. Ong SE, Mann M (2007) Stable isotope
ing reciprocal relationship. Nat Rev Genet labeling by amino acids in cell culture for
13:271–282 quantitative proteomics. Methods Mol Biol
4. Wang X (2014) Composition of seed sequence 359:37–52
is a major determinant of microRNA targeting 13. Bauer KM, Hummon AB (2012) Effects of the
patterns. Bioinformatics 30:1377–1383 miR-143/-145 microRNA cluster on the
5. http://www.targetscan.org colon cancer proteome and transcriptome.
6. http://www.microrna.gr/webServer J Proteome Res 11:4744–4754
7. Doran J, Strauss WM (2007) Bio-informatic 14. Huang TC, Renuse S, Pinto S, Kumar P, Yang
trends for the determination of miRNA-target Y, Chaerkady R et al (2015) Identification of
interactions in mammals. DNA Cell Biol miR-145 targets through an integrated omics
26:353–360 analysis. Mol Biosyst 11:197–207
8. Thomson DW, Bracken CP, Goodall GJ (2011) 15. Selbach M, Schwanhäusser B, Thierfelder N,
Experimental strategies for microRNA target Fang Z, Khanin R, Rajewsky N (2008)
identification. Nucleic Acids Res 39:6845–6853 Widespread changes in protein synthesis
induced by microRNAs. Nature 455:58–63
9. Vinther J, Hedegaard MM, Gardner PP,
Andersen JS, Arctander P (2006) Identification 16. Schwanhäusser B, Gossen M, Dittmar G,
of miRNA targets with stable isotope labeling Selbach M (2009) Global analysis of cellular
by amino acids in cell culture. Nucleic Acids protein translation by pulsed SILAC. Proteomics
Res 34:e107 9:205–2099
Chapter 12
Blood Bio-Sampling Procedures for Multiplex

Biomarkers Studies
Paul C. Guest and Hassan Rahmoune
Abstract
A major challenge in single or panel of biomarker discovery and validation is the inherent biological com-
plexity underlying disease heterogeneity and inconsistent responses to treatment. Moreover, the lack of
standardization in the sampling, processing, and storage of biological fluids such as plasma and serum
disrupts the discovery and validation of blood-based biomarker tests in preclinical and clinical settings.
This chapter presents a reproducible sample collection and handling procedure that aims to enhance ana-
lyte stability and ensure compatibility with the corresponding multiplex biomarker profiling platforms.
The importance of defining bio-sample acquisition and processing, study design, and profiling platform
guidelines for blood-based biomarker measurements is paramount for the success of personalized health-
care strategy and development of companion diagnostics.
Key words Bio-sampling, Bio-processing, Serum, Plasma, Biomarkers
1 Introduction
Multiplex proteomic and metabolomic techniques have increased

in their use in recent years in the ongoing quest of finding novel
biomarkers for life-threatening diseases. In clinical development,
biomarkers discovery, validation, and translating the findings
into a companion diagnostic requires reproducible sample col-
lection and handling procedures. Whenever possible, it is impor-
tant that bio-fluids such as serum or plasma are chosen for these
studies as blood is readily accessible from most study popula-
tions by standard venous puncture techniques. Therefore, these
are likely to result in discovery and development of biomarker
tests with greater clinical utility and even support point-of-care
use [1, 2] and offer a rapid means of analysis in emergency room
situations [3]. However, it is vital that procedures for blood
draw, plasma and serum preparation and storage are standard-
ized to allow successful comparison of results across different
laboratories in validation and product development studies.
161
162 Paul C. Guest and Hassan Rahmoune
Plasma is derived from blood by the addition of an anticoagu-

lant such as EDTA, heparin, or sodium citrate to inhibit the clotting
process, followed by removal of blood cells and cell debris by cen-
trifugation (Fig. 1). This procedure is carried out rapidly and sam-
ples can be aliquoted in low-binding tubes and stored at −80 °C for
several months or even years. Production of serum is different due
to the fact that no anticoagulant is added and the resulting coagu-
lated material, which is comprised mostly of cells, cell debris, and
clotting factors, is removed by the centrifugation step [4]. This
method requires leaving the samples for approximately 90 min to
allow clotting or adding a clot activator to reduce the clotting time,
prior to aliquotting and storage as above. Here, we present stan-
dard protocols for venous puncture along with plasma and serum
preparation and storage techniques, which can be used in multiplex
immunoassay [5], two-dimensional gel electrophoresis [6], tandem
mass spectrometry [7], selected reaction monitoring mass spec-
trometry [8], proton nucleic acid resonance spectroscopy [9], and
many other multiplex biomarker profiling approaches.
It should be stressed that selecting the adequate bio-sampling
and processing methods can influence the results of biomarker
studies. Therefore, compatibility between the preparation and ana-
lytical phases should be established prior to carrying out studies
[10–12]. To ensure uniformity and consistency of bio-fluid pro-
cessing, standard operating procedures (SOPs) are generally set up
and applied uniformly. Specimens such as serum and plasma should
EDTA plasma tube
Plasma
Place on ice for 60 min Centrifuge 4000g ,
30 min, 4oC
Median
cubital
vein
Leave at room Centrifuge 4000g, 5 min,

temperature for 90 min room temperature
Serum
Serum tube
Fig. 1 Schematic diagram showing blood collection for plasma and serum preparation
Blood Bio-Sampling and Processing 163
be prepared and stored according to the SOP and amendments

added and recorded to trace back any discrepancies that might
influence study outcomes.
2 Materials
2.1 Blood Draw 1. Sterile blood draw needles (14–20 gauge).

2. Holder/adapter for use with the collection system (see Note 1).
3. Tourniquet.
4. Alcohol wipes (70 % isopropyl alcohol).
5. Gloves (Latex, rubber or vinyl).
2.2 Serum and/or 1. Evacuated collection tubes (see Note 3).

Plasma Preparation 2. Sterile serological pipettes of appropriate volumes.
(See Note 2)
3. Benchtop centrifuge (see Note 4).
4. Storage tubes or cryovials (see Note 5).
2.3 Serum 1. Tubes with no additive or clotting agent (see Note 6).
Preparation Only 2. Serum Vacutainer tubes with clot activator +/−gel separator
(see Note 7).
2.4 Plasma 1. Tubes with chelating agent +/−gel separator (see Note 9).
Preparation Only (See
Note 8)
3 Methods
3.1 Blood Draw (See 1. Prior to blood draw, record each donor’s demographic details,
Note 10) physiological status, and other metadata in a Clinical laboratory
worksheet approximately as described in Table 1 (see Note 11).
2. Perform venipuncture selecting the most appropriate arm vein
of the participant (see Note 18).
3. Clean the participant’s arm with alcohol in a circular fashion,
beginning at the site and working outward and allow to air dry.
4. Insert the needle at an angle that is 20–30° of the vein, avoid
trauma and excessive probing (see Note 19).
5. Draw 8–10 mL of whole blood for each 4–5 mL of serum or
plasma required
3.2 Serum 1. Collect up to 8 mL of whole blood in serum tubes.

Preparation 2. Immediately after collection, invert the tube 8–10 times.
3. Allow tubes to clot in the vertical position for 90 min at room
temperature (see Note 20).
Table 1
Typical metadata and dietary/lifestyle that may be collected for a biomarker study (see Note 12)
Samples 1. Type of sample collected (e.g., serum and/or plasma) (see Note 13)
2. Date and time of sample withdrawal
3. Time until freezing
4. Date of collection (see Note 14)
5. Additives to sample (see Note 15)
Demographics 1. Gender
2. Ethnicity (see Note 16)
3. Height, weight and body mass index (BMI)
4. Hip and waist measurement
5. Fasting status
6. Smoking status (number cigarettes/day, duration
7. Alcohol consumption
8. Hormonal status (e.g., menstruation, menopause, hormonal treatment)
9. Pregnancy or breastfeeding
10. Disease onset, duration and current medication (including dosage)
11. Co-morbidities (e.g., presence of respiratory/cardiovascular disease,
diabetes, osteoarthritis, chronic inflammatory diseases, cancer, mental
diseases) including disease duration and regular medications taken
Physiological/ 1. Blood pressure diastolic/ systolic (mm Hg)
biochemical analysis 2. Blood clinical laboratory readouts such as blood count, urea, creatinine,
glucose, lipids
3. Total protein, albumin
4. Sodium, potassium
5. Liver enzymes
6. Glucose tolerance test
7. C-reactive protein
Dietary and lifestyle 1. Type of food and beverages consumed prior to sample collection
2. If possible, participants should fast overnight before specimen
withdrawal (see Note 17)
3. If applicable, record medications type and dose
4. Participants should refrain from heavy exercise, alcohol, tobacco and
nicotine use for 12 h prior to specimen collection
5. Use and levels (when possible) of illicit drugs should be recorded
4. Centrifuge at 1100 × g for 15 min at 4 °C.

5. Transfer 0.5 mL aliquots of the top serum layer to pre-labeled
1.5 mL-capacity Eppendorf Lobind tubes on ice (see Note 5).
6. Record and discard samples that are hemolyzed (red or pink
tingeing) or those that show lipemia (floating milky white sub-
stance formed by the accumulation of lipoprotein particles).
7. Place aliquots immediately on dry ice and transfer to a −80 °C
freezer until analysis (see Note 21).
3.3 Plasma 1. Collect up to 10 mL of whole blood in plasma tubes.

Preparation 2. Immediately after collection, invert the tube 8–10 times.
3. Place tubes on wet ice for 30 min (see Note 22).
4. Centrifuge at 1100 × g for 15 min at 4 °C.

5. Transfer 0.5 mL aliquots of top plasma layer to pre-labeled 1.5 mL-
capacity Eppendorf Lobind tubes on ice (see Notes 5 and 23).
6. Record and discard samples that are hemolyzed or show lipe-
mia as above.
7. Place aliquots immediately on dry ice and transfer to a −80 °C
freezer until analysis as described above for serum.
4 Notes
1. Evacuated systems are available for use with a syringe, single

draw, or butterfly system.
2. Blood proteome-based biomarker profiles can be influenced by
the choice of serum versus plasma. This is due to differences in
both content and stability of the resident molecules [4].
3. Tube type with and without stabilizer should be determined at
the start of the study and kept in constant use throughout.
This is also important for comparison of results across studies.
For example, Hab et al reported marked differences in analyte
measurements using immunoassays of EDTA-, heparin- and
citrate-plasma, and serum [13].
4. Centrifugation at room versus refrigerated temperature can
affect the stability of blood biomarkers (e.g., 4 °C is better
suited for platelet preparation).
5. For volumes less than 1 mL, 1.5 mL-capacity Eppendorf Lo
Bind tubes can be used. For volumes greater than 1 mL, cryo-
vials can be used. It is important to pre-label all tubes.
6. We normally use 10 mL Vacutainer Plus tubes with a clear cap.
These can be purchased with or without a gel separator. We
suggest not to use tubes with a gel separator as this can inter-
fere with some assays (e.g., for Pharmacokinetic and pharma-
codynamic investigations).
7. We normally use the 10 mL BD Vacutainer Plus plastic serum
tubes with either a red or mottled red/gray cap.
8. The choice of chelating agent may be important as this could
affect performance of the chosen biomarker profiling platform
(e.g., the use of EDTA tubes leads to more reproducible
results compared to citrate or heparin tubes with the Luminex
multiplex immunoassay system). We suggest contacting the
manufacturer of the intended platform for advice in the mat-
ter prior to initiation of the study. It should also be noted that
protease inhibitors are not generally needed when storing
plasma in ETDA tubes as this reagent can inhibit most prote-
ases through chelation of metal ion co-factors, required for
protease activity.
9. We normally use 10 mL Vacutainer K2EDTA tubes from BD

Bioscience or equivalent. Tubes are also available containing
sodium citrate or sodium heparin. The choice is dependent on
compatibility of the chelating reagent with the biomarker pro-
filing platform, as described above.
10. It is important to remember that safety always comes first. All
bio-samples and materials should be handled as if capable of
transmitting infection and disposed of with proper precaution
in accordance with state and local regulations. It is important
to avoid contact of bio-samples with skin and mucous mem-
branes. Clinical sample disposals are usually performed in
accordance with the local guidance and rules for the safe use
and disposal at containment level 2 or level 3 as necessary.
11. Consistency is important. Ensure that all measurements are
carried out using the same systems and operators if possible. In
addition, attempt to collect all data from all participants to
avoid missing information. This is important to allow correc-
tions for potential confounding factors during the data analysis
phase. Up to 46 % of laboratory errors are known to be associ-
ated with pre-analytic processing [14] and can even generate
false readings of blood biomarker levels [15].
12. The median cubital vein is used most commonly due to ease of
location and size (Fig. 1).
13. This stage should be performed by a trained phlebotomist.
14. Clotting time can be reduced to approximately 30 min if tubes
containing a clotting factor are used. However, these tubes
should be used consistently throughout the study.
15. There can be problems using temperatures ranging from
−20 °C to −30 °C, such as increased degradation or cryo-
precipitation of the molecular content, and this can affect bio-
marker profiling results [16–19]. Problems can also occur with
multiple freeze-thaw cycles so these should be minimized.
Freezers should be monitored by an automated security alarm
and back-up systems, such as spare freezers, should be in place
for emergencies. Upon first thaw, protease inhibitors can be
added as needed and if these do not interfere with the intended
biomarker profiling platform. Using additives such protease
and phosphatase inhibitors may help to stabilize proteome
profiles [19].
16. Times can vary but the main point is to keep this consistent.
17. When acquiring the plasma, take care not to contact the mono-
nuclear cells and platelets, which occur in a white-colored layer
just under the plasma layer.
18. This is a general guideline and the information required may
vary from study to study. The important point is consistency.
19. Collect samples from test and control cases randomly to avoid
biased statistical outcome.
20. Seasonal effects can sometimes alter the results of biomarker
testing.
21. Additives can also affect profiling results.
22. Participant distribution should be equally represented to facili-
tate successful validation phases or repeat studies at alternate
sites.
23. If participants have not fasted it is important to record this
along with the time/nature of the last meal and potentially
take measurements of glucose and insulin levels.
References
1. Guest PC, Guest FL, Martins-de Souza D mics for biomarker discovery. Angew Chem Int
(2015) Making sense of blood-based pro- Ed Engl 49:5426–5445
teomics and metabolomics in psychiatric 10. Beltran A, Suarez M, Rodríguez MA, Vinaixa
research. Int J Neuropsychopharmacol M, Samino S, Arola L et al (2012) Assessment
pii:pyv138. doi:10.1093/ijnp/pyv138 of compatibility between extraction methods
2. Guest FL, Guest PC, Martins-de-Souza D for NMR- and LC/MS-based metabolomics.
(2016) The emergence of point-of-care blood- Anal Chem 84:5838–5844
based biomarker testing for psychiatric disor- 11. López E, Madero L, López-Pascual J, Latterich
ders: enabling personalized medicine. Biomark M (2012) Clinical proteomics and OMICS
Med 10:431–443 clues useful in translational medicine research.
3. Singhal N, Saha A (2014) Bedside biomarkers Proc Natl Acad Sci U S A 10:35.
in pediatric cardio renal injuries in emergency. doi:10.1186/1477-5956-10-35
Int J Crit Illn Inj Sci 4:238–246 12. Hebels DG, Georgiadis P, Keun HC, Athersuch
4. Alsaif M, Guest PC, Schwarz E, Reif A, Kittel- TJ, Vineis P, Vermeulen R et al (2013)
Schneider S, Spain M et al (2012) Analysis of Performance in omics analyses of blood samples in
serum and plasma identifies differences in long-term storage: opportunities for the exploita-
molecular coverage, measurement variability, tion of existing biobanks in environmental health
and candidate biomarker selection. Proteomics research. Environ Health Perspect 121:480–487
Clin Appl 6:297–303 13. Haab BB, Geierstanger BH, Michailidis G,
5. Fulton RJ, McDade RL, Smith PL, Kienker LJ, Vitzthum F et al (2005) Immunoassay and
Kettman JR Jr (1997) Advanced multiplexed antibody microarray analysis of the HUPO
analysis with the flowmetrix system. Clin Chem Plasma Proteome Project reference specimens:
43:1749–1756 systematic variation between sample types and
6. Unlü M, Morgan ME, Minden JS (1997) calibration of mass spectrometry data.
Difference gel electrophoresis: a single gel Proteomics 5:3278–3291
method for detecting changes in protein 14. Becan-McBride K (1999) Laboratory sam-
extracts. Electrophoresis 18:2071–2077 pling: does the process affect the outcome?
7. Paulo JA, Kadiyala V, Banks PA, Steen H, J Intraven Nurs 22:137–142
Conwell DL (2012) Mass spectrometry-based 15. Bowen RA, Hortin GL, Csako G, Otañez OH,
proteomics for translational research: a techni- Remaley AT (2010) Impact of blood collection
cal overview. Yale J Biol Med 85:59–73 devices on clinical chemistry assays. Clin
8. Ji QC, Rodila R, Gage EM, El-Shourbagy TA Biochem 43:4–25
(2003) A strategy of plasma protein quantita- 16. Zander J, Bruegel M, KleinhempelA BS, Petros
tion by selective reaction monitoring of an S, Kortz L et al (2014) Effect of biobanking
intact protein. Anal Chem 75:7008–7014 conditions on short-term stability of biomark-
9. Griffiths WJ, Koal T, Wang Y, Kohl M, Enot ers in human serum and plasma. Clin Chem
DP, Deigner HP (2010) Targeted metabolo- Lab Med 52:629–639
17. Kang HJ, Jeon SY, Park JS, Yun JY, Kil HN, serum/plasma proteome profiling. Proteomics
Hong WK et al (2013) Identification of clini- 6:3189–3198
cal biomarkers for pre-analytical quality con- 19. Rai AJ, Gelfand CA, Haywood BC, Warunek
trol of blood samples. Biopreserv Biobank DJ, Yi J, Schuchard MD et al (2005) HUPO
11:94–100 plasma proteome project specimen collection
18. Hsieh SY, Chen RK, Pan YH, Lee HL (2006) and handling: towards the standardization of
Systematical evaluation of the effects of sample parameters for plasma proteome samples.
collection procedures on low-molecular-weight Proteomics 5:3262–3277
Chapter 13
Multiplex Immunoassay Profiling

Laurie Stephen
Abstract
Multiplex immunoassays allow for the rapid profiling of biomarker proteins in biological fluids, using less
sample and labor than single immunoassays. This chapter details the methods to develop and manufacture
multiplex assays for the Luminex® platform. Although assay development is not included here, the same
methods can be used to covalently couple antibodies to the Luminex beads and to label antibodies for the
screening of sandwich pairs, if needed. The assay optimization, detection of cross-reactivity, and minimiz-
ing antibody interactions and matrix interferences will be addressed.
Key words Disease, Multiplex assay, Antibody, Biomarker, Luminex® assay
1 Introduction
A large majority of immunoassays rely on antibodies to capture and

detect the analyte of interest in a biological matrix. Traditional
immunoassays detect the presence of a single analyte and most rely
on enzyme-driven detections. New technologies have been devel-
oped that allow multiple analytes to be measured simultaneously
on a single sample, all in a single reaction vessel. The method
described here is for the development of a multiplex assay for the
Luminex® system, although the same principal will apply to other
technologies (Fig. 1). For information on multiplexing technolo-
gies, there are several recent publications that review the latest
developments [1–4].
As with any new technology, there are unique advantages and
disadvantages that the user encounters. For example, the ability to
simultaneously measure multiple analytes in a single sample maxi-
mizes the amount of information that can be obtained from single
sample, reduces laboratory analysis time and sample volume
requirement, and provides cost savings. However, multiplexed
assays also present unique challenges for the user that would not be
encountered if single assays were used for each individual analyte.
Examples of these include different detection ranges, the potential
169
170 Laurie Stephen
Anti-target 1 Anti-target 2 Anti-target 3
Beads filled with red and infrared

dyes at different ratios - fluoresce
at different wavelengths when
Specific antibodies excited with a laser
coupled to distinct beads
Add bead
mixture
Quantity Identity
Target 1
Target 5
Target 4
Add Add biotinylated Read in
sample anti-body detector
DETECTOR
Fig. 1 Overview of multiplex immunoassay technique. Samples are added to dye-coded microbead-antibody
conjugates that capture specific targets. Following incubation with a second antibody containing a biotinylated
label to form a “sandwich” configuration, the mixtures are streamed through the Luminex instrument that uses
lasers for the identification of the antibody-microbead conjugates and quantitation of the bound molecules.
The example shows a triplex assay capable of binding targets 1, 2, and 3. Since the sample only contains
target 1, this is the only target bound and quantified
for cross-reactivity, increased matrix interference, and potential for

false positives due to antibody interactions. These challenges, if not
carefully addressed, can generate misleading results. Several papers
have addressed the issue of increased interference in multiplex
immunoassays [5–7] and others have described the additional chal-
lenges involved in the validation stage [8–10]. This chapter will
describe a method to develop a multiplex immunoassay and strate-
gies for minimizing interference and interactions within the assay.
2 Materials
2.1 Bead 1. Magnetic separator.

Conjugation 2. Sonicating bath.
3. Copolymer tubes and labels.
4. 1–4 mL of magnetic 12.5 × 106 beads/mL Luminex micro-
spheres/beads.
5. 125 μg/mL capture antibodies (see Note 1).
6. Sulfo-NHS, N-hydroxysulfosuccinimide.
7. N-(3-dimethylaminopropyl)-N′-ethylcarbodiimide (EDC).
8. Activation buffer: 100 mM NaH2PO4 (pH 6.0).
9. Coupling buffer: 0.05 M 2-morpholino-ethane-sulfonic acid
mono-hydrate (MES) (pH 5.0).
Multiplex Immunoassay Profiling 171
10. Blocking/storage buffer: 10 mM NaH2PO4 (pH 7.4). 150 mM

NaCl, 0.02 % Tween 20, 0.1 % bovine serum albumin (BSA),
and 0.05 % NaN3 or Proclin (PBS-TBN).
2.2 Biotinylation 1. Sulfo-NHS-LC biotin (ThermoFisher Scientific; Waltham,

of Detection Antibody MA, USA).
2. Dialysis unit.
3. Phosphate-buffered saline (PBS) (pH 7.4).
2.3 Multiplex 1. Assay buffer: PBS, 1 % BSA.

Development 2. Wash buffer: PBS, 0.02 % Tween-20.
3. 100 μg/mL Streptavidin, R-phycoerythrin (SAPE).
4. 96-well plate.
5. Plate washer or magnetic separator for 96-well plate.
6. Recombinant protein standards for each assay in the multiplex.
7. Representative serum and /or plasma samples from rheuma-
toid arthritis (RA) patients or with known rheumatoid factor
and controls.
8. Animals sera (e.g., horse, goat, mouse, rabbit, donkey, fetal calf).
9. Hetorophilic blocking reagents (such as IIR from
Bioreclamation; Tru-Block from Meridian Life Sciences or
HBR from Scantibodies).
3 Methods
3.1 Bead 1. Allow 1–4 mL of stock microspheres, to settle by placing vials

Conjugation upright on a flat magnetic separator for 2 min (see Note 2).
2. Taking care not to disturb settled beads remove and discard
0.8 mL of buffer for every 1 mL stock beads.
3. Transfer and pool the remaining volume into a single copoly-
mer tube.
4. Place tube in the magnetic separator and allow separation to
occur for 30–60 s.
5. With tube in separator remove and discard supernatant.
6. Add 0.5 mL of activation buffer, resuspend with vortexing and
sonication, and place in the magnetic separator for 30–60 s.
7. With tube in separator remove and discard supernatant, and
resuspend in 0.4 mL of activation buffer with vortexing and
sonication.
8. Add activation buffer to Sulfo-NHS for a final concentration of
50 mg/mL and add 50 μL of this to the tube and vortex.
172 Laurie Stephen
9. Add activation buffer to EDC for a final concentration of

10 mg/mL and add 50 μL of this to the tube and vortex.
10. Incubate 20 min in the dark while rotating at room tempera-
ture and prepare the protein during the incubation.
11. Place tube in a magnetic separator for 30–60 s, remove the
supernatant, and add 0.5 mL of coupling buffer.
12. Repeat for a total of two washes and resuspend in 0.45 mL of
coupling buffer.
13. Add 0.2 mL of capture antibody to the activated microspheres
with immediate vortexing (see Note 2).
14. Incubate for 2 h in the dark while rotating at room
temperature.
15. Place tube in a magnetic separator for 30–60 s, remove super-
natant, and add 1.0 mL of blocking/storage buffer.
16. Resuspend with vortexing and sonication and incubate for
30 min in the dark with rotation at room temperature.
17. Place tube in magnetic separator for 30–60 s, remove superna-
tant, and wash twice with 0.25 mL blocking/storage buffer.
18. Resuspend in 0.25 mL blocking/storage buffer, count the
beads with a hemocytometer, adjust to 50 × 106 beads/mL,
and store at 2–8 °C.
3.2 Biotinylation 1. Immediately before use, prepare a 10 mM solution of the bio-

tin reagent.
2. Add 10 mM biotin reagent solution to the antibody solution
at a 20:1 biotin:antibody molar ratio (see Note 3).
3. Incubate reaction on ice for two hours or at room temperature
for 30 min.
4. Remove excess biotin by dialysis in PBS using a minimum of
three buffer exchanges.
5. Add BSA to a final concentration of 1 % and preservative for
long-term stability.
3.3 General Protocol 1. Create a capture bead mini-pool by adding 5 μL of each bead
solution to a final volume of 1.4 mL in assay buffer.
2. Make an 8 standard (S8) mini-pool by adding 0.2 μg of each
recombinant protein to a final volume of 0.2 mL in assay buffer
and do seven 10-fold serial dilutions to create a standard curve.
3. Make a mini-pool mix of detection antibodies by adding 5 μg
of each biotinylated antibody to a final volume of 5 mL in assay
buffer.
4. Produce 1:5, 1:10, 1:100, and 1:1000 serial dilutions of serum
and plasma samples in assay buffer (see Note 4).
5. Add 30 μL of standard or sample to a well of the 96-well plate.
6. Add 10 μL of blocking solution and then add 10 μL of capture

beads.
7. Incubate the plate for 1 h on a plate shaker at room
temperature.
8. Wash three times with 100 μL wash buffer, add 40 μL of the
detection mini-pool to each well, and incubate the plate for
1 h on a plate shaker at room temperature.
9. Add 20 μL SAPE to each well plate and mix for 30 min on a
plate shaker at room temperature (see Note 5).
10. Wash three times with 100 μL wash buffer and add 100 μL
assay buffer.
11. Incubate the plate for 2–5 min on a plate shaker at room tem-
perature and then analyze on the Luminex 100 Analyzer.
3.4 Curve Opti- 1. Examine the standard curve median fluorescence intensities
mization (See Note 6) (MFI) and choose the best four points for each that cover the
range of your sample signals.
2. Make a new standard mini-pool based on the above (Table 1)
(see Note 7).
Table 1
Screen of samples and animal sera
Name Analyte 1 Analyte 2 Analyte 3 Analyte 4

S8 6581 18605 24370 8669
S7 2965 8044 19134 3009
S6 1242 3301 9323 1128
S5 528 1276 2855 430
S4 250 554 805 185
S3 123 257 265 104
S2 86 143 155 78
S1 76 109 121 67
Serum 1 1:5 2894 181 281 224
Serum 1 1:10 1127 80 197 121
Rabbit serum 2689 67 154 145
Goat serum 116 59 116 195
Hamster serum 89 79 138 265
Mouse serum 82 71 112 49
Rat serum 330 71 289 23
Donkey serum 30 65 80 35
174 Laurie Stephen
3. By examining where the samples fall on the curve, several mini-

pools of detections can be made to determine the lowest con-
centration of each antibody needed (see Note 8)
3.5 Blocker Opti- 1. Make various blocker formulations with and without 1 % don-
mization (See Note 9) key serum and include assay buffer as one formulation.
2. Run samples (including RA samples) at two dilutions with
each blocker to determine the optimal blocker formulation.
3. Calculate the sample linearity and choose the blocker that gives
linearity values 80–120 % and the highest sample signals.
3.6 Assessing 1. Make a multiplex standard mini-pool along with single mini-
Cross-Reactivity pools of each standard at the same concentrations as those in
the multiplex.
2. Make a detection mini-pool with single mini-pools of each detec-
tion antibody at the same concentrations as in the multiplex.
3. In order to assess standard cross-reactivity, run the assay with
multiplex beads and compare the results with each single assay
standard curve (see Note 10).
4. In order to assess detection cross-reactivity, run the assay with
multiplex beads and standards, along with several positive
serum or plasma samples and test each single detection pool.
5. Calculate any signals and compare to the multiplex curve to
calculate cross-reactivity (see Note 11).
3.7 Packaging 1. Batch bead and detection mixes: package individually in assay
and Use buffer using the volumes listed in the general protocol above.
2. Batch blocker and S8: package in standard diluent.
3. Standard diluent, assay buffer, SAPE and blocker: package
individually.
4. Store beads, SAPE and assay buffer at 4 °C, and all other com-
ponents at −80 °C (see Note 12).
5. Make two to three levels of assay quality controls (QC) by
spiking recombinant protein into serum samples and package
these individually and store at −80 °C.
6. Proceed with assay validation and the generation of assay
acceptance criteria by testing a minimum of 20 of each QC in
duplicate over a minimum of 3 days.
4 Notes
1. Prior to coupling or biotinylation of the antibodies, they should

be free of any amines and other proteins. If using a Tris-based
buffer, dialyze the antibody into 1× PBS using a minimum of
three buffer exchanges. If the antibody preparation contains sta-

bilizer proteins such as BSA or gelatin, purify with Protein A or
Protein G columns, followed by dialysis.
2. 1 mL of microspheres is sufficient for ~ 40 plates of assays.
3. The molecular weight of immunoglobulin G (IgG) is
150,000 kDa, so the amount of biotin required for a 1 mL of
a 1 mg/mL antibody solution could be calculated as follows:
1 mL IgG × 1 mg/mL × 20 mmol biotin/1 mmol IgG × 1 mmol
IgG/150,000 mg IgG × 1000 ul/mL = 0.133 mmol bio-
tin = 13 μL of the 10 mM biotin solution.
4. The assay will be multiplex based on the dilutions requirement
of the samples. In general, if the expected levels are in the pg/
mL range, they will be at the same dilution; however, it is
unlikely that an analyte in the pg/mL range will have the same
dilution requirement as an analyte in the μg/mL range.
5. The SAPE concentration will vary with the number of analytes
in the multiplex.
6. For optimization of the curve, repeat the assay with the indi-
cated changes.
7. In the example in Table 1, the sample dilution was determined
to be 1:5. In order to match the matrix and minimize curve
background, the standard curve should be made up in 20 %
donkey serum.
8. The detection concentration for Analyte 1 can likely be low-
ered, whereas that of Analyte 2 should be increased.
9. Repeat the assay with optimized detection and standard con-
centrations using the indicated changes.
10. If cross-reactivity is greater than 10 %, a different recombinant
standard may be needed.
11. If cross-reactivity is greater than 10 %, the detection in ques-
tion can be titrated down and the assay repeated. If there is no
improvement, the assay may need to be removed from the
multiplex.
12. The detection buffer can also be stored at 4 °C if space is
limiting.
References
1. Tighe P, Negm O, Todd I, Fairclough L 3. Spindel S, Sapsford KE (2014) Evaluation of
(2013) Utility, reliability and reproducibility of optical detection platforms for multiplexed
immunoassay multiplex kits. Methods detection of proteins and the need for point-
61:23–29 of-care biosensors for clinical use sensors.
2. Ellington AA, Kullo IJ, Bailey RK, Klee GG Sensors (Basel) 14:22313–22341
(2010) Antibody-based protein multiplex plat- 4. Marquette CA, Corgier BP, Blum LJ (2012)
forms: technical and operational challenges. Recent advances in multiplex immunoassays.
Clin Chem 56:186–193 Bioanalysis 4:927–936
176 Laurie Stephen
5. Todd DJ, Knowlton N, Amato M, Frank MB, 8. Jani D, Allinson J, Berisha F, Cowan KJ,
Schur PH, Izmailova ES et al (2011) Erroneous Devanarayan V, Gleason C et al (2016)
augmentation of multiplex assay measurements Recommendations for use and fit-for-purpose
in patients with rheumatoid arthritis due to validation of biomarker multiplex ligand binding
heterophilic binding by serum rheumatoid fac- assays in drug development. AAPS J 18:1–14
tor. Arthritis Rheum 63:894–903 9. Bastarache JA, Koyama Y, Wickersham NE,
6. Fraser S, Soderstrom C (2014) Due diligence Ware LB (2014) Validation of a multiplex elec-
in the characterization of matrix effects in a trochemiluminescent immunoassay platform in
total IL-13 Singulex™ method. Bioanalysis human and mouse samples. J Immunol
6:1123–1129 Methods 408:13–23
7. Krika LJ (1999) Human anti-animal antibody 10. Tighe PJ, Ryder RR, Todd I, Fairclough LC
interferences in immunological assays. Clin (2015) ELISA in the multiplex era: potentials
Chem 45:942–956 and pitfalls. Proteomics Clin Appl 9:1862–8354
www.Ebook777.com
Chapter 14
Multiplex Sequential Immunoprecipitation of Insulin

Secretory Granule Proteins from Radiolabeled Pancreatic
Islets
Paul C. Guest
Abstract
Pulse radiolabeling of cells with radioactive amino acids is a common method for tracking the biosynthesis
of proteins. Specific proteins can then be immunoprecipitated and analyzed by electrophoresis and imag-
ing techniques. This chapter presents a protocol for the biosynthetic labeling of pancreatic islets with
35
S-methionine, followed by multiplex sequential immunoprecipitation of insulin and three other secretory
granule accessory proteins. This provided a means of distinguishing those pancreatic islet proteins with
different biosynthetic rates in response to the media glucose concentrations.
Key words Pulse radiolabeling, Immunoprecipitation, Electrophoresis, Pancreatic islets, Insulin,

Secretory granule proteins
1 Introduction
Pulse radiolabeling of cells followed by immunoprecipitation is a

means of looking at the biosynthesis of targeted proteins [1, 2]. In
this method, a radiolabeled amino acid is usually added to the
medium so this can be incorporated into nascent proteins as they
are being synthesized. Then the newly synthesized proteins can be
immunoprecipitated by direct or indirect means and analyzed at any
time, thereby providing a means of tracking their fate inside the cell.
Direct immunoprecipitations are carried out by covalently linking
an antibody of interest to a solid matrix such as cyanogen bromide
(CNBr)-activated Sepharose [3, 4]. In this approach, antibodies are
coupled directly to the resin through primary amines. Indirect
immunoprecipitations are carried out by non-covalently binding
the antibody to an affinity-based matrix such as Protein A Sepharose.
Then in both direct and indirect protocols, cell lysates can be incu-
bated with the resulting immunoadsorbents for binding, elution,
and subsequent analysis of the target proteins.
177
178 Paul C. Guest
This chapter describes a 20 min pulse labeling of pancreatic

islets with 3%S-methionine and the sequential immunoprecipita-
tion of the hormone insulin and the secretory granule accessory
proteins chromogranin A, prohormone convertase (PC)1 and
PC2, as described in previous studies [5–7]. Enzymological analy-
ses have shown that production of mature insulin requires cleavage
of proinsulin by PC1 and PC2 on the carboxy-terminal side of
Arg31-Arg32 and Lys64-Arg65, respectively [8]. PC1 has a pH opti-
mum of 5.5 and requires mM calcion ion concentration, which
coincides with the environment inside insulin secretory granules
[2]. This fits with finding that final conversion of proinsulin to
insulin does not begin to occur until approximately 30 min after
initial synthesis on the rough endoplasmic reticulum (Fig. 1).
Thus, the 20 min labeling period employed here helps to ensure
that insulin is still present mostly in its precursor form, making
Biosynthesis
of insulin
Nucleus
Time
0 minutes
Rough endoplasmic
reculum
Golgi complex
20 minutes
30 minutes
Trans Golgi network
Insulin secretory
granules
60 minutes Constuve
secreon
(not
regulated)
Regulated secretion
Fig. 1 Diagram of a pancreatic islet cell showing the biosynthesis of insulin and transport through the regu-
lated secretory pathway
Multiplex Analysis of Radiolabelled Islet Proteins 179
quantitative studies more direct. Here, the preparation of immu-

noadsorbents, pulse radiolabeling of islets, immunoprecipitation,
and gel-based analyses is presented.
2 Materials
1. Purified monoclonal antibody for insulin (see Note 1).

2. Polyconal antisera for chromogranin A, PC1 and PC2
(see Note 2).
3. CNBr-Activated Sepharose 4 (GE Healthcare; Little Chalfont,
Bucks, UK).
4. Activation solution: 1 mM HCl.
5. Coupling buffer: 100 mM NaHCO3 (pH 8.3) and 500 mM
NaCl.
6. Quenching buffer: 100 mM Tris–HCl (pH 8.0).
7. Wash buffer 1: 100 mM Tris–HCl (pH 8.0) and 500 mM
NaCl.
8. Wash buffer 2: 100 mM NaOAc (pH 4.0) and 500 mM NaCl.
9. Storage buffer: 10 mM NaH2PO4, 2 mM K2HPO4, 137 mM
NaCl, 2.7 mM KCl.
10. Modified Kreb’s bicarbonate buffer: 25 mM NaHCO3
(pH 7.4), 115 mM NaCl, 5.9 mM KCl, 1.2 mM MgCl2,
1.2 mM NaH2PO4, 1.2 mM Na2SO4, 2.5 mM CaCl2, and
0.1 % bovine serum albumin.
11. Islet lysis buffer: 25 mM Na2B4O7 (pH 9), 3 % BSA, 1 % Tween-
20, 1 mM phenylmethanesulfonyl fluoride, 0.1 mM E-64,
1 mM EDTA, and 0.1 % NaN3 (see Note 3).
12. Post-immunoprecipitation wash buffer: 50 mM Tris-HCl
(pH 7.5), containing 150 mM NaCl, 1 % Triton X- 100, 1 %
deoxycholate, 0.1 % SDS, and 5 mM-EDTA.
13. Insulin elution reagent: 25 % acetic acid (see Note 4).
14. Protein A Sepharose (see Note 5).
15. Protein A Sepharose rehydration buffer: 20 mM NaH2PO4
(pH 8.0) and 150 mM NaCl.
16. Reagent for elution of other islet proteins: 20 mM HCl.
35
17. S-methionine.
18. Alkaline-urea electrophoresis gel: polymerized from 7.5 %
acrylamide and 0.20 NN′-methylenebisacrylamide, containing
12.5 mM Tris/80 mM glycine (pH 8.6) and 8 M urea.
19. Alkaline-urea gel tank buffer: 12.5 mM-Tris/80 mM glycine
(pH 8.6).
20. Alkaline urea gel loading buffer: 2.5 mMTris–HCl (pH 8.6),
8 M urea, and 0.001 % Bromophenol Blue.
180 Paul C. Guest
21. SDS polyacrylamide electrophoresis: gels polymerized from

15 % acrylamide and 0.080 NN′-methylenebisacrylamide in a
Tris-glycine buffer system using the discontinuous buffer sys-
tem of Laemmli [9].
22. Fluorography solution: 20 % 2,5-diphenyloxazole in acetic
acid.
23. MSE Sonifier (Crawley, UK) with a microprobe (see Note 6).
3 Methods
3.1 Preparation 1. Dialyze the antibody into two changes of coupling buffer over
of Anti-Insulin 6 h at 4 °C (see Note 7).
Immunoadsorbent 2. Measure the absorbance at 280 nm of the final antibody solu-
tion and calculate the concentration (see Note 8).
3. Add 20 mL ice-cold activation solution to 1 g of dried resin
and gently mix for 2 h at 4 °C (see Note 9).
4. Centrifuge the resin at 1000 × g for 5 min and remove the
supernatant.
5. Add the dialyzed antibody to the resin at a concentration of
2 mg antibody/mL swollen resin and mix overnight at 4 °C.
6. Centrifuge the resin at 1000 × g for 5 min and remove the
supernatant (see Note 10).
7. Add 20 mL of coupling buffer to the resin and gently mix for
30 min at room temperature.
8. Centrifuge at 1000 × g for 5 min and remove the supernatant.
9. Add 20 mL of quenching buffer and gently mix for 2 h at
room temperature.
10. Centrifuge at 1000 × g for 5 min and remove the supernatant.
11. Suspend the resin in 20 mL wash buffer 1, centrifuge at
1000 × g for 5 min and remove the supernatant.
12. Suspend the resin in 20 mL wash buffer 1, centrifuge at
1000 × g for 5 min and remove the supernatant.
13. Repeat steps 11 and 12 twice and suspend the resin in 20 mL
of storage buffer.
14. Centrifuge at 1000 × g for 5 min, remove the supernatant and
suspend the resulting immunoadsorbent in 8 mL of storage
buffer, and store at 4 °C (see Note 11).
3.2 Preparation 1. Add 20 mL of rehydration buffer to 4 g of Protein A Sepharose

of Immunoadsorbent powder and leave gently rocking for 30 min.
for Accessory 2. Centrifuge at 1000 × g for 5 min in a swinging bucket rotor,
Secretory Granule remove the supernatant, and suspend in 20 mL of the same
Proteins buffer.
3. Repeat the centrifugation and washing steps twice and store in

16 mL of rehydration buffer at 4 °C (see Note 11).
4. Add 200 μL of suspended Protein A Sepharose slurry to micro-
centrifuge tubes containing 15 mL of each polyclonal antiserum
and incubate with gentle rocking overnight at 4 °C (see Note 12).
5. Centrifuge at 1000 × g for 5 min, remove supernatant, and
wash three times by centrifugation and resuspension in 500 mL
of the rehydration buffer.
6. Centrifuge at 1000 × g for 5 min and remove the supernatant,
leaving a packed gel of approximately 50 μL for immediate use.
3.3 Biosynthetic 1. Preincubate 100 isolated islets per experimental condition for
Radiolabeling 40 min in 100 μL of Kreb’s bicarbonate buffer containing 2.6
of Pancreatic Islets or 16.7 mM glucose at 37 °C in microcentrifuge tubes under
95 % O2/5 % CO2 (see Note 13).
2. Recover the islets by centrifugation at 100 × g for 10 s and
resuspend in 100 μL of the same pre-warmed medium con-
taining 150 μCi of 35S-methionine and incubate for 20 min at
37 °C under 95 % O2/5 % CO2 (Fig. 2).
3. Recover the islets by centrifugation at 100 × g for 10 s, carefully
remove the radioactive supernatant, and gently resuspend the
islets in 500 μL of the same ice-cold buffer containing 2 mM
methionine (see Note 14).
4. Repeat this process twice and place the tubes containing the
islet pellets on dry ice.
5. Add 200 μL lysis buffer and sonicate for 15 s at approximately
25 W (see Note 15).
6. Centrifuge the lysates at 13,000 × g for 5 min and retain the
supernatants for immunoprecipitation.
3.4 Immunopre- 1. Incubate islet lysates for 1 h at room temperature in microcen-

cipitation of Insulin trifuge tubes with 50 μL of a100 mg/mL suspension of
and Other Pancreatic Cowan-strain Staphylococcus aureus cells (see Note 16).
Islet Proteins 2. Centrifuge the samples at 13,000 × g for 5 min and retain the
supernatants.
Preincubaon Pulse labeling

2.8 mM glucose 2.8 mM glucose

0 10 20 30 40 50 60
Time (minutes)
Fig. 2 Schematic diagram showing the 35S-methionine pulse labeling protocol in

either low (2.8 mM) or high (16.7 mM) glucose concentrations
182 Paul C. Guest
3. Carrying out immunoprecipitation of insulin-related mole-

cules by adding the lysates to the 50 μL packed gel of respec-
tive immunoadsorbent and incubate overnight at 4 °C.
4. Centrifuge the anti-insulin immunoadsorbent in a swinging
bucket rotor at 500 × g for 1 min and retain the supernatant for
the next immunoprecipitation.
5. Wash the anti-insulin immunoadsorbent by repeated centrifu-
gation and resuspension in 4 × 1 mL of lysis buffer, 2 × 1 mL of
immunoadsorbant wash buffer, and then with 2 × 1 mL of dis-
tilled water.
6. Elute with 2 × 1 mL of 25 % acetic acid, freeze dry and recon-
stitute in 50 mL of alkaline-urea gel loading buffer.
7. Incubate the supernatant obtained after immunoprecipita-
tion of the insulin-related molecules overnight at 4 °C
with 50 mL packed gel of anti-chromgranin A
immunoadsorbent.
8. Wash the anti-chromogranin A immunoadsorbent as above
and retain the supernatant for the next immunoprecipitation.
9. Elute the chromogranin A-related peptides with 2 × 100 μL of
20 mM HCl.
10. Combine the two eluates, freeze dry and reconstitute in 50 μL
of 125 mM Tris–HCl pH 6.8, containing 2 % SDS, 0.25 M
sucrose, 5 mM-EDTA, 65 mM-dithiothreitol, and 0.005 %
Bromophenol Blue.
11. Repeat Subheading 3.4, step 8 through Subheading 3.4, step
10 for immunoprecipitation of PC1 and PC2.
3.5 Electrophoresis 1. Prerun alkaline-urea gels in tank buffer for 600 V-hours,
and Fluorography replace the upper tank buffer, and load the insulin immunopre-
of 3 %S-Labeled cipitates in alkaline-urea gel loading buffer and subject to elec-
Immunoprecipitates trophoresis for 1000 V- hours (see Note 17).
2. Disassemble the gel plates and shake the gels for 2 × 5 min in
acetic acid, 2 h in fluorography solution, and then leave for
30 min under cold running water (see Note 18).
3. Vacuum dry and expose the gel to Cronex 4 X-ray film
(Dupont; Stevenage, Herts, UK) for 6 to 72 h (see Note
19).
4. For the other islet proteins, heat the samples at 95 °C for 5 min
and electrophorese to the point where the dye front just
reaches the bottom of the gel.
5. Perform fluorography as above for 3–14 days (Fig. 3 )
(see Note 20).
Proinsulin Chromogranin A PC1 PC2
2.8 16.7 2.8 16.7 2.8 16.7 2.8 16.7
Fig. 3 Immunoprecipitation of proinsulin, chromogranin A, PC1, and PC2 from 35S-methionine-labeled islets.
The concentrations of glucose (mM) used for the labeling are shown on the top of each track on the images
(see Note 21)
4 Notes
1. We used the clone 3B7 that recognizes epitopes in the rat pro-
insulin structure [8].
2. For chromogranin A, antiserum was raised in guinea pigs to a
beta-galactosidase fusion protein incorporating amino acids
60-234 of the rat chromogranin A sequence as described by
Hutton et al [5]. For PC1 and PC2, antisera were raised in rab-
bits against glutathione S-transferase fusion proteins incorpo-
rating amino acids 111-137 and 162-388 of rat PC1 and PC2
respectively. These fusion proteins were produced using the
bacterial expression vector as described by Bennett et al [6].
3. Large amounts of BSA are added to prevent loss of protein on
tube walls during immunoprecipitation. Insulin is known to be
a “sticky” molecule.
4. Insulin elution requires highly acetic conditions due to the
high affinity of the monoclonal antibody and the poor solubil-
ity of the insulin molecule.
5. To be used in indirect immunoprecipitation of the pancreatic
islet proteins chromogranin A, PC1, and PC2.
6. Any sonication probe device can be used but the probe should
fit inside a 1.5 mL-capacity microcentrifuge tube and therefore
be not more than 2 mm in diameter at the tip.
7. It is important to remove all traces of Tris buffer from the anti-
body because this contains primary amines and will therefore
react with the activated resin.
8. The ideal concentration is 2 mg antibody/mL swollen resin.
9. 1 g of dried resin will swell to give a volume of approximately
4 mL.
10. This should be saved in case the coupling did not work. This
can be measured by reading the optical density at 280 nm in a
184 Paul C. Guest
spectrophotometer and by looking for the loss of the mono-

clonal antibody from the solution. A good coupling efficiency
would be greater than 80 %.
11. Store for up to 1 month if not using preservatives.
12. In step, the immunoglobulin fraction in the serum is bound to
the resin.
13. We have used isolated rat islets in this study although rat pan-
creatic beta cell lines can also be used. If this is the case, 5 × 105
cells would be approximately equivalent to 100 islets as each
islet contains around 5000 cells.
14. Use appropriate precautions when handling and disposing of
the radioactive materials. The addition of ice-cold medium
containing nonradioactive methionine stops the uptake of
35
S-methionine and halts metabolic activity of the islet cells.
15. Adjust power setting accordingly using other sonication probe
devices.
16. This is to preclear the supernatants by removing any immuno-
globulin like molecules that could interfere with immunopre-
cipitation experiments.
17. This is equivalent to approximately 1 and 3/4 dye front
lengths. We stop the electrophoresis after the dye has reached
the bottom of the gel and then add new dye to a blank well at
the end of the gel and restart the electrophoresis for the
remaining three fourths gel length run.
18. The gel turns white during this final step.
19. Other films can be used but check the manufacturer’s specifica-
tions. In addition, obtaining the best exposure may require a
few attempts and adjusting the times accordingly.
20. Longer exposure periods may be necessary as these islet pro-
teins are much lower in abundance compared to insulin.
However, the exposure period should not exceed the half life
of 35S, which is 87 days.
21. The biosynthesis of proinsulin, chromogranin A, and PC1 was
stimulated 10-30 fold at the higher glucose concentration,
consistent with previous studies [4, 7, 10].
References
1. Shanmugam G, Vecchio G, Attardi D, Green 3. Houwen B, Goudeau A, Dankert J (1975)

M (1972) Immunological studies on viral poly- Isolation of hepatitis B surface antigen
peptide synthesis in cells replicating murine (HBsAg) by affinity chromatography on
sarcoma-leukemia virus. J Virol 10:447–455 antibody-coated immunoadsorbents.
2. Hutton JC (1994) Insulin secretory granule J Immunol Methods 8:185–194
biogenesis and the proinsulin-processing endo- 4. Guest PC, Rhodes CJ, Hutton JC (1989)
peptidases. Diabetologia 37(Suppl 2):S48–S56 Regulation of the biosynthesis of insulin secre-
tory granule proteins: co-ordinate translational processing in the diabetic Goto-Kakizaki rat.
control is exerted on some, but not all, granule J Endocrinol 175:637–647
matrix constituents. Biochem J 257:432–437 8. Davidson HW, Rhodes CJ, Hutton JC (1988)
5. Hutton JC, Peshavaria M, Johnston CF, Intraorganellar calcium and pH control proin-
Ravazzola M, Orci L (1988) Immunolocalization sulin cleavage in the pancreatic beta cell via two
of betagranin: a chromogranin A-related pro- distinct site-specific endopeptidases. Nature
tein of the pancreatic B-cell. Endocrinology 333:93–96
122:1014–1020 9. Laemmli UK (1970) Cleavage of structural
6. Bennett DL, Bailyes EM, Nielsen E, Guest PC, proteins during the assembly of the head of
Rutherford NG, Arden SD et al (1992) bacteriophage T4. Nature 227:680–685
Identification of the type 2 proinsulin processing 10. Guest PC, Arden SD, Bennett DL, Clark A,
endopeptidase as PC2, a member of the eukaryote Rutherford NG, Hutton JC (1992) The post-
subtilisin family. J Biol Chem 267:15229–15236 translational processing and intracellular sort-
7. Guest PC, Abdel-Halim SM, Gross DJ, Clark ing of PC2 in the islets of Langerhans. J Biol
A, Poitout V, Amaria R et al (2002) Proinsulin Chem 267:22401–22406
Chapter 15
Two Dimensional Gel Electrophoresis of Insulin Secretory

Granule Proteins from Biosynthetically-Labeled Pancreatic
Islets
Paul C. Guest
Abstract
Pulse-chase radiolabeling of cells with radioactive amino acids is a common method for tracking the bio-
synthesis of proteins. Radiolabeled newly synthesized proteins can be analyzed by a number of techniques
such as two dimensional gel electrophoresis (2DE). This chapter presents a protocol for the biosynthetic
labeling of pancreatic islets with 35S-methionine in the presence of basal and stimulatory concentrations of
glucose, followed by subcellular fractionation to produce a secretory granule fraction and analysis of the
granule protein contents by 2DE. This provides a means of determining whether or not the biosynthetic
rates of the entire granule constituents are coordinately regulated.
Key words Pulse-chase radiolabeling, Pancreatic islets, Subcellular fractionation, Insulin secretory
granules, 2D electrophoresis
1 Introduction
The secretory granules of pancreatic B cells perform a specialized

function in the packaging, storage, and secretion of the hormone
insulin (Fig. 1) [1]. In addition to insulin and the connecting (C-)
peptide, these subcellular organelles contain more than 150 poly-
peptides [2]. This includes the proteases involved in proinsulin-to-
insulin conversion, proinsulin conversion intermediates, other
polypeptide precursor proteins, minor cosecreted peptides, mem-
brane proteins involved in cell trafficking, and ion translocating
proteins involved in regulation of the intragranular environment.
Most of these proteins are likely to be synthesized, transported,
and packaged into nascent granules in a coordinated manner to
ensure correct functioning of the granule. Insulin biosynthesis is
regulated by many circulating nutrients and other factors although
glucose is the most important physiologically [3]. However, it is
187
188 Paul C. Guest
Pancreac Islet Beta Cell Insulin Granule
Fig. 1 Images of (a) pancreatic islet as seen through a light microscope, (b) pancreatic beta cell visualized by
electron microscopy, and (c) insulin secretory granule obtained by high power electron microscopy. Each islet
contains around 5000 cells of which 70–80 % are comprised of the insulin-producing pancreatic beta cells. The
granules contain a dense core of insulin hexamers which comprise around 80 % of the protein mass and this is
surrounded by less dense material containing approximately 150 different granule accessory proteins [1]
not known whether just some or all of the granule constituents are
affected in a similar manner.
This chapter addresses this question by two dimensional gel
electrophoresis (2DE) [4] of secretory granule subcellular frac-
tions prepared from 35S-methionine-labeled rat islets, as described
by Guest and coworkers [5]. The protocol employed is a 1 h
pulse labeling of pancreatic islets with 35S-methionine in the
presence of either low or high glucose concentrations, followed
by a chase period of 3 h in nonradioactive medium containing a
low glucose concentration (Fig. 2). The production of mature
insulin requires cleavage of proinsulin by the endoproteases pro-
hormone convertase 1 (PC1) and prohormone convertase 2
(PC2) on the carboxy-terminal side of Arg31-Arg32 and Lys64-
Arg65, respectively, followed by the removal of the exposed basic
residues by the exopeptidase carboxypeptidase H [6, 7]. This
conversion is optimal in the low pH and high Ca2+ environment
in the late trans Golgi network and secretory granule compart-
ments [1]. This fits with finding that final conversion of proinsu-
lin to insulin does not begin to occur until approximately 30 min
after initial synthesis on the rough endoplasmic reticulum and
transport to these compartments. Thus, the 1 h pulse labeling
and 3 h chase employed here ensures that insulin and most of the
other secretory proteins have had sufficient time to reach the
granule compartment. In addition, the chase under low glucose
conditions minimizes secretion of these newly synthesized pro-
teins, thereby ensuring that they are retained within the gran-
ules. Here, the pulse chase radiolabeling of islets, subcellular
fractionation, and 2DE analyses are presented.
Multiplex Analysis of Secretory Granule Biogenesis 189
Pulse labelling Chase

0 1 2 3 4
Time (hours)
Fig. 2 Schematic diagram showing the 35S-methionine pulse chase labeling

protocol
2 Materials
1. 400 rat pancreatic islets per condition (see Note 1).

2. Post-nuclear fraction from 0.5 g of rat insulinoma tissue
(see Note 2).
3. High glucose modified Kreb’s bicarbonate buffer: 25 mM
NaHCO3 (pH 7.4), 115 mM NaCl, 5.9 mM KCl, 1.2 mM
MgCl2, 1.2 mM NaH2PO4, 1.2 mM Na2SO4, 2.5 mM CaCl2,
16.7 mM glucose, and 0.1 % bovine serum albumin.
4. Low glucose modified Kreb’s bicarbonate buffer: 25 mM
NaHCO3 (pH 7.4), 115 mM NaCl, 5.9 mM KCl, 1.2 mM
MgCl2, 1.2 mM NaH2PO4, 1.2 mM Na2SO4, 2.5 mM CaCl2,
2.8 mM glucose, and 0.1 % bovine serum albumin.
5. Low glucose chase incubation buffer: Dulbecco’s modified
Eagle’s medium, containing 10 % newborn calf serum and
2.8 mM-glucose.
35
6. S-methionine (see Note 3).
7. Islet homogenization medium: 10 mM KMes (pH 6.5) con-
taining 0.3 M sucrose, 1 mM MgSO4, and 1 mM EGTA.
8. 4.4, 8.8, and 17.7 % Nycodenz (Nyegaard Diagnostica; Oslo,
Norway) in homogenization medium.
9. Subcellular fractionation wash buffer: 10 mM KMes (pH 6.5)
containing 0.25 M sucrose.
10. Isoelectric focussing buffer: 9.5 M urea, 5 % 2-mercaptoethanol,
0.40 % pH 3–10 range Pharmalytes (Pharmacia Fine
Chemicals; Uppsala, Sweden), 1.6 % pH 5–7 range Ampholines
(Pharmacia LKB Biotechnology, Bromma, Sweden), and
0.001 % Bromophenol Blue.
11. Isoelectric focussing tube gel buffer: 8 M urea, 20 % pH 3–10
range Pharmalyte and 1 % each of pH 3.5–5.0, 5–7, and 7–9
range Ampholines.
12. Isoelectric focussing lower tank buffer: 0.085 % phosphoric acid.
190 Paul C. Guest
13. Isoelectric focussing upper tank buffer: 0.2 N NaOH.

14. Second dimension equilibration buffer: 50 mM Tris–Cl,
pH 6.8, 6 M urea, 30 % (v/v) glycerol, 2 % (w/v) SDS, 0.01 %
Bromophenol Blue.
15. Second dimension tank buffer: 25 mM Tris/192 mM glycine
(pH 8.3) and 0.1 % SDS.
16. Fluorography solution: 20 % 2,5-diphenyloxazole in acetic
acid.
17. Nunc Cryotubes (Gibco; Paisley, Scotland, UK).
18. 1 mL capacity glass tube homogenizer (see Note 4).
19. 1.2 cm × 5.0 cm polypropylene centrifuge tubes (Beckman
Instruments; Palo Alto, CA, USA) (see Note 5).
20. Swinging bucket rotor (see Note 6).
21. MSE Sonifier (Crawley, UK) with a microprobe (see Note 7).
22. Vertical isoelectric focussing tube gel system using
15 cm × 0.15 cm inner diameter glass tubes (see Note 8).
23. Second dimension gels cast in 15 cm × 15 cm × 0.15 cm glass
plates: linear 5–20 % acrylamide gradient, containing
0.065 % N,N′-methylenebisacrylamide, 0.375 M Tris–HCl
pH 8.8, and 0.2 % SDS (see Note 9).
24. Second dimension electrophoresis tank for running gels cast in
15 cm × 15 cm × 0.15 cm glass plates (see Note 10).
3 Methods
3.1 Biosynthetic 1. Preincubate 400 isolated islets for 60 min in 500 μL of modi-
Radiolabeling fied Kreb’s bicarbonate buffer containing 2.6 or 16.7 mM glu-
of Pancreatic Islets cose at 37 °C in Cryotubes under 95 % O2/5 % CO2.
2. Recover the islets by centrifugation at 100 × g for 10 s and sus-
pend in 200 μL of the same pre-warmed medium containing
200 μCi of 35S-methionine and incubate for 1 h at 37 °C in
Cryotubes under 95 % O2/5 % CO2 (Fig. 2) (see Note 11).
3. Recover the islets by centrifugation at 100 × g for 10 s, carefully
remove the radioactive supernatant, and gently suspend the
islets in 500 μL of chase buffer and incubate for 3 h under 95 %
O2/5 % CO2 (see Note 12).
4. Terminate the incubations by the addition of 1 mL of ice-cold
Kreb’s low glucose incubation medium followed by centrifu-
gation for 10 s at 3300 × g in a swinging bucket rotor and dis-
card the media
5. Wash the islet pellets by two further cycles of resuspension and
centrifugation as above and subject immediately to subcellular
fraction (see Note 13).
3.2 Subcellular 1. Combine radiolabeled slets with the insulinoma cells and
Fractionation homogenize in the glass tube homogenizer using ten strokes
of a Teflon pestle driven at 600 rpm at 4 °C (see Note 14).
2. Centrifuge the homogenates for 5 min at 1700 × g at 4 °C to
remove unbroken cells and nuclei.
3. Transfer the supernatants to the 1.2 × 5 cm centrifuge tubes
and add 1.3 mL portions of each of the 4.4, 8.8, and 17.7 %
Nycodenz solutions (in that order), loading from the bottom
(Fig. 3) (see Note 15).
4. Centrifuge the gradients for 1 h at 100,000 × g in the swinging
bucket rotor.
5. Collect the material at the 8.8/17.7 % Nycodenz interface and sus-
pend this in subcellular fractionation wash buffer (see Note 16).
6. Centrifuge for 20 min at 50,000 × g in the swinging bucket
rotor.
7. Resuspend the particulate material containing enriched secre-
tory granules in the subcellular fractionation wash buffer, cen-
trifuge again, and store the final pellets at −80 °C prior to 2DE
analysis.
3.3 Two Dimensional 1. Homogenize the secretory granule pellets by sonication in

Gel Electrophoresis 100 μL of isoelectric focussing buffer.
of 35S-Labeled 2. Centrifuge at 13,000 × g to remove the particulate material.
Secretory Granule
3. Subject the samples to 2DE analysis by isoelectric focusing in
Proteins the first dimension and SDS polyacrylamide gel electrophoresis
in the second dimension (see Note 17).
4. For isoelectric focussing, prefocus the gels in the assembled
tube gel apparatus for 1 h at 200 V.
5. Add up to 50 μL of sample to the top of each tube using a
Hamilton syringe and carry out isoelectric focussing at 800 V
for approximately 6 h until the Bromophenol Blue has reached
the bottom of the tube (see Note 18).
Sample
4.4 4.4
100,000 x g
% Nycodenz
1 hour
8.8 8.8
Layer containing
secretory granules
17.7 17.7
Fig. 3 Schematic diagram showing subcellular fractionation protocol

192 Paul C. Guest

200- 200-
98- 98-
Molecular weight (kDa)
68- 68-
43- 43-
26- 26-
18- 18-
14- 14-
5 6 7 5 6 7
pH pH
Fig. 4 Islets were labeled with 35S- methionine for 1 h in either 2.8 mM glucose or 16.7 mM-glucose and then
incubated for a further 3 h in nonradioactive medium containing 2.8 mM glucose. Subcellular fractions were
prepared and those enriched in secretory granules were subjected to 2DE followed by fluorography (28 days
exposure). The image shows that the biosynthesis of the most of the proteins was stimulated 10–30 fold at the
higher glucose concentration. Previous pulse chase labeling and immunoprecipitation studies have shown that
this is known to occur for the secretory granule proteins insulin [8], chromogranin A [8], secretory granule
membrane protein 110 [9], and PC1 [10]
6. Using a syringe of the appropriate diameter, apply pressure to

the top of the glass tubes to extrude the gels into the second
dimension equilibration buffer (see Note 19).
7. Incubate the gels for 5 min in the equilibration buffer and then
load on to the top of the second dimension gel using a 0.15 cm
thick spacer to nudge the gels into position (see Note 20).
8. Carry out electrophoresis at 60 V for 1 h followed by 120 V
until the dye front reaches the bottom of the gel.
9. Disassemble the gel plates and immerse the gels for 2 × 5 min
in acetic acid, 2 h in fluorography solution, and then leave for
30 min under cold running water (see Note 21).
10. Vacuum dry and expose the gel to Cronex 4 X-ray film
(Dupont; Stevenage, Herts, UK) for 2–8 days (Fig. 4)
(see Note 22).
4 Notes
1. Rat islets were obtained from 10 to 12-week-old New England

Deaconess Hospital rats by a collagenase digestion technique,
as described by Guest et al [8]. However, other protocols can
be used, providing that these can yield the required number of
large intact islets.
2. This was prepared as described by Hutton et al [2] for combi-

nation with radiolabeled rat islets prior to homogenization and
density gradient centrifugation to facilitate efficient recovery.
Other rat beta cell lines can be used as a substitute.
3. Use appropriate precautions when handling and disposing of
radioactive materials.
4. We used a 1 mL capacity glass tube homogenizer and Teflon
pestle from Jencons Scientific (Leighton Buzzard, Beds, UK)
although similar products available from other suppliers would
work just as well.
5. Other tube and swinging bucket rotor combinations can be
used but ensure these have appropriate specifications.
6. We used a Beckman SW 50.1 rotor. Other swinging bucket
rotors can be use but ensure that these are an adequate match
for the centrifuge tubes.
7. Any sonication probe device can be used but the probe should
fit inside a 1.5 mL-capacity microcentrifuge tube and therefore
be not more than 2 mm in diameter at the tip.
8. Although we used a homemade device for this, tube gel sys-
tems are available from many suppliers. In addition, it is likely
that the instruments designed for use of immobilized pH gra-
dient strips could also be employed. Keep in mind to designate
the chosen device for use with radioactive materials.
9. The gels used for this study were cast in the lab. However,
precast 5–20 % gradient gels are also available from many
suppliers.
10. Again, many systems could be used for this although the size
should be compatible with the isoelectric focussing stage in
terms of tube gel or strip length.
11. Use recommended precautions and dispose of radioactive
material appropriately.
12. The addition of the chase medium containing nonradioactive
amino acids stops the uptake of 35S-methionine and halts meta-
bolic activity of the islet cells.
13. Do not freeze the islets as this will disrupt cellular and intracel-
lular membranes, which will disrupt the subcellular fraction-
ation step.
14. The clearance of the glass tube and pestles allows for disrup-
tion of the cell membranes but leaves most intracellular organ-
elles, such as the secretory granules, intact.
15. The under-layering approach helps to form sharp boundaries
between layers, which may be more difficult to achieve loading
using the over-layering method.
194 Paul C. Guest
16. All fractions and the pellet should be collected for further anal-
yses but we are only presenting the results using the layer most
enriched in secretor granules [5].
17. We used the method as described by Anderson et al [11].
18. This is a continuous isoelectric focussing system as it does not
use an immobilized pH gradient. Care should be taken not to
run samples into the lower tank buffer. As a useful guide,
Bromophenol Blue will turn green at pH 4 and yellow at pH 3
and the run should be timed to terminate when a green-yellow
band reaches the bottom of the tube gel. However, it is best to
carry out a time course study to determine optimum running
time when you are running a new kind of sample.
19. A 200 μL pipette tip works well for this.
20. Be careful not to damage the tube gel and ensure that there are
no air bubbles between the tube gel and the second dimension
gel.
21. The gel turns white during this final step.
22. Other films can be used. Please check the manufacturer’s speci-
fications. Obtaining the best exposure may require multiple
attempts and adjusting the times accordingly.
References
1. Hutton JC (1994) Insulin secretory granule 7. Bennett DL, Bailyes EM, Nielsen E, Guest PC,
biogenesis and the proinsulin-processing Rutherford NG, Arden SD et al (1992)
endopeptidases. Diabetologia 37(Suppl 2):S48– Identification of the type 2 proinsulin process-
S56 ing endopeptidase as PC2, a member of the
2. Hutton JC, Penn EJ, Peshavaria M (1982) eukaryote subtilisin family. J Biol Chem
Isolation and characterisation of insulin secre- 267:15229–15336
tory granules from a rat islet cell tumour. 8. Guest PC, Rhodes CJ, Hutton JC (1989)
Diabetologia 23:365–373 Regulation of the biosynthesis of insulin secre-
3. Hedeskov CJ (1980) Mechanism of glucose- tory granule proteins: co-ordinate translational
induced insulin secretion. Physiol Rev control is exerted on some, but not all, granule
60:42–509 matrix constituents. Biochem J 257:432–437
4. O’Farrell PH (1975) High resolution two- 9. Hutton JC, Bailyes EM, Rhodes CJ, Rutherford
dimensional electrophoresis of proteins. J Biol NG, Arden SD, Guest PC (1990) Biosynthesis
Chem 250:4007–4021 and storage of insulin. Biochem Soc Trans
5. Guest PC, Bailyes EM, Rutherford NG, 18:122–124
Hutton JC (1991) Insulin secretory granule 10. Guest PC, Abdel-Halim SM, Gross DJ, Clark
biogenesis. Co-ordinate regulation of the bio- A, Poitout V, Amaria R et al (2002) Proinsulin
synthesis of the majority of constituent pro- processing in the diabetic Goto-Kakizaki rat.
teins. Biochem J 274:73–78 J Endocrinol 175:637–647
6. Davidson HW, Rhodes CJ, Hutton JC (1988) 11. Anderson NG, Anderson NL, Tollaksen SL
Intraorganellar calcium and pH control proin- (1979) Operation of the isodalt system, publi-
sulin cleavage in the pancreatic beta cell via two cation ANL-BIM-79-2, division of biological
distinct site-specific endopeptidases. Nature and medical research. Argonne National
333:93–96 Laboratory, Argonne, IL, USA
Chapter 16
Depletion of Highly Abundant Proteins of the Human

Blood Plasma: Applications in Proteomics Studies
of Psychiatric Disorders
Sheila Garcia, Paulo A. Baldasso, Paul C. Guest,
Abstract
Psychiatric disorders are complex diseases involving exogenous and endogenous factors. Biomarkers for
diagnosis or prediction of successful treatment are not existent. In addition, the molecular basis of these
diseases is still poorly understood. Blood plasma represents the most complex proteome as it contains
subproteomes from several body tissues. However, the high abundance of some little proteins can obscure
the analysis of hundreds of low abundance proteins, which are potential biomarkers. Therefore, removal of
these high abundance proteins is pivotal in any proteomic study of plasma. Here, we present a method of
depleting these proteins using immunoaffinity liquid chromatography.
Key words Plasma, Protein depletion, Immunoaffinity chromatography, Neuropsychiatric disorders,

Plasma biomarker, Neuroproteomics
1 Introduction
Psychiatric disorders are heterogeneous diseases involving genetic

and environmental components. Due to this complexity, it is diffi-
cult to understand the molecular basis of these diseases and, choose
an adequate treatment [1]. The misdiagnosis rate is high among
patients with these disorders, being as high as 40 % for schizophre-
nia and 69 % for bipolar disorder [2, 3]. This occurs due to the
poor understanding of the biochemical pathways involved in these
diseases [4, 5]. In addition, the number of non-responders to
treatment is high. In schizophrenia, for example, the proportion of
patients who fail to respond adequately is 25–40 % and the treat-
ment dropout rate is around 33 % [6, 7].
Another complicating factor in the majority of psychiatric
disorders is that the choice of drug used for treatment is random
and based on patient clinical history and/or the choice of the
physician [8]. In addition, treatment can sometimes lead to strong
195
196 Sheila Garcia et al.
side effects. Therefore, it is essential to find biomarkers that

improve treatment outcomes [9] using blood proteome profiling
methods such as mass spectrometry [10].
Research has shown that some proteins present in blood of
individuals with schizophrenia are altered in abundance compared
to healthy individuals. Some examples are the cytokines and inter-
leukins associated with the inflammation response and circulating
hormones such as insulin, cortisol, and follicle-stimulating hor-
mone (FSH) that can affect brain function [11, 12]. This suggests
that although psychiatric disorders appear in the brain, the effects
can be observed in other parts of the body including the peripheral
blood system [5, 13, 14].
Body fluids have been used in molecular biology studies of
psychiatric disorders for more than a decade. Serum and plasma are
easier and more accessible to work with than solid tissues such as
the brain and their use allows the follow-up of patients during the
disease course or following treatment [15]. However, most of the
blood proteins are present only at low levels, suggesting the need
for more sensitive technologies [15]. Approximately 99 % of blood
proteins are represented by 30 highly abundant proteins, with con-
centrations spanning at least 12 orders of magnitude. Human
serum albumin (HSA) and immunoglobulin G (IgG) are among
the highly abundant proteins and together comprise 70 % of the
total blood protein mass [14, 16]. This is a problem as the current
dynamic range of most mass spectrometry instruments is lower
than five orders of magnitude. Furthermore, many of the highly
abundant proteins obscure those of lower abundance. As a conse-
quence, the direct analysis of crude plasma or serum is not possible
until these high abundance proteins are depleted [17]. Thus,
methods of fractionation, such as immunoaffinity chromatography,
are often used to deplete the levels of these abundant proteins
prior to analysis [16, 18].
Here, we present a protocol using the commercially available
Multiple Affinity Removal System (MARS®; Agilent Technologies;
Santa Clara, CA, USA) that can be used in conjunction with most
high-performance liquid chromatography (HPLC) systems. The
column matrix contains polyclonal antibodies that target the 14
most abundant proteins in plasma or serum. This method can be
applied prior to most proteomic investigations of serum or plasma
to enrich the low abundance proteins.
2 Materials
1. Human plasma or serum samples.

2. Buffer solution A: 0.2 M sodium phosphate pH 7.4, 0.5 M
sodium chloride, 0.02 % sodium azide, store for up to 30 days
after preparation.
Immunoaffinity Depletion of Plasma 197
3. Buffer solution B: 2 M Urea and 0.5 M glycine (pH 2.25),

store for up to 30 days after preparation.
5. HPLC system (we use a Waters 2487 Dual λ Absorbance Detector)
and a manual injection system using a Hamilton syringe.
6. Agilent MARS Human 14 column (4.6 mm inner diameter,
length 100 mm).
7. Concentration centrifuge tubes with 3000 Da molecular
weight cutoff and 6 mL capacity.
8. Microcentrifuge and large centrifuge rotor for up to 15 mL
capacity centrifuge tubes.
9. Eppendorf LoBind tubes 0.5 mL and other 1.5 mL capacity
microcentrifuge tubes as needed.
10. Protein assay quantitation equipment and reagents.
11. SDS-PAGE gel equipment.
12. Gel reagents: acrylamide (30 %), sodium dodecyl sulfate (SDS),
dithiothreitol (DTT), Tris-Base, chlorine hydroxide (HCl),
bromophenol blue, glycine, commasie blue dye, methanol,
ammonium persulfate, tetramethylenediamine (TEMED).
13. Sample buffer: 10 % sodium dodecyl sulfate (SDS), 10 mM
dithiothreitol (DTT), 20 % glycerol, 0.2 M Tris–HCl (pH 6.8),
and 0.05 % bromophenol blue.
14. Stock solution of running buffer (10×): 250 mM Tris–HCl,
2 M glycine, and 10 % of SDS.
1. Dilute 30 μL of plasma or serum four times with buffer solu-

tion A in a 1.5 mL tube (see Note 2).
2. Centrifuge at 21,000 × g for 15 min, then recover the superna-
tant without disrupting the pellet (see Note 3).
3. Transfer the supernatant to a new 0.5 mL LoBind tube and
proceed with step 9 or keep the sample on ice if performing
the experiment on the same day (see Note 4).
4. With the pumps connected in a bottle containing water,
purge HPLC lines at a flow rate of 1 mL/min for 5 min
each (see Note 5).
5. With the pumps connected in buffers solution A and B, purge
HPLC lines with buffer solution A at a flow rate of 1 mL⁄min
for 5 min each.
6. Connect the MARS human 14 column in the HPLC system
and make sure to drip test when installing each extremity of
the column to avoid air bubbles in the system (see Note 6).
www.Ebook777.com
Table 1
LC wash and equilibration conditions
Time % Buffer B Flow rate (mL/min) Max pressure Curve

0.00 100 1.00 60 6
15.00 100 1.00 60 6
16.00 0 1.00 60 6
31.00 0 1.00 60 6
32.00 0 0.2 60 6
Table 2
LC method
Flow rate Max

Phase Time % Buffer B (mL/min) pressure Curve
Low abundant fraction 0.00 0 0.125 60 6
elution 18.00 0 0.125 60 6
Washing 18.01 0 1.0 60 6
20.00 0 1.0 60 6
High abundant 20.01 100 1.0 60 6
fraction elution 27.00 100 1.0 60 6
Column regeneration 27.01 0 1.0 60 6
38.00 0 1.0 60 6
7. Set up the LC method and run a wash method to equilibrate

the column (Table 1) (see Note 7).
8. Inject ~120 μL of prepared sample into HPLC sample loop
(see Note 8) and set up the LC method below as shown in
Table 2.
9. Set up to initiate the depletion run method (see Note 9).
10. Collect the fractions containing low abundant proteins that
should elute within the first 2.25 mL/18 min of the method
and then collect the fractions containing high abundant
proteins that should elute between 7.25 and 11.25 mL/23–
27 min (the high abundance proteins will bind to the antibodies
on the column matrix) (Fig. 1) (see Note 10).
11. Collect low and high abundance protein fractions for buffer
exchange into sodium bicarbonate and storage at −80 °C
(see Note 11).
12. Wash the concentration centrifuge tubes using a wash bottle
containing MilliQ water.
2.8
2.4
2.0
AU (280nm)
1.6
1.2
0.8
0.4
0
2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36
Minutes
Fig. 1 Representative chromatogram of depletion runs on the MARS Hu14 col-

umn. The first peak contains the low-abundance proteins (flowthrough fraction),
the second peak contains the high abundance proteins (bound fraction), and the
last peak is due to the buffer solution
13. Add 3 mL of 50 mM of ammonium bicarbonate buffer solution

(natural pH 7.5) and centrifuge at 6600 × g for 40 min or until
remaining volume is 300–600 μL and discard the remaining
volume.
14. Add the low abundance protein fractions (~1.5 mL) to these
tubes. (see Note 12).
15. Centrifuge at 6600 × g for approximately 30 min or until the
remaining volume is 300–600 μL.
16. Complete the volume of tubes to 6 mL and pipette up and
down to wash proteins off the filter membrane.
17. Centrifuge at 6600 × g for 90 min or until remaining volume is
300–600 μL.
18. Pipette up and down several times to resuspend proteins and
transfer to new 1.5 mL tubes.
19. Measure protein concentrations in duplicates using a protein
assay.
20. Digest the proteins in each sample using a protease such as
trypsin and/or store at −80 °C.
21. After the last run, wash the column with buffer A in isocratic
mode at 1 mL⁄min over 20 min.
22. Annotate the MARS column usage in a logbook to check col-
umn performance in accordance with the manufacturer’s guar-
antee of 200 runs (see Note 13).
23. Flush the HPLC system with water at 1 mL⁄min over 30 min.
MWM P1 FT1 BF1 P2 FT2 BF2 P3 FT3 BF3

225-
150-
102-
76-
52-
MW (kDa)
38-
31-
24-
17-
12-
Fig. 2 SDS-PAGE of human plasma protein fractions from the MARS Hu14 column.
An equal amount (15 μg) of crude plasma (P), flow-through (FT), and bound
fraction (BF) were separated on 12 % SDS-polyacrylamide gel. The proteins were
stained with Comassie blue R-250 dye. MWM: molecular weight markers
Table 3
Standard concentrations of acrylamide used for resolving proteins of
specific molecular weights
Acrylamide (%) M.W. Range (kDa)

7 50–500
10 20–300
12 10–200
15 3–100
24. To use the MARS Hu14 column more than 200 runs, it is impor-
tant to check column efficiency in each set of sample runs.
25. Prepare or purchase a 12 % sodium dodecylsulfate (SDS)
polyacrylamide gel, load 10–20 μg protein from each fraction,
carry out electrophoresis, and stain proteins with Coomasie
Blue R-250 (Fig. 2; Table 3).
26. If preparing the gel, make 5 mL of stacking gel and 10 mL of
running gel according to Tables 4 and 5, respectively.
27. Pipette the running gel solution between glass plates and wait
approximately 30 min for polymerization and then repeat
this procedure for the stacking gel, which is layered on the top
(see Note 14).
Table 4
Composition of the stacking gel
Component Volume (mL)

0.5 M Tris–HCl, pH 6.8 1.25
10 % (w/v) SDS 0.05
Acrylamide/bis-acrylamide (30 %/0.8 % w/v) 0.67
H2O 2.975
Just before pouring the gels, add:
10 % (w/v) ammonium persulfate (APS) 0.05
TEMED 0.005
Table 5
Composition of the running gel
Acrylamide % 8 10 12 15
1.5 M Tris–HCl, pH 8.8 2.6 mL 2.6 mL 2.6 mL 2.6 mL
10 % (w/v) SDS 0.1 mL 0.1 mL 0.1 mL 0.1 mL
Acrylamide/Bis-acrylamide (30 %/0.8 % w/v) 2.6 mL 3.4 mL 4 mL 5 mL
H2O 4.6 mL 3.8 mL 3.2 mL 2.2 mL
Just before pouring the gels, add:
10 % (w/v) ammonium persulfate (APS) 0.1 mL 0.1 mL 0.1 mL 0.1 mL
TEMED 0.01 mL 0.01 mL 0.01 mL 0.01 mL
28. Insert the well comb without introducing air bubbles and
leave in place until the stacking gel has set completely
( see Note 15).
29. Mix the sample with sample buffer.
30. Heat sample for 5–10 min at 95 °C to denature the proteins.
31. Load a protein molecular weight marker into the first well and
the prepared samples into the others, and load empty channels
with sample buffer using the same volume (see Note 16).
32. Load the gel platform with running buffer.
33. Set a voltage or current to run the electrophoresis.
34. Stop the running gel when the final band of protein marker has
reached the bottom of glass plates.
35. Remove the gel from the plates and stain proteins using
Coomasie Blue R-250 dye for at least 2 h.
36. Discard the dye and destain the gel with 10–20 % methanol or
bleach solution for approximately 2 h, changing this solution
approximately three times (see Note 17).
37. Soak the gel briefly in water prior to imaging (see Note 18).
4 Notes
1. All steps should be carried out at 4 °C except the HPLC

method that should be performed at room temperature.
2. In applying 40 μL of crude plasma, we observed the presence
of HSA in the flow through fraction of one sample. Thus, we
suggest using lower amounts (such as 30 μL) to avoid this.
3. We use a step of centrifugation but recommend using spin car-
tridges to avoid introducing particulate matter on the column.
4. Turn on the HPLC system (pumps, controller, recorder, and col-
lector). Before starting a run, place the LoBind tubes in the collec-
tor and set up a method according to manufacturer’s protocol.
5. The buffer solutions do not need degasing. If the HPLC is
purged with ethanol or isopropanol, it is necessary to wash
the HPLC system with water to avoid salt precipitation within
the system.
6. We use the MARS 14 column in a medium temperature room
(17 °C) but the manufacturer suggests room temperature use
(21 °C).
7. This allows the complete elution of the residual matter within
the column.
8. We use a loop of 120–500 μL depending on sample size as the
loop volume should be at least twice the volume of sample.
Before starting each run, wash the sample loop three times the
loop volume.
9. The Waters 2487 Dual λ Absorbance Detector HPLC system
requires a manual injection with a Hamilton syringe and the
sample must be injected within the holder with the sample
valve at load position. After injection, leave the syringe in the
holder and switch the sample valve to the inject position when
starting the run. This step requires care to prevent air bubbles
within sample loop. To set up the LC method and start the run
we used the software Empower Pro 2 (Waters Corporation® v.
6.0). Before removal of the syringe at the end of a run, turn
the sample valve back to the load position. Also, it is important
to wash the syringe at least five times with buffer A between
each run to avoid cross contamination between samples.
10. Use 1.5 mL LoBind tubes for fraction collection to minimize
sample loss. We use 38 LoBind tubes made of nontoxic
polypropylene, nuclease-free and pyrogen free (Axygen®).
11. Store the high abundance protein fraction at −80 °C after

buffer exchange for potential future analysis. This desalting
step can be performed using 6 mL centrifugal concentrator
tubes (e.g., Vivaspin® 6 Centrifugal Concentrator).
12. The buffer exchange should be done at 4 °C to minimize pro-
tein degradation.
13. The manufacturer guarantees 200 runs, but we have experience
of more than 2000 runs with the same column without loss of
efficiency. For this, we test the efficiency of the column by SDS
page electrophoresis of fractions used on each set of runs.
14. Use the manufacturer’s protocol as appropriate.
15. Before removing the comb, make sure that the gel has
polymerized.
16. All wells must be loaded to ensure that the proteins run into
the gel with minimal spreading.
17. Observe if the destaining solution is blue and replace it.
18. The gel profile suggests two possible problems. First, there
may be a low efficiency of immunodepletion, as seen by the
presence of high abundance protein bands in the flow-through
fraction of P2 and P3. Second, the profiles of the different
samples may be different due to biological variation, as seen by
differences in the flow-through fractions. We suggest running
each sample twice to help determine which of these possibili-
ties is most likely. Overall, the results show that the MARS
Hu14 column is efficient (calculations not shown).
Acknowledgments
DMS was funded by FAPESP (São Paulo Research Foundation,

grants 2013/08711-3, 2013/25702-8 and 2014/10068-4).
References
1. Nascimento JM, Martins-de-Souza D (2015) (cochrane diagnostic test accuracy review).
The proteome of schizophrenia. NPJ Schizophr Schizophr Bull 41:792–7944
1:14003 4. Cosgrove VE, Suppes T (2013) Informing
2. Hirschfeld RMA, Lewis L, Vornik LA (2003) DSM-5: biological boundaries between bipolar
Perceptions and impact of bipolar disorder: I disorder, schizoaffective disorder, and schizo-
How far have we really come? Results of the phrenia. BMC Med 11:127
National Depressive and Manic-Depressive 5. Guest PC, Guest FL, Martins-de-Souza D
Association 2000 Survey of individuals with (2015) Making sense of blood-based pro-
bipolar disorder. J Clin Psychiatry 64: teomics and metabolomics in psychiatric
161–174 research. Int J Neuropsychopharmacol.
3. Soares-Weiser K, Maayan N, Bergman H, doi:10.1093/ijnp/pyv138
Davenport C, Kirkham AJ, Grabowski S et al 6. Chan MK, Gottschalk MG, Haenisch F,
(2015) First rank symptoms for schizophrenia Tomasik J, Ruland T, Rahmoune H et al
(2014) Applications of blood-based protein alterations in schizophrenia: clinical status

biomarker strategies in the study of psychiatric and antipsychotic effects. Biol Psychiatry 70:
disorders. Prog Neurobiol 122:45–72 663–671
7. Regier D (2012) Merging categorical and 14. Anderson NL, Anderson NG (2002) The
dimensional diagnoses of mental disorders. human plasma proteome: history, character,
Epidemiol Psychiatr Sci 21:267–269 and diagnostic prospects. Mol Cell Proteomics
8. Hyman SE (2007) Can neuroscience be inte- 1:845–867
grated into the DSM-V? Nat Rev Neurosci 15. Jaros JA, Martins-de-Souza D, Rahmoune H,
8:725–732 Schwarz E, Leweke FM, Guest PC et al (2012)
9. Schwarz E, Guest PC, Rahmoune H, Martins- Differential phosphorylation of serum proteins
de-Souza D, Niebuhr DW, Weber NS et al reflecting inflammatory changes in schizophre-
(2012) Identification of a blood-based bio- nia patients. Eur Arch Psychiatry Clin Neurosci
logical signature in subjects with psychiatric 262:453–455
disorders prior to clinical manifestation. World 16. Jaros JA, Guest PC, Bahn S, Martins-de-Souza
J Biol Psychiatry 13:627–632 D (2013) Affinity Depletion of Plasma and
10. Martins-De-Souza D (2012) Proteomics tack- Serum for Mass Spectrometry-Based Proteome
ling schizophrenia as a pathway disorder. Analysis. Biomarkers: In Medicine, Drug
Schizophr Bull 38:1107–1108 Discovery, and Environmental Health 1002,
11. Potvin S, Stip E, Sepehry AA, Gendron A, Bah (Humana Press, 2013).
R, Kouassi E (2008) Inflammatory cytokine 17. Williamson JC, Edwards AV, Verano-Braga T,
alterations in schizophrenia: a systematic quan- Schwämmle V, Kjeldsen F, Jensen ON et al
titative review. Biol Psychiatry 63:801–808 (2016) High-performance hybrid Orbitrap
12. Schwarz E, Guest PC, Steiner J, Bogerts B, mass spectrometers for quantitative proteome
Bahn S (2012) Identification of blood-based analysis: observations and implications.
molecular signatures for prediction of response Proteomics. doi: 10.1002/pmic.201400545
and relapse in schizophrenia patients. Transl 18. Urbas L, Brne P, Gabor B, Barut M, Strlic M,
Psychiatry 2, e82 Petric TC et al (2009) Depletion of high-
13. Miller BJ, Buckley P, Seabolt W, Mellor A, abundance proteins from human plasma using
Kirkpatrick B (2011) Meta-analysis of cytokine a combination of an affinity and pseudo-affinity
column. J Chromatogr A 1216:2689–2694
Chapter 17
Simultaneous Two-Dimensional Difference Gel

Electrophoresis (2D-DIGE) Analysis of Two Distinct
Proteomes
Adriano Aquino, Paul C. Guest, and Daniel Martins-de-Souza
Abstract
This chapter describes the basics, applications, and limitations of two-dimensional gel electrophoresis
(2DE) and two-dimensional difference gel electrophoresis (2D-DIGE) for multiplex analysis of distinct
proteomes. We also propose a basic protocol for 2D-DIGE, technique that allows the analysis of paired
protein extracts, which are labeled with fluorescent Cy3 and Cy5 dyes and electrophoresed with a Cy2-
labeled standard extract on the same 2DE gels. Scanning the gels at wavelengths specific for each dye
allows direct overlay the two different proteomes and the differences in abundance of specific protein spots
can be determined.
Key words Proteomics, 2D-DIGE, 2DE, 2D difference gel electrophoresis, CyDyes
1 Introduction
It is likely that Marc Wilkins did not expect the term “proteome”
that he coined during his Ph.D. in 1994 would give raise to whole
science that we now know as “proteomics” [1]. In reality, pro-
teomics had already been in action for almost 20 years before this
after Patrick O’Farrell introduced a technique of protein separa-
tion in 1975 called two-dimensional gel electrophoresis (2DE)
[2]. During the genome era, 2DE was the most used method for
comparative global proteome analyses, combined with mass spec-
trometry (MS) for protein identification. Since this time, the
technique has undergone many optimization steps, including the
use of carrier ampholytes, immobilized pH gradients (IPG), and
development of IPG acrylamide strips [3, 4]. Perhaps the most
significant advancement has been the introduction of sample
labeling with fluorophores, which makes 2DE more sensitive,
precise, and replicable.
205
206 Adriano Aquino et al.
2DE involves separation of proteins in two dimensions accord-

ing to their isoelectric point (isoelectric focusing—IEF) and appar-
ent molecular size generally by polyacrylamide gel electrophoresis
(PAGE) (Fig. 1). This enables the simultaneous separation and
display of hundreds of proteins on a single gel. IEF was originally
performed in a capillary-gel format, with a pH gradient generated
by ampholytes in a gel-matrix. These first systems had low repro-
ducibility due to handling difficulties. As a result, IPGs were devel-
oped which are comprised of ampholytes covalently bonded to the
gels. The second dimension of the method separates proteins based
on their apparent molecular weight in a polyacrylamide gel matrix,
in the presence of sodium dodecyl sulfate detergent (SDS). In
SDS-PAGE or 1D-PAGE, large amounts of the SDS are necessary
to mask the charges of the proteins via formation of anionic com-
plexes. The electrophoretic mobility of the proteins in the poly-
acrylamide gel is then dependent on the molecular weight of the
protein and the separation occurs when a current is applied. The
reproducibility of SDS-PAGE and IEF has improved by the devel-
opment of precast gels. Protein visualization or detection after
2DE separation is commonly carried out using organic dyes, silver
staining, radioisotope tagging, or by fluorescent and chemilumi-
nescent labeling agents, as discussed below [5].
After running the gels are digitalized and the resulting images
can be analyzed using a variety of software. This allows comparison
of protein spot densities from test and control samples across dif-
ferent gels. After this, spots of interest can be excised from the gel
and the proteins digested for identification by MS [6, 7]. 2DE-MS
remained as the main proteomic technique for many years due to
advances in protein mass spectrometry, the availability of complete
genome sequences, and the development of computational pro-
grams for correlation and analysis of data [7, 8].
Until the late 1990s, identification of statistically significant dif-
ferences across two or more proteomes required the running and
analysis of many 2DE gels. This was difficult due to technical varia-
tions in sample preparation and gel running conditions [7]. In
1997, Unlu and colleagues presented an effective method of reduc-
ing gel-to-gel variation in a technique that they called difference gel
electrophoresis [9]. This method was later developed by Amersham
Biosciences (now GE Healthcare) and is now called differential in-
gel electrophoresis (DIGE). In a typical experiment, up to three
protein extracts can be compared on a single gel by covalently label-
ing these prior to electrophoresis with size- and charge-matched
spectrally resolvable fluorescent dyes (Cy2, Cy3, and Cy5) [10].
Therefore, fluorescent imaging of the gel at the wavelengths spe-
cific for each CyDye generates separate images for each proteome
and these can be overlaid directly for display of any differentially
expressed proteins and these proteins identified by MS as described
above (Fig. 2). Here we present a typical 2D-DIGE protocol.
pH 3 pH 10
High
1. Add sample to 2. Carry out separation

isoelectric focussing strip according to isoelectric point
3. Transfer strip to top of slab

gel for separation according to
molecular weight
Molecular weight
Low
www.Ebook777.com
Fig. 1 Basic experimental flow in 2D gel electrophoresis
2D-Difference Gel Electrophoresis
207
1. Each sample labelled with a different CyDye
Cy2 Cy3 Cy5
2. Combine into one tube
3. Subject to 2D gel electrophoresis
4. Image at l specific for each CyDye
Cy2 Cy3 Cy5
5. Overlay images
6. Image analysis to identify differentially

expressed protein spots
7. Identify proteins by MALDI-TOF MS
Fig. 2 Experimental flow of 2D-DIGE procedure
2 Materials
1. Protein samples of interest.

2. Extraction buffer: 30 mM Tris pH 8.0, 8 M urea and complete
EDTA-free protease inhibitors (Roche Diagnostics, Mannheim,
Germany).
3. CyDyes (GE Healthcare; Little Chalfont, Bucks, UK):
RPK0272—25 nmol of Cy2™ DIGE Fluor minimal dye;
RPK0273—25 nmol of Cy3™ DIGE Fluor minimal dye;
RPK0275—25 nmol of Cy5™ DIGE Fluor minimal dye.
4. Sonicator with micro-probe that can fit inside a 1.5 mL-
capacity Eppendorf tube.
5. 24 cm IPG strips (pH 4–7) (see Note 1).
2D-Difference Gel Electrophoresis 209
6. Strip rehydration buffer: 30 mM Tris pH 8.0, 7 M urea, 2 M

thiourea, 4 % CHAPS, 2 % DTT, 2 % pH 3–10 IPG buffer.
7. SDS equilibration buffer: 50 mM Tris-Cl, pH 8.8, 6 M urea,
30 % (v/v) glycerol, 2 % (w/v) SDS, 0.01 % bromophenol blue.
8. Strip equilibration buffer 1: SDS equilibration buffer with 2 %
iodoacetamide.
9. Strip equilibration buffer 2: SDS equilibration buffer with 5 %
iodoacetamide, 0.01 % and bromophenol blue (see Note 2).
10. 50 mM Tris–HCl, pH 6.8.
11. IPGbox (GE Healthcare) or similar.
12. First dimension isoelectric focusing (IEF), Ettan™ IPGphor™
3 Isoelectric Focusing System (GE Healthcare) or similar.
13. 12.5 % acrylamide, 0.33 N,N′-methylenebisacrylamide,
0.375 M Tris–HCl pH 8.8, 0.1 % SDS, 0.1 % (w/v) ammo-
nium persulfate, 0.025–0.09 % (v/v) tetramethylethylenedi-
amine (TEMED) (see Note 3).
14. 0.5 % (w/v) agarose in Laemmli SDS-PAGE electrode buffer.
15. Second dimension, Ettan DALT6 Electrophoresis System (GE
Healthcare) or similar.
16. Typhoon FLA 9500 Imager (GE Healthcare).
17. DeCyder™ 2D Software (GE Healthcare).
3 Methods
1. Sonicate tissue samples in 7:1 (w:v) extraction buffer and cen-

trifuge at approximately 13,000 × g for 20 min to pellet insolu-
ble debris.
2. Transfer supernatants to fresh 1.5 mL-capacity microcentri-
fuge tubes.
3. Estimate the protein concentration of the samples using a stan-
dard protein measurement kit (see Note 4).
4. Dilute each CyDye with 25 μL of fresh anhydrous dimethyl
formamide immediately prior to the labeling reaction to obtain
a stock solution of 1 nmol (see Note 5).
5. Add 0.4 μL of 1 nmol Cy2 to 50 μg of sample 1.
6. Add 0.4 μL of 1 nmol Cy3 to 50 μg of sample 2.
7. Add 0.4 μL of 1 nmol Cy5 to 50 μg of a mixture 1:1 (25ug
each) of samples 1 and 2 (see Note 6).
8. Incubate samples on ice for 30 min in the dark.
9. Quench samples by adding 1 μL of 10 mM lysine and leave the
samples on ice for 10 min in the dark.
10. Complete the volume of each the samples to 150 μL with rehy-
dration buffer.
11. Incubate the samples on ice for 15 min in the dark.

12. Mix all three samples together (see Note 7).
13. Add 450 μL of the sample mixture to IPG strips and allow to
hydrate at 20 °C for 12 h in an IPGbox.
14. Run the IPG strip on the IPGphor at 200 V for 1 h, 500 V for
1 h, 1000 V for 1 h, and 8000 V for 8 h at 20 °C using a maxi-
mum current setting of 50 mA/strip (see Note 8).
15. Strips can be frozen at −80 °C or step 14 can be carried out
immediately.
16. Equilibrate the IPG strip in two steps: first, soak the strip in
100 mL equilibration buffer 1 for 10 min.
17. Incubate the strips for 10 min in equilibration buffer 2 (see
Note 9).
18. Prepare resolving acrylamide gel solution as needed (see
Subheading 2, item 12) (see Note 10).
19. Add APS and TEMED last to the solution and poor this
between assembled low fluorescence glass plates to within
approximately 2 cm from the top.
20. Carefully layer butanol on top to help achieve a flat surface on
the gel upon polymerization.
21. Rinse the butanol of the gel surface, apply the equilibrated IPG
strips so it is sitting immediately on top of the gel, and seal with
0.5 % agarose in SDS running buffer on top (see Note 11).
22. Carry out electrophoresis at 60 V for 1 h followed by 30 μA/
gel until the dye front reaches the end of the gel.
23. Scan the gels directly between the glass plates (see Notes 12 and
13) using filters specific for the excitation and emission wave-
lengths of Cy2 (480 and 530 nm, respectively), Cy3 (540 and
590 nm, respectively), and Cy5 (620 and 680 nm, respectively).
24. Export the images as 16-bit tagged image file format (TIFF)
files for analysis.
25. Analyze the images automatically using the DeCyder Batch
Processor and Differential In-Gel Analysis (DIA) software tools.
26. Compare the protein spot volumes on the Cy5-labeled pooled
standard image with matching spots on the Cy2- or Cy3-
labeled images.
27. Match the images from each gel with the biological variation
analysis (BVA) software using the Cy5-labeled pooled standard
image for normalization of each protein spot.
28. Use the software to identify protein spots with differences in
abundance. Within gels, this is achieved by direct overlay of
spots and across gels by land marking, warping, and matching
spots using the Biological Variation Analysis function of the
Decyder software (see Note 14).
2D-Difference Gel Electrophoresis 211
4 Notes
1. The IPG strip pH range should be chosen to maximize resolu-

tion of the proteins of interest. The pH 4–7 strips will provide
better resolution of acidic protein spots.
2. The bromophenol blue is added to provide a visible dye front
during electrophoresis.
3. The acrylamide and N,N′-methylenebisacrylamide concentra-
tions should be chosen based on the desired molecular weight
range separation (low percentages of acrylamide allow resolu-
tion of high molecular weight proteins and high percentages
resolve proteins in the lower molecular weight range).
Ammonium persulfate and TEMED should be added just
before use since these initiate the gel polymerization process.
4. The presence of urea in the extraction buffer might interfere
with some assays, which could lead to false readings.
5. It is best to use DMF as fresh as possible after opening. We
noticed decreased labeling efficiency even if the DMF was only
1 month old.
6. The use of an internal standard helps to minimize false posi-
tives and false negatives since it can serve as a control for each
protein spot on all gels in the analysis. The standard is usually
made by combining equal volumes of each extract.
7. It is possible that the CyDyes may show preferential labeling of
some proteins although this can be accounted for by reversing
the dye:extract combinations.
8. A step voltage gradient was used as we and others have noticed
that this helps to avoid horizontal streaking of protein spots on
the final 2D gel image.
9. This stage serves the same purpose as a stacking gel in 1D gel
electrophoresis. The spots only begin to resolve once they
reach the higher pH in the resolving gel.
10. Lower percentage gels favor resolution of high molecular
weight proteins and higher percentages can resolve proteins of
lower molecular weights.
11. Ensure that there are no air bubbles trapped between the strips
and the gel tops to avoid distortion in the protein spot
patterns.
12. The second-dimension gels should be poured between low
fluorescent Pyrex glass plates to minimize background fluores-
cence during scanning. Furthermore, the Ettan Dalt II system
allowed simultaneous running of multiple plates. This is impor-
tant as it means that all second dimension gels can be run
under approximately the same conditions, which allows for
better matching of the gel images in subsequent stages.
13. A key advantage of the 2D DIGE technique is that gels can be

imaged after electrophoresis without disassembly of the low-
fluorescence glass plates. This ensures that the gels are not
deformed or damaged during imaging and also minimizes the
possibility of contamination. Furthermore, gels can be scanned
for different lengths of time to maximize detection of high-
and low-abundance protein spots.
14. Protein identification can be achieved using any kind of pro-
tein mass spectrometry. We suggest preparing a gel containing
approximately 200 μg of the standard pool followed by col-
loidal Coomassie Blue staining for excision of spots. Spots
must be digested in gel prior to mass spectrometry analysis for
protein identification [11].
Acknowledgments
Dr. Adriano Aquino and Prof. Daniel Martins-de-Souza are funded

by FAPESP (São Paulo Research Foundation), grants 15/09159-8
and 13/08711-3.
References
1. Wilkins MR, Sanchez JC, Gooley AA, Appel 6. Thomas A, Lenglet S, Chaurand P, Deglon J,
RD, Humphery-Smith I, Hochstrasser DF et al Mangin P, Mach F et al (2011) Mass spectrom-
(1996) Progress with proteome projects: why etry for the evaluation of cardiovascular dis-
all proteins expressed by a genome should be eases based on proteomics and lipidomics.
identified and how to do it. Biotechnol Genet Thromb Haemost 106:20–33
Eng Rev 13:19–50 7. Penque D (2009) Two-dimensional gel electro-
2. O’Farrell PH (1975) High resolution two- phoresis and mass spectrometry for biomarker
dimensional electrophoresis of proteins. J Biol discovery. Proteomics Clin Appl 3:155–172
Chem 250:4007–4021 8. Oliveira BM, Coorssen JR, Martins-de-Souza
3. Görg A, Postel W, Weser J, Gunther S, Strahler D (2014) 2DE: the phoenix of proteomics.
JR, Hanash SM et al (1987) Elimination of J Proteomics 104:140–150
point streaking on silver stained two- 9. Unlü M, Morgan ME, Minden JS (1997)
dimensional gels by addition of iodoacetamide Difference gel electrophoresis: a single gel
to the equilibration buffer. Electrophoresis method for detecting changes in protein
8:122–124 extracts. Electrophoresis 18:2071–2077
4. Görg A, Postel W, Günther S (1988) The cur- 10. Knowles MR, Cervino S, Skynner HA, Hunt SP,
rent state of two-dimensional electrophoresis de Felipe C, Salim K et al (2003) Multiplex pro-
with immobilized pH gradients. Electrophoresis teomic analysis by two-dimensional differential
9:531–546 in-gel electrophoresis. Proteomics 3:1162–1171
5. López JL (2007) Two-dimensional electro- 11. Shevchenko A, Tomas H, Havlis J, Olsen JV,
phoresis in proteome expression analysis. Mann M (1996) In-gel digestion for mass
J Chromatogr B Analyt Technol Biomed Life spectrometric characterization of proteins and
Sci 849:190–202 proteomes. Nat Protoc 1:2856–2860
Chapter 18
Selective Reaction Monitoring for Quantitation

of Cellular Proteins
Vitor M. Faça
Abstract
Proteins and proteomes are dynamic and complex. The accurate identification and measurement of their
properties such as abundance, location, and turnover are challenging tasks. Even though high-throughput
proteomics has significantly evolved, the technique still lacks fully quantitative and reproducible qualities.
A mass spectrometry-based targeted proteomic strategy called selective reaction monitoring (SRM) has
emerged in recent years as an important multiplex platform to precisely quantify sets of proteins in multiple
samples. This has several advantages in terms of sensitivity, reproducibility, and sample consumption compared
to classical methods including those based on antibodies. Here, we present a detailed protocol for quantitation
of panels of proteins from cell line extracts using the SRM targeted proteomics approach.
Key words Mass spectrometry, Protein quantitation, Selective reaction monitoring, Targeted
proteomics
1 Introduction
Proteomics has evolved progressively and currently allows identification

and quantification of large sets of proteins with good precision and
across several orders of magnitude. This process reached its highpoint
recently with the publication of the draft of human proteome [1, 2].
However, the proteome is dynamic as it changes constantly and it is
significantly different in both composition and expression levels of
individual components across cells, tissues, organs, and individuals.
Because of this, the capabilities of high-throughput proteomic
methods based on liquid chromatography tandem mass spectrometry
(LC-MS/MS) to analyze multiple samples in a timely and reproducible
manner are still considered to be limited [3]. In order to address this,
targeted proteomics approaches have now focused on precise
quantitation of specific sets of proteins in multiple samples or time
points. In this way, the monitoring of dynamic proteome changes has
emerged as a complementary strategy to complete and validate
proteomic and transcriptomic datasets.
213
214 Vitor M. Faça
The mass spectrometry-based strategy termed selected

reaction monitoring (SRM), also known as multiple reaction
monitoring (MRM), has been applied frequently for accurate
quantitation of small molecules such as drugs or metabolites,
and it is now being applied progressively more to the study of
peptides and proteins [4]. This increasing interest in options for
direct quantitation of proteins or peptides is also driven by limi-
tations and lack of reproducibility of the antibody-based meth-
ods [5]. However, it should be stated that all methods have their
limitations. The SRM method is mainly based on selecting and
quantifying a specific set of peptides (proteotypic peptides)
derived from a target list of proteins of interest using LC-MS/
MS analysis [6]. SRM takes advantage of the ion filtering capa-
bilities of tandem quadrupole-based equipment, allowing selec-
tion of precursors and respective fragment ions produced by
CID (collision-induced dissociation) to detect low abundance
species among complex mixtures and across four to five orders
of magnitude. Due to the rapid tandem quadrupole duty cycle
that occurs over milliseconds and the online chromatographic
separation of peptides, numerous precursor–fragment ion pairs
known as SRM transitions can be monitored simultaneously.
This facilitates quantitative experiments involving several
different samples in a timely manner.
Since SRM methods efficiently quantitate peptides, the
accurate monitoring of proteomic changes in complex samples
requires the inclusion of reproducible and efficient enzymatic
digestion of proteins in the method. Also, the peptides selected as
indicators of protein abundance must meet several characteristics,
including the possibility of chemical synthesis to generate these as
standards for absolute quantitation. Here, a detailed protocol is
presented for quantitative SRM analysis of cellular proteins using,
as an example, identification of the proteomic changes in mammary
epithelial cells following treatment with transforming growth
factor-beta (TGF-β).
2 Materials
2.1 General 1. Use high-performance liquid chromatography (HPLC) or

mass spectrometry grade reagents if possible and prepare fresh
solutions using MilliQ water on the day of use, maintain at
room temperature, unless indicated otherwise.
2.2 Cell Culture 1. Breast cell lineage MCF-10A (acquired from ATCC-
Components CRL-10317™) (see Note 1).
2. Cell culture media: Mammary Epithelial Cell Growth Medium
(MEGM™; Lonza; São Paulo, Brazil), supplemented with
100 ng/mL cholera toxin and MEGM kit CC-3150.
SRM Analysis of Cells 215
2. Tissue culture dishes for adherent cell culture (60.1 cm2 area)
(see Note 1).
3. Cell wash solution: phosphate buffered saline (PBS) (pH 7.4).
2.3 Sample 1. Bradford protein quantification kit (Bio-Rad; Hercules, CA,

Preparation USA) (see Note 2).
Components 2. Denaturation buffer: 8 M urea, 0.15 M Tris-HCl (pH 8.5).
for Targeted Proteomic
3. Sigma Protease Inhibitor Cocktail for use with mammalian
Analysis cell and tissue extracts or similar.
4. Trypsin solution: sequencing grade modified 100 ug/mL
trypsin (Promega; Madison, WI, USA) in 0.1 M ammonium
bicarbonate solution (see Note 3).
5. Reducing solution: 10 mg/mL dithiotreitol in 0.15 M Tris–
HCl (pH 8.5) (see Note 4).
6. Alkylation solution: 20 mg/mL iodoacetamide in 0.15 M
Tris-HCl (pH 8.5) (see Note 4).
7. Solid phase purification: OASIS HLB solid-phase-extraction
columns (Waters Corporation; Milford, MA, USA).
8. Equilibration solution: 95 % water, 5 % acetonitrile, 0.1 % for-
mic acid.
9. Elution solution: 50 % water, 50 % acetonitrile, 0.1 % formic
acid.
2.4 Targeted Mass 1. Equipment: tandem quadruplole Xevo-TQs mass spectrometer

Spectrometry coupled to a Class I Ultra Performance Chromatographic
Components System (UPLC) (Waters Corporation).
2. Chromatographic column: ACQUITY UPLC HSS C18column
(1.8 μm particle size, 1 mm inner diameter × 150 mm length)
(Waters Corporation).
3. Reconstitution buffer: 3 % acetonitrile, 0.1 % formic acid.
4. Solvent A for reverse phase chromatography: 95 % water, 5 %
acentonitrile, 0.1 % formic acid.
5. Solvent B for reverse phase chromatography: 99.9 % acetoni-
trile, 0.1 % formic acid.
3 Methods
3.1 Sample 1. Cultivate cancer cell lineages in 60 cm2 plates in MEGM

Preparation for SRM supplemented with 10 % FBS according to standard protocols
Proteomic Analysis for cell culture and until confluence.
of Cell Lines 2. Remove old media and cultivate cells in fresh media
containing specific treatments and/or for the period of
interest in time-course experiments (see Note 5).
216 Vitor M. Faça
3. If proteins of interest are also present in the cell secretion, save

conditioned media and remove debris first by centrifugation
at 12,000 × g and then by filtration through a 0.22 mm filter
(see Note 6).
4. Collect cells after appropriate time or treatment by washing
three times with ice-cold PBS, drain any residual solution, add
1 mL of denaturation buffer containing freshly added protease
inhibitor cocktail, and remove the cells from the plate surface
using a cell scraper (see Note 7).
5. Carry out cell lysis and total protein extraction performing
three cycles of sonication and then immerse the samples in an
ice bath (see Note 8).
6. Quantitate total proteins in cell extracts by protein assay and
use 50 μg aliquots for SRM analysis (see Note 9).
7. Reduce the protein cysteine residues by adding 5 μL of reduc-
tion solution and maintaining the reaction at 37 °C for 1 h.
8. Perform protein alkylation by adding 10 μL of alkylating
solution and incubate the samples at room temperature for an
additional 1 h.
9. Dilute samples to approximately 0.6 M by the addition of
0.15 mM Tris–HCl (pH 8.5) (see Note 10).
10. Perform trypsin digestion by adding 10 μL trypsin solution to
each sample to a final enzyme:protein ratio of 1:50 (w/w) and
incubate for 2 h at 37 °C, then add an additional 5 μL of tryp-
sin solution to each digest and incubate for an additional 16 h
or overnight at 37 °C.
11. Desalt samples using solid-phase extraction in OASIS columns
as follows: (1) condition the column with 1 mL acetonitrile;
(2) equilibrate the column with 1.6 mL equilibration solution;
(3) apply the sample to the column; (4) wash the column with
1.6 mL of equilibration solution; (5) elute the peptides with
1.2 mL of elution solution; and (6) dry the eluted peptides in
a speedvac (see Note 11).
3.2 SRM Method 1. Select protein targets for SRM analysis and identify the pro-
Development and Data teotypic peptides (Table 1) (see Note 12).
Analysis 2. Synthesize or obtain the selected peptides commercially if
quantitative analysis is of interest (see Note 13).
3. Perform a collision energy and chromatographic separation
standardization for the selected set of proteotypic peptides
(Table 1) (see Note 14).
4. Suspend the samples for SRM analysis in 50 μL of reconstitu-
tion buffer, vortex thoroughly, centrifuge at 12,000 × g for
15 min, and transfer the supernatants to mass spectrometer
compatible injection vials.
Table 1
Proteotypic peptides and analytical parameters utilized in the study with MCF-10A cells. The protein
set was selected based on high-throughput proteomic experiments (unpublished data). Peptides
were synthesized and analyzed individually to maximize efficiency of fragmentation. The selection
of 2 SRM transitions per proteotypic peptide improved method reliability
SwissProt ID Protein Sequence m/z Charge Col. energy Trans.1 Trans. 2
P12830 CDH1 NTGVISVVTTGLDR 716.39 2 22 716.4 > 947.6 716.4 > 1060.7
P19022 CDH2 GPFPQELVR 521.79 2 20 521.8 > 371.4 521.8 > 741.6
P08253 MMP2 AFQVWSDVTPLR 709.88 2 20 709.9 > 787.5 709.9 > 973.6
p14780 MMP9 AFALWSAVTPLTFTR 840.96 2 21 840.9 > 734 840.9 > 835.6
Q13421 MSLN LLGPHVEGLK 531.82 2 18 531.8 > 418.6 531.6 > 836.6
P01033 TIMP1 GFQALGDAADIR 617.32 2 14 617.3 > 404.3 617.3 > 717.6
Q15672 TWIST IIPTLPSDK 492.29 2 26 429.3 > 446.2 429.3 > 757.6
P08670 VIME ILLAELEQLK 585.36 2 18 585.4 > 830.6 585.4 > 943.6
5. Run appropriate calibration curves with standard peptides

(see Note 15).
6. Inject 10 μL of samples in triplicate (or more) to facilitate
good statistical analysis (Fig. 1) (see Note 16).
7. Analyze the data using the MS vendor software or using the
Skyline software [7] (Fig. 1) (see Note 17).
4 Notes
1. Cell culture in MEGM media is standard for MCF-10A cells.

Adapt to appropriate cell culture media following ATCC
specifications (www.atcc.org) for other cell lines as necessary.
In general, cells are cultivated at a density of 2–5 × 106 cells per
plate, producing approximately 200–500 μg of protein in total
cell extracts.
2. Other protein assay reagents can be used but the user must
ensure that it is compatible with the reagents present in the
protein extracts of interest.
3. Once prepared, pass the solution through a 0.22 μm syringe
filter and use the filtrate.
4. Always prepare fresh reducing and alkylation solutions to
guarantee efficient reaction with cysteine residues. Do not
store these solutions.
5. For the purpose of demonstrating the protocol, we performed
experiments with the breast epithelial cell line MCF-10A
treated with10 ng/mL of TGF-β2 for 72 h.
www.Ebook777.com
218 Vitor M. Faça
500 y10 – 1060.5997+ y9 – 947.5156+
Peak area (10^3)

400
300
CHD1
200
100
0
y6 – 741.4254+ y5 – 371.2163++
Peak area (10^3) 600
500
400
CHD2 300
200
100
0
1.2 y8 – 973.5102+ y7 – 787.4308+

Peak area (10^6)
1.0
0.8
MMP2 0.6
0.4
0.2
0
300 y7 – 835.4672+ y6 – 734.4196+

Peak area (10^3)
250
200
MMP9 150
100
50
0
90 y8 – 836.4625+ y6 – 418.7349++
Peak area (10^3)
80
70
60
MSLN 50
40
30
20
10
0
30 y7 – 717.3526+ y3 – 403.2300+
Peak area (10^6)
25
20
TIMP1 15
10
5
0
y7 – 757.4090+ y4 – 446.2245+
160
140
Peak area (10^3)
120
100
TWIST 80
60
40
20
0
140 y8 – 943.5459+ y7 – 830.4618+

Peak area (10^6)
120
100
VIME 80
60
40
20
0
Secreted Secreted Secreted Tot. extract Tot. extract Tot. extract
control TGF-b2 A TGF-b2 B control TGF-b2 A TGF-b2 B
6. The analysis of proteins present in conditioned media requires

special steps for sample processing. Please see Faça et al. for a
detailed protocol [8].
7. The step of cell recovery from culture plate is important for
reproducibility. Since cells may vary significantly in terms of
plate adherence, refine the protocol if necessary.
8. The sonication step can be performed in an ultrasonic bath or
with an ultrasonic probe. Note that sonication can heat
samples and promote protein carbamylation by urea present in
denaturation buffer. Therefore, avoid heating by using short
(e.g., 30 s) cycles of sonication followed by cooling the samples
in an ice bath.
9. Quantification of total protein is important to allow efficient
reduction, alkylation, and enzymatic protein digestion
reactions.
10. Dilution of the solution guarantees good trypsin activity,
which is close to 100 % in urea solutions at concentration less
than 1 M according to the manufacturer.
11. Samples can be stored dried at −80 °C until ready for SRM
analysis.
12. In order to develop a SRM method for target proteins, use the
following workflow: (1) perform a virtual tryptic protein
digestion using for example the PeptideMass tool available in
UNIPROT (www.uniprot.org); (2) select tryptic peptides
containing 10–20 amino acids since these are easier to obtain
by solid-phase peptide synthesis; (3) if possible, select peptides
containing proline residues, which generate intense fragment
peaks; (4) avoid peptides containing methionine or N-terminal
glutamine, as these accumulate in source modifications during
MS ionization; and (5) check previous detection and predicted
suitability for selected peptides using the SRMatlas databank
(www.srmatlas.org) [9].
13. Although the use of synthetic peptides representing the target
proteins is not obligatory, they are useful for method develop-
ment and refinement. Peptides can be synthesized rapidly in
Fig. 1 SRM analysis of MCF-10A cell line treated with TGF-β2. The protein set was selected based on high-throughput
proteomic experiments (unpublished data). Cells were treated with TGF-β2 for 72 h. Both total cell extract (50 μg of
protein) and conditioned media (2 mL) were analyzed for the same set of proteins. The data were collected for two
biological replicates (TGF-β2 A and TGF-β2 B) and samples were injected in triplicate. Secreted proteins such as
TIMP1 were detected only in the media, while the nuclear transcription factor TWIST was detected only in total cell
extracts. The full experiment was performed in approximately 6 h using a 20 min gradient for peptide separation.
The effects on protein levels induced by TGF-β in MCF-10A cells were consistent with the results of previous
experiments and were validated by Western blotting for some proteins
220 Vitor M. Faça
house using standard solid-phase Fmoc chemistry. Preselection

of short peptides (<20 amino acids) facilitates synthesis and
generally produces good yields. If absolute quantitation is of
interest, purify the peptides by reverse phase chromatography
and use an accurate method to quantify these, such as amino
acid analysis. Also, the use of isotopically labeled peptides
(e.g., Lys13C6) allows more precise quantification since this
approach incorporates internal standards.
14. The study of ideal collision energy for SRM studies is essential for
good sensibility. Also, it is important to perform these studies in
the local instrument since several tuning parameters can affect the
fragmentation efficiency. The optimization of a chromatographic
gradient with the specific set of peptides is also important to
guarantee quick runs and good reproducibility. For these steps,
consider using the software Skyline [7] which is designed to
facilitate method development in addition to performing efficient
data analysis across multiple platforms.
15. Chromatography in micro or nano-scale is susceptible to
micro-particles originating from insoluble material or solid-
phase extraction column leaking. This simple centrifugation
step improves the column lifetime significantly.
16. The injection of 10 μL corresponds to 10 μg of protein digest.
This amount is below the loading capacity of the UPLC HSS
C18 Column [1 mm (inner diameter) × 150 mm length] to
obtain best chromatographic separations.
17. The use the multi-platform Skyline software to integrate,
calibrate, and quantify samples analyzed by SRM streamlines
the development and application of the method.
Acknowledgments
This research was supported by FAPESP (Young Scientist Grant –

Proc. No. 2011/0947-1), Center for Cell Based Therapy-CTC-
CEPID (Proc. FAPESP 2013/08135-2), and CISBi-NAP. V.M.F.
received fellowships from CNPq, Proc. No. 308561/2014-7.
References
1. Kim MS, Pinto SM, Getnet D, Nirujogi RS, 3. Cox J, Mann M (2011) Quantitative, high-
Manda SS, Chaerkady R et al (2014) A draft map resolution proteomics for data-driven systems
of the human proteome. Nature 509:575–581 biology. Annu Rev Biochem 80:273–299
2. Wilhelm M, Schlegl J, Hahne H, Moghaddas 4. Picotti P, Aebersold R (2012) Selected reaction
Gholami A, Lieberenz M, Savitski M et al monitoring-based proteomics: workflows,
(2014) Mass-spectrometry-based draft of the potential, pitfalls and future directions. Nat
human proteome. Nature 509:582–587 Methods 9:555–566
5. Baker M (2015) Reproducibility crisis: blame it 8. Faça VM, Palma CS, Albuquerque D, Canchaya
on the antibodies. Nature 521:274–276 GN, Grassi ML, Epifânio VL et al (2014) The
6. Mallick P, Schirle M, Chen SS, Flory MR, Lee secretome analysis by high-throughput
H, Martin D et al (2007) Computational pre- proteomics and multiple reaction monitoring
diction of proteotypic peptides for quantitative (MRM). Methods Mol Biol 1156:323–335
proteomics. Nat Biotechnol 25:125–131 9. Picotti P, Lam H, Campbell D, Deutsch EW,
7. MacLean B, Tomazela DM, Shulman N, Mirzaei H, Ranish J et al (2008) A database of
Chambers M, Finney GL, Frewen B et al (2010) mass spectrometric assays for the yeast proteome.
Skyline: an open source document editor for Nat Methods 5:913–914
creating and analyzing targeted proteomics
experiments. Bioinformatics 26:966–968
Chapter 19
Characterization of a Protein Interactome

by Co-Immunoprecipitation and Shotgun Mass
Spectrometry
Giuseppina Maccarrone, Juan Jose Bonfiglio, Susana Silberstein,
Christoph W. Turck, and Daniel Martins-de-Souza
Abstract
Identifying the partners of a given protein (the interactome) may provide leads about the protein’s func-
tion and the molecular mechanisms in which it is involved. One of the alternative strategies used to char-
acterize protein interactomes consists of co-immunoprecipitation (co-IP) followed by shotgun mass
spectrometry. This enables the isolation and identification of a protein target in its native state and its
interactome from cells or tissue lysates under physiological conditions. In this chapter, we describe a co-IP
protocol for interactome studies that uses an antibody against a protein of interest bound to protein A/G
plus agarose beads to isolate a protein complex. The interacting proteins may be further fractionated by
SDS-PAGE, followed by in-gel tryptic digestion and nano liquid chromatography high-resolution tandem
mass spectrometry (nLC ESI-MS/MS) for identification purposes. The computational tools, strategy for
protein identification, and use of interactome databases also will be described.
Key words Mass spectrometry, Co-immunoprecipitation, Interactome, Proteome, Computational tools
1 Introduction
The interactome may be defined as the complete set of molecular

interactions of a molecule in a given cell or other biological envi-
ronment [1]. The interactome of a protein represents the proteins
and molecules that interact with it to promote, regulate, and
inhibit its activity or expression. Alternatively, members of the
interactome may be substrates or molecular components regulated
by the protein of interest.
Identifying the interactome of a given protein may provide leads
about the protein’s function and the molecular mechanisms in which
it is involved. The protein interaction map of Drosophila melanogas-
ter encompasses 556 protein complexes and can assign functional
223
224 Giuseppina Maccarrone et al.
processes to almost 600 protein-coding genes, many of which previ-

ously lacked annotation. The data gained from the Drosophila mela-
nogaster interactome may also provide information about human
cell interactomes [2] and enrich genome databases by generating
annotations to non-annotated protein-coding genes. An extensive
proteomic survey that used affinity tag purification of E. coli strains
identified 5,993 protein interactions and allowed proteins to be
assigned functions such as protein synthesis, amino acid metabolism,
biofilm formation, and motility [3]. More specific to a protein, novel
partner proteins in brain tissue were unraveled to collapsin response
mediator protein-2 (CRMP2/DPYSL2) (Reference: Martins-de-
Souza D, Cassoli JS, Nascimento JM, Hensley K, Guest PC, Pinzon-
Velasco AM, Turck CW. The protein interactome of collapsin
response mediator protein-2 (CRMP2/DPYSL2) reveals novel
partner proteins in brain tissue. Proteomics Clin Appl. 2015
Oct;9(9–10):817–31). The study of protein interactomes in diseases
may reveal dysfunctional pathways, further insights into regulation,
and the possible role that protein partners play in the disease [4].
Many interactome studies have been performed with yeast
two-hybrid screening (Y2H) and Tandem Affinity Purification
(TAP) [5], although co-immunoprecipitation (co-IP) is also a reli-
able method for studying protein interactomes with its own advan-
tages [6]. Briefly, the co-IP of intact protein complexes employs an
immobilized antibody that targets a specific protein. The sample of
interest is passed through or incubated with the antibody under
non-denaturating conditions. Next, the target protein (bait) binds
to the antibody along with its partners (prey), while noninteracting
proteins pass through or do not bind to the matrix. The captured
protein complex may be analyzed further by proteomic analysis
(Fig. 1). Co-IP and subsequent mass spectrometry (MS) have been
applied to proteins of interest in diseases. Recently, the B-Raf inter-
actome was characterized in mouse hippocampal cells [7, 8]. This
protein is involved in a cellular signaling cascade, known to play a
role in cancer [9]. The interactomes of the G-protein beta subunit
[10], neurocalcin delta [11], dopamine transporter [12], and
dynamin-1 [13], which are involved in neurological disorders,
have also been characterized by co-IP and MS.
Here, we describe a protocol (shown in Fig. 1) that can be
used to investigate the interactome of a protein of interest.
2 Materials
2.1 Immunoprecipi 1. Ice-cold phosphate buffered saline, PBS: 137 mM NaCl,

tation Sample 10 mM phosphate, 2.7 mM KCl, pH 7.4.
Preparation 2. Modified RIPA buffer: 50 mM Tris–HCl pH 7.4, 100 mM
NaCl, 1 mM EDTA, 1 % NP-40, 1/25 protease inhibitor cock-
tail (Roche Molecular Biomedicals, Mannheim, Germany),
1 mM phenylmethylsulfonil fluoride, and 1 mM activated
sodium orthovanadate.
Interactome Analysis 225
Proteome of interest Sample pre-clearing Immuno affinity capture

= Bait protein = Resin Y = Isotype anbody Y = Specific Anbody
1D SDS-PAGE Eluon by denaturang Immobilized Washed out

(or 2-DE PAGE) condions interactome (no-interacng
(Interacng proteins) proteins)
In-gel digeson
Reciprocal IP
&
Western blots
Protein Idenficaon Mining of Interactome Preys

LC-Mass Spectrometry Database Validaon
Computaonal Analysis
Fig. 1 Steps involved in the characterization of a bait protein interactome. Firstly, the cell lysate is precleared
with A/G beads and nonimmune rabbit IgG. The precleared lysate is incubated with the A/G beads and the
antibody against the bait protein to fish protein preys. The precleared lysate is again incubated with the nonim-
mune rabbit IgG to fish nonspecific binders (Mock-IP fraction, not shown). The immobilized interactome is
washed and eluted from the beads, and proteins are separated before identification using LC-ESI-MS/MS. Data
analysis includes the identification of the protein interactome, data filtering considering mock-IP results, min-
ing of the interactome database available in the public domain and validation
2.2 Immunoprecipi 1. Immunoprecipitating antibody: anti B-Raf antibody (Santa

tation Assay Cruz Biotechnology, Santa Cruz, CA, USA).
2. Nonimmune antibody produced in the same species of the
immunoprecipitating antibody: normal rabbit IgG.
3. Protein A/G plus agarose.
2.3 Proteomics 1. Electrophoresis equipment.

Sample Preparation 2. Laemmli sample buffer 2×: 0.125 M Tris–HCl pH 6.8, 4 %
SDS, 20 % glycerol, 10 % 2-mercaptoethanol, 0.004 % bromo-
phenol blue.
3. Coomassie Brilliant Blue R-250.
4. 25 mM ammonium bicarbonate/50 % acetonitrile.
6. 10 mM dithiothreithol.
7. 50 mM iodoacetamide.
8. 50 % acetonitrile/2 % formic acid.
9. Trypsin solution: 5 ng/μL trypsin in 25 mM ammonium
bicarbonate.
2.4 nLC ESI-MS/MS 1. Nano LC system.

2. Trap column RP-C18.
3. In-house packed nano-column (75 μm i.d. × 15 cm).
4. Column packing material RP-C18, 3 μm.
5. Picofrit self-packed column (75 μm i.d, 10 μm tip).
6. Distilled H2O, 0.1 % HCOOH.
7. Eluent A: H2O/ACN (95/5), 0.1 % HCOOH.
8. Eluent B: ACN, 0.1 % HCOOH.
9. Centrifuge filter:0.22 μm.
10. LTQ Orbitrap XL mass spectrometer (Thermo Fisher, Germany).
11. Nano electrospray ionization (ESI) source (Thermo Fisher).
2.5 Mass 1. LC-ESI-MS/MS data acquisition software: Xcalibur v. 2.07

Spectrometry (Thermo Fisher).
Software 2. Nano-LC controlling software: LC-Eksigent v.2.09 (Eksigent,
Dublin, CA, USA).
3. Raw MS and MS/MS spectra processing software: Bioworks v.
3.3 (Thermo Fisher).
4. Protein database search engine: Mascot Demon 2.2.2 (Matrix
Science, UK).
2.6 Western Blot 1. Immobilon PVDF membrane.

Analysis 2. TBS-T buffer: 137 nM NaCl, 20 mM Tris–HCl, 0.05 %
Tween-20, pH 8.0.
3. 5 % nonfat dry milk in TBS-T.
4. ECL Plus reagent kit.
3.1 Immunoprecipi 1. Grow cells to 70 % confluence in a 100 mm cell culture plate

tation Sample (see Note 2).
Preparation 2. Aspirate culture medium and wash cells with 6 mL of ice-cold
PBS.
3. Harvest cells on ice by gently scraping them off with a rubber
policeman in 1 mL of ice-cold complete modified RIPA buffer
(see Note 3).
4. Lyse cells for 2 h at 4 °C with gentle agitation followed by

centrifugation at 3500 × g for 5 min at 4 °C.
5. Collect the supernatant in a 1.5 mL microcentrifuge tube.
6. Take an aliquot of this supernatant for future studies (e.g.,
Western blot validation).
3.2 Co-immunopreci 1. Preclear the resulting lysate by incubation for 1 h at 4 °C with

pitation Assay 1 μg of nonimmune rabbit polyclonal antibody (nonimmune
antibody produced in the same species of the anti-B-Raf anti-
body used).
2. After 1 h, add 30 μL of protein A/G plus agarose for a 2 h
incubation at 4 °C with gentle agitation (see Note 4).
3. Separate nonspecific proteins bound to the antibody and pro-
tein A/G plus agarose beads by centrifugation at 3500 × g for
5 min at 4 °C, and discard the pellet containing nonspecific
proteins, antibody, and protein A/G plus agarose.
4. Collect the supernatant and separate into two equal parts for
the specific and control IPs.
5. The specific co-IP procedure is performed by incubating the
supernatant with 1 μg of anti B-Raf antibody overnight at
4 °C, with gentle agitation (see Note 5).
6. The control IP is carried out by incubating an aliquot of the
supernatant with 1 μg of nonimmune rabbit polyclonal anti-
body overnight at 4 °C with gentle agitation (label the immu-
noprecipitated fraction resulting from the control IP as “mock
IP” (see Note 6)).
7. Fish the B-Raf antibody, the B-Raf protein, and B-Raf interac-
tors by adding 35 μL of protein A/G plus agarose, incubate for
2 h at 4 °C with gentle agitation, and then pellet by centrifuga-
tion at 3500 × g for 5 min at 4 °C (save the supernatant for
future studies).
8. Perform the same procedure for the mock IP.
9. To reduce nonspecific binding, wash the immunoprecipitate
three times with 500 μL of complete modified RIPA buffer
and once with 1 mL of ice-cold PBS (see Note 7).
10. Apply the same wash procedure for the mock IP pellet.
3.3 Proteomics 1. Suspend the pellet/immunoprecipitate in 30 μL 2× Laemmli

Sample Preparation sample buffer (see Note 8).
2. Before electrophoresis, heat the sample for 5 min at 95 °C,
centrifuge and separate from the agarose beads.
3. Resolve the proteins in the total sample by 10 % SDS-PAGE
and then stain in-gel with Coomassie Brilliant Blue R-250 for
1 h at room temperature with gentle agitation (see Note 9).
4. Slice the gel lane is sliced into ~20 fractions, cut into small
pieces and wash twice with 25 mM ammonium bicarbon-
ate/50 % acetonitrile.
5. Reduce sulfhydryl bonds in the proteins within the gel by incu-
bating with 10 mM dithiothreithol for 30 min at 56 °C.
6. Alkylate the reduced sulfhydryl groups by incubating with
50 mM iodoacetamide for 30 min at room temperature, wash
twice, dry and rehydrate in 20 μL trypsin solution for 15 min
on ice.
7. Remove the surplus trypsin solution, cover the gel pieces with
25 mM ammonium bicarbonate, and allow digestion to pro-
ceed for 4–6 h at 37 °C (see Note 10).
8. Extract the resulting peptides twice with 50 % acetonitrile/2 %
formic acid and then lyophilize to dryness.
3.4 nLC ESI-MS/MS 1. Dissolve the dried peptides in 10 μL 0.1 % HCOOH and pass
through the filter before analysis by nLC ESI-MS/MS (see
Note 11).
2. Separate the peptides according to their physiochemical prop-
erties using a nano HPLC system equipped with a trap column
and a picofrit nano-column coupled online to an LTQ-Orbitrap
mass spectrometer via a nano ESI source (see Note 12).
3. Load 5 μL of each sample onto the RP-C18 trap column and
wash for 10 min with eluent A at a flow-rate of 3 μL/min.
4. Separate the peptides on the RP-C18 nano-column by applying
a linear gradient of eluent B from 2 to 10 % in 5 min and
10–40 % in 98 min, at a flow rate of 200 nL/min (see Note 13).
5. Operate the mass spectrometer in positive ion mode, using a
set method that includes a data-dependent acquisition of a full
scan (FS) and MS/MS scans of each peptide.
6. The FS scan is recorded at the Orbitrap analyzer in the mass
range from m/z 380 to 1600 at resolution 60,000 (FMHW,
m/z 400) and the FS data are acquired in profile mode.
7. The fragment spectra (MS/MS) are acquired at the ion trap
(LTQ) in centroid mode (see Note 14).
8. The fragments are generated using collision-induced dissocia-
tion mode, with no fragmentation at the source considered.
9. The five most intensive ions per FS are selected for the frag-
mentation such that each mass ion with intensity greater than
500 counts is fragmented once and inserted into a dynamic
exclusion list for 60 s (see Note 15).
10. Other parameters set for the fragmentation are as follows: (1)
30 s repeat duration time; (2) 2 μm isolation width (m/z); (3)
30 ms activation time; (4) 35 V normalized collision energy;
and (5) Q = 0.250 activation.
www.Ebook777.com
3.5 Protein 1. Convert the MS raw datad to *.XML format using BioWorks
Identification tool or similar before searching against a decoy UniProt mouse
protein database (see Note 16).
2. Set the search parameters as follows: (1) trypsin as enzyme; (2)
one missed cleavage allowed; and (3) cysteine carbamidometh-
ylation (fixed) and methionine oxidation (variable) chosen as
modifications.
3. Precursor and fragment ion mass accuracies are set to 15 ppm
and 0.8 Da, respectively.
4. Peptides are accepted as accurate if the confidence interval (CI) is
95 %, the Mascot score is greater than 30, and if at least two unique
and nonredundant peptides can be matched to the protein.
3.6 Computational 1. In each co-IP MS dataset, check for the presence of the bait
Analysis (See Note 17) protein in terms of peptide signal intensities and sequence cov-
erage (see Note 18).
2. Identify and discard “mock IP” proteins and other proteins
introduced during sample handling (e.g., keratin, IgGs) iden-
tified in multiple co-IPs and MS experiments.
3. Generate a single protein list by determining common proteins
identified in multiple, independent co-IP MS experiments,
with candidates identified in more than one co-IP MS replicate
selected as the most reliable (see Note 19).
4. Sort proteins on the basis of the sequence coverage as those
identified by only a small number of peptides may be weak or
dynamic interactors.
5. Use interactome databases to look for the bait protein and
known interactors (see Note 20).
6. Perform a detailed sequence analysis of the interactors and sort
those that share subdomains as this may be an additional indi-
cator of the validity of the identification.
7. Make a list of the novel interactors.
8. For novel protein-protein interactions, use reverse co-IP fol-
lowing the above steps to verify the specificity of the interac-
tion with the bait protein.
3.7 Western Blot 1. Separate a fraction of the resuspended pellet containing the
(See Note 21) immunoprecipitated proteins by SDS-PAGE and electrotrans-
fer the resulting protein bands to an Immobilon PVDF mem-
brane at 100 V for 1 h at 4 °C.
2. Block the membranes for 1 h at room temperature with 5 %
nonfat dry milk in TBS-T.
3. Incubate with the selected primary antibody against the candi-
date protein overnight at 4 °C and then for 1 h at room tem-
perature with the corresponding secondary antibody.
4. Detect the immune complexes using ECL Plus.
3.8 Reciprocal IP 1. Perform reciprocal IPs followed by Western blot to further

Followed by Western validate a potential interacting molecule that has been iso-
Blot lated and identified with the co-IP methods described above
(see Note 22).
4 Notes
1. This protocol can be used to identify associated molecules of

any protein of interest as long as an antibody against the protein
is available. Likewise, any kind of biological material can be
used, such as cells, a tissue, or extracellular matrix. As an exam-
ple, we describe a workflow for the interactome analysis of
B-Raf performed in HT22 cell lysates. The steps include immu-
noprecipitation of B-Raf and its protein partners under native
conditions using co-IP followed by shotgun mass spectrometry
for the identification of the interactome. B-Raf interacting part-
ners were validated with non-MS-based approaches such as
Western blot and reciprocal IP followed by Western blot.
2. If the bait protein is expressed at low levels, it may be desirable
to use three or more 100 mm plates and harvest them in smaller
volumes of modified RIPA buffer (0.5 or 0.75 mL per plate).
3. Buffer composition (salinity, detergent, and concentration) is
critical for a successful co-IP experiment. Optimal conditions
are empirical and should be determined for each investigated
protein/antibody complex. The buffer composition suggested
in this protocol is a good starting point. Normally, reducing
salt concentration allows for the immunoprecipitation of more
proteins. However, the chance of identifying false positives
(non-true preys) is also increased. If there is a special interest in
identifying phosphorylated proteins, the modified RIPA buffer
should contain higher concentrations of activated sodium
orthovanadate (1/25 of stock solution).
4. Proteins A and G exhibit variation in binding strength to dif-
ferent immunoglobulins (Ig). This variation exists between
different species and between different antibody subclasses
from the same species. The decision to use proteins A, G, or
A/G depends on the species in which the antibody was pro-
duced [14] (see Table 1).
5. The concentration of immunoprecipitation antibody and the
incubation time may be specific for each protein target. Optimal
conditions need to be set up experimentally. However, the
amounts and times suggested here can be considered good
starting points.
Table 1
Protein A and protein G exhibit variation in binding strength to different immunoglobulins (Ig) [14,
15]. This exists both between different species and between different antibodies subclasses from
the same species. The decision of using protein A or G or A/G in an IP experiment depends on the
specie in which the antibody was produced. W = weak binding, M = medium binding, S = strong
binding, NB = no binding
Species Antibody Class Protein A Protein G Protein A/G

Human Total IgG S S S
IgG1, IgG2 S S S
IgG3 W S S
IgG4 S S S
IgM W NB W
IgD NB NB NB
IgA W NB W
IgA1, IgA2 W NB W
IgE M NB M
Fab W W W
ScFv W NB W
Mouse Total IgG S S S
IgM NB NB NB
IgG1 W M M
IgG2a, IgG2b, IgG3 S S S
Rat Total IgG W M M
IgG1 W M M
IgG2a NB S S
IgG2b NB W W
IgG2c S S S
Cow Total IgG W S S
IgG1 W S S
IgG2 S S S
Goat Total IgG W S S
IgG1 W S S
IgG2 S S S
Sheep Total IgG W S S
IgG1 W S S
(continued)
Table 1
(continued)
Species Antibody Class Protein A Protein G Protein A/G

IgG2 S S S
Horse Total IgG W S S
IgG(ab), IgG(c) W NB W
IgG(T) NB S S
Rabbit Total IgG S S S
Guinea pig Total IgG S W S
Hamster Total IgG M M M
Pig Total IgG S W S
Donkey Total IgG M S S
Cat Total IgG S W S
Dog Total IgG S W S
Monkey Total IgG S S S
Chicken Total IgY NB NB NB
6. The mock IP sample can be used both as a negative control in

Western blot for the validation of binding proteins (preys) and
in the identification of the nonspecific binding proteins by
LC-ESI-MS/MS.
7. To reduce the number of nonspecific proteins, washes can be
performed with gentle agitation at 4 °C for 5 min. Furthermore,
additional washes can be implemented to determine the speci-
ficity and strength of the identified partners.
8. The pellet can be stored at −20 °C for a maximum of 7 days.
9. In some cases, loading smaller amounts of sample may result in
a better separation in the gel, especially if the concentration of
immunoprecipitating antibody is high or the immunoprecipi-
tated protein interacts with high numbers of other proteins,
such as molecular chaperones.
10. Overnight trypsin digestion does not yield more peptides but
results in more autolysis trypsin peptides.
11. Before filtering the samples, wet the filter membrane device by
adding 0.1 % HCOOH, then centrifuge and discard it. This
additional step increases sample recovery.
12. The picofrit columns are not coated. In our system, the ioniza-
tion occurs via the liquid junction and is highly stable across
the LC-MS/MS analysis.
13. The use of HPLC-grade bottled water reduces the likelihood
of clogging the nano-column by particulate contamination.
Particulate contamination can be carbon particles released
from the in-house meg-ohm water system.
14. The centroid spectra have a small file size because they contain
less information describing the m/z signal. It is advisable to
reduce the file size of an LC-MS/MS run by running the FS in
profile mode and the MS/MS spectra in centroid mode. The
MS/MS information content in centroid spectra is sufficient to
identify the fragment sequence.
15. Mono-charged ions are rejected for MS/MS.
16. We use an in-house version of the Mascot search engine for this.
17. A crucial part of the interactome analysis is to filter the nonspe-
cific binding molecules from true physiological interactors. A
minimum of two independent co-IP MS analyses should be
run to generate a list of interactors (preys) of a target protein.
Peptides are considered if they are identified with a false dis-
covery rate (FDR) value between 0.5 and 1 %.
18. This is a critical requirement for the subsequent validation
studies.
19. A large bait interactome (>100 interactors) may indicate the
identification of proteins that cross-interact, either directly or
indirectly [6]. The BioVenn tool (http://www.cmbi.ru.nl/
cdd/biovenn/) can be used to compare data from a maximum
of three experiments.
20. The interactome databases IntAct [15], PINA, [16], BioGRID
[17]. Report the methods used for the identification of the
interacting proteins. This information can also confirm the
validity of the identification. Interactors identified by more
than one method are more reliable than those identified by a
single method and any previously identified interactor found in
the dataset serves as an internal validation and an assessment of
the data quality.
21. Selected candidates are validated by Western blot of resolved
co-immunoprecipitated proteins.
22. In this case, the reciprocal IP should be performed in cell lysates
and the proteins should be resolved by SDS-PAGE and analyzed
by Western blot using antibodies against the bait protein (B-Raf
in our example). Furthermore, the reciprocal IP stage can help
to circumvent the problem of immunodetecting proteins that
might be obscured due to their migration close to the IgG heavy
or light chains during SDS-PAGE (e.g., vimentin).
Acknowledgments
D.M.S. is funded by FAPESP (São Paulo Research Foundation,

grants 2013/08711-3, 2013/25702-8, and 2014/10068-4).
References
1. Sanchez C, Lachaize C, Janody F, Bellon B, 9. Garnett MJ, Marais R (2004) Guilty as charged:
Röder L, Euzenat J et al (1999) Grasping at B-RAF is a human oncogene. Cancer Cell
molecular interactions and genetic networks in 6:313–319
Drosophila melanogaster using FlyNets, an 10. Zhang JH, Simonds WF (2000) Copurification
Internet database. Nucleic Acids Res of brain G-protein beta5 with RGS6 and
27:89–94 RGS7. J Neurosci 20:RC59
2. Guruharsha KG, Rual JF, Zhai B, Mintseris J, 11. Ivings L, Pennington SR, Jenkins R, Weiss JL,
Vaidya P, Vaidya N et al (2011) A protein com- Burgoyne RD (2002) Identification of Ca2+-
plex network of Drosophila melanogaster. Cell dependent binding partners for the neuronal
147:690–703 calcium sensor protein neurocalcin delta: inter-
3. Hu P, Janga SC, Babu M, Díaz-Mejía JJ, action with actin, clathrin and tubulin. Biochem
Butland G, Yang W et al (2009) Global func- J 363:599–608
tional atlas of Escherichia coli encompassing 12. Hadlock GC, Nelson CC, Baucum AJ 2nd,
previously uncharacterized proteins. PLoS Biol Hanson GR, Fleckenstein AE (2011) Ex vivo
7, e96. doi:10.1371/journal.pbio.1000096 identification of protein-protein interactions
4. Coulombe B (2011) Mapping the disease pro- involving the dopamine transporter. J Neurosci
tein interactome: toward a molecular medicine Methods 196:303–307
GPS to accelerate drug and biomarker discov- 13. Gorini G, Ponomareva O, Shores KS, Person
ery. J Proteome Res 10:120–125 MD, Harris RA, Mayfield RD (2010) Dynamin-1
5. Williamson MP, Sutcliffe MJ (2010) Protein- co-associates with native mouse brain BKCa
protein interactions. Biochem Soc Trans channels: proteomics analysis of synaptic protein
38:875–878 complexes. FEBS Lett 584:845–851
6. Markham K, Bai Y, Schmitt-Ulms G (2007) 14. Richman DD, Cleveland PH, Oxman MN,
Co-immunoprecipitations revisited: an update Johnson KM (1982) The binding of staphylo-
on experimental concepts and their imple- coccal protein A by the sera of different animal
mentation for sensitive interactome investiga- species. J Immunol 128:2300–2305
tions of endogenous proteins. Anal Bioanal 15. Kerrien S, Aranda B, Breuza L, Bridge A,
Chem 389:461–473 Broackes-Carter F, Chen C et al (2011) The
7. Bonfiglio JJ, Maccarrone G, Rewerts C, IntAct molecular interaction database in 2012.
Holsboer F, Arzt E, Turck CW et al (2011) Nucleic Acids Res 40:D841–D846
Characterization of the B-Raf interactome in 16. Cowley MJ, Pinese M, Kassahn KS, Waddell N,
mouse hippocampal neuronal cells. J Proteomics Pearson JV, Grimmond SM et al (2012) PINA
74:186–198 v2.0: mining interactome modules. Nucleic
8. Bonfiglio JJ, Inda C, Senin S, Maccarrone G, Acids Res 40:D862–D865
Refojo D, Giacomini D et al (2013) B-Raf 17. Stark C, Breitkreutz BJ, Chatr-Aryamontri A,
and CRHR1 internalization mediate bipha- Boucher L, Oughtred R, Livstone MS et al
sic ERK1/2 activation by CRH in hippo- (2010) The BioGRID Interaction Database:
campal HT22 Cells. Mol Endocrinol 2011 update. Nucleic Acids Res 39:
27:491–510 D698–D704
Chapter 20
Using 15N-Metabolic Labeling for Quantitative Proteomic

Analyses
Giuseppina Maccarrone, Alon Chen, and Michaela D. Filiou
Abstract
Quantitative proteomics has benefited from the application of stable isotope labeling-based approaches.
Using stable isotopically labeled material as an internal standard in proteomic comparisons allows an unbi-
ased and accurate quantification of protein expression level changes. Here, we describe the use of in vivo
15
N metabolic labeling to generate labeled protein standards from mice. We then present a protocol includ-
ing sample preparation, mass spectrometry, and data analysis workflows using these standards to compare
unlabeled proteomes. We focus on mouse brain tissue and plasma samples, although this conceptual frame-
work can be applied to most organisms.
15
Key words N metabolic labeling, Quantitative proteomics, Mass spectrometry, Peptide
quantification
1 Introduction
Quantitative proteomics enables the comparison of proteomes to

identify altered protein expression levels at a given time point.
State-of-the-art quantitative proteomic methodologies include the
use of stable nonradioactive isotopes to generate labeled internal
standards for subsequent comparisons of unlabeled proteomes.
The advantage of this approach is the high quantification accuracy
due to the fact that handling biases during sample preparation are
eliminated [1, 2]. Here, we describe a quantitative proteomic
workflow using the 15N nitrogen isotope, which contains one more
neutron compared to the most abundant 14N form found in nature.
To generate 15N-labeled material, mice are fed with a
15
N-labeled, bacteria-based diet [3]. Then, 15N-labeled biosamples
such as brain tissue or plasma are used as internal standards for
quantitative proteomic comparisons of two unlabeled groups of
interest (hereafter named A and B). Briefly, an unlabeled (14N) ver-
sion of proteome A is mixed at a 1:1 ratio based on tissue weight
or protein content with the 15N-labeled internal standard (A/15N).
235
The A/15N protein mix is then resolved according to molecular

weight by gel electrophoresis and the protein bands are excised
and enzymatically digested to produce peptides, which are ana-
lyzed by mass spectrometry. The same procedure is applied to the
second group B (B/15N). In both cases, analysis of the 14N/15N
peptide mixtures by liquid chromatography-tandem mass spec-
trometry (LC-MS/MS) results in the unlabeled (14N) and the
15
N-labeled versions of the same peptide being separated in the
mass spectrometer due to their different mass to charge ratios
(m/z), leading to generation of two distinct signals. In silico quan-
tification of the 14N and 15N peptide signals reveals their relative
abundance in the mixture. Next, the data from the 14N/15N A and
B comparisons are combined in silico to provide an indirect com-
parison between the two proteomes. The methodological concept
followed in this study is shown in Fig. 1.
14N_A 15N 14N_B
14N_A:15N (1:1) 14N_B:15N (1:1)
Sample processing Sample processing
LC-ESI-MS/MS LC-ESI-MS/MS
14N_A 14N_B
15N Quantification 15N
m/z m/z
14N_A/15N 14N_B/15N
14N_A/14N_B
Fig. 1 Principle of peptide/protein quantification using 15N-labeled internal standards

15
N Metabolic Labeling-Based Proteomics 237
Protein quantification by 15N metabolic labeling is character-

ized by high repeatability that depends on the complexity of the
biological material analyzed [4]. Importantly, by using a 15N-labeled
internal standard, any existing 15N isotope effect on the labeled
proteome will not influence the quantification and data interpreta-
tion steps. We have previously used in vivo 15N metabolic labeling
to identify protein expression changes in mouse models of neuro-
psychiatric phenotypes [5–11].
2 Materials
15
2.1 15N Metabolic 1. N-labeled, bacteria-based diet U-15 N-SILAM-Mouse
Labeling: Sample (Silantes GmbH; Munich, Germany) (see Note 1).
Acquisition 2. Standard mouse diet (Altromin; Lage, Germany).
3. 0.9 % NaCl solution for perfusion.
2.2 Sample 1. Protein content estimation assay reagents (see Note 2).
Preparation 2. Reagents and equipment for carrying out electrophoresis of
for Proteomic Analysis biosamples (see Note 3).
3. 4× sodium dodecyl sulfate polyacrylamide gel electrophoresis
(SDS-PAGE) loading buffer: 125 mM Tris–HCl (pH 6.8),
9.2 % SDS (w/v), 6 % dithiothreitol (w/v), 40 % glycerol (v/v),
and 0.01 % Bromophenol blue (w/v).
4. Costar SpinX tubes (Corning Inc; Glendale, AZ, USA).
5. Multiple Affinity Removal Spin cartridge mouse 3 (MARS3;
Agilent Technologies; Santa Clara, CA, USA) (see Note 4).
6. Buffer A and Buffer B (Agilent Technologies).
7. Coomassie Brilliant Blue R-250.
8. Ammonium bicarbonate (Merck Millipore; Darmstadt,
Germany).
9. Dithiothreitol (BioRad).
10. Iodoacetamide (BioRad).
11. Formic acid (Merck Millipore) (see Note 5).
12. Acetonitrile (Merck Millipore) (see Note 6).
13. Porcine trypsin (Promega; Madison, WI, USA).
2.3 Mass 1. LC buffer: 95 % acetonitrile/0.1 % formic acid.

Spectrometry: 2. In-house packed fused silica 3 μm RP-C18 (0.075 mm × 15-20 cm)
Proteomics Data chromatographic columns (Maisch; Monheim, Germany).
Analysis
3. LC-MS/MS: nano HPLC-2D system (Eksigent; Dublin, CA,
USA) coupled online to an LTQ-Orbitrap mass spectrometer
via a nano electrospray ion source (ThermoFisher Scientific;
Bremen, Germany).
4. Databases: BioWorks (v3.3.1) and SEQUEST (v28) (Thermo

Fischer Scientific; San Jose, CA, USA).
3 Methods
3.1 15N-Metabolic 1. Mice are mated and once pregnancy is detected, remove the
Labeling: Sample male mice from the cages and subject the pregnant females to
Collection an ad libitum diet containing a 1:1 mixture of standard food
and the 15N-labeled bacteria-based diet for 4 days.
2. Following this 4 day period, feed the pregnant mice exclusively
with the 15N-labeled bacteria-based diet throughout pregnancy
until weaning and then feed the offspring with the same diet
for a total of 56 days post-partum [3] (see Note 7).
3. Anesthetize the 15N-labeled 2-month-old mice and the brain
following perfusion with the NaCl solution.
4. Dissect brain areas of interest according to the mouse brain
atlas [12].
5. Acquire plasma by centrifuging blood samples at 13,000 × g for
10 min at 4 °C.
6. Collect the plasma supernatants, snap-freeze these in liquid
nitrogen, and store at −80 °C.
3.2 Sample 1. Plasma samples (Fig. 2a): For depletion of highly abundant
Preparation plasma proteins, combine 400 μg of 14N and 15N-labeled
for Proteomic Analysis plasma protein samples at a 1:1 (w/w) ratio (see Note 8).
2. Dilute with 200 μL of buffer A and filter using the Costar
SpinX tubes according to the manufacturer’s instructions.
3. Apply the flow through to the MARS3 cartridge according to
the manufacturer’s instructions.
4. Collect the cartridge flow through volume corresponding to
the depleted proteome and estimate the protein concentration
by Bradford assay (see Note 9).
5. Wash the cartridge to elute the bound proteins using buffer B
(see Note 10).
6. Brain samples (Fig. 2b): Enrich brain region subproteomes if
required using appropriate homogenization buffers and stan-
dard procedures.
7. For the enriched brain and depleted plasma proteomes, mix
50–150 μg of the 14N/15N protein extracts in the 4× SDS-
PAGE sample buffer to achieve a final 1x concentration, heat
at 95 °C for 5 min, and leave to cool at room temperature.
8. Load the 14N/15N protein extracts onto mini 12.5 % SDS-
PAGE gels and resolve the proteins by gel electrophoresis
(see Note 11).
15
14N 15N 14N 15N
14N Plasma 15N 15N Brain 14N
14N:15N (1:1, w/w)

Brain protein extraction
14N : 15N (1:1, w/w)
Plasma depletion
LC-ESI-MS/MS LC-ESI-MS/MS
14N 14N
15N 15N
m/z m/z
14N/15N 14N/15N
A B
Fig. 2 14N/15N relative protein quantification workflow for mouse plasma and brain. (a) Plasma samples are
mixed 1:1 based on protein content and the 14N/15N mixture is depleted of three highly abundant plasma pro-
teins. (b) Brain samples are either mixed 1:1 based on tissue weight and then the subproteome of interest is
enriched or the subproteome of interest is extracted first and then the protein extracts are mixed 1:1 based on
protein content. Protein mixtures are separated based on molecular weight by gel electrophoresis and ana-
lyzed by LC-MS/MS
9. Stain the protein bands in the gel with Coomassie.

10. Slice each gel lane into approximately 20 fractions and cut
these further to produce smaller pieces.
11. Wash the gel pieces twice with 25 mM ammonium bicarbonate
in 50 % acetonitrile.
12. Perform reduction with 10 mM dithiothreitol in 25 mM
ammonium bicarbonate for 30 min at 56 °C (see Note 12).
13. Perform carboxymethylation with 50 mM iodoacetamide in
25 mM ammonium bicarbonate for 30 min at room tempera-
ture (see Note 13).
14. Wash the gel pieces twice 25 mM ammonium bicarbonate in
50 % acetonitrile and dry them in a vacuum centrifuge.
15. Rehydrate the gel pieces in 5 ng/μL trypsin in 25 mM ammo-

nium bicarbonate using a volume that is just enough to cover
the pieces.
16. Leave for 15 min on ice, remove the surplus trypsin solution,
and cover the gel pieces with 25 mM ammonium
bicarbonate.
17. Incubate the samples for 4–6 h at 37 °C (see Note 14).
18. Extract the resulting peptides with 50 % acetonitrile/2 % for-
mic acid twice (see Note 15).
19. Lyophilize the peptide extracts and store them at −80 °C for
mass spectrometry analysis (see Note 16).
3.3 Mass 1. Dissolve the peptides in 10 μL 0.1 % formic acid.

Spectrometry 2. Load 5 μL per fraction onto an in-house packed column and ana-
(LC-ESI-MS/MS): lyze using a 2.5 h gradient that includes washing with 0.1 % for-
Proteomics Data mic acid for 20 min followed by elution with 95 % acetonitrile/0.1 %
Analysis formic acid from 2 to 45 % at a flow rate of 200 nl/min.
3. Operate the LTQ-Orbitrap in positive ion mode, using a data-
dependent automatic scan switch between the MS and MS/
MS acquisitions.
4. Record full scans in the mass analyzer with a m/z range of
380–1600 in profile mode.
5. Use a TOP5 method during which the MS/MS analysis of the
five most intense peptide ions for each scan is recorded in the
LTQ mass analyzer in centroid mode (see Note 17).
6. Search each raw file generated by mass spectrometry twice:
once against a 14N and once against a 15N decoy mouse data-
base using BioWorks (v3.3.1) and SEQUEST (v28) software
(Thermo Fischer Scientific, San Jose, CA). Use the following
parameters: peptide mass tolerance = 20 ppm; fragment ion
mass tolerance = 1 Da; enzyme = trypsin; missed cleavage
sites = 2; only tryptic peptides allowed; static modification = cys-
teine carboxyamidomethylation; variable modification = methi-
onine oxidation.
7. Export 14N and 15N identified peptides as DTA files and use
DTAselect [13] to filter the peptide identifications and
assemble the peptides into proteins using the following filter-
ing parameters: minimum delta CN = 0.08; cross correlation
score (Xcorr) = 1.90 (z = 1+), 2.70 (z = 2+), 3.50 (z ≥ 3+); purg-
ing duplicate spectra on the basis of Xcorr (−t = 2); minimum
charge state = 1; maximum charge state = 6; minimum of 1
identified peptide per protein (see Note 18).
8. Use ProRata software [14] for ion chromatogram extraction
and relative 14N/15N peptide and protein quantification
(see Note 19).
15
9. Combine the two indirect 14N/15N comparisons (A/15N and

B/15N) to generate protein quantification differences between
groups A and B.
4 Notes
1. The 15N-labeled bacteria-based diet should be stored long

term at 4 °C.
2. We use the Bradford assay although many protein assay kits can
be used. The user should ensure compatibility of the reagents
in the kit with those in the sample extracts.
3. We use reagents and equipment from Bio-Rad (Hercules, CA,
USA) although many others can be used.
4. This is for depletion of the three abundant plasma proteins
(albumin, IgG, and transferrin).
5. Other suppliers can be used although all reagents should be
HPLC grade.
6. Other suppliers can be used although all reagents should be
HPLC grade.
7. This feeding protocol typically results in greater than 90 % 15N
incorporation rates into mouse brain and plasma. 15N incorpo-
ration rates in a sample can be assessed by using the QuantiSpec
[15] or Atomizer [16] software.
8. This can be done based on original tissue weight in the case of
brain samples or on protein content for both brain and plasma
samples. Protein concentrations can be determined using the
Bradford assay. When 14N and 15N-labeled samples are mixed
according to protein content, homogenization buffers com-
patible with the Bradford assay should be used. In case of ana-
lyzing a brain subproteome (e.g., synaptosomes), it is also
advised to enrich the same subproteome for the 15N-labeled
brain to avoid dilution effects when using a whole 15N-labeled
brain region extract.
9. This step is recommended to expand the dynamic range of the
plasma proteome analysis.
10. This step regenerates the column for future use.
11. In case a higher resolution is required for proteins of a certain
molecular weight, gels with different percentages of acrylamide
can be used.
12. This step reduces the disulfide bonds in proteins.
13. This step alkylates the reduced sulfide groups.
14. Digestion with trypsin overnight does not increase the number
of generated peptides but results instead in a higher number of
autolysis (self-digestion of trypsin) products.
15. To reduce the mass spectrometry run time if required, peptide

extracts from individual gel slices can be combined and ana-
lyzed together in a single run.
16. Storage of lyophilized extracts in −80 °C is the most reliable
long-term storage option to preserve sample quality.
17. All other parameters should be set as described previously
[17, 18].
18. Manual validation of differential expression is necessary to
avoid false positives.
19. In order to report protein groups instead of individual pro-
teins, the Trans-Proteomic Pipeline (TPP) should be used for
peptide filtering and grouping [19].
Acknowledgments
This work was funded by the Max Planck Society and a BMBF
Quant Pro grant. The authors thank Prof. Daniel Martins de Souza
for insightful comments and discussions.
References
1. Filiou MD, Martins-de-Souza D, Guest PC, 7. Zhang YY, Filiou MD, Reckow S, Gormanns P,
Bahn S, Turck CW (2012) To label or not to Maccarrone G, Kessler MS et al (2011)
label: applications of quantitative proteomics in Proteomic and metabolomic profiling of a trait
neuroscience research. Proteomics 12:736–747 anxiety mouse model implicate affected path-
2. Filiou MD (2013) The potential of 15N meta- ways. Mol Cell Proteomics 10:M111.008110.
bolic labeling for schizophrenia research. Arch 8. Lopes S, Teplytska L, Vaz-Silva J, Dioli C,
Clin Psychiatr 40:51–52 Trindade R, Morais M et al. (2016) Tau dele-
3. Frank E, Kessler MS, Filiou MD, Zhang Y, tion prevents stress-induced dendritic atrophy
Maccarrone G, Reckow S et al (2009) Stable in prefrontal cortex: Role of synaptic mito-
isotope metabolic labeling with a novel chondria. Cereb Cortex Apr 12. pii: bhw057
15
N-enriched bacteria diet for improved pro- (Epub ahead of print)
teomic analyses of mouse models for psychopa- 9. Filiou MD, Turck CW, Martins-de-Souza D
thologies. PLoS ONE 4, e7821 (2011) Quantitative proteomics for investigat-
4. Filiou MD, Soukupova M, Rewerts C, ing psychiatric disorders. Proteomics Clin Appl
Webhofer C, Turck CW, Maccarrone G (2015) 5:38–49
Variability assessment of 15N metabolic 10. Kao CY, He Z, Henes K, Asara JM, Webhofer
labeling-based proteomics workflow in mouse C, Filiou MD et al (2016) Fluoxetine treat-
plasma and brain. Mol Biosyst 11:1536–1542 ment rescues energy metabolism pathway alter-
5. Filiou MD, Teplytska L, Otte DM, Zimmer A, ations in a posttraumatic stress disorder mouse
Turck CW (2009) Myelination and oxidative model. Mol Neuropsychiatry 2:46–59
stress alterations in the cerebellum of the G72/ 11. Nussbaumer M, Asara JM, Teplytska L,
G30 transgenic schizophrenia mouse model. Murphy MP, Logan A, Turck CW et al (2016)
J Psychiatr Res 46:1359–1365 Selective mitochondrial targeting exerts anxio-
6. Filiou MD, Zhang Y, Teplytska L, Reckow S, lytic effects in vivo. Neuropsychopharmacology
Gormanns P, Maccarrone G et al (2011) 41:1751–1758
Proteomics and metabolomics analysis of a trait 12. Paxinos G, Franklin K (2001) The mouse brain
anxiety mouse model reveals divergent mito- in stereotaxic coordinates, 2nd edn. Academic,
chondrial pathways. Biol Psychiatry 70:1074–1082 San Diego, CA. ISBN 13: 978-0123887214
15
13. Tabb DL, McDonald WH, Yates JR 3rd 16. MacCoss MJ, Wu CC, Matthews DE, Yates JR
(2002) DTASelect and contrast: Tools for 3rd (2005) Measurement of the isotope enrich-
assembling and comparing protein identifica- ment of stable isotope-labeled proteins using
tions from shotgun proteomics. J Proteome high-resolution mass spectra of peptides. Anal
Res 1:21–26 Chem 77:7646–7653
14. Pan C, Kora G, McDonald WH, Tabb DL, 17. Filiou MD, Bisle B, Reckow S, Teplytska L,
VerBerkmoes NC, Hurst GB et al (2006) Maccarrone G, Turck CW (2010) Profiling of
ProRata: a quantitative proteomics program mouse synaptosome proteome and phosphopro-
for accurate protein abundance ratio estima- teome by IEF. Electrophoresis 31:1294–1301
tion with confidence interval evaluation. Anal 18. Filiou MD, Turck CW (2012) Psychiatric dis-
Chem 78:7121–7131 order biomarker discovery using quantitative
15. Haegler K, Mueller NS, Maccarrone G, proteomics. Methods Mol Biol 829:531–539
Hunyadi-Gulyas E, Webhofer C, Filiou MD 19. Keller A, Eng J, Zhang N, Li XJ, Aebersold R
et al (2009) QuantiSpec – quantitative mass (2005) A uniform proteomics MS/MS analysis
spectrometry data analysis of 15N-metabolically platform utilizing open XML file formats. Mol
labeled proteins. J Proteomics 71:601–608 Syst Biol 1:2005.0017, Epub 2005 Aug 2
Chapter 21
Multiplex Measurement of Serum Folate Vitamers

by UPLC-MS/MS
Sarah Meadows
Abstract
The implementation of a liquid chromatography–tandem mass spectrometry (LC-MS/MS) method to
measure six folate vitamers in serum samples allows a more individual and accurate measurement than the
commonly used immunoassays. In the described method, serum samples undergo solid phase extraction fol-
lowed by liquid chromatography coupled with electrospray ionization tandem mass spectrometry with a run
time of 3.5 min. Recovery is 95 % for the most important folate metabolite, 5-methyltetrahydrofolate (MTHF),
and greater than 78 % for other minor folate forms. The limit of detection ranges from 0.2 to 0.4 nmol/L with
a intra-batch imprecision of less than 7 % for all analytes and calibration ranges of 1–100 nmol/L for MTHF
and 0.5–20 nmol/L for the minor folate forms, with greater than 0.99 R2 linearity.
Key words Folate, 5-Methyltetrahydrofolate, Folic acid, Solid phase extraction, LC-MS/MS
1 Introduction
Folate is the general term for a water-soluble B vitamin naturally

found in foods such as leafy vegetables, legumes, egg yolks, liver
and some citrus fruits. Folic acid itself does not occur naturally but
can be obtained from vitamin supplements or by eating fortified
foods [1, 2]. A number of critical cellular pathways depend on
folate, including DNA, RNA, and protein methylation as well as
DNA synthesis and maintenance [2, 3]. Because of this, there are
many health consequences of folate deficiency among all age
groups. These include megaloblastic anemia, depression, cognitive
impairment, low birth weight, risk of placental abruption, neural
tube defects, and other birth-related abnormalities, including oro-
facial clefts and heart defects [4, 5].
The most widely publicized issue surrounding folate is that of
low levels during pregnancy causing birth defects associated with
the nervous system. Folate is an essential micronutrient during
fetal development due to its importance in transmethylation
reactions and synthesis of DNA in growing cells [3]. A significant
245
246 Sarah Meadows
portion of the 300,000 neural tube defects (NTDs) that occur

yearly worldwide continue to be a great public health burden
although these are preventable by the periconceptual consumption
of folic acid [2]. NTDs occur early in pregnancy (neural tube clo-
sure is completed by embryonic day 21–28) [6, 7] and arise when
the tube fails to close properly leading to anencephaly, encephalo-
cele and spina bifida. The demonstration that periconceptional
supplementation with folic acid dramatically reduces the incidence
of NTDs has generated considerable clinical and public health
interest, which has led to fortification with folic acid of the food
supply in the USA and some other countries [8]. The recent WHO
guidelines for optimal serum and red blood cell folate levels in women
of reproductive age for prevention of neural tube defects, states that
in 2012 an estimated 270,358 deaths globally were attributable to
congenital abnormalities during the first 28 days of life.
In addition to neural tube closure, B vitamins including folate
are required for brain metabolic pathways and are fundamental in
all aspects of brain development and maintenance of brain health
throughout life. Observational and animal evidence appears to
support a role for maternal folate status in later cognitive perfor-
mance of the resulting children and there are also studies linking
low maternal folate status with a higher incidence of behavioral
and emotional problems, inattention, and hyperactivity in the
offspring [7]. Recent studies have also shown links between low
plasma folate and poor cognitive performances in children and
adolescents, as well as a positive association between higher dietary
folate intake and academic achievement [7].
Mental dysfunction in the elderly ranging from cognitive
impairment to dementia is also a matter of concern. Brain changes
occur long before the diagnosis of dementia is made and, given the
current increases in life expectancy, the numbers of individuals suf-
fering from this condition is expected to double by 2025. Therefore
it is important to find early biomarkers which would enable timely
interventions to delay the onset or slow the progression of the
disease. There is emerging evidence suggesting that suboptimal
status of folate and metabolically related B vitamins may be linked
with cognitive dysfunction and dementia. If cognitive dysfunction
can be slowed or prevented by improving the B vitamin status in
healthy older people it could offer a cost effective preventative
public health strategy in aging populations [7].
Evidence showing that supplementation with folic acid pro-
tects against NTDs has led to worldwide governmental recom-
mendations advising all women planning a pregnancy to consume
400 μg of folic acid per day from the time of preconception until
the end of the first trimester of pregnancy [2, 7, 9–11]. Even with
this knowledge public health campaigns remain largely unsuccessful
[9] since up to 50 % of all pregnancies are unplanned in countries
such as the USA [2]. Mandatory fortification programs have been
implemented in many countries to improve folate status and reduce
Multiplex Folate Measurement 247
the current high costs associated with prevention programs such as

education campaigns and other interventions that require behav-
ioral change [2]. The Scientific Advisory Committee on Nutrition has
called for mandatory fortification in the UK to replace voluntary
fortification in a bid to increase the country’s folate status [10].
Despite the unequivocal success of folic acid in reducing NTD
rates, several studies have questioned the harmful role of unme-
tabolized folic acid in blood [3, 11]. Concerns have been raised
that due to fortification, the subsequent increase in folic acid
intakes across the population may have harmful effects on health,
such as the masking of pernicious anemia, colorectal cancer pro-
motion in people with preexisting lesions or adverse cognitive
effects in the elderly with low vitamin B12 status [4, 9, 11, 12].
Measurement of unmetabolized folic acid has been suggested as a
way of monitoring folic acid intake [2, 5]. At a time when there are
still questions around the safety of high levels of folic acid in the
blood, the ability to differentiate between this and endogenous
folates is valuable.
Folate has traditionally been measured using a microbiological
assay but since the 1970s commercial protein binding assays on
automated clinical analysers have been widely employed due to the
ease of use of these platforms and the increased throughput they
offer. The microbiological assays are considered more accurate as
they can recover folate vitamers equally [13]. In contrast the pro-
tein binding assays generally underestimate folate concentrations
due to the different affinities of the folate vitamers for the antibody
used [1, 6]. Serum folate is considered to be an indicator of recent
folate intake whereas red blood cell folate concentrations indicate
long term status [6, 10]. Red cell folate is normally calculated
using whole blood folate, serum folate, and hemocrit. However,
repeated low concentrations of serum folate occurring within an
individual over the course of a month are also indicative of folate
depletion [6, 14]. It has proven to be more technically challenging
to measure whole blood folate by liquid chromatography tandem
mass spectrometry (LCMS/MS) than it is to measure serum folate.
This is because red blood cells require hemolysis to release the
polyglutamate folates, which then need to be deconjugated to
monoglutamates without any folate loss prior to analysis [6, 8].
This has led to whole blood tests only being carried out in specialist
laboratories which mostly use microbiological assays.
In contrast to both the microbiological assay and the radioiso-
tope protein-binding assay, chromatography techniques are able to
differentiate between individual folate species [6] and are now
often coupled to mass spectrometers with high sensitivity, specific-
ity, and selectivity compared to other detection methods [6, 8, 15].
The advantage of measuring the different folate species is likely to
become greater in the future as more information on genetic poly-
morphisms that affect nutritional status and folate distributions
become available [1]. Initial steps to standardize folate methods
248 Sarah Meadows
began with the development of higher order reference methods

that use isotope dilution with LC-MS/MS analysis. In addition,
recent advances in sample clean up procedures have led to increas-
ing use of LC/MS or LC-MS/MS procedures [5].
The method described here is for a routine LC-MS/MS analy-
sis for determination and quantitation of six folate vitamers in
serum: 5-methyltetrahydrofolate (MTHF), folic acid, tetrahydro-
folate (THF), 5-formyltetrahydrofolate (FTHF), 5, 10 methenyl-
tetrahydrofolate (CH + THF), and pyrazino-s-triazine derivative of
4α-hydroxy-5-methyl tetrahydrofolate (MeFox).
2 Materials
2.1 Specialized 1. Waters HSS T3 1.8 μ 2.1 × 100 mm with Waters Acquity HSS
Equipment T3 Vanguard precolumn (see Note 1).
2. AB Sciex 5500 Qtrap mass spectrometer.
3. 96-well plate positive pressure processor.
4. 96-well plate vacuum manifold and pump.
5. 50 μg phenyl 96-well SPE Plates.
6. PVDF 96-well filter plates.
2.2 Reagents 1. Stock 10 % ammonium formate, pH 3.2 (adjust using ammonium

hydroxide). Store at room temperature for up to 6 months
(see Note 2).
2. Conditioning buffer: 1 % ammonium formate pH 3.2. Make
up fresh for each assay.
3. Sample diluent: 1 % ammonium formate pH 3.2, 0.5 % ascorbic
acid. Make up fresh for each assay.
4. Wash buffer: 0.05 % ammonium formate pH 3.4, 0.1 % ascorbic
acid. Make up fresh for each assay.
5. Mobile phase solvent: water–methanol–acetonitrile
(4.95/4.0/1.0). Degas using sonication. Keep for up to 2 weeks
at room temperature.
6. Working mobile phase: 5 mM ammonium formate, 0.5 % ace-
tic acid in mobile phase solvent. Make up fresh for each assay.
7. Elution buffer: 0.5 % acetic acid, 0.5 % ascorbic acid in mobile
phase solvent. Make up fresh for each assay.
8. Working standard diluents: phosphate buffered saline (PBS)
pH 7.2, 4 % albumin, 0.5 % ascorbic acid. Make up fresh
each assay.
9. Strong needle wash: 100 % methanol.
10. Weak needle wash: 65 % acetonitrile.
11. Seal wash: 10 % methanol.
2.3 Standards 1. 0.1 M potassium phosphate pH 7.2. Store at room tempera-

and Internal Standards ture for up to 1 month.
(See Notes 2–6 2. 20–40 μg/mL folic acid/13C5 folic acid (Stock I). Remove
and Table 1) 1 mL for spectrophotometric analysis and store the remaining
solution in 5 mL aliquots in cryovials at −80 °C for up to 5
years. Calculate the exact concentration from the absorbance
reading on the spectrophotometer and make 20 μM solutions
(Stock II). Store in 300 μL aliquots in cryovials at −80 °C for
up to 1 year (see Notes 3–5).
3. 200–400 μg/mL MeFox / 13C5 MeFox (Stock I). Remove
1 mL for spectrophotometric analysis and store the remain-
ing solution at 4 °C. Calculate the exact concentration from
the absorbance reading on the spectrophotometer and make
100 μg/mL (Stock II) and 20 μM (Stock III) solutions
(see Table 2) Store in 1.2 mL aliquots in cryovials at −80 °C
for up to 5 years.
4. 200–400 μg/mL MTHF/13C5 MTHF, FTHF/13C5 FTHF,
THF/13C5 THF, and 5,10-methenylTHF/13C5
5,10-methenylTHF (Stock I). Remove 1 mL for spectro-
photometric analysis. To the remaining solution add 1 %
ascorbic acid and store at 4 °C. Calculate the exact concen-
tration from the absorbance reading on the spectrophotometer
(see Notes 6 and 7) and make 100 μg/mL (Stock II) and
20 μM (Stock III) solutions (see Table 2). Store in 1.2 mL
aliquots in cryovials at −80 °C for up to 5 years.
5. Mixed standard: 2 μM MTHF and 0.4 μM of other folates in
sample diluents. Make up fresh for each assay, using a new set
of stock aliquots each time.
Table 1
Diluents required for standard preparation
Analyte/labeled analyte Stock I diluent Stock II diluent Stock III diluent

MTHF 20 mM phosphate buffer 1 % aqueous 0.5 % aqueous
pH7.2 + 0.1%cysteine ascorbic acid ascorbic acid
Folic Acid 20 mM phosphate buffer pH7.2 Deionized water N/A
FTHF 20 mM phosphate buffer 1 % aqueous 0.5 % aqueous
THF 20 mM phosphate buffer 1 % aqueous 0.5 % aqueous
CH + THF 1 M HCl 1 M HCl + 1 % 0.5 M HCl + 0.5 %
ascorbic acid ascorbic acid
MeFox 0.1 M NaOH Deionized water 0.1 % aqueous
ascorbic acid
www.Ebook777.com
250 Sarah Meadows
Table 2
Composition of stock II and III standard solutions
Analyte Volume of Stock II (μL)

MTHF 919
FTHF 947
THF 891
CH + THF 911
MeFox 947
13
C5 MTHF 929
13
C5 FTHF 957
13
C5 THF 901
13
C5 CH + THF 923
13
C5 MeFox 957
Table 3
Volume of stock II required to make 20µM stock III solutions
Conc of MTHF/other Standard to Volume Volume of PBS + 0.4 %

Standard folates (nmol/L) dilute standard (μL) albumin (μL)
S1 Top Standard 100/20 Mixed standard 50 950
S2 75/15 Mixed standard 30 770
S4 25/5 Mixed standard 12.5 987.5
S6 5/1 S1 50 950
S7 1/0.2 S1 10 990
6. Working standards. Make up fresh each assay as shown in

Table 3.
7. Working internal standard: 60 nM 13C5 MTHF and 20 nM all
other 13C5 folates in 0.1 % ascorbic acid. Make up fresh for
each assay, using a new set of stock aliquots each time (Table 4).
2.4 Quality Control 1. In house pooled anonymized serum/plasma: (a) 10–15 nM

(QC) Material and 1–2 nM MTHF and folic acid, respectively; and (b)
20–30 nM and 8–12 nM MTHF and folic acid respectively.
Dispense into 350 μL aliquots in amber glass UPLC vials
Table 4
Parameters for calculating folate concentration from absorbance readings
Folic
MTHF/13C5 acid/13C5 FTHF/13C5 CH + THF/13C5 MeFox/13C5
MTHF folic acid THF/13C5 THF FTHF CH + THF MeFox
MW 459.4/464.4 441.4/446.4 445.4/450.4 473.4/478.4 455.5/460.5 473.4/478.4
λmax (nm) 290 282 and 346 298 285 288 and 348 280
Absorption 31700 27600@282 25000 37200 13500@288 19365
coefficient 7200@346 26500@348
(excess can be stored in glass scintillation vials and aliquoted as

required) and store at −80 °C for up to 5 years.
2. QC spike solution: 4 μM MTHF and 0.4 μM all other folates,
0.1 % ascorbic acid.
3. Spiked serum/plasma from anonymized donor samples: Create
a pool with low endogenous levels of folates and use the spike
mix to create approximate spiked concentrations of 50 nM
MTHF and 5 nM all other folates. Store in 350 μL aliquots in
amber glass UPLC vials (excess pooled plasma can be stored in
glass scintillation vials and aliquoted as required) at −80 °C for
up to 5 years.
4. Standard reference material 1950: purchased from the National
Institute of Standards and Technology (NIST) in 1 mL vials
with known values of MTHF and folic acid of 27 nM and
3.5 nM respectively. Store at −80 °C (see Note 8).
3 Methods
3.1 Sample 1. Make up a worksheet to include the reagent blank, diluent

Extraction (S0), working standards (S1–S7), QCs, and samples (see
Notes 9 and 10).
2. Use separate 12 × 75 mm borosilicate glass tubes for reagent
blank, standards, QCs, and samples.
3. Add 150 μL standard, QC or sample into the tube (see Note 11).
4. Add 50 μL working internal standard into each tube except the
reagent blank.
5. Add 200 μL water into the reagent blank tube.
6. Add 100 μL methanol into each tube.
7. Cover the tubes with Parafilm and mix for 5 min on a multi-
tube vortexer.
8. Centrifuge at 1650 × g for 5 min.
252 Sarah Meadows
9. Decant the supernatant from each tube into a second set of

labeled borosilicate glass tubes.
10. Pipette 350 μL sample diluent into each tube.
11. Cover the tubes with Parafilm and mix for 30 s on a multi-tube
vortexer.
12. Let stand for 20 min at room temperature (see Note 12).
13. Condition SPE columns on the positive pressure processor
with 500 μL acetonitrile, 500 μL methanol, and 2 × 650 μL
conditioning buffer. Push each addition through the column
(see Note 13).
14. Load 550 μL of sample into each well of the SPE plate. Stand
for 2 min
15. Push the samples through using minimal pressure (see Note 14).
16. Wash the samples with 2 × 650 μL of wash buffer. Push through
using minimal pressure.
17. Elute the samples into a 96-well filter plate using 2 × 150 μL
elution buffer. Turn the pressure down to zero and up slowly
until the liquid just starts to drip through (see Note 14). The
elution buffer needs to drip through slowly for optimum
elution.
18. Filter the samples under vacuum into a 2 mL 96-well plate.
19. Seal the plate.
20. Mix on the plate shaker for 10 min.
21. Load onto the UPLC and inject 15 μL into the LC-MS instru-
ment (see Fig. 1 for an example chromatogram).
3.2 UPLC Setup Set running conditions as follows:

for Waters Acquity H 1. Autosampler temp: 7 °C.
Class
2. Column oven: 30 °C.
3. Mobile phase: isocratic.
4. Flow: 250 μL/min.
5. Run time: 3.5 min (see Note 15).
3.3 Mass 1. Analyte parameters (Table 5) (see Note 16).

Spectrometer Setup 2. Polarity: ESI in positive mode.
for AB Sciex 4000
3. Curtain gas: 10.
4. Collision gas: high.
5. Ion spray voltage: 5500.
6. Temperature: 650.
7. Ion source gas 1: 55.
8. Ion source gas 2: 45.
9. Interface heater: on.
2.4-
2.2-
2.0-
1.8-
1.6-
Intensity (cps x 104)
1.4-
1.2-
1.0-
8.0-
6.0-
4.0-
2.0-
0.0-
0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.6
Time (min)
Fig. 1 Example chromatogram
4 Notes
1. Comparable C8 LC columns manufactured by Phenomenex

and Waters with the same particle size, diameter, and length
yield slightly poorer chromatography for THF. Only consider
using these if the minor folate forms are of no interest.
2. Make each analyte and its corresponding internal standard up
on the same day. Only make one solution at a time and clean
the bench area thoroughly with water between each analyte
and its labeled internal standard to prevent cross-contamination.
Use designated volumetric flasks for all the analytes and internal
standards to prevent contamination. These should be rinsed
254 Sarah Meadows
Table 5
Analyte parameters
Declustering Collision Collision cell Entrance

Analyte Transition potential energy exit potential potential
MTHF 460.2 > 131.2/180.1 56 29/51 18/12 10
Folic Acid 442.3 > 295.1/176.0 61 21/57 18/8 10
FTHF 474.4 > 299.2/165.8 61 45/65 16/28 10
THF 446.2 > 299.2/120.1 61 27/55 22/6 10
CH + THF 456.2 > 412.3/482.0 96 41/69 10/14 10
MeFox 474.2 > 284.2/132.1 61 53/73 15/22 10
13
C5 MTHF 465.2 > 313.1 81 25 8 10
13
C5 Folic Acid 447.2 > 295.4 56 19 16 10
13
C5 FTHF 479.2 > 299.2 76 47 16 10
13
C5 THF 451.1 > 299.1 51 29 17 10
13
C5 CH + THF 461.3 > 416.3 106 41 12 10
13
C5 MeFox 479.2 > 284.4 76 53 22 10
with deionized water after use, left to drain and stored until
needed again. DO NOT USE DETERGENT.
3. The absorbance of the stock I standards is measured at ½ and
1/5 dilution for folic acid and 1/20 and 1/50 dilution for all
the other folates using the same diluent as in the stock solution
spectrophotometrically at their respective λmax and concentra-
tion is calculated using the following formula (Table 4):
Conc ( m g / mL ) = ( abs ´ Dilution factor ´ 1000 ´ MW ) ´ molar absorption coefficient
For MTHF, THF, FTHF, CH + THF and MeFox and their

corresponding internal standards, once the concentration has
been determined, make as much of Stock I into Stock II as
possible and discard the rest.
4. Folate solids that come as a calcium salt may need glacial acetic
acid to aid dissolution in the buffer. Add 5 μL at a time and
mix well after each addition until the salt is dissolved.
5. Folic acid is already oxidized and not very soluble at low pH so
ascorbic acid is not required when making the standard
solutions.
6. 5,10 methenylTHF stock I dissolves better if the flask is heated.
We find using a beaker of hot water sufficient.
7. 5-Methyl THF and its labeled analog should be measured at

290 nm and 245 nm and the ratio calculated (290/245) to
ensure there is no oxidation to dihydrofolate. The ratio should
not exceed 3.3
8. We run SRM 1950 along with in house QCs as there is no
commercial QC material available with LC-MS/MS assigned
values.
9. We run QCs after the standards and at the end of the plate with
the exception of the SRM which is only run once on the plate,
at the beginning as an accuracy check. No drift was observed
across the plate so duplicate analysis is unnecessary.
10. All standards, QCs, and samples are run singly. With the QC
protocol described above this leaves enough space for 80 sam-
ples on a full 96-well plate.
11. Minimize the time samples are out at room temperature
defrosting as there is a loss of folates in the sample at pro-
longed times.
12. Do not leave the diluted samples standing on the bench for
longer than 30 min as there is a loss of folates in the sample at
prolonged room temperature.
13. Do not keep the air / nitrogen running any longer than neces-
sary when pushing liquids through the columns to prevent the
columns from drying out.
14. The samples should slowly drip through the column, I look for
about one drip per second. If the samples are eluted too quickly
the folates will not bind to the columns reducing the sensitivity
of the assay.
15. Washing the column with 65 % acetonitrile for 1.5 min after
each injection prolongs the column life and minimizes back
pressure increases during the run.
16. A second transition is only routinely measured for MTHF as
the sensitivity is too low for the other analytes. If a peak of
>5 nmol/L for folic acid, THF, FTHF, and CH + THF or
>10 nmol/L for MeFox is observed, then the sample is imme-
diately reinjected along with the standards and a set of QCs
using a method with the full set of transitions for all analytes.
In this way a quantifier peak is obtained with which to double
check the initial result.
Acknowledgments
This work was funded by the Medical Research Council

MRC_MC_U105960384
256 Sarah Meadows
References
1. Shane B (2011) Folate status assessment his- 9. Hopkins SM, Gibney MJ, Nugent AP, McNulty
tory: implications for measurement of bio- H, Molloy AM, Scott JM et al (2015) Impact
markers in NHANES. Am J Clin Nutr of voluntary fortification and supplement use
94:337S–342S on dietary intakes and biomarker status of
2. Crider KS, Bailey LB, Berry RJ (2011) Folic folate and vitamin B-12 in Irish adults. Am
acid food fortification-its history, effect, J Clin Nutr 101:1163–1172
concerns, and future directions. Nutrients 10. Clarke R, Bennett D (2014) Folate and preven-
3:370–384 tion of neural tube defects. BMJ 349:g4810
3. Obeid R, Kasoha M, Kirsch SH, Munz W, 11. http://www.efsa.europa.eu/en/events/
Herrmann W (2010) Concentrations of unme- event/corporate090121
tabolized folic acid and primary folate forms in 12. Bailey RL, Mills JL, Yetley EA, Gahche JJ, Pfeiffer
pregnant women at delivery and in umbilical CM, Dwyer JT et al. (2012) Serum unmetabo-
cord blood. Am J Clin Nutr 92:1416–1422 lized folic acid in a nationally representative sam-
4. Smith AD (2007) Folic acid fortification: the ple of adults >/=60 years in the United States,
good, the bad, and the puzzle of vitamin B-12. 2001–2002. Food Nutr Res 56. doi: 10.3402/
Am J Clin Nutr 85:3–5 fnr.v56i0.5616. Epub 2012 Apr 2
5. de Benoist B (2008) Conclusions of a WHO 13. Pfeiffer CM, Hughes JP, Lacher DA, Bailey
Technical Consultation on folate and vitamin RL, Berry RJ, Zhang M et al (2012) Estimation
B12 deficiencies. Food Nutr Bull 29(2 of trends in serum and RBC folate in the U.S.
Suppl):S238–S244 population from pre- to postfortification using
6. WHO (2015) Guideline: optimal serum and assay-adjusted data from the NHANES 1988-
red blood cell folate concentrations in 2010. J Nutr 142:886–893
women of reproductive age for prevention of 14. McDowell MA, Lacher DA, Pfeiffer CM,
neural tube defects. WHO, Geneva. ISBN Mulinare J, Picciano MF, Rader J et al (2008)
9789241549042 Blood folate levels: the latest NHANES results.
7. McGarel C, Pentieva K, Strain JJ, McNulty H NCHS Data Brief 6:1–8
(2015) Emerging roles for folate and related 15. Wang X, Zhang T, Zhao X, Guan Z, Wang Z,
B-vitamins in brain health across the lifecycle. Zhu Z et al (2014) Quantification of folate
Proc Nutr Soc 74:46–55 metabolites in serum using ultraperformance
8. Bailey LB (2009) Folate in health and disease, liquid chromatography tandem mass spectrom-
2nd edn. CRC, Boca Raton, FL. ISBN 10: etry. J Chromatogr B Analyt Technol Biomed
1420071246 Life Sci 962:9–13
Chapter 22
UPLC-MS/MS Determination of Deuterated

25-Hydroxyvitamin D (d3-25OHD3) and Other Vitamin D
Metabolites for the Measurement of 25OHD Half-Life
Shima Assar, Inez Schoenmakers, Albert Koulman, Ann Prentice,
and Kerry S. Jones
Abstract
Plasma 25-hydroxyvitamin D (25OHD) half-life (25OHDt1/2) is a dynamic marker of vitamin D metab-
olism that can be used to assess vitamin D expenditure and help inform vitamin D requirements. Our
group recently established an approach to determine the 25OHDt1/2 as an alternative biomarker of
25OHD expenditure in humans. The approach uses a small oral dose of stable isotope labeled 25OHD3
[3-2H-25OHD3 (6,19,19-d3)] (d3-25OHD3) (tracer), which is distinguishable from endogenous
25OHD by liquid chromatography tandem-mass spectrometry (LC-MS/MS). We report here the
method, which relies on protein precipitation, purification with solid phase extraction, derivatization
with 4-phenyl-1,2,4-triazoline-3,5-dione (PTAD), and determination of the compounds by isotope-
dilution UPLC-MS/MS. The method proved to be rapid and sensitive (LOQ 0.2 nmol/L) for the
quantification of this tracer as well as the other vitamin D metabolites: 25OHD3, 25OHD2, and
24,25(OH)2D3 in human plasma.
Key words Vitamin D, Stable isotope of 25OHD3, 25OHD3, 25OHD2, 24,25(OH)2D3, Biomarker,
PTAD derivatization, Liquid chromatography mass spectrometry
1 Introduction
25-hydroxyvitamin D (25OHD) is the most widely assessed bio-

marker of vitamin D status in biological samples. This metabolite
has relatively high serum concentration that is not under tight
homeostatic control and a relatively long half-life (2–3 weeks) [1–
4]. 25OHD concentration reflects vitamin D supply from cutane-
ous synthesis, diet, and usage through metabolism, excretion, and
sequestration into body tissues [5, 6]. To investigate the link
between the vitamin D supply and tissue requirements with vita-
min D metabolism, our group has proposed the plasma half-life of
25OHD as a dynamic biomarker of vitamin D expenditure [5].
The plasma half-life of 25OHD has been measured previously
257
258 Shima Assar et al.
using radiolabeled and unlabeled compounds and has been

reviewed by Jones et al. [7]. The use of radiolabeled compounds is
no longer considered acceptable for human use, particularly in
children, pregnant women, and other vulnerable groups. The use
of unlabeled compounds is not recommended because it is not
possible to distinguish the administered dose from endogenous
vitamin D metabolites and large doses are required that may per-
turb normal vitamin D metabolism [7]. The use of stable isotope
labeled compounds together with mass spectrometry is the pre-
ferred choice for research of vitamin absorption and metabolism
[8]. Recently, stable isotope labeled vitamin D compounds have
become available for research. We designed and performed studies
to measure 25OHD half-life (25OHDt1/2) using an oral dose of
stable isotope labeled 25OHD (2H-25OHD3 (6,19,19-d3), (d3-
25OHD3), as a tracer in humans [7, 9, 10].
A number of experimental factors may affect the usefulness of
tracer methods to determine half-life [7]. Two points are of rele-
vance here. Firstly, quantification of the tracer compound should
be highly specific and analytical methods should distinguish tracer
from endogenous compounds. Secondly, in order not to perturb
normal metabolism, the tracer dose should not exceed more than
10 % of endogenous levels and thus high sensitivity is required.
This can be achieved through the coupling of ultra-performance
(sub 2 μm particle size columns) liquid chromatography with
electrospray-based tandem mass spectrometry (UPLC-MS/MS).
LC-MS/MS is frequently referred to as the “gold standard”
method for the determination of the 25OHD concentration in
biological samples. However, the lipophilic nature, low ionization
efficiency of the vitamin D metabolites, low abundance in human
plasma, and potential interferences from co-eluting isobaric com-
pounds make development of a LC-MS/MS method challenging
[11, 12]. To overcome these issues, derivatization with Cookson-
type triazoline-diones (TADs) reagent has been reported for the
analysis of vitamin D metabolites. These reagents introduce polar
groups into the analytes and therefore increase the ionization effi-
ciency, sensitivity, and specificity for the analysis of vitamin D com-
pounds [13, 14]. 4-phenyl-1,2,4-triazoline-3,5-dione (PTAD)
(Fig. 1) is a readily available and cost-effective derivatizing reagent
applicable to the analysis of vitamin D.
Here, we describe a fast and sensitive LC-MS/MS method that
simultaneously determines low abundance tracer (2H-25OHD3) and
25(OH)D3, 25(OH)D2 and 24,25(OH)2D3 in human plasma and is
applicable to tracer experiments of vitamin D metabolism. In sum-
mary, plasma samples were subjected to protein precipitation with
acetonitrile and followed by purification with solid-phase extraction.
The purified samples were derivatized with PTAD solution prepared
in acetonitrile and derivatized samples were analyzed on an UPLC
system interfaced to a triple quadropole mass spectrometer.
UPLC-MS/MS Analysis of 25OHD Stable Isotope Tracer Concentration 259
OH
OH O
H
N
O N
N
N
N O
N
HO
O PTAD-25(OH)D3 (6S-isomer)
PTAD
HO
OH
O
H
N
N
N
HO
PTAD-25(OH)D3 (6R-isomer)
Fig. 1 Derivatization of 25OHD3 with PTAD derivatization reagent. Adapted from [15]
2 Materials
2.1 Sample 1. Human plasma samples (see Note 1).

Pretreatment 2. Acetonitrile (AcN) (see Note 2).
3. Ethyl acetate.
4. Methanol (MeOH).
5. 30 % methanol (see Note 3).
6. 4-phenyl-1,2,4-triazoline-3,5-dione (PTAD) (Sigma-Aldrich,
Gillingham, UK) solution: 0.5 mg in 1 mL AcN (see Note 4).
7. Oasis HLB SPE cartridge (1 mL, 30 mg sorbent, Waters,
Elstree, UK).
8. SPE vacuum manifold.
9. N2 sample concentrator.
2.2 Standards 1. 25OHD3, 25OHD2, 24,25(OH)2D3, d3-25OHD3

(Sigma-Aldrich).
2. d6-25OHD3, d6-24,25(OH)2D3 (Sigma-Aldrich).
www.Ebook777.com
2.3 UPLC-MS/MS 1. 0.1 % formic acid (FA) in solvent A and B (below).

2. 5 mM monomethylamine (MMA) in solvent A and B (below).
3. Solvent A: 0.1 % FA, 5 mM MMA in H2O.
4. Solvent B: 0.1 % FA and 5 mM MMA in mixture of
MeOH:H2O:AcN (97:2:1, v:v:v).
5. Strong/weak and needle wash solvent: AcN:H2O (1:1, v/v).
6. 0.2 μm Supelco Nylon 66 membrane (Supelco, Bellefonte, PA,
USA).
7. 200 μL utosampler glass micro insert.
8. 2 mL autosampler amber glass vials with cap and septa.
9. Acquity UPLC™ BEH C18 column (2.1 × 100 mm, 1.7 μm)
(Waters Corporation; Milford, MA, USA).
10. Acquity UPLC module (Waters Corporation).
11. 5500 QTRAP quadruple-linear ion trap mass spectrometer
(AB Sciex; Concord, ON, Canada).
3 Methods
3.1 Preparation 1. Prepare main stock solutions of each standard (i.e., 25OHD3,
of Standard 25OHD2, 24,25(OH)2D3, and d3-25OHD3) and internal
and Internal Standard standard (i.e., d6-25OHD3 and d6-24,25(OH)2D3) by
Solutions dissolving the compounds in AcN and store at −20 °C
until use.
2. Prepare working solutions of each standard and internal stan-
dard by making serial AcN dilutions of the main stock solu-
tions to obtain the required concentration for the calibration
curves (see Note 5).
3. Prepare six external standards containing the analytes of inter-
est with different ranges of concentrations (see Note 5) and
their corresponding internal standards (see Note 6) to generate
the standard curves.
4. Dry the standard solutions under gentle stream of nitrogen gas
using sample concentrator and apply for derivatization process.
5. Prepare quality control material (see Note 7).
3.2 Pretreatment 1. Transfer 250 μL plasma to a 2 mL Eppendorf tube and spike

of Plasma with internal standard solution containing d6-25OHD3 and d6-
24,25(OH)2D3 (see Note 6).
2. Mix for a few seconds using a mixer vortex and incubate the
spiked plasmas at room temperature for 1 h on a multi-tube
vortexer.
3.3 Protein 1. Add 600 μL AcN to spiked plasma and mix for approximately
Precipitation 1 min using a vortexer.
2. Centrifuge the mixture at 10,000 × g for 15 min to precipitate
the protein.
3. Transfer the supernatant to a 2 mL Eppendorf tube and evapo-
rate AcN under a gentle flow of nitrogen gas at room tempera-
ture to a volume of around 200 μL.
4. Add 800 μL of deionized H2O to remaining solution and mix
using a vortexer.
3.4 Solid Phase 1. Precondition Oasis HLB SPE cartridges with 1 mL ethyl ace-
Extraction tate and allow to pass through the cartridges and then add
1 mL MeOH.
2. Equilibrate the cartridges with 1 mL deionized H2O.
3. Load the prepared samples onto the cartridge and allow the
solution to pass through the cartridges to the waste vessel.
4. Add 1 mL deionized H2O onto cartridges and allow to
pass through to the waste vessel to remove the unwanted
compounds.
5. Repeat step 4 with 1 mL aqueous 30 % MeOH.
6. Replace waste vessels with 2 mL Eppendorf tubes.
7. Add 1 mL AcN followed by 500 μL ethyl acetate to elute the
analytes of interest.
8. Dry the eluate under gentle stream of nitrogen gas at room
temperature in the preparation for the derivatization step.
3.5 Derivatization 1. Add 50 μL 0.5 mg/mL PTAD solution to the eluate residue of
plasmas, in-house quality control material and the six prepared
standard solutions (see Note 8).
2. Mix gently using a multi-channel vortexer for 1 h at room
temperature.
3.6 LC-MS/MS 1. Transfer 50 μL derivatized plasma and calibration standard

Analysis solutions into 200 μL inserts and load in a 2 mL amber color
capped vial (see Note 9).
2. Load the samples on an LC plate and perform MS run
(see Note 10).
3. LC parameters: LC gradient programme: solvent B; initial 2 %
for 0.5 min and increase to 100 % at 0.5 min, remain at 100 %
for 4.5 min, and decrease to 2 % for 0.5 min (Fig. 2); flow rate
at 350 μL.minute–1; run time of 5.5 min; target sample tem-
perature 5 °C; column temperature 45 °C; injection mode:
UPLC Gradient
100
80
Solvent B (%)
60
40
20
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5
Time (min)
Fig. 2 LC Gradient programme used to elute vitamin D metabolites over 5.5 min.
The initial and end concentrations selected for solvent B were 2 %. The Y-axis
indicates the solvent B percentage and the X-axis is the elution time
partial loop; loop size: 20 μL; injection volume: 10 μL; equili-

bration time: 5 min.
4. MS operating parameters (see Note 11): electrospray ioniza-
tion (ESI) positive polarity; ion spray voltage, 5500 V;
source temperature, 550 °C; curtain gas 30 psi; ion source
gas1, 70 psi and ion source gas2, 60 psi; selected mass tran-
sitions for analysis of PTAD-derivatized vitamin D metabo-
lites were as 607 > 298 for 25OHD3, 613 > 298 for
d6-25OHD3, 610 > 301 for d3-25OHD3, 623 > 298 for
24,25(OH)2D3, and 629 > 298 for d6-24,25(OH)2D3 with
dwell time 100 ms.
3.7 Data Analysis 1. Multiple ways exist to carry out the data analysis and the quanti-
fication method (see Fig. 3 as an example for the chromatograms
obtained from analysis of a plasma sample) (see Note 12).
2. Derive the external standard curves by plotting the peak area
ratio against their amount ratio (i.e., analyte: corresponding
internal standard) (see Fig. 4 as an example for the calibration
curve) (see Note 13).
3. Assess the precision and quality of the analysis by calculation of
the coefficient variation percentage (CV %) between the results
obtained from analysis of the set of standards and in-house
quality control material (see Note 10).
Fig. 3 LC-MS/MS chromatograms of 25OHD3, 25OHD2, d3-25OHD3 (tracer), and 24,25(OH)2D3 in pooled human
plasma
0.12
0.10
0.08 y = 0.8861x - 0.0006

R² = 0.9993
Area Ratio
0.06
0.04
0.02
0.00
0.00 0.05 0.10 0.15
Amount Ratio
Fig. 4 Calibration curve for the determination of the tracer (d3-25OHD3)

constructed by plotting the peak area ratio against amount ratio for
d3-25OHD3:d6-25OHD3
4 Notes
1. Methods for plasma sample collection were carried out as

described elsewhere [3]. Details of the study design and half-
life determination are reported by [3]. In summary, tracer
(d3-25OHD3) was dissolved in vegetable oil and an aliquot
equal to 40 nmol was given to healthy volunteers on a small
piece of bread. Fasting LiHep-blood samples of volunteers was

collected and stored at −80 °C until use.
2. Use analytical HPLC grade for all reagents.
3. Prepare all aqueous solutions using deionized water to attain a
sensitivity of 18 MΩ-cm at 25 °C.
4. Prepare PTAD solution in AcN freshly on the day of sample
LC-MS/MS analysis.
5. The concentration ranges were selected for the standard curves
based on the expected concentration range in human plasma
samples from our studies: 25OHD3 (5–200 nmol/L),
d3-25OHD3 (0.25–4 nmol/L), 24,25(OH)2D3 (1–20 nmol/L).
6. The concentrations of internal standards, i.e., d6-25OHD3 and
d6-24,25(OH)2D3, were calculated to be 20 nmol/L in final
50 μL PTAD solutions for both external standard solutions
and plasma samples.
7. Pooled plasma, as in-house quality control material, was col-
lected from healthy volunteers under a method development
ethics approval and informed, written consent was obtained.
In-house quality control sample was treated and prepared as
mentioned above and analyzed with each batch of plasma
samples.
8. Optimization of the PTAD solution concentration and mixing
time have been reported elsewhere [16].
9. 50 μL sample allows for two runs accounting for duplicate
injections using partial loop injection mode and 10 μL injec-
tion volume.
10. Run the set of external standards and in-house quality control
material at the beginning and end of the batch to assess the
quality of the run.
11. MS parameters were optimized to give the highest response for
the determination of the compounds of interest.
12. The system control and peak integration were performed using
Analyst version 1.6.2 (Applied Biosystems, Concord, ON,
Canada).
13. Stable isotope labeled (d6-25OHD3) was used as an internal
standard for both the tracer (d3-25OHD3) and 25OHD3 and
d6-24,25(OH)2D3 for 24,25(OH)2D3 quantification.
Acknowledgments
We would like to thank the UK Medical Research Council for

funding this study (programme number U105960371).
References
1. Holick MF (2009) Vitamin D status: measure- Predictors of 25(OH)D half-life and plasma
ment, interpretation, and clinical application. 25(OH)D concentration in The Gambia and
Ann Epidemiol 19:73–78 the UK. Osteoporos Int 26:1137–1146
2. van den Ouweland JM, Vogeser M, Bacher S 11. Volmer DA, Mendes LR, Stokes CS (2015)
(2013) Vitamin D and metabolites measure- Analysis of vitamin D metabolic markers by
ment by tandem mass spectrometry. Rev mass spectrometry: current techniques, limita-
Endocr Metab Disord 14:159–184 tions of the “gold standard” method, and
3. Jones KS, Assar S, Harnpanich D, Bouillon R, anticipated future directions. Mass Spectrom
Lambrechts D, Prentice A et al (2014) 25(OH) Rev 34:2–23
D2 half-life is shorter than 25(OH)D3 half-life 12. Qi Y, Geib T, Schorr P, Meier F, Volmer DA
and is influenced by DBP concentration and gen- (2015) On the isobaric space of
otype. J Clin Endocrinol Metab 99:3373–3381 25-hydroxyvitamin D in human serum: poten-
4. Aghajafari F, Field CJ, Rabi D, Kaplan BJ, tial for interferences in liquid chromatogra-
Maggiore JA, O’Beirne M et al (2016) Plasma phy/tandem mass spectrometry, systematic
3-Epi-25-hydroxycholecalciferol can alter the errors and accuracy issues. Rapid Commun
assessment of vitamin D status using the cur- Mass Spectrom 29:1–9
rent reference ranges for pregnant women and 13. Higashi T, Yamauchi A, Shimada K (2003)
their newborns. J Nutr 146:70–75 Application of 4-(4-nitrophenyl)-1,2,4-
5. Prentice A, Goldberg GR, Schoenmakers I triazoline-3,5-dione to analysis of 25-hydroxyvi-
(2008) Vitamin D across the lifecycle: physiology tamin D3 in human plasma by liquid
and biomarkers. Am J Clin Nutr 88:500S–506S chromatography/electron capture atmospheric
pressure chemical ionization-mass spectrome-
6. Seamans KM, Cashman KD (2009) Existing try. Anal Sci 19:941–943
and potentially novel functional markers of
vitamin D status: a systematic review. Am J Clin 14. Aronov PA, Hall LM, Dettmer K, Stephensen
Nutr 89:1997S–2008S CB, Hammock BD (2008) Metabolic profiling
of major vitamin D metabolites using Diels-
7. Jones KS, Schoenmakers I, Bluck LJ, Ding S, Alder derivatization and ultra-performance liq-
Prentice A (2012) Plasma appearance and uid chromatography-tandem mass spectrometry.
disappearance of an oral dose of Anal Bioanal Chem 391:1917–1930
25-hydroxyvitamin D2 in healthy adults. Br
J Nutr 107:1128–1137 15. Higashi T, Suzuki M, Hanai J, Inagaki S, Min
JZ, Shimada K et al (2011) A specific LC/
8. Bluck LJ (2009) Recent progress in stable iso- ESI-MS/MS method for determination of
tope methods for assessing vitamin metabo- 25-hydroxyvitamin D3 in neonatal dried blood
lism. Curr Opin Clin Nutr Metab Care spots containing a potential interfering metab-
12:495–500 olite, 3-epi-25-hydroxyvitamin D3. J Sep Sci
9. Braithwaite V, Jones KS, Assar S, Schoenmakers 34:725–732
I, Prentice A (2014) Predictors of intact and 16. Ding S, Schoenmakers I, Jones K, Koulman A,
C-terminal fibroblast growth factor 23 in Prentice A, Volmer DA (2010) Quantitative
Gambian children. Endocr Connect 3:1–10 determination of vitamin D metabolites in
10. Jones KS, Assar S, Vanderschueren D, Bouillon plasma using UHPLC-MS/MS. Anal Bioanal
R, Prentice A, Schoenmakers I (2015) Chem 398:779–789
Chapter 23
iTRAQ-Based Shotgun Proteomics Approach for Relative

Protein Quantification
Erika Velásquez Núñez, Gilberto Barbosa Domont,
and Fábio César Sousa Nogueira
Abstract
Shotgun proteomics has a key role in quantitative estimation of proteins from biological systems under
different conditions, which is crucial in the understanding of their functional roles. Isobaric tagging for
relative and absolute quantitation (iTRAQ) mass spectrometry is based on pre-labeling of peptides with
mass tags which allows the multiplex analysis of up to eight proteomes simultaneously. We describe here a
detailed protocol for sample preparation and iTRAQ 4-plex labeling for relative quantification of multiple
samples from human and plant tissues. We also present two strategies for peptide fractionation after the
iTRAQ labeling protocol.
Key words iTRAQ, Animal tissues, Plant tissues, Stable isotope labeling, Quantitative proteomics,
Mass spectrometry
1 Introduction
Shotgun proteomics is no longer merely descriptive, but has

become quantitative [1]. The estimation of protein abundances in
different biological conditions provides more information than in
qualitative studies, in which only the presence or posttranslational
modifications of proteins are determined. One of the turning
points in quantitative proteomics was the implementation of iso-
baric methods for relative protein quantification. Different from
other stable isotope labeling methods in which identical peptides
with a mass variance from two or more samples can be distin-
guished in the mass spectrometry (MS) 1 spectrum, isobaric meth-
ods add isotope tags of the same mass to each peptide. In MS1
mode, each peptide labeled with different isobaric tags is undistin-
guishable but in MS2 mode reporter ions of distinct masses are
released such that peptides from different samples can be differen-
tiated and quantified, based on the intensity of these reporter ions
[2]. Two of the most used isobaric labeling methods are isobaric
267
268 Erika Velásquez Núñez et al.
tags for relative and absolute quantitation (iTRAQ) [3] and tan-
dem mass tags (TMT) [4, 5].
The widely employed iTRAQ approach is a reliable labeling
technique employed in the analysis of up to four (iTRAQ 4-plex)
[3] or eight samples (iTRAQ 8-plex) [6]. Compared to label-free
mass spectrometry techniques, the multiplexing potential of isoto-
pic labeling increases the statistical relevance and accuracy of results
through the simultaneous analysis of different biological samples
and through normalization to an internal standard. In the iTRAQ
4-plex approach, the reagent contains reporter (N-methylpiperazine),
balance (carbonyl) and reactive groups (NHS ester). Each reagent
has the same mass achieved by a combination of 13C, 15 N, and
18O in the reporter (m/z 114-117) and balance groups (28–
31 Da). The labeled peptides have identical retention times in liq-
uid chromatography and, since the tags are isobaric, the peptides
appear as single peak with the same m/z in a MS1 spectrum.
However, selection of the precursor ion for fragmentation pro-
duces a MS2 spectrum with reporter ion peaks at low mass region
(114, 115, 116, and 117) and peptide backbone fragmentation
peaks. Intensity of the reporter ion peaks directly reflects the abun-
dance of the peptide in each sample (Fig. 1).
Here, we describe a detailed sample preparation and iTRAQ
4-plex labeling method for relative quantification of multiple sam-
ples from human and plant tissues. Additionally, we detail two
strategies for peptide fractionation after iTRAQ labeling.
2.1 Protein 1. Human tissues: RapiGest® (Waters Corporation).

Extraction of Human 2. 1 M triethylammonium bicarbonate (TEAB; Sigma-Aldrich)
Tissues (see Note 2).
2.2 Protein 1. 50 mM pyridine (pH 5.0), 10 mM thiourea, 1 % SDS, 5 %

Extraction of Plant polyvinylpolypyrrolidone (PVPP).
Tissues 2. 10 % trichloroacetic acid (TCA) in acetone.
3. 7 M urea/2 M thiourea solution containing 5 % TEAB.
QUANTITATION
114 117
114
Intensity
Intensity
115 Combine MS MS/MS 116

116 115
117
m/z m/z
Fig. 1 Principle of iTRAQ mass spectrometry technique

TRAQ for Shotgun Proteomics 269
2.3 Enzymatic 1. Qubit® 2.0 fluorimetric assay kit (Invitrogen).

Digestion 2. Reducing solution: 100 mM dithiothreitol (DTT) or 50 mM
tris(2-carboxyethyl) phosphine (TCEP) (see Notes 3 and 4).
3. Alkylating solution: 400 mM iodoacetamide (see Note 5).
4. Sequencing grade modified yrypsin (Promega).
5. Acetic acid.
6. 10 % trifluoroacetic acid (TFA).
2.4 iTRAQ Peptide 1. 0.1 % TFA.

Labeling 2. 0.1 % TFA/50 % acetonitrile (ACN).
3. 0.1 % TFA/70 % ACN.
4. iTRAQ reagent 4-plex Kit (Applied Biosystems Sciex).
5. C18 macro-spin column (Harvard Apparatus) (see Note 6).
6. Strong cation exchange (SCX) macro-spin column (Harvard
Apparatus) (see Note 7).
7. Buffer A: 5 mM KH2PO4/25 % ACN (pH 3).
8. Buffer B: 1 M KCl stock solution.
9. Shimadzu LC-20AT high performance liquid chromatography
(HPLC) instrument for hydrophilic interaction chromatogra-
phy (HILIC) using a 3 μm × 5 cm × 2 mm TSKgel® amide-80
column (Sigma-Aldrich) (see Note 8).
10. Solvent A (HILIC-A): 90 % ACN/0.1 % TFA.
11. Solvent B (HILIC-B): 0.1 % TFA.
2.5 Labeled Peptide 1. 2 cm length, 200 μm inner diameter EASY II trap-column

Analysis by Nano (ThermoFisher Scientific).
LC-MS/MS 2. 18 cm length, 100 μm inner diameter, 5 μm resin (ReproSil-Pur
C18) pulled analytical capillary column (Dr. Maisch GmbH).
3. Phase A: 0.1 % formic acid, 5 % ACN.
4. Phase B 0.1 % formic acid, 95 % ACN.
5. LTQ Orbitrap Velos mass spectrometer (ThermoFisher
Scientific).
2.6 Data Analysis 1. Data inspection: Xcalibur 2.1 software (ThermoFisher

Scientific).
2. Database searches: Proteome Discoverer 2.1 software
(ThermoFisher Scientific) with the SEQUEST algorithm.
3. Databases: Uniprot (www.uniprot.org/), NCBI (www.ncbi.
nlm.nih.gov/), and neXtProt (www.nextprot.org/).
3 Methods
3.1 Protein 1. Pulverize and macerate the tissues in liquid nitrogen [7].
Extraction in Human 2. Add 0.1 % RapiGest in 50 mM TEAB (see Note 9).
Tissues
3. Vortex the samples and centrifuge for 30 min at 20,000 × g at
4 °C.
4. Transfer the supernatant to another tube and take one aliquot
for protein quantification.
3.2 Protein 1. Mix the sample at a 1:40 ratio with pyridine buffer [8].
Extraction in Human 2. Stir for 2 h at 4 °C and centrifuge at 10,000 × g for 40 min.
Tissues
3. Precipitate proteins in ice-cold 10 % TCA in acetone.
4. Wash the pellet with ice-cold acetone three times and dry
under vacuum.
5. For protein quantification, dissolve the pellet in urea/thiourea
solution.
3.3 Protein Digestion 1. Quantitate proteins using the Qubit 2.0 fluorometric assay kit
with Trypsin according to the manufacturer’s instructions.
2. Disulfide bond reduction in proteins: incubate the samples
with DTT or TCEP solution at a final concentration of 10 mM
for 1 h at 30 °C.
3. Thiol group alkylation in proteins: incubate the samples with
iodoacetamide solution at a final concentration of 40 mM for
30 min at room temperature in the dark.
4. Trypsin digestion: add trypsin at a trypsin:protein ratio of 1:50
and incubate for 12–18 h at 37 °C (see Note 10).
5. Stop the reaction by adding 10 % TFA to give a final concentra-
tion of 0.1 % (see Note 11).
3.4 iTRAQ Peptide 1. Peptide cleaning: incubate C18 spin columns with 500 μL
Labeling [9, 10] 100 % ACN for 15 min and centrifuge at 2000 × g for 1 min.
2. Add the same amount of ACN and repeat the centrifugation step.
3. Equilibrate the columns with 150 μL 0.1 % TFA and centrifuge
at 2000 × g.
4. Repeat this step three times, add 75–150 μL sample, and cen-
trifuge at 2000 × g.
5. Wash the columns using 0.1 % TFA and centrifuge at 2000 × g.
6. Repeat the wash/centrifugation three times in total and elute
the peptides in two successive steps into the same collection
tube with 0.1 % TFA/50 % ACN and 0.1 % TFA/70 % ACN
followed by centrifugation at 2000 × g.
7. Dry the peptides by vacuum centrifugation.

8. Peptide labeling: suspend peptides in 30 μL of 20 mM TEAB
(pH 8.5) (see Note 12) and quantify using the Qubit 2.0 fluo-
rometric assay to normalize peptide amounts.
9. Briefly centrifuge the iTRAQ reagent solution at room tem-
perature so that the content is collected in the bottom of the
tube and add 70 μL of ethanol to each vial.
10. Vortex the vials and centrifuge briefly as above.
11. Transfer the contents of each vial to a specific sample tube,
vortex and centrifuge again (see Note 13).
12. Incubate the samples at room temperature for 1 h.
13. Stop the reaction by adding formic acid at a final concentration
of 1 %, vortex and centrifuge as above (see Note 14).
14. Combine the contents of all samples labeled with different
iTRAQ tags into one tube, vortex and centrifuge.
15. Dry the contents in a vacuum centrifuge but stop before com-
plete dryness is reached (see Note 15).
3.5 iTRAQ-Labeled 1. Suspend the semidry pellets at approximately 1 μg peptides/

Peptide Fractionation μL in 100 μL of 5 mM KH2PO4/25 % ACN solution and
vortex.
2. Incubate the SCX spin column (Harvard Apparatus) with
500 μL of the same solution for 15 min at room temperature.
3. Centrifuge at 2000 × g until all solution has passed through the
column and repeat this step.
4. Add the sample to the spin column, centrifuge at 2000 × g and
collect the column flow-through.
5. Carry out four sequential elution steps using 150 μL of the
5 mM KH2PO4/25 % ACN solution, containing first 75 mM
KCl, second 150 mM KCl, third 250 mM KCl, and last
500 mM KCl, followed by centrifugation each time at 2000 × g
and collecting the eluates in separate tubes.
6. Desalt the samples using the peptide cleaning step above.
7. Suspend the samples in 0.1 % formic acid and quantify as above.
8. Suspend the samples at approximately 1 μg/μL in 100 μL of
HILIC-A solution, vortex and centrifuge briefly to collect the
sample at the bottom of the tube.
9. Load samples at a flow rate of 0.2 mL/min into the TSKgel
Amide-80 column on the LC-20AT HPLC system.
10. Fractionate the peptides by applying 100 % HILIC-A for
10 min, 12 % HILIC-B for 2 min, 20 % HILIC-B for 30 min,
30 % HILIC-B for 30 min, 100 % HILIC-B for 5 min, and
return to 100 % HILIC-A for 5 min.
3.6 Labeled Peptide 1. Load 1 μg labeled peptides onto the trap and capillary columns
Analysis on the nano LC system coupled online to the LTQ Orbitrap
by nanoLC-MS/MS Velos mass spectrometer.
2. For peptide elution, apply a gradient from 100 % phase A to
35 % phase B over 120 min at a flow rate of 200 nL/min.
3. After each run, wash the column with 90 % phase B and re-
equilibrate with phase A.
4. Acquire spectra in positive mode applying a data-dependent
automatic survey MS scan and MS/MS.
5. Set the resolution of the Orbitrap mass analyzer at 60,000 at
m/z 400, automatic gain control target at 1 × 106, and maxi-
mum ion injection at 500 milliseconds (see Note 16).
6. Acquire MS/MS spectra with a resolution of 7500 at 400 m/z, a
signal threshold of 30,000, normalized collision energy of 40, and
dynamic exclusion enabled for 30 s with a repeat count of 1.
7. Place an Eppendorf tube covered with 5 % ammonia water
solution under the nano ESI needle (see Note 17).
3.7 Data Analysis 1. Inspect the raw data using the Xcalibur software.
2. Perform database searches against target and decoy (reverse)
databases from Uniprot, NCBI, and neXProt using the follow-
ing search parameters: MS accuracy = 10 ppm; MS/MS accu-
racy = 0.1 Da; trypsin digestion with two missed cleavages
allowed; fixed carbamidomethyl modification of cysteine; and
variable modification of oxidized methionine.
3. For identification of iTRAQ labeled peptides, also include the
iTRAQ 4-plex monoisotopic mass = 144.102 and variable
modification for N-terminus, lysine and tyrosine.
4. Accept false discovery rates less than 1 % and peptide rank = 1.
4 Notes
1. Reagents should be of analytical grade, solvents HPLC or

LC-MS grade and solutions should be prepared with ultrapure
water (18 MΩ-cm at 25 °C). LC-MS solutions require LC-MS
grade water.
2. It is advisable to use protease inhibitors in this step to prevent
degradation of proteins caused by proteases in the sample. In
addition, for phosphoproteomics, it is necessary to use phos-
phatase inhibitors to prevent dephosphorylation during prepa-
ration and handling of sample.
3. It is recommended to use a fresh DTT stock solution.
4. TCEP has the advantage of being a more powerful reducing

agent than DTT, providing an irreversible reaction. In addi-
tion, it is more hydrophilic, active in alkaline and acidic condi-
tions, and more resistant to air oxidation. Also, it does not
reduce metals and is significantly more stable than DTT in the
absence of a metal chelator.
5. It is necessarily to prepare the iodoacetamide solution immedi-
ately before use and keep it protected from light because it is
unstable and light-sensitive. Also, the alkylation step must be
performed in darkness.
6. The C18 Macro-Spin Column has a binding capacity of 30–300 μg
of sample, accepting a sample volume of 70–150 μL. Review the
manufacturer’s specifications before use.
7. The SCX Macro-Spin Column has an ion capacity of 0.18–
0.25 mmol (Cl-)/ml and a binding capacity of 30–300 μg of
protein sample, accepting a sample volume of 70–150 μL. Review
the manufacturer’s specifications before use.
8. HILIC is recommended to remove excess iTRAQ reagent from
iTRAQ-labeled peptides because the solutions involved in this
procedure are compatible with mass spectrometric analysis.
This eliminates an additional step of sample cleaning [11, 12].
9. We recommend addition of 200 μL of 0.1 % RapiGest for
100 mg of tissue.
10. Samples in 8 M urea/ 2 M thiourea must be diluted to final
concentrations lower than 1 M urea using 100 mM TEAB and
avoid heating. Check the pH to ensure that it is close to 8.
11. For RapiGest samples, acidify the samples with TFA to a final
concentration of 1 % to stop the reaction and incubate for
40 min at room temperature. Centrifuge for 30 min at
20,000 × g to remove insoluble material.
12. Before peptide labeling, ensure that the pH is close to 8.5.
13. It is advisable to use commercial peptides (e.g., Glu-1-
fibrinopeptide B) at a known concentration at the time of labeling
to serve as internal controls and facilitate data normalization.
14. It is advisable to analyze a peptide sample aliquot by MS before
making the final mix of all iTRAQ labels to confirm the pres-
ence of labeled peptides with the appropriate m/z peaks for
each reporter ion. If the labeling process was not successful,
repeat the labeling procedure.
15. The peptide pellets are easier to resuspend if not completely
dry.
16. Choose a method consisting of fragmentation of the ten most
intense ions by high-energy collision dissociation.
17. The presence of 5 % ammonia during analysis avoids the super-
charge effect of the iTRAQ 4-plex tag [13].
Acknowledgments
This work was funded by the CNPq (Grant number: 477325/2013-

0) and FAPERJ (Grant number: E-26/202.801/2015).
References
1. Nogueira FC, Domont GB (2014) Survey of 8. Vasconcelos ÉAR, Nogueira FCS, Abreu EFM,
shotgun proteomics. Methods Mol Biol Gonçalves EF, Souza PAS, Campos FAP
1156:3–23 (2005) Protein extraction from cowpea tissues
2. Rauniyar N, Yates JR 3rd (2014) Isobaric for 2-D gel electrophoresis and MS analysis.
labeling-based relative quantification in Chromatographia 62:447–450
shotgun proteomics. J Proteome Res 9. Nogueira FC, Palmisano G, Schwämmle V,
13:5293–5309 Campos FA, Larsen MR, Domont GB et al
3. Ross PL, Huang YN, Marchese JN, Williamson (2012) Performance of isobaric and isotopic
B, Parker K, Hattan S et al (2004) labeling in quantitative plant proteomics.
Multiplexed protein quantitation in J Proteome Res 11:3046–3052
Saccharomyces cerevisiae using amine-reac- 10. Nogueira FC, Palmisano G, Schwämmle V,
tive isobaric tagging reagents. Mol Cell Soares EL, Soares AA, Roepstorff P et al
Proteomics 3:1154–1169 (2013) Isotope labeling-based quantitative
4. Thompson A, Schäfer J, Kuhn K, Kienle S, proteomics of developing seeds of castor oil
Schwarz J, Schmidt G et al (2003) Tandem seed (Ricinus communis L.). J Proteome Res
mass tags: a novel quantification strategy for 12:5012–5024
comparative analysis of complex protein mix- 11. Palmisano G, Lendal SE, Engholm-Keller K,
tures by MS/MS. Anal Chem 75:1895–1904 Leth-Larsen R, Parker BL, Larsen MR (2010)
5. Dayon L, Hainard A, Licker V, Turck N, Kuhn Selective enrichment of sialic acid-containing
K, Hochstrasser DF et al (2008) Relative quan- glycopeptides using titanium dioxide chroma-
tification of proteins in human cerebrospinal tography with analysis by HILIC and mass
fluids by MS/MS using 6-plex isobaric tags. spectrometry. Nat Protoc 5:1974–1982
Anal Chem 80:2921–2931 12. Melo-Braga MN, Verano-Braga T, León IR,
6. Choe L, D’Ascenzo M, Relkin NR, Pappin D, Antonacci D, Nogueira FC, Thelen JJ et al
Ross P, Williamson B et al (2007) 8-plex quan- (2012) Modulation of protein phosphoryla-
titation of changes in cerebrospinal fluid pro- tion, N-glycosylation and Lys-acetylation in
tein expression in subjects undergoing grape (Vitis vinifera) mesocarp and exocarp
intravenous immunoglobulin treatment for owing to Lobesia botrana infection. Mol Cell
Alzheimer’s disease. Proteomics 7:3651–3660 Proteomics 11:945–956
7. Aquino PF, Lima DB, de Saldanha da Gama 13. Thingholm TE, Palmisano G, Kjeldsen F,
FJ, Melani RD, Nogueira FC, Chalub SR et al Larsen MR (2010) Undesirable charge-
(2014) Exploring the proteomic landscape of a enhancement of isobaric tagged phosphopep-
gastric cancer biopsy with the shotgun imaging tides leads to reduced identification efficiency.
analyzer. J Proteome Res 13:314–320 J Proteome Res 9:4045–4052
Chapter 24
1
H NMR Metabolomic Profiling of Human and Animal Blood
Serum Samples
João G.M. Pontes, Antonio J.M. Brasil, Guilherme C.F. Cruz,
Rafael N. de Souza, and Ljubica Tasic
Abstract
Nuclear magnetic resonance (NMR) spectroscopy techniques allow the acquisition of a large amount of
data and when combined with multivariate statistical analysis, it is possible to process and interpret the
obtained NMR data in accordance with the biological problem being investigated. In this chapter, the
search for biologically relevant biomarkers is addressed using NMR spectroscopy-based metabolomics, due
to their clinical relevance for either diagnosis or monitoring of diseases and disorders.
Key words Nuclear Magnetic Resonance, Metabolomics, Serum, Chemometrics, Biomarkers
1 Introduction
The search for biomarkers that characterize metabolic conditions,

disorders or diseases, has grown in the last years. Biotechnology
tools such as genomics, proteomics, and metabolomics are the
most used techniques to find biomarkers [1–3] and are comple-
mentary to each other. Metabolomics is a scientific approach that
allows qualitative and quantitative identification of metabolites.
Metabolites are products of cellular metabolic pathways. The first
researchers to use the term “metabolome” to designate the set of
all low-molecular-mass compounds synthesized by an organism
were Oliver et al. in 1998 [4]. However, it was only in 2002 that
Fiehn suggested the classification of analytical approaches and,
thus, introduced the word “metabolomics” [4–6]. Metabolomics
brought to science a new perspective for diseases diagnosis, since
biofluids can be collected noninvasively and in small amounts, as is
the case for urine and blood sampling. These biofluids are useful
because they reflect the biochemical status of an organism and the
potential homeostasis changes [7].
275
276 João G.M. Pontes et al.
With non-destructive analysis, easier sample preparation, and

selective experiments for biological samples, NMR has some advan-
tages in analysis of biofluids compared to other techniques such as
those based on mass spectrometry (MS). Although MS is more
sensitive and requires only small sample amounts, the analysis is
destructive [8]. NMR data interpretation sometimes can be simpler
when compared to MS. However, a two-dimensional (2D) NMR
analyses are needed for complex samples because they give more
details about sample composition. Besides, there are more data
indexed in databases, which provide more accurate and reliable
analysis. However, 2D spectra acquisition requires a longer proce-
dure, when compared to one-dimensional (1D) methods [8].
The NMR applications in metabolomics are based on the iden-
tification of small and macromolecules that caused group separa-
tion, which are present in one and absent in other group, or have
different concentrations in the investigated groups. The chemical
shift, coupling constants (splitting patterns) and different peak
intensities in 1H NMR spectra reveal details about qualitative and
quantitative relationships between intramolecular and intermolec-
ular resonances. This information is acquired due to nuclei mag-
netic spin properties. When nuclei are submitted to a strong
magnetic field their spin energy levels changes and this can be
observed through radiofrequency waves (λ between 102 to 104 m).
Pulsed NMR generates a Free Induction Decay (FID) as an observ-
able signal of nonequilibrium nuclear spin magnetization. However,
the signal output is in the time domain, which is not explainable.
Therefore, the transformation to frequency domain through
Fourier Transformer (FT) is needed [9, 10].
Spectra obtained by NMR offer the raw data to be processed in
the next step in the search for biomarkers, known as chemometrics.
This is a crucial method for obtaining useful information from the
raw NMR data. Usually, statistical procedures include multivariate
methods such as principal component analysis (PCA) [11] and par-
tial least squares (PLS) analysis [12]. Even so, biofluid 1H NMR
spectra may show variability, probably caused by differences in pH,
metal-ion concentrations, and chemical exchange phenomena [13].
This requires baseline correction, alignment, binning, normalisa-
tion and scaling of the NMR data before applying chemometrics.
2 Materials
2.1 Collecting 1. Serum samples (see Notes 1 and 2).

Blood Serum 2. Serum vacutainer tubes.
3. Sodium azide (NaN3).
4. Polypropylene tubes.
5. Biofreezer at −80 °C (see Notes 3 and 4).
Metabolomics by NMR 277
2.2 Samples 1. Deuterium oxide (D2O, 99.9 %).

Preparation 2. NMR tubes (18 cm × 5 mm).
3. Phosphate buffered saline (PBS) (see Note 5)
2.3 One- and Two- 1. Bruker® Avance III 600 MHz spectrometer (see Note 6).
Dimensional NMR 2. Triple resonance Broadband Inverse (TBI) probe (see Note 7).
Spectra Acquisition
3. FID data interpreting software Bruker® TopSpin version 3.1.
2.4 Chemometric 1. Spreadsheet software, such as Microsoft® Office Excel and

Analysis Origin®.
2. Chemometrics analysis software, such as Pirouette®, MATLAB®.
3 Methods
3.1 Sera Sample 1. Collect the blood sample using a peripheral vein access in the
Collection morning (7–10 AM) after fast of 12 h (see Note 8).
2. Immediately put the sample recipient on ice and leave 1 h for
coagulation.
3. Centrifuge, collect the supernatant fraction and add 0.05 mM
sodium azide to avoid bacterial contamination [14].
4. Store the vacutainer at −80 °C (see Note 9).
3.2 1H NMR Spectra 1. To a 250 μL of animal/human serum sample add 250 μL of

Acquisition deuterium oxide or PBS with 10 % of deuterium oxide.
and Processing 2. Transfer the solution into an NMR tube.
3. For analysis of non-lyophilized blood serum samples, the
WATERGATE pulse sequence should be used [15], which is a
specific and appropriate pulsed gradient field technique for
suppression of the water signal (see Notes 10 and 11).
4. Acquisition of 1H NMR spectra is performed in triplicate for
monitoring reproducibility.
5. For 1H NMR acquisition use following parameters shown in
Table 1 (see Note 12).
6. Adjust the baseline of all 1H NMR spectra acquired (see Note 13).
7. Reference the chemical shift using the lactate doublet
(see Note 14).
8. For HSQC acquisition, use the parameters shown in Table 2.
3.3 Chemometrics 1. After acquiring and preprocessing 1H NMR spectra, these are
Data Processing exported as American Standard Code for Information
Interchange (ASCII) files, so intensity and chemical shift val-
ues are listed and the data organized using a spreadsheet
software (see Note 15).
Table 1
Steps to 1D spectra acquisition
A Solvent (D2O).
B. Lock (commonly one can use automatic correcting)
C Probe match/tune (manual or automatic correcting)
D Shimming (automatic/manual command). Sometimes it is necessary to activate more than one
time and adjust according to reference signal
E Acquisition pars. This can be a pulse sequence that in this case was either WATERGATE or
cpmg1d in T2 analysis. The number of scans can change according to spectra precision.
Generally, 2–4 scans are used to check the sample and 128 scans are used for data acquisition
F Prosol pars. Use the same number of scans utilized in acquisition pars
G Receiver gain (automatic command)
H Start acquisition
Table 2
Acquisition parameters 2D NMR–HSQC spectra acquisition
Pulse sequence hsqcedetedgpsp.3

Number of scans 64
Number of dummy scans 16
Matching/tuning Manual
2. The chemometric analysis in data preprocessing is carried out

according to the method of metabolomics analysis that is
employed. For example: target analysis, metabolic profiling or
metabolic fingerprinting.
3. Choose autoscale as preprocessing for Principal Component
Analysis to data analysis at selected spectral range.
4. Exclude outliers and repeat the analysis.
3.4 Biomarkers 1. The choice of a spectral region for chemometrics analysis in the
Identification search for biomarkers should be made in accordance with the
biochemical and spectroscopical knowledge necessary to inter-
pret the biological problem at question.
2. After the chemometrics analysis run, it is necessary to identify
“outlying” samples, i.e., data that do not fit in the model. It is
important to remove the outliers if any [16].
3. For biomarkers searching and comparison use any of the following
databases: Human Metabolome Database (HMDB), Biological
Magnetic Resonance Bank (BMRB) and Madison-Qingdao
Metabolomics Consortium Database.
4. Finally, once the spectral range is determined, the peaks are

assigned and the contour maps correlations used to com-
pare experimental results with available data banks (Fig. 1)
(see Note 16).
Fig. 1 Summary of key steps in the metabolomics analysis. Step 1 — Place sample into an NMR tube, insert
this into the NMR instrument and record the 1H NMR spectra according to the instruction manual. Step 2 —
After Fourier Transformation of the data, each NMR spectrum must be referenced and exported as an ASCII file.
Target the parts of the spectra that include differences between the groups of samples. Step 3 — Run PCA and
PLS-DA (or other chemometrics tool according to the model) over selected parts of the spectra and analyze
groupings. Identify the NMR data responsible for the differences between the groups of samples. Step 4 — If
there are no rough deviations between the model and reality, assign and interpret the NMR data, using princi-
pally 2D NMR and search the appropriate NMR databases for identification of the metabolites
4 Notes
1. While handling animal and human biofluids it is necessary wear

a lab coat, eyewear, and gloves for self-protection during all
processes, up to the point of sample storage after NMR
analysis.
2. It is necessary to obtain authorization of local ethics commit-
tee and a document with experiment details, such as start and
final dates and procedures to be included in the research.
3. Sample freezing and thawing procedures should be minimized
as much as possible as to avoid subtle changes in the spectro-
scopic profiles of certain metabolites that may occur and lead
to a systematic error among the samples [17].
4. It is important to mention that each metabolite present in
blood serum sample has its own degradation time according to
its chemical nature and the temperature that it was stored. For
instance, l-proline immediately undergoes chemical changes in
serum samples when stored at −20 °C, while other metabo-
lites, such as, uric acid and cholesterol are stable at this tem-
perature for a period of 3 months [18].
5. In analysis of tissues and some biofluids like urine, phosphate
buffered saline (PBS) is commonly used as to provide pH sta-
bilization and to simulate native conditions, thus keeping
cells/fluids as close as possible to the physiological conditions
[19, 20].
6. The magnetic field frequency of the NMR spectrometer is
essential to a gain better sensitivity, so high-field NMR (greater
than 500 MHz) provides a better resolution and high-quality
information.
7. In 1H NMR spectra analysis for metabolomics, it is preferable
to use the TBI probe in order to have higher sensitivity in
analysis because the inner coil of this probe is designed for
optimized 1H nucleus analysis.
8. It is important to collect the blood during the morning and
after 12 h of fasting to ensure that other metabolites or
xenobiotics present in the patient organism resulting from
daily activities (to lunch, to take medications) do not interfere
or mask the biological process in question.
9. Once fasting blood is collected, serum is obtained and the
sample prepared, it is necessary to store it at an ultra-low tem-
perature (−70 or −80 °C). Do not have abrupt changes of
conditions and do not unfreeze them more than two times
before submitted to the NMR analysis [21, 22].
10. In the WATERGATE method, a frequency-selective RF pulse
is an applied, which gives a broad excitation in the small
selected region that corresponds to an intensity reduction of
www.Ebook777.com
about 20 % plus some signals nearer to the solvent region are

lost. In practice, these peaks correspond to anomeric carbons
of carbohydrates, which is a method limitation [23].
11. Ideal water suppression takes in to account the following criteria:
high effectiveness (suppression of water signals in the range of 105
to 109 times more intense than metabolites peaks); high selectiv-
ity; and high efficiency to suppress peaks in a short time [24].
12. NMR serum sample analysis can sometimes be difficult to ana-
lyze due to high concentrations of proteins and lipids that can
obscure and overlap the low-molecular-mass signals. To solve
this problem, different pulse sequences have been used that
suppress resonances in a spin-echo mode through phase mod-
ulation and relaxation effects [25]. A commonly used pulse
sequence is Carr–Purcell–Meiboom–Gill (CPMG) [26] which
suppresses signals of large molecules (peptides and lipids)
improving spectral resolution and consequently improving
multiplicity identification and peak assignment.
13. In metabolomics, it is not appropriate to integrate the signals of
1
H NMR spectra, because there are small distortions in the base-
lines that do not a allow reliable data quantification [27].
14. Besides lactate (δ 1.33, 3H, doublet, 3J = 7 Hz), some reference
compounds can be used in aqueous medium, such as 3-(tri-
methyl)silylpropionic acid-2,2,3,3-d4 (TMSP-d4) referenced at
δ 0.00 (9H, singlet) and 2,2-dimethyl-2-silapentane-5-sulfonate
sodium salt (DSS) (main signal at δ 0.00, singlet). In organic
solvents, tetramethylsilane (TMS) is widely used [19, 28].
15. T2-edited 1H NMR spectra should not be used as data for che-
mometrics due to fact that differences in the relaxation times
of metabolites might invalidate their quantification. In T2-
edited 1H NMR spectra the intensities of peaks one variable of
the PCA are altered, but chemical shifts and/or coupling con-
stant values are not.
16. Sometimes it is necessary to employ Dynamic Nuclear Magnetic
Resonance Spectroscopy for the search of complementary data
that give information about fluxomics, studies about protein–
ligand interactions, pharmacokinetics, thereby contributing to
a greater understanding of biochemical mechanisms and pro-
cesses involved in the biological system under study.
Acknowledgments
We thank the Fundação de Amparo à Pesquisa do Estado de São

Paulo (FAPESP, São Paulo, Brazil) and Conselho Nacional de
Desenvolvimento Científico e Tecnológico (CNPq, Brasília, Brazil)
for financial supports (FAPESP Process No 2014/18938-8 and
CNPq Process Number: 454234/2014-7) and fellowships.
References
1. Horgan RP, Kenny LC (2011) ‘Omic’ tech- detection in chemometrics. Crit Rev Anal
nologies: genomics, transcriptomics, pro- Chem 36:221–242
teomics and metabolomics. Obstet Gynecol 17. Teahan O, Gamble S, Holmes E, Waxman J,
13:189–195 Nicholson JK, Bevan C et al (2006) Impact of
2. Bonne NJ, Wong DTW (2012) Salivary bio- analytical bias in metabonomic studies of
marker development using genomic, pro- human blood serum and plasma. Anal Chem
teomic and metabolomic approaches. Genome 78:4307–4318
Med 4:82. doi:10.1186/gm383 18. Yin P, Lehmann R, Xu G (2015) Effects of pre-
3. Kovac JR, Pastuszak AW, Lamb DJ (2013) The analytical processes on blood samples used in
use of genomics, proteomics, and metabolo- metabolomics studies. Anal Bioanal Chem
mics in identifying biomarkers of male infertil- 407:4879–4892
ity. Fertil Steril 99:998–1007 19. Beckonert O, Keun HC, Ebbels TMD, Bundy
4. Oliver SG, Winson MK, Kell DB, Baganz F J, Holmes E, Lindon JC et al (2007) Metabolic
(1998) Systematic functional analysis of the profiling, metabolomic and metabonomic pro-
yeast genome. Trends Biotechnol 16:373–378 cedures for NMR spectroscopy of urine,
5. Fiehn O (2002) Metabolomics – the link plasma, serum and tissue extracts. Nat Protoc
between genotypes and phenotypes. Plant Mol 2:2692–2703
Biol 48:155–171 20. Parab GS, Rao R, Lakshminarayanan S, Bing YV,
6. Villas-Bôas SG, Rasmussen S, Lane GA (2005) Moochhala SM, Swarup S (2009) Data-driven
Metabolomics or metabolite profiles? Trends optimization of metabolomics methods using rat
Biotechnol 23:385–386 liver samples. Anal Chem 81:1315–1323
7. Pinto J, Domingues MRM, Galhano E, Pita C, 21. Zheng P, Gao HC, Li Q, Shao WH, Zhang
Almeida MC, Carreira IM et al (2014) Human ML, Cheng K et al (2012) Plasma metabo-
plasma stability during handling and storage: nomics as a novel diagnostic approach for
impact on NMR metabolomics. Analyst major depressive disorder. J Proteome Res
139:1168–1177 11:1741–1748
22. Rankin NJ, Preiss D, Welsh P, Burgess KEV,
8. Emwas AHM, Salek RM, Griffin JL, Merzaban
Nelson SM, Lawlor DA et al (2014) The emer-
J (2013) NMR-based metabolomics in human
gence of proton nuclear magnetic resonance
disease diagnosis: applications, limitations, and
metabolomics in the cardiovascular arena as
recommendations. Metabolomics 9:1048–1072
viewed from a clinical perspective.
9. Bharti SK, Roy R (2012) Quantitative 1H Atherosclerosis 237:287–300
NMR spectroscopy. TrAC Trend Anal Chem
23. Adams RW, Holroyd CM, Aguilar JA, Nilsson
35:5–26
M, Morris GA (2013) “Perfecting”
10. Fourier JBJ (1822) Théorie Analytique de la WATERGATE: clean proton NMR spectra from
Chaleur. Paris: Chez Firmin Didot, père et fils, aqueous solution. Chem Commun 49:358–360
p. 525
24. Tang H, Wang Y (2008) High-resolution
11. Wold S (1987) Principal component analysis. NMR spectroscopy in human metabolism and
Chemom Intell Lab 2:37–52 metabonomics. In: Webb GA (ed) Modern
12. Wold S, Sjöström M, Eriksson L (2001) PLS- magnetic resonance – Part III: Applications in
regression: a basic tool of chemometrics. materials science and food science. Springer,
Chemom Intell Lab 58:109–130 New York, pp 1623–1630
13. Trygg J, Holmes E, Lundstedt T (2007) 25. Wevers RA, Engelke U, Heershap A (1994)
Chemometrics in metabonomics. J Proteome High-resolution 1H NMR spectroscopy of
Res 6:469–479 blood plasma for metabolic studies. Clin Chem
14. Verwaest KA, Vu TN, Laukens KA, Clemens 40:1245–1250
LE, Nguyen HP, Gasse BV et al (2011) 1H 26. Meiboom S, Gill D (1958) Modified spin-echo
NMR based metabolomics of CSF and blood method for measuring nuclear relaxation times.
serum: a metabolic profile for a transgenic rat Rev Sci Instrum 29:688–691
model of Huntington disease. Biochim Biophys 27. Smolińska A (2012) Chemometrics and NMR
Acta 1812:1371–1379 spectroscopy for metabolomics analysis of neu-
15. Piotto M, Saudek V, Sklenár V (1992) rological disorders. Radboud University,
Gradient-tailored excitation for single- Nijmegen. ISBN 978-94-6191-358-6
quantum NMR spectroscopy of aqueous solu- 28. Gebregiworgis T, Powers R (2012) Application
tions. J Biomol NMR 2:661–665 of NMR metabolomics to search for human
16. Debruyne M, Engelen S, Hubert M, disease biomarkers. Com Chem High T Scr
Rousseeuw PJ (2006) Robustness and outlier 15:595–610
Chapter 25
Lab-on-a-Chip Multiplex Assays

Harald Peter, Julia Wienke, and Frank F. Bier
Abstract
Lab-on-a-chip multiplex assays allow a rapid identification of multiple parameters in an automated manner.
Here we describe a lab-based preparation followed by a rapid and fully automated DNA microarray hybrid-
ization and readout in less than 10 min using the Fraunhofer in vitro diagnostics (ivD) platform to enable
rapid identification of bacterial species and detection of antibiotic resistance. The use of DNA microarrays
allows a fast adaptation of new biomarkers enabling the identification of different genes as well as single-
nucleotide-polymorphisms (SNPs) within these genes. In this protocol we describe a DNA microarray
developed for identification of Staphylococcus aureus and the mecA resistance gene.
Key words Bacterial infection, Early diagnosis, Point-of-care, Lab-on-a-chip, Antimicrobial drug
resistance, Antibiotic resistance detection, MRSA detection, Genotyping, SNP detection, Automated
DNA microarray hybridization
1 Introduction
There is an urgent need for rapid diagnostics in the field of infectious

diseases where fast decisions need to be made for an effective therapy.
Traditional microbiological methods take around 48 h for species
identification and antibiogram results. The use of molecular tests
built on multi-parameter lab-on-a-chip platforms are a good option
to speed up the time taken between sampling and results. Here we
describe a protocol which makes use of the Fraunhofer integrated
lab-on-a-chip in vitro diagnostic (ivD) platform. The platform can be
used to run multiplex immunoassays and serological assays to identify
antibodies directly from blood samples [1]. In this protocol we dem-
onstrate another type of application, consisting of a fully automated
DNA microarray assay and readout, together with a lab-based prepa-
ration method. This protocol allows rapid bacterial species identifica-
tion and the genotyping of relevant antibiotic resistance genes.
The advantage of microarrays is the possibility to analyze a large
amount of parameters at one time, making these assays particularly
useful in the field of antibiotic resistance determination [2–7].
283
284 Harald Peter et al.
Nonetheless, there are still some disadvantages regarding current

DNA microarray protocols that hinder the broad use of this approach
in routine analysis. These include long hybridization times and com-
plex laboratory based procedures, as well as the need for skilled oper-
ational personnel. The use of lab-on-a-chip based systems may help
to overcome these limitations since they offer a rapid and automated
solution, which integrates many of the lab-based procedures.
The combination of microarrays with microfluidics constitutes an
elegant solution to automate and speed up microarray hybridization.
The ivD platform consists of a microfluidic cartridge and a base unit.
The credit card sized cartridge contains all relevant elements neces-
sary for DNA microarray hybridization. These are mainly reservoirs
for all of the reagents, integrated pumping systems, the specific
microarray, integrated temperature control for hybridization and an
optical transducer to allow integrated sensing [1, 8, 9]. The base unit
contains all necessary electronics to control the cartridge, an optical
readout system to analyze the microarray directly after hybridization
and a touch screen to control the assay and monitor the results. With
this setup, hybridization times of less than 5 min can be achieved,
yielding equivalent results as found by lab-based microarray hybrid-
izations of 1 h. Therefore, the total time to generation of results is
less than 10 min for hybridization, washing and readout.
Here, we describe all necessary steps from sample preparation
to microarray analysis for a lab-based and lab-on-a-chip-based pro-
cedure. As an example, we use a species identification microarray
for the detection of methicillin-resistant Staphylococcus aureus
(MRSA). This microarray applies the principle of allele-specific
hybridization and can therefore be used for identification of single-
nucleotide polymorphisms (SNPs). It consists of four oligonucle-
otide probesets which differ only in one central nucleotide position,
based on principles published previously [3, 4, 10]. Figure 1 shows
Fig. 1 SNP detection principle used on DNA microarray. (Left) Array layout of a probe set in duplicate for SNP
detection with the expected perfect match signal marked in black. (Middle) Fluorescent signals on the microarray
(false color image). (Right) Quantified fluorescent signal intensities of one probe set, normalized to the perfect
match signal
Lab-on-a-Chip Multiplex Assays 285
a SNP detection probe set as an example. The assay can be per-

formed using conventional hybridization or automated using the
Fraunhofer lab-on-a-chip system. The use of the latter will shorten
the hybridization, washing and readout times from approximately
2 h to less than 10 min. Both procedures are listed in the protocol
as shown in Fig. 2. Alternatively to this complete protocol, an even
faster assay time can be achieved by carrying out the labeling PCR
amplification directly from a colony, followed by DNA fragmenta-
tion and automated hybridization, washing, and readout using the
Fraunhofer ivD platform (Fig. 3). Species identification and antibi-
otic resistance information can be obtained with this (fast) proto-
col directly from a colony with a PCR time of 60 min and an ivD
Fig. 2 Schematic of the lab-based microarray protocol. For hybridization, washing, readout and data analysis,
one can either use a manual variant (2 h processing time) or a lab-on-a-chip system, like the automated
Fraunhofer ivD platform, allowing a processing time of less than 10 min
Fig. 3 Schematic of the rapid, semi-automated ivD platform protocol, allowing a microarray analysis directly
from a colony within 75 min (60 min PCR, 5 min fragmentation, and 10 min ivD platform)
step of 10 min (see Note 1). Our practical experience has shown
that the results of the rapid protocol are comparable to those pro-
duced by the full protocol.
2 Materials
2.1 DNA Extraction 1. Müller-Hinton agar with sheep blood (Thermo Fisher Scientific
Oxoid; Braunschweig, Germany).
2. 30 g/L CASO-Bouillon medium (ROTH; Karlsruhe, Germany).
3. Ethanol at −20 °C.
4. 2 mg/mL lysostaphin (Sigma-Aldrich; Taufkirchen, Germany).
5. 10 mg/mL ribonuclease A (Thermo Fisher Scientific;
Darmstadt, Germany).
6. Enzymatic lysis buffer: 20 mM Tris base (pH 8), 2 mM Na-EDTA,
1.2 % Triton X-100.
7. Commercially available kit: DNeasy Blood & Tissue Kit, including
buffers AW1 and AW2 (Qiagen; Hilden, Germany).
2.2 PCR 1. Standard PCR reaction mixture for amplification of 23S

Amplification rDNA, 16S rDNA, and mecA gene: 10 ng chromosomal
and Labeling DNA, 0.8× Taq buffer, 0.6 mM dNTPs (Qiagen), 0.6 mM
23S rDNA and 0.4 mM 16S rDNA/mecA oligonucleotide
primers (Metabion; Martinsried, Germany), 3.75 U HotStar
Taq polymerase (Qiagen), and 2 mM MgCl2 in a final volume
of 12.5 μL.
2. PCR labeling dNTP mixture for amplification of 23S rDNA,
16S rDNA, and mecA: 10 ng chromosomal DNA, 0.8× Taq
buffer, dATP, dGTP, dCTP, and 0.36 μM dTTP/0.24 μM
DY-647P1-aadUTP (Dyomics; Jena, Germany), 0.6 μM 23S
rDNA, and 0.4 μM 16S rDNA/mecA oligonucleotide primers,
3.75 U HotStar Taq polymerase, and 2 mM MgCl2 in a final
volume of 12.5 μL.
3. Thermal cycler PTC200 with alpha block (MJ Research; St.
Bruno, Canada) (see Note 2).
2.3 Agarose Gel 1. 1 % agarose gel in 100 mM Tris–HCl (pH 8.3), 50 mM acetic
Electrophoresis acid, 1 mM EDTA (TAE) buffer containing peqGreen dye
(peqLab; Erlangen, Germany).
2. GeneRuler 1 kb Ladder and 6× DNA Gel Loading Dye
(Thermo Fisher Scientific).
2.4 PCR Purification 1. QIAquick PCR Purification Kit (Qiagen).

2. NanoDrop ND-1000 spectrophotometer (NanoDrop Techno-
logies; Wilmington, DE, USA) (see Note 3).
2.5 DNA 1. 10× DNase buffer (Promega; Mannheim, Germany).

Fragmentation 2. DNase: 1 U/μL (Promega).
3. DNase stop buffer: 20 mM EGTA (pH 8).
4. Agilent Bioanalyzer 2100 (Agilent Technologies; Palo Alto,
CA, USA).
2.6 DNA Microarray 1. DNA spotting buffer: Nexterion Spot 2× (Schott; Jena,
Fabrication Germany).
2. Oligonucleotides: each 100 μM (Metabion).
3. Epoxysilane slides (3D-Epoxy Glass Slides; PolyAn; Berlin,
Germany).
4. Rinse buffer 1: 0.1 % Triton X-100.
5. Rinse buffer 2: 6 mM HCl.
6. Rinse buffer 3: 0.1 mM KCl.
7. Blocking solution: 0.4 M Tris base, 50 mM ethanolamine
(pH 9).
8. Microarray spotter: sciFLEXARRAYER S11 (Scienion AG;
Berlin, Germany) (see Note 4).
2.7 Microarray 1. 20× saline sodium citrate buffer (SSC): 3 M sodium chloride,
Hybridization, Washing 0.3 M sodium citrate (pH 7.0).
and Scanning 2. Wash buffer 1: 2× SSC, 0.2 % SDS (freshly added).
3. Wash buffer 2: 2× SSC.
4. Wash buffer 3: 0.2× SSC.
5. ProPlate Multi-array system (Grace Bio-Labs; Bend, OR,
USA).
6. ProPlate adhesive seal-strips (Grace Bio-Labs).
7. Hybridization oven or temperature controlled orbital shaker
(see Note 5).
8. Laser scanner (see Note 6).
9. GenePixPro (Molecular Devices LLC; Sunnyvale, CA, USA)
or alternative quantification software.
2.8 Automated 1. Fraunhofer ivD cartridge (e.g., with MRSA-Detect microar-

Hybridization, Washing ray) (BiFlow Systems; Chemnitz, Germany).
and Readout 2. Fraunhofer ivD platform (base unit).
(Fraunhofer ivD
Platform)
2.9 Microarray Data 1. Spreadsheet software such as Microsoft Excel (Microsoft,

Analysis Redmond, WA, USA).
3 Methods
3.1 Genomic DNA 1. Inoculate 5 mL of CASO-Bouillon with a single colony of

Extraction Staphylococcus species and incubate at 37 °C overnight while
shaking.
2. Pour 1.8 mL of the culture into a 2 mL Eppendorf tube and
centrifuge at ≥6000 × g for 5 min at 4 °C.
3. Remove as much of the medium as possible and proceed
immediately to the next step or freeze the cell pellet at −20 °C
overnight if needed (see Note 7).
4. Suspend the pellet in 200 μL enzymatic lysis buffer containing
freshly added 100 μg/mL lysostaphin and 100 μg/mL ribo-
nuclease A.
5. Incubate the tube at 37 °C for 20 min until it becomes more
transparent and viscous (see Note 8).
6. Add 25 μL proteinase K and 200 μL Buffer AL (provided in the
Quiagen DNase kit), vortex, and incubate at 56 °C for 1 h.
7. Add 200 μL of ethanol at -20 °C and mix by inverting the tube
several times.
8. Pour the whole suspension into the spin column provided in
the kit and centrifuge at ≥6000 × g for 1 min at 4 °C.
9. Transfer the column to a new 2 mL Eppendorf tube, add
200 μL buffer AW1 and centrifuge at ≥6000 × g for 1 min at
4 °C.
10. Add 200 μL buffer AW2 and centrifuge at ≥8000 × g for 1 min
at 4 °C.
11. Transfer column to a new 2 mL Eppendorf tube and centri-
fuge at 14,000 × g for 2 min at 4 °C.
12. Transfer column to a new 1.5 mL Eppendorf tube and elute
the DNA by adding 50 μL water, incubating for 1 min at room
temperature and centrifuging at ≥6000 × g for 1 min at room
temperature (see Note 9).
13. Estimate the concentration of extracted DNA using a UV/Vis
spectrophotometer and check for the correct size of about
20 kb by running an agarose gel (see Note 10).
3.2 Standard PCR 1. Carry out PCR to amplify as follows: hot start at 95 °C for
Amplification (See 15 min and then 30 cycles of denaturation at 95 °C for 0.5 min,
Note 11) annealing at 50 °C for 0.5 min, and elongation at 68 °C for
1.5 min, followed by a final elongation step at 68 °C for 4 min
(see Note 12).
2. For verification of PCR products, carry out an agarose gel
electrophoresis on 1 % agarose gels in TAE buffer using 2 μL
of the PCR product (see Note 13).
3.3 Labeling PCR 1. Follow the protocol for a standard PCR amplification using
Amplification the labeling dNTP mixtures (see Note 15).
2. Confirm amplification by agarose gel electrophoresis as above
using 0.5 μL of the PCR product.
3.4 Purification 1. Use the PCR Purification Kit to purify the labeled PCR prod-
of PCR Products (See uct following the standard protocol and elute in 30 μL water.
Note 14) 2. Analyze the purified labeled PCR products for DNA yield and
fluorescent dye incorporation by spectrophotometry using the
ND-1000 UV/Vis spectrophotometer.
3. Calculate the incorporation rate of the fluorescent dye as
shown in the equation below (see Note 16).
R = DNA concentration (ng/μL) × 1000/dye concentration
(pmol/μL) × 330 g/mol (see Note 17).
3.5 DNA 1. For each microarray hybridization use approximately 750 ng

Fragmentation DNA generated from the PCR step and add 1× DNase buffer
and 0.6 U DNase (see Note 18).
2. Digest for 5 min (±2 s) at room temperature and stop the reac-
tion by adding 6 μL of DNase stop buffer and incubate at
65 °C for 10 min (see Note 19).
3. Keep the digested DNA solution at 4 °C until further use but
do not store longer than a few hours.
3.6 Microarray 1. Dilute each oligonucleotide probe in 1× DNA spotting buffer

Fabrication to a final concentration of 20 μM and pipette 30 μL of probe
solution into designated wells of a 384-well microtiter plate.
2. Use a non-contact spotter to print the probes onto epoxy
coated slides, setting the grid to 400 μm distance between the
spots (see Note 20).
3. In order to immobilize the probes after spotting, incubate the
slides at 60 °C for 30 min in a drying oven (see Note 21).
4. To remove unbound probes, incubate the slides in rinse buffer
1 for 5 min, in rinse buffer 2 for 4 min, in rinse buffer 3 for
10 min and finally in water for 1 min at room temperature,
with constant steering.
5. To inactivate free epoxy groups block the slides for 15 min in
blocking solution at 50 °C and rinse for 1 min in water.
6. Dry the slides by centrifugation or using a flow of nitrogen.
3.7 Microarray 1. For hybridization use the labeled and fragmented target DNA
Hybridization, solution in a total of volume of 46 μL, add 0.5 μL of 0.05 μM
Washing, and Image labeled hybridization control (see Note 22), dilute to a final 2×
Acquisition SSC concentration in a total volume of 80 μL.
2. Assemble the hybridization chamber on the slide and add the
hybridization solution into the chamber, ensuring that no air
bubbles form.
3. Seal the chamber with adhesive seal strips and incubate for 1 h
at 48 °C while shaking (see Note 23).
4. Following hybridization, wash the slides in wash buffer 1, 2,
and 3 each for 10 min at room temperature with constant
steering, then dip the slides in water for 1 s and dry under a
flow of nitrogen (see Note 24).
5. For fluorescence image acquisition, image the slides using a laser
scanner at 635 nm and using appropriate PMT/gain settings
depending on signal and background intensity (see Note 25).
6. Quantify the fluorescent signals with the software provided
with the scanner.
3.8 Automated 1. Prepare the hybridization solution as described above and add
Hybridization, to reservoir 5 of the ivD cartridge (see Note 26) (Fig. 4).
Washing, and Readout 2. Insert the cartridge into the ivD platform base unit and start
(Fraunhofer ivD the hybridization program (see Note 27).
Platform) 3. The fluorescence image data can either be analyzed automatically
within the base unit or exported for external analysis.
3.9 Data Analysis 1. After image acquisition and fluorescent signal quantification,
(See Note 28) subtract the local background of each spot from the raw spot
intensity value, and calculate the mean net signal intensity (NI)
and standard deviation (SD) of the replicates.
2. Within each probe set, which is interrogating one mutation site,
the probe with the highest signal intensity is termed perfect match
(PM) and the remaining probes are marked as mismatch (MM).
3. In order to evaluate the performance of each probe set, calcu-
late the ratios between the MM and PM signal intensities, as
the relative signal intensity RImax(MM) (see Note 29).
Fig. 4 Fraunhofer ivD platform for fully automated DNA microarray hybridization and analysis. (Left) Lab-on-a-
chip cartridge (size: 60 × 40 mm) with 9 reservoirs and integrated micropumps, microfluidic channels, thermal
control elements, electronics and a sensor area for microarrays with up to 400 spots. (Right) Base-unit (size:
14 × 14 × 14 cm) for control, readout and data analysis of the lab-on-a-chip cartridge. Results can either be
transferred to a computer or analyzed directly and presented on the display
4. The MM probe with the highest signal intensity is used for

calculation of the relative signal intensity (RImax(MM) = NImax(MM)/
NIPM) (see Note 30).
5. In addition to the RI value, the limit of detection (LOD) is used
to evaluate performance and this is calculated based on the max-
imum signal intensity (NImax) obtained within each probe set
based on a no template control (NTC) hybridization plus 3
times the highest standard deviation (LOD = NImax + 3 × SDmax)
(see Note 31).
6. In addition, the coefficient of variation (CV) is calculated for
each set of replicate probes (CV = SD/NIPM).
7. Probe sets with a CV > 30 % of the signal intensity of PM should
be flagged and excluded from analysis to ensure that only those
probe signals with a high reproducibility are used for the analysis
(see Note 32).
4 Notes
1. For the rapid protocol, the DNA extraction can be omitted

and a labeling PCR can be performed directly from a colony
[11]. The subsequent purification can also be omitted.
2. Other thermal cyclers can be used but it is recommended to
optimize the cycling conditions on a machine-by-machine
basis.
3. Other spectrophotometers can be used.
4. Other spotters can be used but compatibility with other mate-
rials and reagents is essential.
5. We used the Array Plate Multi-Well Microarray Hybridization
Station (ArrayIt, Sunnyvale, CA, USA).
6. We used the GenePix 4300A (Molecular Devices LLC).
7. Freezing improves the breaking down of the cells, especially of
gram positive species.
8. If the suspension does not become transparent, add an addi-
tional aliquot of lysostaphin (100 μg/mL) and keep the tube
at 37 °C until this occurs.
9. Heating the water for elution to 60 °C, as well as increasing
the incubation time up to 5 min can improve the yield.
10. The expected yield can differ from 15 to 120 ng/μL. The
DNA concentration can be increased by using less water for
the elution (although there is a minimum amount of 30 μL).
11. This non-labeling protocol can be used during the establish-
ment of a new single or multiplex PCR but can be omitted
once a new PCR is established. Then the user can continue
Fig. 5 Example gel electrophoresis image of three Staphylococcus strains (MRSA,

MSSA, and S. epidermidis with mecA). Results are shown for a triplex PCR analy-
sis for 23S rDNA (1840 bp), 16S rDNA (1485 bp) and mecA (364 bp)
directly with the labeling PCR step. In this example we use a

triplex PCR to amplify the 23S rDNA, 16S rDNA, and mecA
gene.
12. Once completed keep samples at 4 °C until further use or store
at −20 °C.
13. The PCR reaction mix contains three primer pairs to identify
the three genes: 23S rDNA, 16S rDNA, and mecA gene. The
DNA fragments corresponding to each of the genes are shown
in Fig. 5.
14. Purification of the PCR product is important to remove resid-
ual fluorescent dye.
15. Decreasing the salt concentration by using 0.8× Taq buffer can
improve amplification of longer amplicons. In contrast, increas-
ing the salt concentration can improve the yield of smaller ampli-
cons and using an elongation temperature of 68 °C improves
the amplification of longer fragments in multiplex PCRs [12].
16. The incorporation rate R is defined as the average distance of
each labeled nucleotide calculated as shown the equation. For
an efficient hybridization, the incorporation rate R should be
in the range between 80 and 200.
17. The average mass of a nucleotide is 330 g/mol.
18. The enzymatic digestion of the amplified DNA molecules is nec-
essary to improve the hybridization efficiency. The amount of
DNase needed for digestion should be optimized if a new target
is to be amplified, although 0.1 U DNase for 125 ng DNA is a

good amount to start with. The labeled PCR amplicons should
be digested nonspecifically to a range of 20–200 nucleotides.
The fragmentation efficiency can be monitored with an Agilent
Bioanalyzer 2100 using a DNA 1000 LabChip kit.
19. This is for heat inactivation of the enzyme. For optimization
purposes or development of a new fragmentation protocol the
fragmentation efficiency can be monitored with an Agilent
Bioanalyzer 2100 using a DNA 1000 LabChip kit.
20. We use the sciFLEXARRAYER for spotting and PolyAn slides
for the array surface. Using the indicated settings, the spots
will have a diameter of about 150 μm.
21. After this stage, the slides can be stored for several months
before use.
22. This is complementary DNA to the immobilized positive con-
trol probe.
23. Make sure the slides are protected from light during the
hybridization as the fluorophores are light sensitive.
24. For the rapid protocol, the purification step can be omitted.
The labeled Amplicon can then be directly fragmented using
the standard protocol. Constant stirring of all solutions during
washing is needed. Slides should not run dry between washing
steps.
25. We use the GenePix 4300A scanner with the corresponding
Cy5 filter, as the dye DY-647P1-aadUTP has similar spectral
properties as Cy5.
26. All other reservoirs are already delivered with the necessary
hybridization and wash buffers. If an empty cartridge is being
used, each reservoir can also be filled with custom buffers and
reagents individually.
27. After 5 min of hybridization the cartridge will automatically
start with the washing and imagine acquisition and the results
are obtained after 10 min.
28. The following protocol is recommended to analyze DNA
microarrays containing genotyping SNP identification probe
sets. The algorithms need to be adjusted for the specific DNA
microarrays used.
29. The larger the relative difference between MM and PM signal,
the better the discriminative power of the probe set.
30. Only probe sets that show a performance with RImax(MM) < 0.7
are used for the analysis. The use of this threshold has been
shown to result in high quality discriminations [3, 4, 13].
31. Only probe sets with a perfect match signal intensity above the
limit of detection (NIPM > LOD) should be used for analysis.
32. For convenience it is recommend to fully automate the

algorithms described above (e.g., by using a spreadsheet pro-
gram, using the raw quantification file from the quantification
software as input).
References
1. Schumacher S, Nestler J, Otto T, Wegener M, detection of fluoroquinolone-resistant
Ehrentreich-Förster E, Michel D et al (2012) Escherichia coli from urine samples using a
Highly-integrated lab-on-chip system for genotyping DNA microarray. Int J Med
point-of-care multiparameter analysis. Lab Microbiol 297:417–429
Chip 12:464–473 8. Streit P, Nestler J, Shaporin A, Schulze R, Gessner
2. Antwerpen MH, Schellhase M, Ehrentreich- T (2016) Thermal design of integrated heating
Förster E, Bier FF, Witte W, Nübel U (2007) for lab-on-a-chip systems. Proceedings of the 17th
DNA microarray for detection of antibiotic International Conference on Thermal,
resistance determinants in Bacillus anthracis Mechanical and Multi-Physics Simulation and
and closely related Bacillus cereus. Mol Cell Experiments in Microelectronics and
Probes 21:152–160 Microsystems (EuroSimE), April 18–20, pp. 1–6
3. Peter H, Berggrav K, Thomas P, Pfeifer Y, 9. Schumacher S, Ludecke C, Ehrentreich-Förster
Witte W, Templeton K et al (2012) Direct E, Bier FF (2013) Platform technologies for
detection and genotyping of Klebsiella pneu- molecular diagnostics near the patient’s bedside.
moniae carbapenemases from urine by use of a Adv Biochem Eng Biotechnol 133:75–87
new DNA microarray test. J Clin Microbiol 10. Barl T, Dobrindt U, Yu X, Katcoff DJ,
50:3990–3998 Sompolinsky D, Bonacorsi S et al (2008)
4. Leinberger DM, Grimm V, Rubtsova M, Weile Genotyping DNA chip for the simultaneous
J, Schröppel K, Wichelhaus TA et al (2010) assessment of antibiotic resistance and patho-
Integrated detection of extended-spectrum- genic potential of extraintestinal pathogenic
beta-lactam resistance by DNA microarray- Escherichia coli. Int J Antimicrob Agents
based genotyping of TEM, SHV, and CTX-M 32:272–277
genes. J Clin Microbiol 48:460–471 11. Jonas D, Speck M, Daschner FD, Grundmann
5. Strommenger B, Schmidt C, Werner G, Roessle- H (2002) Rapid PCR-based identification of
Lorch B, Bachmann TT, Witte W (2007) DNA methicillin-resistant Staphylococcus aureus
microarray for the detection of therapeutically from screening swabs. J Clin Microbiol
relevant antibiotic resistance determinants in 40:1821–1823
clinical isolates of Staphylococcus aureus. Mol 12. Henegariu O, Heerema NA, Dlouhy SR, Vance
Cell Probes 21:161–170 GH, Vogt PH (1997) Multiplex PCR: critical
6. WeileJ SRD, Bachmann TT, Susa M, Knabbe C parameters and step-by-step protocol.
(2007) DNA microarray for genotyping Biotechniques 23:504–511
multidrug-resistant Pseudomonas aeruginosa 13. Grimm V, Ezaki S, Susa M, Knabbe C, Schmid
clinical isolates. Diagn Microbiol Infect Dis RD, Bachmann TT (2004) Use of DNA micro-
59:325–338 arrays for rapid genotyping of TEM beta-
7. Yu X, Susa M, Weile J, Knabbe C, Schmid RD, lactamases that confer resistance. J Clin
Bachmann TT (2007) Rapid and sensitive Microbiol 42:3766–3774
Chapter 26
Multiplex Smartphone Diagnostics

Juan L. Martinez-Hurtado, Ali K. Yetisen, and Seok-Hyun Yun
Abstract
Increasing computing power in smartphones allows for their transformation into point-of-care diagnostic
devices. Mobile medical diagnostic applications enable utilization of the processing capabilities of smart-
phones through their cameras. Hardware attachments or stand-alone versions of smartphone diagnostics
have the capability to revolutionize quantitative readouts. Here, we describe a protocol for quantifying
commercial colorimetric diagnostic tests with a stand-alone smartphone application. This approach can be
used in the multiplexed analyses of biomarker readouts.
Key words Smartphone, Diagnostics, Mobile, Medical, Application, Quantitative assays
1 Introduction
Smartphone applications utilizing their built-in cameras have been

developed for dermatology [1], microscopy [2, 3], ophthalmol-
ogy [4], chemical analyses [5], and paper-based diagnostic devices
[6–8]. Although they are normally optimized for photography,
smartphone camera algorithms can be modified so that sufficient
information can be extracted to enable their use as diagnostic tools.
Here, we describe the use of a stand-alone smartphone application
that compensates for lighting variability and fully utilizes the sen-
sor capabilities of smartphone cameras. This mobile application is
capable of analyzing multiple colorimetric tests allowing for multi-
plex molecular analyses in scenarios such environmental monitor-
ing, veterinary screening, and medical diagnostics. Specifically, we
describe the detailed methodology for a three-analyte commercial
test (glucose, protein, and pH) that can be extrapolated for 12 or
more analytes. The device uses test strips similar to commercial
kits that are utilized for monitoring kidney and liver functions, or
screening for diabetes [9]. Although these analytes are commonly
measured in laboratories, there is an increasing demand for home
295
296 Juan L. Martinez-Hurtado et al.
or workplace monitoring due to expanding healthcare expendi-

tures. It is important for diagnosed patients or individuals at risk to
continuously monitor their conditions to limit disease progression.
2.1 Artificial Urine 1. Prepare artificial urine stock solutions by varying the concen-
trations of protein and glucose, and the pH level following
previously reported protocols [10] and as shown in Table 1.
2. Prepare 200 mL of each stock solution in individual 10 mL
vials.
3. The protein concentrations should be 0 mg/dL; 30 mg/dL;
100 mg/dL; and 500 mg/dL.
4. Adjust the pH values to 5, 6, 7, 8, and 9.
Table 1
Stock concentrations of reagents in artificial urine
Component Quantity (g) Concentration (mM)

Peptone L 37 1
Yeast extract 0.005 N/A
Lactic acid 0.1 1.1
Citric acid 0.4 2
Sodium bicarbonate 2.1 25
Urea 10 170
Uric acid 0.07 0.4
Creatinine 0.8 7
Calcium chloride (2H2O) 0.37 2.5
Sodium chloride 5.2 90
Iron II sulphate (7H2O) 0.0012 0.005
Magnesium sulphate (7H2O) 0.49 2
Sodium sulphate (10H2O) 3.2 10
Potassium dihydrogen phosphate 0.95 7
Dipotassium hydrogen phosphate 1.2 7
Ammonium chloride 1.3 25
Distilled water to 1 L
Hydrochloric acid to specific pH
Sodium hydroxide to specific pH
Multiplex Smartphone Diagnostics 297
5. The glucose values should be 0 mg/dL; 50 mg/dL; 100 mg/

dL; 300 mg/dL; and 1000 mg/dL.
3.1 Basic Protocol 1. Calibrate the smartphone using three test strips dipped into
each of the 15 vials prepared as above.
2. Dip the test strip into artificial urine for 1 s and ensure that all
test areas are wetted.
3. Wipe the strip against the edge of the recipient to remove
excess liquid.
4. After 60 s reaction time, analyze the test strips through the
smartphone application by selecting the appropriate type of
test in the application menu.
3.2 Experimental 1. Light source: white light or fluorescent.

Conditions 2. Environmental conditions: 24 °C, 60 % relative humidity.
3.3 Technical 1. Phone model: Samsung Galaxy 5.

Variables 2. OS version: Android 2.4.
3. Mobile application: Colorimetrix v1.0 (Fig. 1).
3.4 Calibration 1. Open the app and select “ADD CALIBRATION” (Fig. 2).
2. Follow the on-screen instructions and add information such as
test name, analyte, units of concentration, and number of cali-
bration points (see Note 3).
Fig. 1 Ideal positioning of the smartphone and test strip. Ensure all test zones are
well lit with the ambient light kept as constant as possible. To allow for multiplex-
ing, increase the distance from the phone to the test strip until all areas are
covered
Fig. 2 Calibration steps 1 and 2
Fig. 3 Calibration steps 3–7
3. Proceed to record each calibration point by aligning the center

of the capture screen to the test zone (Fig. 3) (see Note 4).
4. Take a photograph of the calibration assay (or reference chart)
and approve the image (see Note 5).
5. Click “ANALYZE” and wait for a calibration circle to appear
on the screen.
6. Click “USE” and enter the concentration value for the first
point.
7. Repeat steps 4–6 until all calibration points are saved.
8. Return to the main screen and click “SELECT CALIBRATION”
and select the newly recorded calibration (Fig. 4) (see Note 6).
3.5 Measurement 1. Click “MEASURE” on the main screen to analyze the test
region.
2. Take a photograph and approve the photograph.
Fig. 4 Calibration step 8
Fig. 5 Measurement, steps 9 and 10
3. Click “EVALUATE,” wait for circle to appear, click “USE”

and record the value as displayed or access it from the database
(Fig. 5) (see Note 7).
4. Repeat measurement for all test zones (see Note 8).
5. Example calibration data are shown in Tables 2, 3, and 4 (see
Notes 9 and 10).
Table 2
Example calibration for pH test
Sample Real value Meas.1 Meas.2 Meas.3 Instrument value

1 5 5.44 4.76 5.38 N/A
2 6 5.77 6.32 6.43 N/A
3 7 6.76 7.35 7.56 N/A
4 8 8.44 8.36 7.64 N/A
5 9 9.24 9.33 8.66 N/A
N/A = not applicable
Table 3
Example calibration for protein test

1 0 23 13 18 N/A
2 50 14 56 67 N/A
3 100 77 167 154 N/A
4 500 567 435 534 N/A
5 1000 # # # N/A
Table 4
Example calibration for glucose test

1 0 8 24 3 N/A
2 30 25 81 78 N/A
3 100 156 63 189 N/A
4 300 367 266 376 N/A
5 1000 1078 922 945 N/A
4 Notes
1. All solutions should be prepared with purified deionized water

and analytical grade reagents, and stored and used at room
temperature unless indicated otherwise. In this study, we pre-
pared an artificial urine solution for practical purposes although
native body fluids can also be used. Waste disposal regulations
should be followed for used solutions and materials.
www.Ebook777.com
2. To test clinical urine samples, instructions from the test pro-

vider should be followed as described in the methods section.
3. Record a minimum number of five calibration points. The
application allows for recording any number of points but it
performs optimally with five calibration points. Additional cali-
bration points will increase the accuracy of the measurements,
which are limited by the sensitivity of the colorimetric test.
4. The mobile application uses a 100 pixel area per test zone and
it should be entirely covered by the test zone color.
5. The calibration photograph can be retaken if outside of the
test area as many times as needed.
6. For repeat measurements, it is not necessary to recalibrate the
application. Recorded calibrations can be selected directly
from the main menu.
7. For multiplexed assays, select “MULTIPLEXING” and align
test zones with multiplex test zones on the capturing screen.
Autodetection can also be used to determine the test zones
and capture the images.
8. Versions of the application allow selection of other types of
analytes for different commercial test strips. Such multiplex
tests may target applications such as drug abuse or assimilation,
urine adulteration, nutritional factors, and hormone levels.
9. It is suggested that multiple measurements should be per-
formed to validate the use of the smartphone application.
10. Each triplet allows for accuracy comparison of the smartphone
application to standard techniques. The presented mobile
medical application can also be applied for quantifying pho-
tonic crystal arrays and holographic sensors [11–21]. The
described application has been designed for laboratory use and
for clinical testing. The next stage will be to seek regulatory
approval for each test and its intended use [22].
References
1. Kroemer S, Frühauf J, Campbell TM, Massone 4. Pamplona VF, Mohan A, Oliveira MM, Raskar
C, Schwantzer G, Soyer HP et al (2011) R (2010) Dual of Shack-Hartmann optometry
Mobile teledermatology for skin tumour using mobile phones. Frontiers in Optics,
screening: diagnostic accuracy of clinical and Optical Society of America, Rochester, NY,
dermoscopic image tele-evaluation using cellu- paper FTuB4.doi:10.1364/FIO.2010. FTuB4
lar phones. Br J Dermatol 164:973–979 5. Zhu H, Mavandadi S, Coskun AF, Yaglidere O,
2. Breslauer DN, Maamari RN, Switz NA, Lam Ozcan A (2011) Optofluidic fluorescent imag-
WA, Fletcher DA (2009) Mobile phonebased ing cytometry on a cell phone. Anal Chem
clinical microscopy for global health applica- 83:6641–6647
tions. PLoS ONE 4:e6320 6. Martinez AW, Phillips ST, Whitesides GM
3. Smith ZJ, Chu K, Espenson AR, Rahimzadeh (2008) Three-dimensional microfluidic devices
M, Gryshuk A, Molinaro M et al (2011) Cell- fabricated in layered paper and tape. Proc Natl
phone-based platform for biomedical device Acad Sci USA 105:19606–19611
development and education applications. PLoS 7. Wang S, Zhao X, Khimji I, Akbas R, Qiu W,
ONE 6:e17150 Edwards D et al (2011) Integration of cellphone
imaging with microchip ELISA to detect ovarian (2014) Light-directed writing of chemically
cancer HE4 biomarker in urine at the point-of- tunable narrow-band holographic sensors. Adv
care. Lab Chip 11:3411–3418 Opt Mater 2:250–254. doi:10.1002/
8. Pollock NR, Rolland JP, Kumar S, Beattie PD, adom.201300375
Jain S, Noubary F et al (2012) A paper-based 16. Yetisen AK, Qasim M, Nosheen S, Wilkinson
multiplexed transaminase test for low-cost, TD, Lowe CR (2014) Pulsed laser writing of
point-of-care liver function testing. Sci Transl holographic nanosensors. J Mater Chem C
Med 4:152ra29 2:3569–3576. doi:10.1039/C3TC32507E
9. Yetisen AK, Akram MS, Lowe CR (2013) 17. Tsangarides CP, Yetisen AK, Vasconcellos FC,
Paper-based microfluidic point-of-care diag- Montelongo Y, Qasim MM, Lowe CR et al
nostic devices. Lab Chip 13:2210–2251 (2014) Computational modelling and charac-
10. Martinez AW, Phillips ST, Butte MJ, Whitesides terisation of nanoparticle-based tuneable pho-
GM (2007) Patterned paper as a platform for tonic crystal sensors. RSC Adv 4:10454–10461
inexpensive, low-volume, portable bioassays. 18. Martinez-Hurtado JL, Lowe CR (2014)
Angew Chem Int Ed Engl 46:1318–1320 Ammonia-sensitive photonic structures fabri-
11. Yetisen AK, Volpatti LR, Humar M, Kwok SJJ, cated in nafion membranes by laser ablation.
Pavlichenko I, Kim KS et al (2016) Photonic ACS Appl Mater Interfaces 6:8903–8908
hydrogel sensors. Biotechnol Adv 34:250–271 19. Martínez-Hurtado JL, Davidson CA, Blyth J,
12. Yetisen AK, Naydenova I, Vasconcellos FC, Lowe CR (2010) Holographic detection of
Blyth J, Lowe CR (2014) Holographic sensors: hydrocarbon gases and other volatile organic
three-dimensional analyte-sensitive nanostruc- compounds. Langmuir 26:15694–15699
tures and their applications. Chem Rev 20. Martinez-Hurtado JL, Lowe CR (2015) An
114:10654–10696 integrated photonic-diffusion model for
13. Yetisen AK, Montelongo Y, Qasim MM, Butt H, holographic sensors in polymeric matrices.
Wilkinson TD, Monteiro MJ, Yun SH (2015) J Membr Sci 495:14–19. doi:10.1016/j.
Photonic nanosensor for colorimetric detection memsci.2015.07.064
of metal ions. Anal Chem 87:5101–5108 21. Martinez-Hurtado JL, Akram MS, Yetisen AK
14. Yetisen AK, Montelongo Y, Vasconcellos FC, (2013) Iridescence in meat caused by surface
Martinez-Hurtado JL, Neupane S, Butt H et al gratings. Foods 2:499–506. doi:10.3390/
(2014) Reusable, robust, and accurate laser- foods2040499
generated photonic nanosensor. Nano Lett 22. Yetisen AK, Martinez-Hurtado JL, Vasconcellos
14:3587–3593 FC, Simsekler MCE, Akram MS, Lowe CR
15. Yetisen AK, Butt H, Vasconcellos FC, (2014) The regulation of mobile medical appli-
Montelongo Y, Davidson CAB, Blyth J et al cations. Lab Chip 14:833–840
Chapter 27
Development of a User-Friendly App for Assisting

Anticoagulation Treatment
Johannes Vegt
Abstract
Blood coagulation time is an important factor to consider for postoperative and cardiac disorder patients
who have been prescribed anticoagulant coagulant medications. This chapter describes a patient self-
management system for assessment of blood coagulation times and determining appropriate anticoagulant
dosages using a test strip device and the Coagu app. This app can also be used as a patient reminder of
treatment times and to monitor treatment and effects over time.
Key words Blood coagulation, Clotting cascade, Anticoagulant test strips, Coagu app, Cardiac disease
1 Introduction
Coagulation is the process in which blood changes from a liquid to a

gel. The mechanism involves a cascade of reactions in which enzyme
precursors are successively and rapidly activated to catalyze the next
reaction, ultimately resulting in cross-linking of fibrin polymers and
clot formation [1]. The extrinsic coagulation cascade ensues in the
following sequence: tissue damage → activation of tissue factor →
activation of factor VII → activation of factor X → cleavage of pro-
thrombin to produce thrombin → cleavage of fibrinogen to produce
fibrin → and polymerization of fibrin. Anticoagulants are often used
in medicine to help prevent blood clots from forming and to reduce
the risk of heart attack, stroke, and blockages in arteries and veins [2].
When anticoagulants are prescribed for a patient, it is essential to
carry out periodic monitoring of the time it takes the blood to clot
using the international normalized ratio (INR) to adjust the dose as
necessary [3]. Many patients are advised to carry out the relevant
checks themselves using coagulation test strips combined with a suit-
able reader [4]. Based on these results, the appropriate dosage of
anticoagulant should be administered to achieve the required target
range. This management process would be facilitated by the use of
Coagu app, developed by Appamedix UG in Berlin, Germany [5].
303
304 Johannes Vegt
In 2013, the Coagu app was recognized for its usability by the
International Design Centre Berlin and as an example of a particu-
larly user-friendly product on the international Funkausstellung
(IFA) [6]. The app draws on a universal design, which means that it
can be used by all ages alike. Currently, it is now used by patients in
over 70 countries. The evidence shows that greater use of self-mon-
itoring offers clinical and patient benefits and is likely to result in
reductions in heart attacks and strokes caused by blood clots [7].
2 Materials
1. Smartphone or tablet (see Note 1).

2. Coagu app.
3. Test strips and meter (see Note 2).
3 Methods
3.1 Preparation: 1. Insert the test strip into the coagulation meter (Fig. 1).
Determination of INR 2. Lightly pierce the tip of a finger.
3. Immediately apply the resulting blood drop to the test strip.
4. After 60 s, the measured value appears on the meter display.
Fig. 1 (A) The figures of the last INR (International Normalized Ratio) measurement gradually fade away over
seven days. This reminds the patient to take the next measurement. (B, C) Lightly pierce the top of a finger. (D)
Immediately apply resulting blood drop to the test strip that is in the measurement device (see Note 2). (E) After
60 s, the measured value appears on the meter display. (F) The patient enters the INR Value, using a number
picker and stores it. (G) The measured and saved Value appears with the actual date on the start page of the
Coagu app
Anticoagulation App 305
3.2 App Usage: 1. Open the Coagu app on the smartphone or tablet.
Determination 2. First use: each patient may configure the app for their specific
of Anticoagulant needs including target INR range, drug varieties, and dosage
Dosage (Fig. 1) (see Note 4).
(See Note 3) 3. Input the measured value for storage and display in the calen-
dar and histogram of the app (Figs. 2 and 3) (see Note 5).
4. The patient determines which medication and what dosage he
should take based on the current INR reading (Fig. 4).
5. Measurement history: the entered values are stored and can be
visualized as a trend over a period of up to 6 months in a his-
togram (Fig. 5) (see Note 6).
6. Notifications: the patient is reminded up to twice a day via a
notification which occurs as an audible and visual signal advis-
ing them to set the time for taking their medication or some
other actions (see Note 7).
7. Patient comments: the patient may add a comment on the
entry each day individually (see Note 8).
Fig. 2 Anticoagulation patients have to take their medication every day. The app
reminds them of the dose to be taken that day. The patient confirms having taken
the tablets with a tap on the display. This is then registered in the histogram and
in the calendar
Fig. 3 Calendar showing the input daily INR values. Red values indicate a high reading
Fig. 4 The patient chooses a prescribed medication from a list manually. A confirmation that the medication
has been taken is registered automatically in the calendar. With a tap on the calendar, a daily summary
appears (not shown). Here, the entered notes can be read or new notes can be entered
Anticoagulation App 307
Fig. 5 The histogram shows the measured INR values and medication use and
relates these over time. Tendencies can be seen over a period of 6 months
4 Notes
1. iOS (Apple) and Android (GooglePlay) smartphones and tab-

lets can be used. It should be noted that there is no connection
between the developer of the Coagu App Johannes Vegt
(Appamedix UG i.Gr.) and the producer of the CoaguChek
measurement device (Hoffmann-La Roche AG).
2. A lancing device comes with the CoaguChek instrument,
although other sources can be used.
3. Data protection is paramount to the functioning of the app.
For the Coagu app, the data on the smartphone or tablet is
being managed by the patient alone and it is not stored on any
external server or as cloud data. The patient makes the decision
about data sharing. It should be noted that in its present form
the Coagu app is not a medical product. Since the app is sold
worldwide, this would require testing and approval by the reg-
ulatory authorities in the relevant countries. The escalation of
the app to medicinal product would be possible if a health
insurance company accepts liability.
308 Johannes Vegt
4. The target range is the maximum and minimum amount INR

values that patients receive from their doctor.
5. The average age of the audience is about 60 years. Accordingly,
relatively large touch-sensitive surfaces were created to com-
pensate for possible motor inaccuracies. In addition, larger
font sizes are used at crucial points to compensate for potential
visual impairments.
6. Here the user can see the correlation between drug dosage and
reading.
7. For example, if the medication is to be interrupted, this can be
listed in the app.
8. The comment is stored in the calendar and is available for 6
months. Through the app, the patient can also contact their
physician by phone, SMS, or email. Technically, it would be
possible for the patient to share their information with their
doctor so that their history could be updated, if required. This
is only possible in cooperation with the health insurance and
higher authorities.
References
1. Luchtman-Jones L, Broze GJ Jr (1995) The 4. van den Besselaar AM, Meeuwisse-Braun J,

current status of coagulation. Ann Med Schaefer-van Mansfeld H, van Rijn C, Witteveen
27:47–52 E (2000) A comparison between capillary and
2. Chakrabarti R, Das SK (2007) Advances in venous blood international normalized ratio
antithrombotic agents. Cardiovasc Hematol determinations in a portable prothrombin time
Agents Med Chem 5:175–185 device. Blood Coagul Fibrinolysis 11:559–562
3. Jespersen J, Hansen MS, Dyerberg J, Ingerslev 5. http://www.coagu.com/en/
J, Jensen G, Jørgensen M et al (1991) 6. http:// www.inr-austria.at/index.php?article_id=6
Standardized prothrombin time determinations 7. Tideman PA, Tirimacco R, St John A, Roberts
and optimal anticoagulant therapy. Ugeskr GW (2015) How to manage warfarin therapy.
Laeger 153:355–360 Aust Prescr 38:44–48
Part IV
Future Directions
Chapter 28
Multiplex Biomarker Approaches to Enable Point-of-Care

Testing and Personalized Medicine
Paul C. Guest
Abstract
This chapter describes how current and future innovations driven by application of multiplex biomarker
techniques can help in earlier and more efficacious treatment of patients, suffering from the world’s most
devastating and costly diseases. The application of new miniaturized biosensors and transducers will enable
point-of-care testing by facilitating analysis of a single drop of a blood within the time span of a visit to the
doctor’s office. It is anticipated that the scoring algorithms used with future tests will incorporate both
biochemical and clinical data, resulting in specific profiles for each patient or tested subject to enable
personalized medicine approaches.
Key words Disease, Multiplex biomarkers, Lab-on-a-chip, Smartphone apps, Personalized medicine
Diseases such as cardiovascular disorders, type 2 diabetes, obesity,

cancer, and mental health disorders can affect people of both sexes
at different ages and seriously impair medical health, quality of life,
social well-being, and productivity, with a significant negative
impact on society and the economy. According to the World Health
Organization (WHO), the global burden of noncommunicable
diseases is expected reach approximately exceed 30 trillion US dol-
lars within the next 20 years and account for almost half of the
global GDP [1]. Given the urgent medical need and the impor-
tance of counteracting these negative effects, it is now vital that
point-of-care testing gains wider acceptance on the marketplace.
The existing methods are simply not working and may also be
equally draining on the economy due to the large-scale instrumen-
tation required along with long processing times and the delay in
receiving, communicating, and acting upon diagnostic results.
One means of overcoming these issues which has been emerging
over the last decade is through clinically validated lab-on-a-chip
(LOC) approaches. LOCs provide many advantages over existing
methods, such as the need for lower biosample volumes, less waste,
lower fabrication and reagent costs, improved process control due
to faster system response times, and compactness due to integration
311
312 Paul C. Guest
and high parallelization of functionality [2, 3]. Most importantly,

this translates to a reduced waiting time for the results and tests can
even be performed right there in the doctor’s office.
One emerging LOC approach involves a new way to diagnose
and manage HIV infections. Around 40 million people are infected
with HIV in the world today, yet scarcely more than one million of
these receive the correct anti-retroviral treatment. In fact, most of
the people with HIV have never even been tested for the disease.
Currently, measuring the circulating levels of CD4 positive lym-
phocytes in a person’s blood is the best way to determine if they
have HIV and this can also be used for tracking the infection. This
is typically achieved using a technique called flow cytometry that is
not available in most developing countries where the presence of
HIV is disproportionately high since the instrumentation is large,
expensive, and complicated and requires trained technicians in the
operation and interpretation of data. Recently, a company called
ClonDiag developed a LOC device that employs similar static
image analysis and counting of CD4 positive cells as in flow cytom-
etry but in a compact and transportable package that does not
require extensive laboratory training [4]. Furthermore, it requires
only 25 μL of blood and can deliver results within only 20 min.
Another LOC device that was developed more recently consists of
a printed flexible plastic microchip that can capture and quantify
viruses such as HIV in several types of biosamples, including blood
through an electrical sensing approach [5]. Another group has
developed tuberculosis diagnostic LOC device that is 96 % accurate
for detection of tuberculosis [6]. The device uses an immunofluo-
rescence-based microtip sensor that can detect tuberculosis com-
plex cells in sputum within 30 min. Concentration mechanisms
based on flow circulation and electric field are combined at differ-
ent scales to concentrate target bacteria in 1 mL samples onto the
surfaces of microscale tips. Multiplex antibody-based biomarker
tests have also been developed on LOC devices that are the size of
a credit card [7]. One specific application is the detection of pros-
tate cancer using either surface-enhanced RAMAN scattering [8]
or voltage-based [9] readouts. In general, the procedure involves
the application of a blood drop into a chamber in the card, fol-
lowed by the insertion of the card into a small table top analyzer.
The diagnosis is then read out as a “score” in less than 15 min.
One of the anticipated major benefits of all of these LOC tests is
that the rapid diagnosis will help to cut down on waiting times for
results of laboratory tests which can often take several days or even
weeks using the customary methods. Furthermore, these devices
can be manufactured to contain a universal serial bus (USB)
that would enable connection to a computer and transmission
of the data to other devices such as smartphones via near-field
communication.
Multiplex Biomarkers for Personalized Medicine 313
Point-of-care testing now encompasses a variety of devices

ranging from larger table-top instruments to implanted, wearable,
and handheld equipment. The handheld systems typically consist
of disposable strips incorporated into a cassette that allows addition
of the sample, conduction of the test, and signal generation that is
usually interpreted visually or through an inexpensive reader. There
are now such devices that can be used to indicate the presence of a
heart attack, diabetes, blood clotting capacity, some infectious dis-
eases, or even to measure the levels of substances such as alcohol
and drugs of abuse. The most familiar example is most likely the
Clearblue Digital Pregnancy test® produced by SPD Swiss Precision
Diagnostics GmbH [10] which gives both a readout and an indica-
tion of the number of weeks since conception.
Growth in the LOC diagnostic and monitoring systems reflects
patient preferences for being seen in a doctor’s office or in a clinic
rather than a central laboratory. This not only requires solutions
that use lower sample volumes with reduced effort and time spent
carrying out the tests, and lower costs, it also calls for an increase
in connectivity solutions. Furthermore, massive companies such as
Apple and Google have shown an escalating interest in the diag-
nostic market. This has been mainly driven by opportunities to
connect an app result with a diagnostic answer via smart software.
Likewise, pharmaceutical companies now consider that low cost
and time-effective biomarkers are essential for decision-making in
clinical trials and drug discovery efforts. Therefore, an important
requirement of new point-of-care devices involves incorporation of
mobile communication and internet capabilities so that the associ-
ated data can be formatted for ease of presentation and interpre-
tation by the users. There has also been an increased move to
incorporate networked computing to aid in such functions as dis-
ease prediction, diagnosis, prognosis and even for monitoring
medication compliance. This may lead to the development of bio-
profiles or biomarker fingerprints for individuals that combines
genomic, proteomic, transcriptomic, metabolomic, and/or imag-
ing data with patient physiometrics and histories.
In the twenty-first century, mobile phones have become virtu-
ally ubiquitous, with an estimated six billion subscriptions at the
end of 2011 [11]. There has now been a convergence in the entire
technological concept which has resulted in development of the
smartphone. This combines general voice and text messaging ser-
vices with computation functions to support applications (apps),
sensors, as well as wireless internet accessibility and connectivity
with other smart devices. These latter features make the smart-
phone an attractive and user-friendly platform for health and dis-
ease management [12, 13]. Basic mobile phone-based interventions
have already shown promise and have led to improved outcomes in
a variety of health conditions, diseases, and control of certain
habits. A review of clinical trials that incorporated health care
314 Paul C. Guest
interventions assisted through smartphone apps indicated

improvements in 61 % of the outcomes such as better attendance of
patients at appointments, faster diagnosis and treatment, improved
communication [14]. In addition, the benefits included behavioral
improvements such as cessation of smoking and better medication
compliance, as well as clinical factors such as better blood sugar
control, decreased symptoms of asthma, and lower behavioral
stress levels.
Along with the theme of this book, there have now been
developments of multiplex biomarker tests on hand-held and
smartphone-based devices, which use a 3D-printed opto-
mechanical interface to illuminate each reaction well on a multiwell
plate [15]. The resulting images can be transmitted to databases
for analysis and the results can be returned to the user within
1 min. This mobile platform has already been tested with a success-
ful outcome in a clinical setting using multiplex assays for mumps,
measles, and herpes simplex I and II virus immunoglobulins using
the smartphone camera optics function for accumulation and
transmission of the readouts. Other optical-based smartphone
detection assays include the measurement of nucleic acids by PCR
[16], avian flu virus subtyping by immunoassay [17], detection of
kaposi’s sarcoma by solar thermal PCR [18], hemoglobin and HIV
levels by immunoassay [19], prostate-specific antigen using enzy-
matic amplification, and a chromogenic substrate [20]. It is easy to
imagine that other LOC and smartphone-based system will be
developed for other important diseases such as various types of
cancer, cardiovascular conditions, diabetes, and mental disorders.
In conclusion, the movement has been toward improving patient
care through the initial study of various diseases using multiplex bio-
marker approaches and then deploying these technologies as hand-
held microfluidic devices linked with smartphone apps for ease of use
in the clinical setting. This is currently our best approach of achieving
a paradigm shift personalized medicine. This will allow persons to be
treated based on their individual biomarker profiles rather than as one
of many with a particular disease using a standard blockbuster drug.
In addition, the use of multiplex biomarker tests on LOC and/or
handheld devices that are capable of distinguishing disease subtypes
may be useful for rapid identification of patients who are most likely
to respond to specific medications either alone or in combination
with other drugs. This general approach has been used increasing in
the field of cancer treatment. For example, the measurement of
human epidermal growth factor receptor 2 (HER2) at both the gene
and transcript level can be used to identify those breast cancer patients
who are most likely to benefit from treatment with Trastuzumab
(also known as Herceptin), which was developed by Genentech in
the 1990s [21]. It obvious that such an approach could result in
more effective treatment of patients with fewer side effects and, thus,
a decreased proportion of patients who decide to discontinue medi-
cation because of severe side effects.
Multiplex Biomarkers for Personalized Medicine 315
References
1. http://www3.weforum.org/docs/WEF_ 12. Klasnja P, Pratt W (2012) Healthcare in the
Harvard_HE_GlobalEconomicBurdenNon pocket: mapping the space of mobile-phone
CommunicableDiseases_2011.pdf health interventions. J Biomed Inform 45:
2. Pawell RS, Inglis DW, Barber TJ, Taylor RA 184–198
(2013) Manufacturing and wetting low-cost 13. Ventola CL (2014) Mobile devices and apps
microfluidic cell separation devices. Biomicro- for health care professionals: uses and benefits.
fluidics 7:056501. doi:10.1063/1.4821315 P T 39:356–364
3. Yager P, Edwards T, Fu E, Helton K, Nelson K, 14. Krishna S, Boren SA, Balas EA (2009) Health-
Tam MR et al (2006) Microfluidic diagnostic care via cell phones: a systematic review.
technologies for global public health. Nature Telemed J E Health 15:231–240
442:412–418 15. Berg B, Cortazar B, Tseng D, Ozkan H, Feng
4. Ermantraut E, Bickel R, Schulz T, Ullrich T S, Wei Q et al (2015) Cellphone-based hand-
Tuchscheerer J (2011) Device and method for held micro-plate reader for point-of-care test-
the detection of particles. USPTO Patent ing of enzyme-linked immunosorbent assays.
US8040494. Company: Clondiag GmbH ACS Nano 9:7857–7866
5. Shafiee H, Kanakasabapathy MK, Juillard F, 16. Liao SC, Peng J, Mauk MG, Awasthi S, Song J,
Keser M, Sadasivam M, Yuksekkaya M (2015) Friedman H (2016) Smart cup: a minimally-
Printed flexible plastic microchip for viral load instrumented, smartphone-based ooint-of-care
measurement through quantitative detection molecular diagnostic device. Sens Actuators B
of viruses in plasma and saliva. Sci Rep 5:9919. Chem 229:232–238
doi:10.1038/srep09919 17. Yeo SJ, Choi K, Cuc BT, Hong NN,
6. Kim JH, Yeo WH, Shu Z, Soelberg SD, Inoue Bao DT, Ngoc NM et al (2016) Smartphone-
S, Kalyanasundaram D (2012) Immunosensor based fluorescent diagnostic system for highly
towards low-cost, rapid diagnosis of tuberculo- pathogenic H5N1 viruses. Theranostics 6:
sis. Lab Chip 12:1437–1440 231–242
7. Schumacher S, Nestler J, Otto T, Wegener M, 18. Snodgrass R, Gardner A, Jiang L, Fu C,
Ehrentreich-Förster E, Michel D et al (2012) Cesarman E, Erickson D (2016) KS-Detect –
Highly-integrated lab-on-chip system for validation of solar thermal PCR for the diagno-
point-of-care multiparameter analysis. Lab sis of Kaposi’s sarcoma using pseudo-biopsy
Chip 12:464–473 samples. PLoS One 11:e0147636
8. Gao R, Cheng Z, deMello AJ, Choo J (2016) 19. Guo T, Patnaik R, Kuhlmann K, Rai AJ, Sia SK
Wash-free magnetic immunoassay of the PSA (2015) Smartphone dongle for simultaneous
cancer marker using SERS and droplet micro- measurement of hemoglobin concentration
fluidics. Lab Chip 16:1022–1029 and detection of HIV antibodies. Lab Chip
9. Parra-Cabrera C, Samitier J, Homs-Corbera A 15:3514–3520
(2016) Multiple biomarkers biosensor with 20. Barbosa AI, Gehlot P, Sidapra K, Edwards AD,
just-in-time functionalization: application to Reis NM (2015) Portable smartphone quanti-
prostate cancer detection. Biosens Bioelectron tation of prostate specific antigen (PSA) in a
77:1192–1200 fluoropolymer microfluidic device. Biosens
10. Johnson S, Cushion M, Bond S, Godbert S, Bioelectron 70:5–14
Pike J (2015) Comparison of analytical sensitiv- 21. Demonty G, Bernard-Marty C, Puglisi F,
ity and women’s interpretation of home preg- Mancini I, Piccart M (1997) Progress and new
nancy tests. Clin Chem Lab Med 53:391–402 standards of care in the management of HER-2
11. http://mobithinking.com/mobile-marketing- positive breast cancer. Eur J Cancer 43:
tools/latest-mobile-stats/a#subscribers 497–509
INDEX
A Drug discovery
clinical trial ............................................................. 5, 313
Algorithm efficacy ......................................................................8, 14
clinical score ........................... 90, 93, 103, 104, 108, 109, FDA ...............................................................................4
111, 112, 116, 118 imaging .......................................................................313
machine learning .................................. 90, 104, 111, 119 marketing........................................................................3
patient data ............................................. 51, 93, 117, 118 side effect ......................................................................27
statistics ........................................................ 50, 104, 108 stratification ..................................................................69
surrogate biomarker ........................................................8
B
Biomarker F
genomics ......................................................... 5, 275, 313 Flow cytometry..........................................76, 80, 88, 91, 312
metabolomics ...................................5, 46, 47, 49–51, 313
proteomics ...................................5, 7, 28, 41–49, 57, 313 L
transcriptomics .......................................................5, 313
Lab-on-a-chip
Biosample
immunoassay .............................................................. 283
blood ................................................................... 161–167
micrifluidics ........................................................ 284, 290
brain ................................................................... 235, 241
microarray ........................................................... 284, 290
cells ...............................................................................86
cerebrospinal fluid.........................................................67 M
heart............................................................................313
plasma .....................................................................13, 67 Metabolomics
serum ......................................................................13, 67 mass spectrometry .................................................... 8, 11
spittle ............................................................................13 NMR ..........................................................................276
tissue ................................................................. 6, 69, 209
P
C Personalized medicine ........................................ 15, 311–314
Cytomics ............................................................................ 12 Proteomics .................................... 7, 41, 48, 57–71, 149, 150,
152–158, 205–220, 267–273
D 2D gel electrophoresis
2D-DIGE ........................................ 41, 48, 205–212
Disease
mass spectrometry
anxiety .......................................................................... 27
ITRAQ .........................................................267–273
bipolar disorder ......................................... 21, 22, 70, 195
label-free ...........................................................57–71
cancer...........................11, 15, 67–69, 75, 76, 93, 94, 104,
MS/MS ............................................................57–71
116, 117, 128, 143, 150, 215, 247, 311, 312, 314
shotgun .................. 7, 41, 58, 60–61, 69, 70, 267–273
diabetes ........................ 14, 26, 37–52, 295, 311, 313, 314
SILAC .......................................... 149, 150, 152–158
heart disease.............................................. 45, 46, 50, 143
SRM .............................................................213–220
major depression ................................................... 27, 245
multiplex immunoassay.......................................169–175
schizophrenia ..................... 19–31, 70, 106, 117, 195, 196
DNA/RNA S
microarray ................................. 5, 13, 106, 135, 143, 283,
284, 287, 290 Smartphone
RT-PCR .....................................................................131 app ................................................................ 30, 303–308
SNP analysis ....................................................... 143, 284 tablet ...................................................................304–307
DOI 10.1007/978-1-4939-6730-8, © Springer Science+Business Media LLC 2017
317
www.Ebook777.com

Biomarkers PDF

Uploaded by

Copyright:

Available Formats

Biomarkers PDF

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

Biomarkers PDF

Uploaded by

Copyright:

Available Formats

Free ebooks ==> www.Ebook777.

Paul C. Guest Editor

METHODS IN MOLECULAR BIOLOGY

For further volumes:

Methods and Applications

ISSN 1064-3745 ISSN 1940-6029 (electronic)

Library of Congress Control Number: 2016958277

© Springer Science+Business Media LLC 2017

Printed on acid-free paper

This Humana Press imprint is published by Springer Nature

Campinas, Brazil Paul C. Guest

PART II STATISTICAL CONSIDERATIONS

PART III PROTOCOLS

11 Pulsed SILAC as a Approach for miRNA Targets Identification

25 Lab-on-a-Chip Multiplex Assays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283

PART IV FUTURE DIRECTIONS

JULIANA DE ALMEIDA-FARIA • University of Cambridge Metabolic Research Laboratories

CHRISTOPH W. TURCK • Department of Translational Research in Psychiatry, Max Planck

Application of Multiplex Biomarker Approaches

Key words Pharmaceutical company, Drug discovery, Biomarker, Genomics, Transcriptomics,

Pharmaceutical companies are under pressure to improve their

companies in this process, they have encouraged the incorporation

2 A Brief History of Failed Drugs

The need for biomarker tests to guide drug development is per-

3 Biomarker Impact in the Drug Discovery Process

Estimates are that the total number of potential biomarkers is

4 Multiplex Biomarker Techniques

Biomarkers are physical characteristics that can be measured in bio-

changes in physiological states are dynamic in nature, these are

4.2 Two-Dimensional Two-dimensional gel electrophoresis (2DE) works as follows: (1)

5 Use of Multiplex Biomarker Profiling in the Drug Discovery Process

Biomarker profiling can be used at multiple stages of drug discov-

Target Lead Drug toxicity Clinical studies Approval and

Biomarker identification Development of Clinical validation and development of

Pre-clinical studies Clinical studies

Metabolomics Safety Metabolomics Safety

physiological levels [23]. However, many of these networks have

can accelerate development of the best lead compounds by facili-

aspartate aminotransferase were not detected until day three, pro-

5.4 Clinical Studies Clinical applications of multiplex biomarker approaches include

6 Conclusions and Future Perspectives

This chapter describe the emerging use of multiplex biomarker

29. Meneses-Lorente G, Watt A, Salim K, approach to understanding elevated serum ala-

The Application of Multiplex Biomarker Techniques

Key words Schizophrenia, Blood-based biomarkers, Proteomics, Multiplex immunoassay, Mass

Schizophrenia is a debilitating, mental health disorder which can

Select drug Treat with right drug

the symptoms of schizophrenia without pointing to the underlying

secondary services such as hospitals, community groups, and crisis

2 Current Diagnosis of Schizophrenia

Most psychiatrists and clinicians agree that schizophrenia is a gen-

for making a diagnosis. In many cases, diagnosis may be made

possible phase, as described above This is important as numerous

Another criterion of biomarker tests that is often overlooked is

3 Biomarkers Identified for Schizophrenia

Although genomic studies are able to identify genes conferring

3.1 Inflammation A multiplex immunoassay profiling study which used cytokine

Cortisol Thymus Transport

Pancreas Transport Interleukins