Expert Review of Proteomics

ISSN: 1478-9450 (Print) 1744-8387 (Online) Journal homepage: https://www.tandfonline.com/loi/ieru20

Detecting post-translational modification

signatures as potential biomarkers in clinical mass
spectrometry

Ruzanna Mnatsakanyan, Gerta Shema, Mark Basik, Gerald Batist, Christoph

H. Borchers, Albert Sickmann & René P. Zahedi

To cite this article: Ruzanna Mnatsakanyan, Gerta Shema, Mark Basik, Gerald Batist, Christoph
H. Borchers, Albert Sickmann & René P. Zahedi (2018) Detecting post-translational modification
signatures as potential biomarkers in clinical mass spectrometry, Expert Review of Proteomics,
15:6, 515-535, DOI: 10.1080/14789450.2018.1483340

To link to this article: https://doi.org/10.1080/14789450.2018.1483340

© 2018 The Author(s). Published by Informa View supplementary material

UK Limited, trading as Taylor & Francis
Group.

Published online: 03 Jul 2018. Submit your article to this journal

Article views: 9299 View related articles

View Crossmark data Citing articles: 39 View citing articles

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=ieru20
EXPERT REVIEW OF PROTEOMICS
2018, VOL. 15, NO. 6, 515–535
https://doi.org/10.1080/14789450.2018.1483340

REVIEW

Detecting post-translational modification signatures as potential biomarkers in

clinical mass spectrometry
Ruzanna Mnatsakanyana, Gerta Shemaa, Mark Basikb, Gerald Batistb, Christoph H. Borchers b,c,d,e
,
Albert Sickmann a,f,g and René P. Zahedi a,b,e
a
Protein Dynamics, Leibniz-Institut für Analytische Wissenschaften – ISAS – e.V, Dortmund, 44227, Germany; bGerald Bronfman Department of
Oncology, Jewish General Hospital, McGill University, Montreal, Quebec H4A 3T2, Canada; cUniversity of Victoria-Genome British Columbia
Proteomics Centre, University of Victoria, Victoria, British Columbia V8Z 7X8, Canada; dDepartment of Biochemistry and Microbiology, University of
Victoria, Victoria, British Columbia, V8P 5C2, Canada; eSegal Cancer Proteomics Centre, Lady Davis Institute, Jewish General Hospital, McGill
University, Montreal, Quebec H3T 1E2, Canada; fMedizinische Fakultät, Medizinische Proteom-Center (MPC), Ruhr-Universität Bochum, 44801
Bochum, Germany; gDepartment of Chemistry, College of Physical Sciences, University of Aberdeen, Aberdeen AB24 3FX, Scotland, United Kingdom

ABSTRACT ARTICLE HISTORY

Introduction: Numerous diseases are caused by changes in post-translational modifications (PTMs). Received 23 October 2017
Therefore, the number of clinical proteomics studies that include the analysis of PTMs is increasing. Accepted 29 May 2018
Combining complementary information—for example changes in protein abundance, PTM levels, with KEYWORDS
the genome and transcriptome (proteogenomics)—holds great promise for discovering important Clinical proteomics;
drivers and markers of disease, as variations in copy number, expression levels, or mutations without biomarkers;
spatial/functional/isoform information is often insufficient or even misleading. post-translational
Areas covered: We discuss general considerations, requirements, pitfalls, and future perspectives in modification (PTM); LC-MS;
applying PTM-centric proteomics to clinical samples. This includes samples obtained from a human phosphorylation;
subject, for instance (i) bodily fluids such as plasma, urine, or cerebrospinal fluid, (ii) primary cells such glycosylation; proteolysis
as reproductive cells, blood cells, and (iii) tissue samples/biopsies.
Expert commentary: PTM-centric discovery proteomics can substantially contribute to the under-
standing of disease mechanisms by identifying signatures with potential diagnostic or even therapeutic
relevance but may require coordinated efforts of interdisciplinary and eventually multi-national con-
sortia, such as initiated in the cancer moonshot program. Additionally, robust and standardized mass
spectrometry (MS) assays—particularly targeted MS, MALDI imaging, and immuno-MALDI—may be
transferred to the clinic to improve patient stratification for precision medicine, and guide therapies.

1. PTMs and their role in health and disease archaea [1]) but also in their prevalence. Every human protein has
the potential to be post-translationally modified at least once
Cellular homeostasis is primarily maintained through adaptive
during its life span. The fundamental role of PTMs as switches of
changes of the proteome and metabolome. On the proteome
cellular homeostasis therefore implies that they represent critical
level, this mainly involves changes in (i) protein abundance levels,
and sensitive nodes for pathogenesis, both as drivers and markers,
for example through altered expression, degradation, or both, (ii)
and thus aberrant PTM patterns have been associated with numer-
protein localization, (iii) protein–protein interactions, (iv) protein
ous diseases.
function, and (v) protein activity—all of which can be modulated
Mass spectrometry (MS) is an unrivaled tool for detecting
by post-translational modifications (PTMs). PTM involves the
and quantifying PTMs and their changes in a broad variety of
attachment, removal, exchange, and rearrangement of functional
samples, including purified proteins and protein complexes,
groups to amino acid side chains and protein N-termini, but also
organelles, cells, tissues, or biofluids. Consequently, more than
the hydrolysis of peptide bonds (proteolysis). Indeed, PTMs are a
95% of current data on PTMs is derived from MS-based proteo-
most elegant, energetically efficient, and rapid strategy to modu-
mic studies [2] and—with the ongoing advances in instrumen-
late and maximize the functionality of a single gene product—
tation, methods, and bioinformatics—the number of known
depending on the type of PTM, this even can occur in a highly
PTMs is likely to increase in the future [3,4]. The largest and
reversible manner, such as in case of phosphorylation or redox
manually curated PTM database available, PhosphoSitePlus
modifications. Hence, throughout all species, PTMs are important
(www.phosphosite.org) contains more than 400,000 non-
for regulating fundamental biochemical processes. This is reflected
redundant PTMs on more than 20,000 proteins, mainly from
in both the high number of known PTMs (>400 reported in the
human and mouse [2] (Figure 1).
Uniprot database, 326 in eukaryotes, 250 in bacteria, and 80 in

CONTACT René P. Zahedi rene.zahedi@ladydavis.ca Lady Davis Institute - Jewish General Hospital, JGH Proteomics Centre, E-615, 3755 Côte Ste-Catherine
Road, Montreal H3T 1E2Quebec
*These authors contributed to this paper equally
supplementary material can be accessed here
© 2018 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group.
This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License (http://creativecommons.org/licenses/by-nc-nd/4.0/),
which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited, and is not altered, transformed, or built upon in any way.
516 R. MNATSAKANYAN ET AL.

Figure 1. Distribution of the >400,000 PTM sites reported at www.phosphosite.org, October 2017. Due to historical and technical reasons five PTMs—Ser/Thr/Tyr
phosphorylation, ubiquitination, and acetylation—account for 92% of these modifications, but this may not necessarily mirror the biological relevance and
abundance of individual PTMs.

Figure 2. Frequently reported post-translational protein modifications (adapted from Pagel et al., Expert Rev Proteomics; 12(3):235-53; with permission).

Importantly, a protein sequence can have many different that particular protein. Even for a specific residue, there can be
modifiable amino acids (Figure 2), but not necessarily all will a competition between different modifications, as shown, for
be modified at the same time point or within the same copy of example, for Lys acetylation and ubiquitination [5], for different

lipid modifications [6], and also for O-phosphorylation and
O-GlcNac [7,8]. For each modifiable amino acid, there is an
equilibrium between the free and different modified versions
within a biological system, which can be shifted to precisely
drive cellular response in a specific direction. Particularly for
p53, it has been demonstrated that function and activity
depend on the complete PTM pattern rather individual PTMs,
as they can synergistically modulate protein structure (PTM
crosstalk and PTM code) [9-11]. Analyzing complete protein
PTM patterns is a major challenge for MS, as currently PTM
research is mainly restricted to the identification of peptides
after proteolytic digestion (‘bottom-up’ proteomics) [12]. Thus,
information about the full modification state of individual pro-
tein molecules is typically lost and would require the analysis of
intact proteins (‘top-down’ proteomics) which, unfortunately,
might not be applicable to the analysis of PTMs from clinical
samples in the near future, as discussed elsewhere [12,13]. In
this review, we define ‘clinical samples’ as samples obtained
from a human subject. This includes (i) bodily fluids such as
serum, plasma, urine, nasal fluid, cerebrospinal fluid, etc., and
stool; (ii) primary cells, including reproductive cells, blood cells
such as erythrocytes, leukocytes, platelets, etc.; (iii) tissue sam-
ples/biopsies; whereas studies including (primary) cell culture,
patient derived xenografts or cell lines, and animal models are
not considered as ‘clinical’ in this review.
Numerous diseases are driven by the dysregulation of the
proteome, either due to changes in expression levels, changes
in the primary sequence, and/or changes in PTMs. Non-synon-
ymous single nucleotide polymorphisms (SNPs) can alter the
activity or specificity of enzymes that attach/remove a certain
PTM to proteins and thereby can change for example phos-
phorylation, glycosylation, or acetylation patterns in a path-
way-specific or even proteome-wide manner. SNPs can also
affect the range of substrates of these enzymes by altering
consensus/targeting/recognition motifs on common sub-
strates or by incorporating such motifs into novel (non-phy-
siological) substrates. Both cases, mutations of enzymes and
substrates, can have severe consequences for cellular home-
ostasis [14,15]. For many types of cancer, a dysregulation of
kinases has been reported, which can, for example, lead to
constitutive kinase activity and hyperactive signaling path-
ways [16-18]. SNPs can also induce mistargeting of proteins
such as has been reported for the renal Fanconi syndrome,
where the mutation E3K in EHHADH produces a de novo
mitochondrial import sequence [19,20]. Upon mitochondrial
import the mitochondrial targeting sequence is proteolytically
removed and the truncated and mislocalized protein impairs
energy production [19,20]. Differential antibody glycosylation
has been associated with autoimmunity [21], and murine
cytomegalovirus has been reported to activate intracellular
PTM-based signaling pathways through expression of the
viral G protein-coupled receptor M33 [22]. Owing to these
fundamental roles of PTMs in health and disease, their analysis
is important from two different aspects: First, to understand
disease mechanisms on the molecular level. Second, in the
context of precision medicine: To identify PTM signatures that
can be correlated with disease (progression), and that may
help to diagnose and stratify patients and to predict the
effectiveness of a treatment and guide therapy.
EXPERT REVIEW OF PROTEOMICS 517
In this review, we discuss general considerations, require-
ments, pitfalls, and future perspectives in applying PTM-centric
proteomics to clinical samples—from experimental design to
data analysis. We focus on four classes of PTMs: phosphorylation,
glycosylation, proteolytic cleavage and oxidative Cysteine mod-
ifications and provide examples of relevant studies and methods.
2. Can PTMs be used as biomarkers?
Studying disease mechanisms has been mainly done with the
help of models, such as cell culture and animal models.
However, more and more, omics studies involve the analysis
of clinical samples. For precision medicine, PTM signatures may
either be identified in disease models and then validated in
clinical samples or directly identified using clinical samples.
Although, for example, specific changes in phosphorylation
and glycosylation are increasingly reported as potential markers
of disease, and many PTM assays have been developed so far
(for instance reviewed in [23-25]), only few are really routinely
used for diagnostic purposes to date, due to a variety of
reasons [26]. We are still in the early days of PTM-specific
research and detection and many potential PTM biomarkers
and the corresponding assays still need to be validated exten-
sively. Consequently, most clinical assays are genomics-based
or involve the use of antibodies, which are mainly available for
proteins rather than for specific PTMs. For instance, the expert
curated Cancer Biomarkers database (https://www.cancergen
omeinterpreter.org/biomarkers) reports around 200 different
genes as potential cancer biomarkers, ~50 of those as
approved. Still, whether pre-clinical, from case reports and
early trials, or approved, virtually all of them are either point
mutations, gene amplifications, or expression changes, whereas
no specific PTM markers are listed. Nevertheless, some PTM-
specific assays are currently used. Glycated hemoglobin (HbA1c)
is formed non-enzymatically upon contact with plasma glucose
and is increased in diabetes [27]. Whereas diabetes can be
diagnosed by measuring fasting glucose, HbA1c has several
advantages such as higher repeatability [28]. HbA1c can also
be measured in the non-fasting state, and elevated levels may
be a risk factor for macrovascular disease [28]. The phosphor-
ylation levels of vasodilator stimulated phosphoprotein (VASP)
are measured as indicator of P2Y12 receptor inhibition during
antiplatelet treatment [29]. Different glycan structures on the
surface of red blood cells define the blood type. Besides that,
there is huge list of potential biomarkers reported in the litera-
ture, though their diagnostic or predictive value remains to be
proven in many cases (see supplemental table 1).
Considering the complexity of human health and disease,
in many cases a single molecule (e.g. a single gene, protein,
glycan or phosphorylation site) may be insufficient as specific
biomarker for diagnosis and for predicting effective treatment.
Recent studies indicate, that combining multiple ‘omics’ levels
such as (i) the analysis of protein expression and PTMs, (ii)
proteomics, metabolomics and lipidomics, or (iii) proteoge-
nomics integrating genomics, transcriptomics and proteomics
data is beneficial for identifying specific disease patterns and
signatures. Particularly in cancer, it has emerged that informa-
tion from next-generation sequencing or RNAseq alone is not
sufficient to understand the mechanisms underlying the
518 R. MNATSAKANYAN ET AL.
development of resistance in certain tumors [30], such that
additional data on protein expression and PTM levels is
needed. Thus, mutations in the receptor tyrosine kinase epi-
dermal growth factor receptor (EGFR) are apparent in many
cancer types and lead to aberrant phosphorylation signaling.
EGFR mutations are often used to select treatment with spe-
cific EGFR inhibitors for non-small cell lung cancer. Response
rates, however, are still too low (40-80%) while concurrently
10-15% of patients without mutation do respond [31].
Obtaining data on protein phosphorylation levels for down-
stream targets of the EGFR signaling pathway might comple-
ment the non-conclusive genomics data. Indeed, in a recent
study by the Clinical Proteomic Tumor Analysis Consortium
(CPTAC), only integration and clustering of DNA, RNA, protein,
and protein phosphorylation profiles allowed distinguishing
subtypes in 77 breast cancer tumors [32].
3. General considerations for sampling, sample
preparation and data analysis
When conducting PTM (and also global) proteome analyses on
clinical samples, several important aspects have to be consid-
ered. While bodily fluids (such as blood, urine, or tear fluid) are
usually available in high amounts and (almost) non-invasively,
tissue samples are often limited in amount and/or are often
formalin-fixed and paraffin-embedded (FFPE). For biopsies,
most of the sample is often needed for histological purposes,
or is stored in dedicated biobanks. Consequently, the amounts
that remain for proteomic analyses are often strictly limited
(low to medium µg-range of protein) and downstream com-
patibility with sensitive proteomics workflows can be poor.
Another key challenge when analyzing PTMs from clinical
samples is heterogeneity, which may have a multitude of
preanalytical sources that are usually not relevant or less
relevant for cell culture experiments and are only partially
relevant for animal models [33]. First, even for healthy control
groups, biological variation between individuals can be sub-
stantial. This is due not only to generally occurring genetic
variations, but, more importantly can be due to considerable
differences in life style, diet, environment, and age, as well as
between genders. While it may already be challenging to
clearly state what defines a ‘healthy’ control, the variation
across patient cohorts can be substantially higher, and often
patients present with multiple diseases. Particularly for cancer,
during the past decade we have learned that even for specific
subtypes of cancer such as breast cancer, tumors are indivi-
dual and may require specific, personalized, treatment [34].
Therefore, a thorough and clear subtyping of patient cohorts
is as important as maintaining a sufficiently high number of
individuals for both patient and matching (age, gender, etc.)
control groups [35].
Another major source of variation lies in the sampling proce-
dure [36-38]. Clinical samples have to be collected and stored over
time. While here the need for well-defined standard operating
procedures (SOPs) that define how, when, and under exactly
which conditions samples should be taken, is well-accepted,
their implementation into daily clinical practice is often not
straightforward. This could be due not only to different personnel
being involved in the sampling and the inevitable daily stress in
clinical environments, but also due to underestimating how small
variations in SOPs, such as changing incubation times by only a
few minutes or buffers by a few µL, different lengths of time
between sampling and freezing, etc., can severely affect the pro-
teome and—particularly—PTM signaling pathways. Therefore, a
close collaboration between clinicians and analytical scientists is
mandatory to jointly (i) design realistic and robust SOPs [39-44]
and (ii) (ideally) to monitor their correct implementation, particu-
larly in the beginning of a project but also throughout its progress.
Because minor deviations from the agreed-upon sampling SOPs
cannot be completely excluded, it is also important to track and
document these thoroughly [45] so that later inconsistent results
may be correlated with such changes. An important issue, in the
case of tissue, is how the samples are provided. Fresh-frozen tissue
might be preferable, but often clinical specimen are provided as
FFPE samples. Although sample preparation protocols for FFPE
tissues have been published [46-48], this type of sample prepara-
tion considerably complicates the entire analysis, and may reduce
both sensitivity and reproducibility. It should be kept in mind that
fixation processes can potentially alter labile PTM signatures, such
as phosphorylation, if not proteome-wide at least for certain path-
ways that quickly respond to cell stress.
Finally, variation can be introduced during the actual proteo-
mic sample preparation and LC-MS analysis. Therefore, here also it
is mandatory to strictly follow well-designed SOPs detailing all
individual steps, from lysis to sample analysis [49-51], including
quality control of proteolytic digestion prior to LC-MS analysis [52],
as well as controls for labeling efficiency in case chemical stable
isotope labeling (SIL) is used [53,54]. Lysis and homogenization
should be harsh and fast to preserve the biological state of the
sample and avoid artificial changes, particularly of fast and rever-
sible PTMs. Tissues should be kept on ice and ideally, inhibitors
should be used where applicable (e.g. phosphatase inhibitor cock-
tails for phosphoproteomics), including the use of protease inhi-
bitors (particularly when analyzing proteolytic cleavage). Protein
digestion should be controlled prior to extensive analytical work-
flows, as unreproducible and inefficient digestion can substantially
affect and ruin sample analysis, including strong variation
between samples and artificial fold-changes. Digestion can be
controlled in multiple ways that differ in precision, effort, and
sensitivity [52]. Gel electrophoresis of samples pre- and post-
digestion does not require special equipment but is rather slow,
cannot be automated and poor resolution in the low MW range
and comparably poor sensitivity do not allow to reliably evaluate
the reproducibility of individual digests. MALDI MS spectra can be
used for a quick and very sensitive overall pattern comparison
using only small aliquots of the digested samples, but due to the
huge sample complexity may lack the sensitivity to detect differ-
ences apart from major components that are easy to digest.
Monolithic HPLC columns require a dedicated HPLC system, ide-
ally with nL/µL UV detector, but are fast, sensitive, automated, and
robust. Complete UV traces allow a reliable assessment of digest
efficiency and reproducibility [52]. Finally, digested (but not pre-
digest) samples can be analyzed by short LC-MS gradients, pre-
ferably using not the newest instrument generations in the labora-
tory, to assess the overall reproducibility. However, samples with
poor digest efficiency may ruin the column or contaminate the
HPLC. LC-MS instruments have to be well-maintained, including
regular cleaning, calibration, and (ideally) the analysis of a daily

quality-control analysis of well-designed samples [55]. For this
purpose, specifically designed peptide mixtures [44,50,56] and
software tools [57-60] allow monitoring the reproducibility of
retention times, intensities, sensitivity, and mass deviation with
and without background.
For clinical samples, data analysis is substantially more
challenging than for common proteome analyses, as it
requires the analysis of larger sample cohorts as well as
more sophisticated data analysis and statistical approaches.
Analyzing a large number of samples, however, leads to an
additional challenge: missing values [61-63], in both label-free
and label-based quantitation approaches. Thus, even for tech-
nical replicates analyzed consecutively under highly reprodu-
cible conditions, the overlap of confidently identified and even
more so quantified peptides and proteins is usually far below
100% in discovery experiments. Generally, when analyzing
large cohorts, one should (i) only place samples on the auto-
sampler immediately prior to injection, (ii) run daily quality
control samples, (iii) analyze samples in randomized order, to
avoid systematic biases derived from changing instrument
performance, and (iv) use SIL peptides for precise quantitation
for all steps following discovery.
4. From discovery to validation: sample analysis by
quantitative MS
For clinical samples, three different kinds of analysis plat-
forms have to be considered. First, discovery would rely on
non-targeted methods that provide high sensitivity and cov-
erage (in-depth analysis), but which do not necessarily pro-
vide high throughput. Here the goal is to identify potential
biomarker panels or targets from a manageable number of
representative samples. This is mainly achieved through rela-
tive quantitation, where fold-changes between samples are
detected, however without knowing analyte concentrations
or amounts. Second, there is validation with larger cohorts
focusing on a limited number of candidate proteins/pep-
tides, usually applying targeted MS techniques [64,65] such
as multiple reaction monitoring (MRM) [66,67] or parallel
reaction monitoring (PRM) [68], and usually employing abso-
lute quantitation where the concentration or amount of an
analyte in a sample is precisely determined using a specific
internal standard (discussed more in detail in the following
chapter). This not only allows obtaining more reliable fold-
changes, but in case of PTMs enables determining site occu-
pancy, for example which share of EGFR molecules in a cell
is phosphorylated at a specific position. This information
may be more biologically relevant then mere fold-changes
usually obtained in MS-based proteomics. Third, there is
translation to the clinic. Here, sensitivity is still an issue,
but robustness, throughput, and precise quantitation of a
limited number of candidates are even more important, so
that absolute quantification is essential.
As aforementioned, there is usually an equilibrium—in the
case of reversible PTMs often very dynamic—between mod-
ified and non-modified states of amino acid residues and
proteins, which can be shifted in order to induce certain
biochemical effects. For instance, for the best-studied
EXPERT REVIEW OF PROTEOMICS 519
signaling PTM, phosphorylation, there might be only minor
qualitative changes (i.e. presence of phosphorylation sites)
between control and stimulated or diseased conditions,
whereas quantitative differences (i.e. degree of phosphoryla-
tion) can be vast. It is therefore unlikely that a specific phos-
phorylation site can be solely detected under a specific
disease state and therefore just used as a biomarker. It is,
however, more likely that this phosphorylation is significantly
more or less prominent under disease conditions, for instance
representative of an increased or reduced activity of the cor-
responding protein, while protein expression may be
unchanged. Importantly, in some cases a phosphorylation
might be only detectable due to a disease-relevant SNP: For
instance when single amino acid substitutions lead to the
presence/exposure of either novel Ser/Thr/Tyr residues or
novel kinase consensus motifs inducing a non-standard phos-
phorylation site. In both cases, the novel phosphorylated pep-
tide might only be detectable using protein databases that
consider the respective mutation. Therefore, as long as a
specific PTM is not present only due to SNPs, a mere qualita-
tive analysis of clinical samples is not sufficient, and well-
defined and carefully-conducted quantitative experiments
are mandatory to subtract relevant signaling from back-
ground. This is particularly relevant as clinical tissue samples
represent cell populations that may differ in their respective
PTM (e.g. phosphorylation) status, such that disease markers
may be suppressed in the final analysis. For instance, hetero-
geneous tumors may contain certain cells with a significantly
regulated PTM, while surrounding cells may mask the extent
of this regulation. The often transient and sub-stoichiometric
nature of PTMs renders an enrichment step prior to analysis
mandatory to make them accessible for MS analysis, with the
most efficient methods typically applied on the peptide level.
This substantially improves the identification and quantitation
of modified peptides, however, at the expense of more elabo-
rate workflows and—consequently—further increases in tech-
nical variation (as well as additional shortcomings of bottom-
up PTM analysis discussed elsewhere [13]). Some examples of
enrichment strategies that have been successfully applied to
clinical samples will be given in the following chapters. There,
we will focus on four classes of PTMs of high abundance and
relevance (phosphorylation, glycosylation, proteolytic clea-
vage, and redox modifications) that can be enriched without
antibody IP, as such methods often require relatively large
amounts of sample and are usually not applied to clinical
samples.
5. Methods for quantifying PTMs in clinical samples
MS-based discovery is still mostly done using data-dependent
acquisition (DDA) methods, where the MS continuously applies
analysis cycles in which usually the top 5-20 most intense peptides
are subjected to fragmentation, slight variations in (i) retention
times, (ii) signal intensities, as well as (iii) co-isolation of peptides,
and (iv) poor MS/MS spectrum quality, can cause so-called ‘under-
sampling’—in other words, although a peptide is clearly present in
a sample, it will not be identified and therefore not quantified. This
affects both label-free and label-based approaches, although the
latter usually include numerous multiplexed experiments in order
520 R. MNATSAKANYAN ET AL.

to analyze sufficiently high numbers of samples (e.g. ten 10-plex
experiments). In label-free quantitation (discovery), this has been
compensated for by various software tools that use so-called
matching or alignment of LC-MS runs [69-71]. Here, different
LC-MS runs can be aligned based on unique characteristics of
‘features’ (i.e. ions represented by e.g. m/z, retention time window,
elution profile, isotope pattern). Thus, it is sufficient that a specific
peptide (feature) is only reliably identified in a single LC-MS run to
trace it in all other analyses, as long as these are sufficiently
reproducible. Thus, quantitation becomes almost independent
of identification, allowing a considerably improved coverage
across different samples. Still, inconsistent behavior during
LC-MS analysis limits the alignment capabilities of the different
software tools. Even when identified and clearly present in multi-
ple runs, the same peptide may still be represented by different
features, so that the quantitative information is incomplete with-
out manual inspection (Figure 3). In large datasets, missing values
can be predicted by imputation [62], which facilitates statistical
data analysis but which is certainly not ideal.
For modified peptides, quantitation is even more challen-
ging. First, single-peptide quantitation in discovery approaches
is less robust than protein quantitation based on multiple pep-
tide measurements. Second, the issue of missing values is more
pronounced because, in addition to alignment problems, for
PTMs the position of the modified amino acid has to be unam-
biguously identified within the peptide sequence [72] (Figure 3).
This can be complicated, as multiple modifiable amino acids
can be in close proximity within a sequence (for example, the
unambiguous localization of the phosphorylation site in the
sequence SSSLPAYGR in human dermatin is challenging) and,
in addition, MS/MS spectra quality can be severely impaired
due to the modification(s) [73,74]. Consequently, label-free
PTM-centric discovery studies suffer even more from missing
values and inconsistent quantitation (Figure 3). Recent studies
demonstrated that data-independent acquisition (DIA) strate-
gies allow a more consistent, reproducible, and precise quanti-
tation of proteomes and particularly of larger sample sets and
cohorts [75-77]. Here, in contrast to DDA, individual ions are not
targeted for isolation and fragmentation, but instead a mass
range of interest (e.g. 300-1500 m/z) is divided in multiple
smaller m/z windows that are consecutively selected for frag-
mentation, leading to hybrid MS/MS spectra often including
several peptides. Individual fragment ion spectra are then
later assembled based on elution profiles. In DIA peptides are
continuously fragmented throughout their elution profile, and
quantitation can be conducted using fragment ions rather than
precursor ions, reducing interference and allowing a more
robust quantitation than in DDA. A potential shortcoming of
DIA methods is that, for optimal performance, they require the
generation of comprehensive spectral libraries, which may not
be optimal, particularly for discovering novel candidates under
disease conditions. Thus, spectral libraries would be ideally
required for representative samples of all cohorts included in
the study.
Generally, MS methods must be thoroughly validated and
the transfer to a clinical environment for daily use as routine
procedure has to be realistic and feasible. This may include a
shift from nano-LC typically used in discovery proteomics to
larger columns and flow rates that are considerably more
robust at the expense of slight sensitivity losses [78]. There is
always a trade-off between proteomic coverage and analysis
time on one hand, and precision and costs on the other.
Sample fractionation prior to LC-MS will substantially improve
coverage, not only for global proteomics, but also with regard
to PTM-centric workflows where fractionation can boost the
identification of modified peptides. However, fractionation in
label-free approaches is challenging as—even under ideal and
reproducible conditions—it will add another level of technical
variation which is extremely challenging to handle with data
analysis software tools. An alternative strategy, one that is
particularly relevant for PTM-centric studies due to more com-
plex and susceptible workflows, is the use of chemical SIL
strategies allowing multiplexing of samples. The most popular
commercial labels (iTRAQ [79,80] and TMT [81,82]) allow

Figure 3. Issues with alignment of LC-MS runs in label-free discovery experiments, exemplified by theoretical extracted ion chromatograms leading to inconsistent
quantitation of peptides. (A) Alignment can be generally impaired by inconsistent retention times due to non-linear shifts of peptide and background peaks. A
substantial retention time shift of 20 s in one of the LC-MS runs might lead to misassignment of the peptide to another feature (feature 2). B) For PTM peptides
inconsistent assignment of PTM modification sites within an identified peptide can also impair alignment; preferentially if various modifiable amino acid residues are
in close proximity and if MS/MS spectra quality is insufficient (‘s’ indicates the phosphorylation site assigned by the database search). C) Indeed, both effects can
superimpose, leading to multiple features that may represent a single peptide. In targeted approaches this can be compensated for by the use of stable isotope
labeled (SIL) reference peptides.
 EXPERT REVIEW OF PROTEOMICS 521

multiplexing of eight and ten samples, respectively. Though this
multiplexing capacity is still way too low for virtually any clinical
setting, several such experiments could be combined in order
to analyze sufficient numbers of patient samples. Ideally, this
would imply the use of a mixture of samples (all or a subset)
that is applied as one channel to every single multiplexing
experiment for normalization purposes. Because, after labeling,
the same peptide derived from different samples is always
virtually isobaric, even multiplexing 10 samples does not con-
siderably increase sample complexity. Consequently, the differ-
ently labeled versions of a single peptide will be isolated and
fragmented just like a single precursor in DDA experiments,
releasing reporter ions that reflect the relative amounts of the
peptide in the different samples in the low m/z region of the
MS/MS spectrum, which furthermore contains the usual peptide
backbone information (Figure 4). Once samples are multiplexed,
they can undergo various enrichment and fractionation strate-
gies without largely compromising precision. Thus, iTRAQ and
TMT are well-suited for highly sensitive PTM-centric analysis, as
multiplexing allows quantitative PTM analysis even from as little
as 10 µg of protein start material per sample or condition [83]
and has been applied to quantify e.g. phosphorylation [84],
glycosylation [85], proteolysis [86], Lys acetylation [87], cysteine
modifications [88], or combinations thereof. However, TMT and
iTRAQ suffer from ratio compression (Figure 4), an effect caused
by the potential co-isolation of background peptides and con-
sistent, low-level background fragmentation [89], as virtually
every peptide in the sample releases the same reporter ions
(labeling efficiency >97% can be readily achieved) [90,91].
Notably, this effect renders reporter ion-based methods incom-
patible with DIA methods as it cannot be deduced how much
of a specific reporter ion signal in a hybrid spectrum derives
from a specific peptide (Figure 4). However, selective PTM
enrichment and the usage of dedicated DDA analysis strategies,
particularly involving narrow m/z isolation windows with mod-
ern instruments of 0.4 to 0.8 m/z can largely compensate for
ratio compression. Importantly, iTRAQ and TMT can alter the
physicochemical properties of peptides and thus affect the
efficiency of PTM enrichment and MS detection, but also chro-
matographic behavior. Consequently, the validation of iTRAQ/
TMT derived target peptides for example by targeted methods
such as parallel reaction monitoring (PRM) [92] with unlabeled
samples may not always be successful, even for sequences
which should be well detectable. Recent improvements

Figure 4. Problems with reporter ion-based quantitation, demonstrated for iTRAQ 8plex. A) In DDA experiments, all differentially labeled versions of an iTRAQ/TMT
labeled peptide will contribute to (basically) a single isotope pattern in MS survey scans. In this case, an isolation window of 1.6 m/z (dashed lines) is sufficient to
prevent co-isolation. Upon fragmentation, the reporter ions reflecting the relative amounts of the peptide in the eight samples will be released. B) iTRAQ 8plex
reporter ion area of an MS/MS spectrum, exemplified by 4 diseased samples (red) and four healthy controls (blue). Note that the higher m/z range of the MS/MS (not
shown here) contains the fragment ions representing the peptide backbone. From, the reporter ions it can be deduced that the peptide is significantly
downregulated in the diseased samples. C) If another peptide is in close m/z proximity, an isolation window of 1.6 m/z (grey) may be too wide, leading to co-
isolation and ratio compression (representative MS/MS spectrum shown in D), whereas a narrower isolation window of 0.8 m/z (black) may prevent this;
representative MS/MS spectrum shown in B). D) Ratio compression due to co-isolation distorts quantitation: The peptide no longer appears to be significantly
downregulated. E) In DIA experiments, wider windows are selected for continuous fragmentation of the entire mass range, as represented here by 4.0 m/z. F) In DIA,
the elution profiles of the fragment ions are assembled from individual MS/MS scans, and matching elution profiles point to fragment ions from the same precursor
(as demonstrated by the black traces). In contrast, reporter ions will be constantly present in all MS/MS spectra with varying intensities, rendering it impossible to
infer reporter ion intensities for individual peptides.
522 R. MNATSAKANYAN ET AL.
exploiting mass defects and ultra-high resolution MS, such as
the NeuCode approach developed by the Coon lab (for a
detailed explanation of the approach we refer to [93,94])
show great promise that one of the drawbacks of current
reporter ion methods, that is the limited multiplexing capacity,
will be overcome in the future, allowing for more comprehen-
sive studies. Importantly, this would massively reduce the pro-
blem of missing values in large scale studies, because a
quantitative value will be obtained for every peptide as long
as the peptide is really present in the sample above the detec-
tion limit.
While these approaches allow the discovery-based quantifi-
cation of thousands of PTMs across multiple samples, validation
of interesting targets will be usually conducted with MS-inde-
pendent technologies, for instance immunoassays such as
ELISA. These are straightforward to implement and provide
high throughput, but suffer from low to medium multiplexing
capacities and are relatively expensive [95], although substantial
improvements have recently been achieved for antibody-based
assays. For instance Treindl et al. have developed a complex
workflow that allows antibody-based quantitation of hundreds
of proteins from a single sample [96]. Proteins are first sepa-
rated by gel electrophoresis, blotted on a membrane and bio-
tinylated. Next, the membrane lane is cut into 96 bands and
proteins are eluted in a 96-well plate. Next, Neutravidin-coated
and differentially color-coded Luminex beads are added to each
well, so that different colors represent the different molecular
weight areas of the original gel. The beads can be pooled and
small aliquots can be incubated with hundreds of antibodies
when using 5-20 μg of protein, comparable to a standard
Western blot analysis. A Phycoerythrin-labeled secondary anti-
body is used for signal generation, and the samples are ana-
lyzed on a flow cytometer, generating discrete signals for 96
bead populations which can be used for visualization.
Still, completely relying on antibody specificity can be an
issue, even for validated antibodies, and even more so for PTM-
specific antibodies (Figure 5). Notably, antibodies are not even
available for many candidates of interest that appear in system-
wide quantitative PTM-centric discovery studies. Therefore,
recently, MRM and PRM (discussed in detail elsewhere [97-99]),
have gained increasing attention as powerful alternative
Figure 5. Antibody specificity as an issue. Depicted are a full length anti-Zyxin
antibody and two antibodies against phosphorylated Zyxin (p-Zyxin) from
different vendors showing poor specificity. Importantly, there can be substantial
differences between antibodies from different vendors.
validation approaches. They have the advantage of always com-
bining peptide (or protein) quantitation with identification,
instead of just detecting an increase in the signal as in anti-
body-based procedures. In MRM, the MS is set up to specifically
quantify a pre-defined set of candidates in a very specific man-
ner, allowing higher precision and throughput, as sample analy-
sis can often be performed in one hour or less. However, the non-
targeted bulk of the proteome remains a black box. Ideally,
MRM/PRM assays would involve the usage of SIL reference pep-
tides. Here, the peptides-of-interest are synthesized with incor-
porated stable isotopes and their sequences, such as 13C, 15N or
18
O. The label is usually introduced in the C-terminal amino acid
and therefore the y-ion series shows a shifted m/z. A SIL refer-
ence and the corresponding endogenous peptide share the
same sequence and physicochemical properties, but differ by
mass—ideally by at least 4 Da. Thus, they have the same dose/
response rates in MS. Consequently, a well-defined amount (e.g.
50 fmol) of SIL reference peptide can be spiked into a sample,
allowing precise absolute quantitation of endogenous peptide
(and consequently protein) amounts/concentrations by compar-
ing their signals (i.e. intensity or area under the curve) over
several orders of magnitude, down to amol amounts on column.
The usage of SIL peptides further increases the confidence in
peptide identification, as the SIL and endogenous peptides co-
elute and have identical fragmentation patterns, after the inten-
tionally-introduced mass shift has been taken into account. The
key to precise protein quantitation using SIL reference peptides
is the selection of appropriate peptides, that are proteotypic and
not prone to degradation or modifications artificially induced
during sample preparation.
Once set up and optimized, MRM and PRM enable a highly
sensitive quantitation of 100s of peptides per LC-MS run in
‘scheduled mode’, which relies on reproducible chromatogra-
phy with stable retention times. For targeted MS, peptides-of-
interest are typically evaluated for their suitability, namely
detectability, stability, and uniqueness [100]. Well-generated,
quality controlled, and validated targeted assays [64] enable
the analysis of larger sample cohorts in accordance with the
demands of clinical research, including good inter-laboratory
reproducibility and precision [101]. To this end, more than 1500
characterized targeted assays are available through the CPTAC
Assay Portal [102] (https://proteomics.cancer.gov/assay-portal),
including >200 for phosphopeptides.
If the sensitivity of MRM/PRM is not sufficient to quantify
the peptide of interest from whole cell or tissue lysate, as
often the case for modified peptides, these methods can be
combined with fast enrichment procedures to reduce the
huge background of non-modified peptides [103,104]. If 90%
of the non-modified background can be removed, a larger
proportion of the PTM-proteome of interest can be loaded
on-column to improve sensitivity. If this is still not sufficient,
combining anti-peptide antibody immunoprecipitation (IP)
with MS, such as ‘Stable Isotope Standards and Capture by
Anti-Peptide Antibodies’ (SISCAPA) [105-107] and immuno
MALDI (iMALDI) [108-111], or TXP antibodies [112] that recog-
nize peptides with a common epitope of 3-4 amino acids at
the N- or C-terminus, prior to (LC-)MS analysis can consider-
ably improve the accessibility of target peptides. The unique
capabilities of MS can even compensate for low antibody

specificity, while the use of SIL reference peptides allows
absolute quantitation from bodily fluids, cells, and tissues
[42]. In a recent project which shows great promise for clinical
settings, Popp et al. used iMALDI (a technique that combines
anti-peptide antibody immuno-enrichment with on target elu-
tion and MALDI MS) for the accurate quantitation of AKT1 and
AKT2 out of cell lysate, flash frozen and FFPE tissues [111].
Translation to clinical settings requires robust methods that
are precise, cheap and provide high throughput, among those
immunoassays (given the previously mentioned limitations),
targeted MS with robust LC conditions, as well as SISCAPA
[105-107] and iMALDI [108-111].
6. Analyzing protein glycosylation and
phosphorylation in clinical samples
Protein glycosylation and phosphorylation are the two most abun-
dant PTMs. While phosphorylation is generally reversible and
allows rapid signaling, glycosylation represents a wide range of
heterogeneous sugar structures of different complexity, which can
make up more than 90% of a protein’s molecular weight and may
be reversible or irreversible. Phosphorylation and glycosylation
play key roles in many clinically relevant processes such as immu-
nity [113,114], inflammation [115,116], or metastasis [117,118].
Aberrant modification patterns can be linked to multiple diseases,
including Alzheimer’s disease [20,119-121], diabetes [122-125],
cancer [126-129], and cardiovascular diseases [130-132]. Both
PTMs have been extensively studied by MS. However, while
numerous strategies for the direct analysis of phosphopeptides
from complex samples after specific enrichment have been devel-
oped, because of the enormous complexity of glycan structures
and glycopeptide fragmentation patterns, in-depth analysis stra-
tegies for protein glycosylation usually employ the release of
glycans, so that either previously-glycosylated peptides (site map-
ping) or the glycan structures by themselves (without the peptide
backbone) can be analyzed in detail. To date, most studies on
PTMs in clinical samples have focused on either protein phosphor-
ylation or glycosylation, mostly with the goal of discovering bio-
markers (panels of proteins or affected pathways rather than
single proteins) and stratifying patients to improve risk assess-
ment, precision medicine, and therapy guidance (reviewed
by [133,134]).
For glycan profiling, different approaches have been applied to
the study of clinical samples, e.g. focusing on either on glycomes
by LC-MS after glycan release and capture, or glycan spatial dis-
tributions by MALDI imaging. MS-based profiling of glycans in
bodily fluids had already started in the 2000s [135-140], as those
samples can be easily obtained in rather large amounts. In 2012,
Alley et al. studied N-glycans in blood sera obtained from ovarian
cancer patients and controls [141]. They were able to observe
differences that allowed distinguishing patient samples from con-
trols. Sonneveld et al. used nano-LC-MS to study aberrant glyco-
sylation in autoimmune hemolytic anemia (AIHA) using a total of
103 plasma samples [142]. In AIHA, red blood cells (RBC) are
destroyed by IgG anti-RBC autoantibodies, which can lead to
hemolysis. They were able to demonstrate aberrant galactosyla-
tion patterns in IgG1-anti-RBC autoantibodies. More recently, gly-
coproteomics has also been increasingly applied to the study of
tissue samples. Thus, Jin et al. studied mucin o-glycosylation from
EXPERT REVIEW OF PROTEOMICS 523
10 individuals with and without gastric disease [143]. Mucins were
isolated from frozen gastric specimen using isopycnic density
centrifugation and pooling of mucin-containing fractions.
O-glycans were released and analyzed by LC-MS. The authors
found a large diversity in gastric O-glycosylation patterns,
and concluded that the high individual variation may render it
difficult to identify cancer-specific patterns, though they did
observe higher sialylation and sulfation patterns on O-glycans
identified from cancerous tissue. Hinneburg et al. reported a
novel approach to study N- and O-glycomes simultaneously
from FFPE tissue [144]. Cells were isolated from FFPE tissue
through laser capture microdissection, proteins were extracted,
blotted on PVDF membranes, and the N- and O-glycans were
sequentially released, followed by nano-LC-MS using porous gra-
phitized carbon columns. Using this method, the authors were
able to characterize the N- and O-glycomes of tissue specimens,
and even to differentiate and relatively quantify structural isomers.
Recently, Everest-Dass et al. used MALDI imaging to study
N-glycans in ovarian cancer FFPE tissues [141]. N-linked glycans
were first released and analyzed by porous graphitized carbon
LC-MS and the identified glycan profiles were then used to deter-
mine the spatial distribution of released N-glycans on FFPE ovar-
ian-cancer sections by MALDI imaging. Interestingly, the authors
were able to detect differences between the tumor region which
had high levels of high mannose glycans, and the intervening
stroma which had high levels of complex/hybrid N-glycans.
Compared to glycosylation, phosphorylation-based signal-
ing can be very rapid, dynamic, and easily altered during
sampling and sample processing. Thus, strategies applicable
to clinical samples were developed only after the develop-
ment of robust and highly sensitive strategies that allowed
the system-wide precise quantitation of phosphorylation
changes (extensively reviewed elsewhere [145-148]), and
therefore phosphoproteomics of clinical samples has recently
gained more attention. The first quantitative phosphoproteo-
mics studies of patient samples included the temporal profil-
ing of phospho-signaling in human platelets obtained from
fresh blood donations, after specific treatment with inhibiting
and activating compounds. This analysis revealed proteins in
platelets that might be relevant for hemostasis and thrombo-
sis [104]. Here, chemical stable-isotope labeling, multiplexing,
and TiO2/HILIC-based phosphopeptide enrichment/fractiona-
tion [149] were combined with nano-LC-MS to quantify thou-
sands of phosphorylation sites from approximately 100 µg of
protein per sample. Schweppe et al. used strong cation
exchange (SCX) chromatographic fractionation followed by
TiO2 phosphopeptide enrichment to quantify non-small lung
cancer (NSLC) tissue from human individuals using a super-
SILAC approach [150]. For this study, NSLC cancer cell lines
were grown in SILAC medium with heavy arginine and lysine
and were spiked in as reference standard, allowing the quan-
titation of two individual tumor samples against the super-
SILAC mix and thus against each other [151]. Dazert et al.
analyzed primary tumor and control non-tumor biopsies
from a sorafenib-treated hepatocellular carcinoma (HCC)
patient, 7 weeks before treatment and 7.5 weeks after a non-
effective treatment [152]. They used a super-SILAC standard
consisting of five different cell lines for normalization,
SCX-based fractionation, and TiO2-based enrichment followed
524 R. MNATSAKANYAN ET AL.
by nano-LC-MS. A small portion of each SCX fraction was used
for global proteome analysis. The authors were able to
observe reduced phosphorylation of the Raf-Erk-Rsk pathway,
indicating a successful inhibition of the target kinase Raf, but
downstream targets did not show reduced phosphorylation. A
pathway analysis combining both proteome and phosphopro-
teome data revealed the potential role of two processes in
developing resistance: epithelial-to-mesenchymal transition,
and cellular adhesion molecule-dependent drug resistance.
The validation of phosphoproteomic changes obtained from
discovery experiments in sample cohorts is still mainly done using
antibodies. However, because of the aforementioned issues of
antibody availability and specificity (which is particularly true for
phosphospecific antibodies, see (Figure 5)), a combination of phos-
phopeptide enrichment and targeted MS is increasingly used
instead of antibody-based approaches. Targeted MS is flexible,
can be multiplexed, and allows the highly specific and sensitive
quantitation of phosphopeptides by single-shot LC-MS analyses,
e.g. after TiO2 [103,104] or IMAC-based enrichment [153,154]. In
contrast to Western blot or ELISA, MS-based methods allow the
unambiguous connection between qualitative (peptide sequence
and phosphorylation site) and quantitative information with high
precision, and, if SIL peptides are used, even absolute quantitation,
including the assessment of phosphorylation levels (stoichiome-
try). Thus, in order to quantify a specific phosphorylation-site of
interest, PHD Ser125, in fresh frozen colorectal cancer biopsies,
Di Conza et al. applied HPLC-based electrostatic repulsion-hydro-
philic interaction liquid chromatography (ERLIC) [155,156] for com-
bined phosphopeptide enrichment and fractionation [157]. In
ERLIC, the majority of the unphosphorylated peptides are found
in the flow-through and the early fractions [158], while phospho-
peptides are efficiently retained [159] and separated according to
the number of phosphoamino acids [160]. Using a SIL reference for
the phosphopeptide AKPPADPAAAAsPCR, the authors first deter-
mined a retention-time window, which then allowed them to
specifically collect the enriched phosphopeptide from only 32 µg
of total protein from the tissue samples. Absolute quantitation of
the phosphopeptide, and the non-phosphorylated counterpart in
complete digest, using PRM, allowed them to determine the phos-
phorylation stoichiometry in cancerous and healthy tissue samples
from ten individuals. Their results showed significantly reduced
phosphorylation levels in the cancer tissue, in line with the hypoth-
esis based on from cell culture experiments [157].
CPTAC has conducted several relevant phosphoproteomics
studies in cancer tissues. Mertins et al. addressed the afore-
mentioned issue that tissue collection cannot always be stan-
dardized, and consequently ischemic time can vary between
tissue samples collected for the same study [161]. This, how-
ever, can alter phosphorylation levels of biopsy samples as the
authors demonstrated. They collected human ovarian tumor
and breast cancer xenograft tissue from different patients
without vascular interruption (0 min) and then allowed
defined ischemic time intervals of 5, 30, and 60 min. Samples
were labeled with 4-plex iTRAQ, fractionated, and enriched for
phosphorylated peptides using IMAC. A total of ~3 mg protein
was used per experiment, 10% of which was used for global
proteome quantitation. Interestingly, the global proteome and
most of the more than 25,000 phosphorylation sites did not
show significant changes after 60 min. However,
approximately one quarter of the phosphoproteome showed
rapid changes, many in critical cancer pathways related to
stress response, transcriptional regulation, and cell death.
This underscores the need for well-standardized and docu-
mented sampling procedures for clinical proteomics. This
was further confirmed by Gajadhar et al. who analyzed high-
grade serous ovarian carcinoma tissue and colon adenocarci-
noma biopsies from five patients each [162]. After vessel liga-
tion, specimens were surgically removed and the first biopsy
was taken immediately and snap frozen (0 min). Further biop-
sies were taken after 10, 30, and 60 min of cold ischemia.
Samples were labeled with the iTRAQ-4plex reagents, and
phosphotyrosine (pTyr) peptides were enriched using a pTyr
antibody mixture, followed by IMAC enrichment and LC-MS.
Ten percent of each sample was further fractionated and
analyzed by LC-MS for global proteome quantitation. The
authors observed a rapid, unpredictable, and widespread
impact on phosphotyrosine patterns in which up to 50% of
pTyr sites were changed more than 2-fold, in addition to
observing a considerable spatial heterogeneity within the tis-
sue. Notably, both spatial heterogeneity and the dynamic
response to ischemia varied substantially between the differ-
ent patients. Huang et al. applied quantitative (phospho)pro-
teomics to study 24 breast cancer-derived xenografts (PDX)
models [30]. They were able not only to confirm predicted
genomic targets, but also to find protein expression and
phosphorylation changes that could not be explained based
on genomic data, showing the potential of MS-based proteo-
mics for identifying novel drug targets.
7. Analyzing endogenous proteolytic cleavage
(proteolysis) in clinical samples
Proteolysis is the irreversible cleavage of proteins into smaller
polypeptides, and is usually catalyzed by proteases. In addition
to its role in protein degradation and apoptosis, proteolysis also
affects protein localization, function, activity, and protein-
protein interactions. Aberrant proteolytic cleavage is connected
to a variety of diseases, including cardiovascular and neurode-
generative diseases [163,164], inflammation and impaired
wound healing [165], as well as tumor metastasis [166]. This
can be caused by, for example, dysregulation of proteases, or
SNPs that suddenly render a protein a non-physiological sub-
strate of a protease. To better understand the role of proteolysis
in disease, it is important to not only identify the substrates that
are cleaved under pathophysiological conditions, but also to
determine where they are cleaved and how they are differen-
tially regulated between healthy and diseased states. While the
protease responsible for a specific cleavage might be a promis-
ing drug target, its reliable identification is rather challenging
and not straightforward. This may usually require cell culture-
based differential control experiments either with specific pro-
tease inhibitors or the usage of specific protease knock outs.
Analysis may be conducted by MS or by Western blot-based
migration shifts.
As is the case for phosphorylation and glycosylation, it can be
anticipated that specific proteolytic signatures indicate aberrant
cleavage, perhaps even by a specific protease, and—importantly
—may serve as markers for diagnosis and treatment.

Proteolysis leads to the formation of novel (‘neo’) protein
N-termini and MS-based proteomics has become an indispen-
sable tool to identify these neo N-termini. This is preferentially
done through the system-wide specific enrichment of
N-terminal peptides, which include not only the original but
also neo protein N-termini, followed by LC-MS. This approach
has been referred to as N-terminomics. A variety of methods
have been developed for this purpose, among these are com-
bined fractional diagonal chromatography (COFRADIC) [167],
terminal-amine isotopic labeling of substrates (TAILS) [168],
charge-based fractional diagonal chromatography
(ChaFRADIC) [169], and subtiligase N-terminal enrichment
[170]. All of these methods include the specific labeling of
protein N-termini prior to in vitro digestion in order to distin-
guish the specifically labeled N-terminal from internal pep-
tides. These methods allow the identification of protease
substrates through the differential formation of neo
N-terminal peptides between different conditions—usually,
protease knockout vs. wild-type, but increasingly also patients
vs. controls. Notably, the identification of neo N-terminal pep-
tides appearing under specific conditions directly reveals the
respective cleavage sites, and also allows the researcher to
deduce consensus motifs. In addition to these direct strate-
gies, there are also more indirect methods such PROTOMAP,
which combines SDS-PAGE with LC-MS to globally identify
shifts in protein migration that are indicative of proteolytic
processing [171]. This method, however, cannot readily reveal
cleavage sites or motifs. In general protease inhibitors should
be used during sample homogenization/lysis, especially in
case of tissues, to ensure that ongoing endogenous protease
activity does not alter the cleavage patterns.
To date, there are still only few studies that have applied
quantitative N-terminomics to patient-derived samples; in the
large majority of studies, these techniques have been used to
study fundamental and clinically relevant processes in cell cul-
ture or animal models. Lange et al. used TAILS and LC-MS to
study erythrocytes obtained from 6 mL of whole blood per
healthy donor. The extremely high abundance of hemoglobin
in erythrocytes makes their proteomic analysis quite challenging.
Using TAILS and thus specifically focusing on protein N-termini
allowed the researchers to substantially reduce sample complex-
ity in order to improve the detection of low-abundance proteins.
In this way, the authors were able to identify 1369 N-terminal
peptides from 1234 proteins, 281 of which were novel erythro-
cyte proteins while six previously-missing proteins [172,173]
were identified by MS for the first time [174]. The same group
also analyzed human dental pulp as a unique tissue that could
lead to the identification of additional missing proteins, and
found that most of the identified N-termini represented proteo-
lytic cleavage sites [175]. In another study, TAILS was used to
compare B cells from the only known living patient worldwide
with a genetic MALT1 deficiency, which leads to immunodefi-
ciency and immune dysregulation, with heterozygous family
members presenting no immunological disorders. MALT1 is a
paracaspase that plays an important role in regulating lympho-
cyte responses in NF-κB activation. The authors were able to
identify HOIL1, a member of the linear ubiquitin chain assembly
complex, as a novel MALT1 substrate, inhibiting reactivation of
the NF-κB pathway [176].
EXPERT REVIEW OF PROTEOMICS 525
Subtiligase labeling and N-terminal enrichment [170] was
used for enriching protein N-termini in plasma and serum
obtained from healthy donors, in order to elucidate the roles
of proteases in blood. The authors identified 772 unique
N-terminal peptides from 222 proteins, ranging over six orders
of magnitude in abundance [177]. They were able to validate
known proteolytic processes, including peptide cleavage in
coagulation and complement activation, and also found a
large number of cleavages that had not previously been
reported. Recently, using the same method and combining
discovery with subsequent quantitation by MRM, Witta et al.
analyzed plasma from patients after chemotherapy to detect
proteolytic signatures released from apoptotic cells which may
serve as indicators of chemotherapy-induced cell death [178].
The experiment was performed in 1.5-mL cell-free plasma
samples from 5 different patients, which included two acute
myeloid leukemia (AML) patients and one patient with non-
Hodgkin lymphoma (NHL) of diffuse large B-cell lymphoma
subtype. The authors reported 153 proteolytic peptides which
may serve as an initial library of proteolytic biomarkers.
Solari et al. applied quantitative iTRAQ-8plex-based ChaFRADIC
and (phospho)proteomics to platelets isolated from the only Scott
syndrome patient available worldwide. The Scott syndrome is a
rare but likely underdiagnosed bleeding disorder associated with
mutations in ANO6 that lead to an impaired procoagulant
response [179]. From only 20 µg of protein per sample, they
quantified 1596 N-terminal peptides between activated patient
and control platelets, 180 of which were confirmed as calpain-
regulated (corresponding to 106 proteins) and 23 (corresponding
to 23 proteins) were confirmed as caspase-regulated. The authors
detected reduced calpain-dependent cleavage of cytoskeleton-
linked and signaling proteins in the Scott patient, in agreement
with increased phosphorylation states. While in this study, the
high sensitivity was achieved using a rather elaborate HPLC-
based ChaFRADIC workflow, the authors have recently developed
a tip-based improvement of the procedure that allows quantita-
tive N-terminomics with high sensitivity [180].
8. Analyzing oxidative Cys modifications in clinical
samples
A high diversity of protein PTMs are formed as a result of
direct or indirect interaction with reactive oxygen/nitrogen
species (ROS/RNS) [181,182]. Small reactive radical or non-
radical ROS/RNS, such as superoxide (O2·−2) produced by the
NAD(P)H oxidase family, and nitrogen monoxide (NO) pro-
duced by nitric oxide synthases (NOS), are byproducts of
cellular respiration [183] and primarily form redox modifica-
tions on cysteine (Cys), due to the high reactivity of its thiol
side chain [182,184]. Reversible cysteine redox PTMs include
S-nitrosylation (SNO), S-sulfenylation (SOH), disulfide forma-
tion (S–S), S-glutathionylation (SSG), S-cysteinylation (S-Cys),
and S-sulfhydration (SSH), while sulfinylation (SO2H) and
sulfonylation (SO3H) are generally considered to be irrever-
sible [181,184]. Cys modifications are key regulators of cel-
lular redox homeostasis [185]. At low ROS levels, reversible/
protective PTMs are formed that are associated with cell
proliferation, differentiation, and redox signaling, while
increasing ROS levels result in irreversible modifications
526 R. MNATSAKANYAN ET AL.
and damage to proteins [186]. Importantly, specific redox-
reactive Cys residues can be present in their reduced forms
or may compete for different types of redox modifications,
making them into ‘redox-switches’ which sense the redox
microenvironment and allow reversible and—if required—
irreversible fine tuning of biochemical processes [181,185].
ROS-mediated signaling is involved in many key metabolic
processes, and affects a wide variety of upstream and
downstream protein targets, thereby influencing the regula-
tion of other PTMs such as phosphorylation, acetylation, and
ubiquitination [88,187,188].
Although the functional relevance of Cys oxidative PTMs in
human diseases, such as cancer [189], cardiovascular [188,190],
inflammatory [191], and neurodegenerative diseases [191,192]
is clear, the molecular mechanisms of redox signaling still
remain unclear. A number of proteomics techniques have
been developed to study the targets and modulations of
redox signaling [193]. Because redox proteomics is still an
emerging field with continuing method development,
research has mainly focused on basic method development
and the application of the newly-developed methods to cell
culture and animal models [193,194]; only a few redox pro-
teomics studies have involved human subjects [195,196]. A
major challenge in transferring redox proteomics methods to
clinical samples is the necessity for blocking free Cys residues
and preserving endogenous levels as much and as soon as
possible. In contrast to purely enzymatically driven PTMs, Cys
redox states can even change in denatured samples and dur-
ing sample preparation, which means that protocols have to
be specifically optimized to reduce the occurrence of new
(and exchange of existing) redox modifications. This involves
avoiding exposure to UV light and atmospheric oxygen, avoid-
ing lysis of the cells before the addition of blocking reagent,
avoiding common pathogen inactivation procedures [197] etc.
While this is already complicated for cell culture and requires
careful use of controls, it represents a major challenge for
clinical samples, particularly tissues. In the following section,
we will briefly discuss the most relevant Cys redox modifica-
tions, their roles in health and disease, and the current state of
proteomic investigations.
The important role of NO in cardiac function and neurotrans-
mission has been suggested by multiple studies [186,188,192],
and the cardioprotective potential of S-nitrosylation has been
attributed to its ability to shield thiol groups from irreversible
oxidation [186-188]. The neuronal isoform of NOS produces NO
in the brain, which was first thought to be a neurotoxin associated
with neurodegenerative diseases. Recent studies, however, also
indicate the protective function of SNO in neurotransmission and
synaptic plasticity [192].
Direct analysis of SNO is highly challenging as it is not only
present at low levels, but—more importantly—also labile.
Currently, phenylmercury-resin-based enrichment approaches
allow the direct capture of SNO via formation of stable thiol-
Hg bonds. Using this technique Raju et al. identified 269
endogenous SNO sites in wild-type mouse brain [192]. A
broad spectrum of indirect methods to enrich and analyze
SNO and other cysteine PTMs utilize different types of mod-
ified biotin switch techniques (BST) such as iodoTMT [198],
resin-assisted capture (RAC) [199], and isotope-coded affinity
tags (ICAT) [200]. These assays include blocking of the free
thiols, followed by selective reduction of modified cysteines
and introduction of functional groups enabling stable-adduct
formation, enrichment, and (preferably) also quantitation of
the switched PTMs. Using a IodoTMT switch assay, including
specific reduction of SNO by sodium ascorbate, Pan et al.
identified 12 significantly regulated SNO proteins in hypoxic
cardiomyocytes [201]. Another modified BST, disulfide
exchange based RAC, showed high specificity (~95 %) and
superior sensitivity for proteins higher than 100 kDa compared
to biotin-based enrichment [199,202]. Using another RAC
approach in combination with iTRAQ labeling, Forrester et al.
demonstrated the progressive decay of SNO signal over time
as a result of CysNO treatment of HEK293 cells. Roughly 300
negatively-regulated SNO sites were quantified [202], possibly
indicating the presence of regulatory mechanisms for removal
of SNO or exchange of the transient SNO to more stable
disulfide [203]. A modified form of ICAT: SNOxICAT, allowed
for quantitative analysis of SNO site occupancy in vivo and
demonstrated that the exposure to NO2 leads to widespread
SNO formation only in ischemic mouse hearts, but not in
ischemia or NO2 exposure solely [204]. Ibáñez-Vea et al.
applied titanium dioxide (TiO2) chromatography enrichment
after SNO reduction by ascorbate followed by labeling with
cysteine specific phosphonate adaptable tag (CysPAT) and
identified high number of SNO proteins (569 endogeneous
and 795 chemically induced) in macrophages [205].
SOH is a highly reactive and unstable modification [206].
Depending on the local redox-environment, it can be
further converted to other reversible and irreversible
cysteine PTMs. As an intermediate form of cysteine oxida-
tion, it is a key player in redox signaling [207]. Notably, the
unique reactivity of SOH to dimedone compounds allows its
use for direct labeling in living cells. Thus, stable isotope
coded alkyne-functionalized dimedone (DYn-2 and d6-DYn-
2) probes enable the enrichment and relative quantitation
of SOH peptides. In a recent study of human RKO cells,
about 1000 SOH-modified Cys sites from more than 700
proteins were identified in response to epidermal growth
factor (EGF) stimulation and H2O2 treatment, from which
215 and 360 SOH sites could be quantified, respectively
[208]. Indirect analysis of SOH by modified BST and specific
reduction of SOH by arsenite was used for the study of a
SILAC-labeled THP-1 human monocytic cell line treated with
H2O2 or activated human platelet releasate. More than 200
SOH peptides were quantified [209].
One of the important functions of cellular disulfide bonds is
the stabilization of protein structure via intramolecular and inter-
molecular covalent linkages [210]. Nevertheless, specific Cys resi-
dues form redox-active disulfides as a result of reduction by
oxidoreductases, interaction with small-molecule thiols such as
glutathione (GSH) and free cysteine, or by replacement of less-
stable SNO/SOH. Allosteric redox-active disulfides induce a
functional change in the protein and are involved in redox
signaling [210]. A recent study has demonstrated the role of
redox-active S-S in cancer: In vascular endothelial growth factor
receptor (VEGFR2), Cys1199 and Cys1206 form an intra-molecular
disulfide bond at elevated ROS levels, suppressing VEGFR2 activ-
ity [189]. Importantly, knockdown of the key antioxidant enzyme

peroxiredoxin III resulted in inactivation of VEGFR2 and repres-
sion of tumor angiogenesis in vivo [211].
Proteome-wide analysis of intact disulfides represents a sig-
nificant challenge. For analysis of native disulfides, Lu et al. used
a combination of proteases, including Lys-C, trypsin, subtilisin,
Glu-C, elastase, proteinase K, and/or Asp-N for protein diges-
tion, extensively fractionated the resulting digests, performed
LC-MS, and used their pLink-SS data analysis software for the
identification of cross-linked peptides. In this way, more than
550 disulfide bonds were mapped in the secretome of human
umbilical vein endothelial cells (HUVECs) [212].
SSG has been suggested to provide more persistent protec-
tion against oxidative damage than transient SNO or SOH.
Murdoch et al. showed that SSG is increased in ischemic
muscle even several days after hindlimb ischemia, and that it
facilitates revascularization [213]. SSG of specific cysteines in
the upstream and downstream targets of VEGFR2 were shown
to considerably increase cardiac angiogenesis [187]. As there
are no direct methods available for the analysis of endogen-
ous SSG, most studies are based on modified BST and the
specific reduction of SSG by glutaredoxin [214,215].
The free intracellular cysteine concentration of ~100 µM is
the source of S-Cys [210]. An analysis of the cysteinylation of
human serum albumin (HSA) at Cys34 in human plasma sam-
ples obtained from 229 patients (139 of whom had liver disease,
38 of whom had kidney disease, and 52 of whom had diabetes
mellitus), showed a strong correlation of increased S-Cys34 with
these three oxidative stress-related diseases [195].
All these examples clearly demonstrate the high relevance
of Cys modifications for human health, and the need to further
improve the sensitivity and robustness of current methods so
that they can be applied to clinical samples in the future. As
demonstrated for phosphorylation, it can be anticipated that
additional information on Cys redox status will be another
important layer of information to stratify patient samples,
and may lead to important markers of disease.
9. Study design for biomarker discovery
With the continuous advancement of proteomic instrumenta-
tion, methods, algorithms, and data analysis strategies, there
has been also a paradigm shift with regard to study design. In
the early days of clinical proteomics, two-dimensional PAGE was
a widely-used but tedious and limited technology (with regard
to quantitation, precision and coverage), and studies usually
included only a few patients; nowadays, there is a general
trend for the application of sophisticated gel-free workflows
to larger cohorts. Cross-sectional comparisons between differ-
ent cohorts are complemented more and more by longitudinal
studies that follow patients over time [216]. Particularly for
biofluids which tend to show a large variation between
patients, this is a promising strategy for the identification bio-
marker panels. Still today, most studies follow the ‘biomarker
pipeline’ and include smaller initial cohorts for identifying
potential markers and then expand this to larger cohorts for
validation [217]. If sufficient instrumentation and computational
power are available, Geyer et al. propose an alternative strategy,
which utilizes a large cohort in the discovery phase, and a
EXPERT REVIEW OF PROTEOMICS 527
second alternative cohort of the same size for the validation
phase, both in a non-targeted rather than targeted mode [218].
Obtaining well-classified samples for a sufficient number of
patients and matched controls that have been collected under
well-defined conditions that are compatible with downstream
proteomic analysis, is often a bottleneck. However, beside
stratification sample size is a major determinant of the relia-
bility of the biomarkers and targets which have been identi-
fied in clinical samples, as sample size directly affects the
statistical power [219,220]. Skates et al. calculated the prob-
abilities for a biomarker passing through the biomarker pipe-
line from discovery to validation, and to then clinical
validation as a function of the number of biospecimens and
for different scenarios [221]. In this study, they hypothesized
that approximately 8000 proteins would be quantified during
discovery, leading to 20/50/100 candidates for measurement
in the validation state, and then 2/5/10 candidates for mea-
surement in the clinical validation state. Thus, if 10 controls are
compared against 10 patients during discovery, the likelihood
that an authentic biomarker present in 50% of the cases with a
median distance of 3 standard deviations is passed to the
validation step is only 15% when 20 targeted assays are
performed. In the case where 50 (100) targeted assays are
performed, this likelihood increases to 24% (35%). However,
increasing the number of samples has a much stronger effect.
Under the aforementioned conditions, with 20 targeted assays
and 25 (50) patients and controls, the likelihood increases to
60% (93%).
These theoretical considerations underline the importance
of cohort size for biomarker discovery, but sample normal-
ization and statistical tests to actually identify potential bio-
markers in the discovery phase are also relevant. The
conventional two-sided t-test to estimate the probability that
a value is significantly different is not sufficient—if many
individual tests are conducted (i.e. for thousands of proteins),
a substantial share will yield significant p-values just by chance
(false positives) [222]. Therefore, corrections such as Benjamini
Hochberg are typically applied in order to take false discovery
rates into account for multiple comparisons. In addition to the
statistical significance, the level of regulation must also be
considered. A statistically highly significant 1.1-fold regulation
may not be relevant—the median standard deviation of the
entire dataset needs to be considered, too, in order to apply
realistic cutoffs that take into account the quality of the indi-
vidual dataset (e.g. 3 times the standard deviation as in the
aforementioned example from Stakes et al.).
Sometimes sample size is inherently limited, for instance in
case of rare disorders with low prevalence, rendering the
estimation of the variance uncertain. Nevertheless, such rare
disorders may still be analyzed by proteomics [223]. For small
sample sizes, the t-test may consider that a potential marker
showing strong regulation but (concurrently) high intra-group
variance is not significant [224] (Figure 6). Large intra-group
variations can particularly appear between patients. Thus, one
strategy to compensate for the inaccuracy of the estimation of
variance in small sample sizes is ‘empirical Bayes method
shrinking’ [224] which uses a moderated t-test where the
p-value is assessed by the fold-change rather than the initially
estimated variance. Notably, empirical Bayes methods have
528 R. MNATSAKANYAN ET AL.

Figure 6. Limits of t-test based statistics for small sample sizes. (A) Two simulated potential markers 1 and 2. Although the first marker shows a considerable difference
between patients and controls, the strong intra-group variance leads to a poor p-value. In contrast, the second marker yields a low p-value and might be considered as
statistically significant, but the difference between patients and groups is negligible. (B) Volcano plot of the two markers in the background of an iTRAQ-based
phosphoproteomics experiment based on a t-test. (C) Volcano plot of the same data based on the moderated t-test provided in the Limma package [226-228]. Only marker
1 is actually significant. A detailed description for the use of this package was recently published by Kammers et al. [224]. (Figure reprinted with permission, Pagel et al., Expert
Rev Proteomics 2015).

been successfully applied in the microarray field for approxi-
mately 10 years [225] and can be adapted for use in proteo-
mics studies [224,226,227].
Despite these promising alternative data analysis strategies,
the major important determinant of successful biomarker dis-
covery still remains: The most sophisticated data analysis stra-
tegies cannot compensate for poor sample or data quality.
There is a continuously increasing number of clinical proteo-
mics studies facing these issues, but there are also continuous
improvements in available methods and instrumentation.
Thus, it can be anticipated that improved statistical methods
and tools will be developed to improve the detection and
verification of biomarker candidates in proteomics studies.
10. Expert commentary
In recent years (i) technologies for proteomic sample preparation,
PTM enrichment, chromatographic separation, and (ii) LC and MS
instrumentation, as well as (iii) data analysis strategies, have been
considerably improved. Concurrently, there has been an increas-
ing awareness of the importance of quality control and standardi-
zation in MS-based proteomics in order to produce more reliable
and robust quantitative data. Particularly for phosphorylation,
glycosylation, proteolytic cleavage, and Cys modifications, highly
sensitive methods have been developed that can even be applied
to limited sample amounts, as often the case for biopsies.
Consequently, an increasing number of clinical proteomics studies
that focus on PTMs are emerging. Combining different and com-
plementary layers of information, e.g. changes in (i) protein abun-
dance and PTM levels, (ii) differences in PTMs, or (iii) changes in the
genome, transcriptome, or proteome (proteogenomics [229,230])
holds great potential for discovering important drivers and mar-
kers of disease. Mere knowledge of protein abundance without
spatial information and without full isoform resolution (both of
which are missing in conventional proteomics studies) might, in
some cases, be insufficient or even misleading. Therefore, the
search for specific PTM signatures of disease is a promising com-
plementary strategy, as demonstrated for potential proteolytic
signatures in postchemotherapeutic blood [178], for example.
Compared to genomics and transcriptomics, proteomic work-
flows are still considerably more challenging and less standar-
dized. Thus, a general transfer of discovery proteomics workflows
into routine clinics might be unrealistic in the near future.
Nevertheless, interdisciplinary collaborations between clinicians,
proteomics researchers, biochemists, bioinformaticians, and sta-
tisticians can provide important and novel insights into disease
mechanisms, biomarkers, and therapeutic targets. Pilot projects
such as CPTAC have made great progress in standardization and
identifying potential pitfalls for in-depth (phospho)proteomics
analyses of cancer tissues [102,161], and different cancer tissues
have been analyzed using proteogenomic workflows, leading to
the relative quantitation of thousands of proteins and phospho-
peptides [30,32].
In addition to discovery, the transfer of MS-based assays into
the clinic is an exciting opportunity. In cases where, for example
current immunohistochemistry assays have limited predictive
value, technologies such as MALDI imaging or iMALDI may be
promising alternatives for the future [111,231]. Standardized
MRM and PRM assays can be used to quantify defined sets of
proteins and modified peptides. For phosphopeptides, this has
already been demonstrated, but existing workflows are still too
complex and time-consuming. Thus, Di Conza et al. were able
to use a tailored PRM workflow for absolute quantitation of
phosphorylation levels for PHD2 Ser125 in colorectal cancer
biopsies [157]. Although this is an exciting example which
demonstrates the unique capabilities of MS for the absolute
quantitation of phosphorylation levels from small amounts of
tissue, the analysis required a specific ERLIC enrichment step;
although this could probably be automated, it still might be too
laborious to transfer into clinics. Also, it will be very important
to ensure that such workflows work for FFPE samples. An
alternative indirect workflow for determining phosphorylation

stoichiometry is by spiking in the non-phosphorylated SIL refer-
ence peptide and splitting the sample in two aliquots, one of
which treated with phosphatase [232]. Individual measurement
then allows assessment of phosphorylation stoichiometry.
However, phospo-isomers cannot be distinguished after phos-
phatase treatment.
11. Five-year perspective
The increasing awareness of quality measures in the field of
proteomics, initially with regard to reliable peptide and pro-
tein identification through the implementation of false discov-
ery rate assessments [233] and site localization algorithms [72],
but (importantly) also for system performance [50] and more
precise quantitation, have been critical milestones in the move
from cell biology-driven proteomics to clinical proteomics. It
can be anticipated that instrumentation will further improve in
the next 5 years, leading to more robust, more rapid, and
more sensitive LC-MS analyses with higher resolution. (Semi-)
automated and parallelized sample preparation workflows are
increasingly being developed and applied to reduce technical
variation and increase analysis throughput. These develop-
ments will have an impact on DDA, DIA, and targeted meth-
ods, and will further improve the coverage, precision, and
reproducibility of MS-based identification and quantitation.
In this way, precise and deep proteomic profiling will become
less time-consuming and more cost-effective, facilitating the
implementation of studies on large sample cohorts as well as
complex longitudinal studies. It will be important that data
analysis pipelines are able to match the speed of the experi-
mental progress, particularly regarding data amounts, compre-
hensiveness, and complexity. In addition to label-free analysis
of samples, the increased multiplexing capacities of NeuCode
will allow multi-level analysis of PTMs even from clinical
samples [234]. As instrumentation for targeted MS is compara-
tively robust, inexpensive, and widespread, it can be antici-
pated that clinics will increasingly consider establishing MS-
based assays for the quantitation of proteins and PTM levels
[217]. MALDI imaging also holds a great potential for improv-
ing or complementing current histopathology, with the ulti-
mate goal of having powerful algorithms that can help with
diagnosis [235]. As a rapid and straightforward approach,
iMALDI may allow the MS-based precise quantitation of pro-
tein panels that may help to stratify patients [110,111].
Large multinational consortium-driven proteogenomic
studies will increasingly exploit the unique possibilities of
synergizing genomic, transcriptomic, proteomic, metabolo-
mic, and lipidomic datasets [236]. Thus, the ‘cancer moon-
shot’ program initiated by former US vice-president Joe
Biden [237,238] led to the formation of the International
Cancer Proteogenomics Consortium (ICPC), bringing
together ‘some of the world’s leading cancer and proteo-
genomic research centers’ in order to improve our under-
standing of cancer. Such large worldwide consortia hold a
great potential, as they enable access to larger patient
cohorts, complementary technologies, and improved data
sharing. It may be expected that similar large coordinated
efforts will be undertaken for other widespread diseases, as
EXPERT REVIEW OF PROTEOMICS 529
well as for rare diseases where the increase in the number
of available patients is even more critical.
In summary, MS has just recently began to reveal its real
potential for clinical research, and the near future will be an
exciting era where researchers will have the opportunity to
demonstrate that the unique capabilities of MS-based pro-
teomics can be translated into clinical environments in
order to improve human health and healthy aging.
Key issues
● To discover PTM signatures in clinical samples that may
serve as biomarkers or targets, it is vital to have an optimal
and elaborate study design, including the use of matching
controls and thorough stratification of patient and control
samples. Including inappropriate samples may lead to
increased intra-group variation and therefore could mask
potential markers. On the contrary, having well-character-
ized samples may allow reducing the number of samples.
● Also, to keep preanalytical variance as low as possible it is
important to strictly follow elaborate SOPs during sampling
and sample processing. Ideally the sample collection should
be commonly planned by clinicians and proteomics experts
together and all steps should be followed strictly. In case
small deviations from agreed-upon SOPs cannot be
excluded, these should be well-documented so that later
inconsistent results may be correlated with such changes.
● PTM enrichments procedures should first be tested and
optimized for sensitivity and robustness using reference
samples such as HeLa or other cell culture models. As
these procedures are often complex, the use of chemical
labeling procedures that allow multiplexing prior to enrich-
ment should be considered in order to reduce technical
variation. Additionally, multiplexing may be used to
improve sensitivity as in total more sample material can
be used for the analysis.
● For data analysis, it is important to consider how to deal
with missing values. The selection of interesting targets
should not be solely based on p-values and should addi-
tionally require reasonable fold-changes that take into
account the standard deviation and distribution of the
whole dataset at hand.
● The validation of targets identified in the discovery phase
should include a larger independent sample cohort, ideally
collected using the same SOPs. The use of solely antibody-
based assays has to be carefully considered, as particularly
modification-specific protein antibodies often show poor
specificity. More promising alternatives may be (i) targeted
LC-MS in case of moderate to high-abundance targets using
whole lysates and (ii) after enrichment e.g. of phosphopep-
tides, or (iii) using a combination of anti-peptide antibodies
and (targeted) MS such as SISCAPA and iMALDI.
Compatibility with FFPE tissue samples should be evaluated
thoroughly.
● For transfer into the clinic promising assays should be
straightforward, robust, ideally fast and precise and there-
fore should have potential for (partial) automation and
provide a simple and clear read-out.
530 R. MNATSAKANYAN ET AL.
Funding
The authors R. Mnatsakanyan, G.Shema, A. Sickmann were supported by the
Ministerium für Innovation, Wissenschaft und Forschung des Landes
Nordrhein-Westfalen, the Senatsverwaltung für Wirtschaft, Technologie und
Forschung des Landes Berlin, and the Bundesministerium für Bildung und
Forschung. C.H. Borchers and R.P.Zahedi are grateful to Genome Canada and
Genome British Columbia for financial support (project codes 204PRO for
operations and 214PRO for technology development). CHB is also grateful for
support from the Leading Edge Endowment Fund (University of Victoria) and
for support from the Segal McGill Chair in Molecular Oncology at McGill
University (Montreal, Quebec, Canada). C.H.Borchers, A.Spatz, M.Basik and
G.Batist are grateful for support from the Warren Y. Soper Charitable Trust
and the Alvin Segal Family Foundation to the Jewish General Hospital
(Montreal, Quebec, Canada). A.Spatz, M.Basik and G.Batist are also grateful
for support from Genome Canada's Business-Led Networks of Centres of
Excellence progran, Exactis Innovations, and the Fonds de recherche du
Québec - Santé, Axe cancer du sein et de l'ovaire (FRSQ-ACSO).
Declaration of interest
C.H.Borchers is the Chief Scientific Officer of MRM Proteomics, Inc. The
other authors have no other relevant affiliations or financial involvement
with any organization or entity with a financial interest in or financial
conflict with the subject matter or materials discussed in the manuscript
apart from those disclosed.
Reviewer disclosures
Peer reviewers on this manuscript have no relevant financial or other
relationships to disclose.
ORCID
Christoph H. Borchers http://orcid.org/0000-0003-2394-6512
Albert Sickmann http://orcid.org/0000-0002-2388-5265
René P. Zahedi http://orcid.org/0000-0002-4960-5460
References
Papers of special note have been highlighted as either of interest (•) or of
considerable interest (••) to readers.
1. UniProt C. UniProt: a hub for protein information. Nucleic Acids Res.
2015;43(Database issue):D204–212.
2. Hornbeck PV, Zhang B, Murray B, et al. PhosphoSitePlus2014:
mutations, PTMs and recalibrations. Nucleic Acids Res. 2015;43
(Database issue):D512–520.
3. Vaudel M, Burkhart JM, Zahedi RP, et al. PeptideShaker enables
reanalysis of MS-derived proteomics data sets. Nat Biotechnol.
2015;33(1):22–24.
4. Lemeer S, Heck AJ. The phosphoproteomics data explosion. Curr
Opin Chem Biol. 2009;13(4):414–420.
5. Zhao Y, Brickner JR, Majid MC, et al. Crosstalk between ubiquitin
and other post-translational modifications on chromatin during
double-strand break repair. Trends Cell Biol. 2014;24(7):426–434.
6. Hentschel A, Zahedi RP, Ahrends R. Protein lipid modifications–
More than just a greasy ballast. Proteomics. 2016;16(5):759–782.
7. Ruan HB, Nie Y, Yang X. Regulation of protein degradation by
O-GlcNAcylation: crosstalk with ubiquitination. Molecular & cellular
proteomics: MCP. 2013;12(12):3489–3497.
8. Hart GW, Greis KD, Dong LY, et al. O-linked N-acetylglucosamine:
the “yin-yang” of Ser/Thr phosphorylation? Nuclear and cytoplas-
mic glycosylation. Adv Exp Med Biol. 1995;376:115–123.
9. Creixell P, Linding R. Cells shared memory and breaking the PTM
code. Molecul sys bio. 2012;8:598.
10. Benayoun BA, Veitia RA. A post-translational modification code for
transcription factors: sorting through a sea of signals. Trends Cell
Biol. 2009;19(5):189–197.
11. Gu B, Zhu WG. Surf the post-translational modification network of
p53 regulation. Int J Biol Sci. 2012;8(5):672–684.
12. Venne AS, Kollipara L, Zahedi RP. The next level of complexity: crosstalk
of posttranslational modifications. Proteomics. 2014;14(4–5):513–524.
13. Solari FA, Dell’Aica M, Sickmann A, et al. Why phosphoproteomics
is still a challenge. Mol Biosyst. 2015;11(6):1487–1493.
14. Ren J, Jiang C, Gao X, et al. PhosSNP for systematic analysis of
genetic polymorphisms that influence protein phosphorylation.
Molecular & cellular proteomics: MCP. 2010;9(4):623–634.
15. Ruggles KV, Krug K, Wang X, et al. Methods, Tools and Current
Perspectives in Proteogenomics. Molecular & cellular proteomics:
MCP. 2017;16(6):959–981.
• Excellent overview of the state-of-the art in proteogenomics.
16. Slamon DJ, Clark GM, Wong SG, et al. Human breast cancer: corre-
lation of relapse and survival with amplification of the HER-2/neu
oncogene. Science. 1987;235(4785):177–182.
17. Yuste L, Montero JC, Esparis-Ogando A, et al. Activation of ErbB2 by
overexpression or by transmembrane neuregulin results in differ-
ential signaling and sensitivity to herceptin. Cancer Res. 2005;65
(15):6801–6810.
18. Paez JG, Janne PA, Lee JC, et al. EGFR mutations in lung cancer:
correlation with clinical response to gefitinib therapy. Science.
2004;304(5676):1497–1500.
19. Klootwijk ED, Reichold M, Helip-Wooley A, et al. Mistargeting of
peroxisomal EHHADH and inherited renal Fanconi’s syndrome. N
Engl J Med. 2014;370(2):129–138.
20. Ohno M. Roles of eIF2alpha kinases in the pathogenesis of
Alzheimer’s disease. Front Mol Neurosci. 2014;7:22.
21. Seeling M, Bruckner C, Nimmerjahn F. Differential antibody glyco-
sylation in autoimmunity: sweet biomarker or modulator of disease
activity? Nat Rev Rheumatol. 2017;13(10):621–630.
22. Sherrill JD, Stropes MP, Schneider OD, et al. Activation of intracel-
lular signaling pathways by the murine cytomegalovirus G protein-
coupled receptor M33 occurs via PLC-{beta}/PKC-dependent and
-independent mechanisms. J Virol. 2009;83(16):8141–8152.
23. Powers AD, Palecek SP. Protein analytical assays for diagnosing,
monitoring, and choosing treatment for cancer patients. J Healthc
Eng. 2012;3(4):503–534.
24. Chandler K, Goldman R. Glycoprotein disease markers and single pro-
tein-omics. Molecular & cellular proteomics: MCP. 2013;12(4):836–845.
25. Van Scherpenzeel M, Willems E, Lefeber DJ. Clinical diagnostics and
therapy monitoring in the congenital disorders of glycosylation.
Glycoconj J. 2016;33(3):345–358.
26. Abdel-Kahaar E, Kabakchiev M, Hartmann B, et al. Performance of a
phosphoflow assay to determine phosphorylation of S6 ribosomal
protein as a pharmacodynamic read out for mTOR inhibition. Clin
Biochem. 2016;49(15):1181–1187.
27. Rahbar S, Blumenfeld O, Ranney HM. Studies of an unusual hemo-
globin in patients with diabetes mellitus. Biochem Biophys Res
Commun. 1969;36(5):838–843.
28. Selvin E, Steffes MW, Zhu H, et al. Glycated hemoglobin, diabetes,
and cardiovascular risk in nondiabetic adults. N Engl J Med.
2010;362(9):800–811.
29. Schwarz UR, Geiger J, Walter U, et al. Flow cytometry analysis of
intracellular VASP phosphorylation for the assessment of activating
and inhibitory signal transduction pathways in human platelets–
definition and detection of ticlopidine/clopidogrel effects. Thromb
Haemost. 1999;82(3):1145–1152.
30. Huang KL, Li S, Mertins P, et al. Proteogenomic integration reveals
therapeutic targets in breast cancer xenografts. Nat Commun.
2017;8:14864.
31. Heist RS, Christiani D. EGFR-targeted therapies in lung cancer:
predictors of response and toxicity. Pharmacogenomics. 2009;10
(1):59–68.
32. Mertins P, Mani DR, Ruggles KV, et al. Proteogenomics connects
somatic mutations to signalling in breast cancer. Nature. 2016;534
(7605):55–62.

•• Impressive proof-of-concept proteogenomics study quantita-
tively analyzing the proteomes and phosphoproteomes of 105
genomically annotated breast cancers.
33. Salvagno GL, Danese E, Lippi G. Preanalytical variables for liquid
chromatography-mass spectrometry (LC-MS) analysis of human
blood specimens. Clin Biochem. 2017;50(10–11):582–586.
34. Longo DL. Tumor heterogeneity and personalized medicine. N Engl
J Med. 2012;366(10):956–957.
35. Maes E, Mertens I, Valkenborg D, et al. Proteomics in cancer
research: are we ready for clinical practice? Crit Rev Oncol
Hematol. 2015;96(3):437–448.
36. Burkhart JM, Vaudel M, Gambaryan S, et al. The first comprehensive
and quantitative analysis of human platelet protein composition
allows the comparative analysis of structural and functional path-
ways. Blood. 2012;120(15):e73–82.
37. Skold K, Alm H, Scholz B. The impact of biosampling procedures on
molecular data interpretation. Molecular & cellular proteomics: MCP.
2013;12(6):1489–1501.
38. Aasebo E, Mjaavatten O, Vaudel M, et al. Freezing effects on the
acute myeloid leukemia cell proteome and phosphoproteome
revealed using optimal quantitative workflows. J Proteomics.
2016;145:214–225.
39. Lalmahomed ZS, Coebergh van den Braak RR, Mh O, et al.
Multicenter fresh frozen tissue sampling in colorectal cancer: does
the quality meet the standards for state of the art biomarker
research? Cell tissue banking 2017;18(3):425–431.
40. Ivanov AR, Colangelo CM, Dufresne CP, et al. Interlaboratory studies
and initiatives developing standards for proteomics. Proteomics.
2013;13(6):904–909.
41. Percy AJ, Yang J, Chambers AG, et al. Protocol for Standardizing
High-to-Moderate Abundance Protein Biomarker Assessments
Through an MRM-with-Standard-Peptides Quantitative Approach.
Adv Exp Med Biol. 2016;919:515–530.
42. Percy AJ, Yang J, Hardie DB, et al. Precise quantitation of 136
urinary proteins by LC/MRM-MS using stable isotope labeled pep-
tides as internal standards for biomarker discovery and/or verifica-
tion studies. Methods. 2015;81:24–33.
43. Percy AJ, Chambers AG, Smith DS, et al. Standardized protocols for
quality control of MRM-based plasma proteomic workflows. J
Proteome Res. 2013;12(1):222–233.
44. Percy AJ, Chambers AG, Yang J, et al. Method and platform stan-
dardization in MRM-based quantitative plasma proteomics. J
Proteomics. 2013;95:66–76.
45. Malm J, Fehniger TE, Danmyr P, et al. Developments in biobanking
workflow standardization providing sample integrity and stability. J
Proteomics. 2013;95:38–45.
46. Casadonte R, Caprioli RM. Proteomic analysis of formalin-fixed
paraffin-embedded tissue by MALDI imaging mass spectrometry.
Nat Protoc. 2011;6(11):1695–1709.
47. Ostasiewicz P, Wisniewski JRA. Protocol for Large-Scale Proteomic
Analysis of Microdissected Formalin Fixed and Paraffin Embedded
Tissue. Methods Enzymol. 2017;585:159–176.
48. Kennedy JJ, Whiteaker JR, Schoenherr RM, et al. Optimized Protocol
for Quantitative Multiple Reaction Monitoring-Based Proteomic
Analysis of Formalin-Fixed, Paraffin-Embedded Tissues. J Proteome
Res. 2016;15(8):2717–2728.
49. Dickhut C, Radau S, Zahedi RP. Fast, efficient, and quality-controlled
phosphopeptide enrichment from minute sample amounts using
titanium dioxide. Methods in molecular biology. 2014;1156:417–430.
50. Burkhart JM, Premsler T, Sickmann A. Quality control of nano-LC-
MS systems using stable isotope-coded peptides. Proteomics.
2011;11(6):1049–1057.
51. Muller T, Winter D. Systematic Evaluation of Protein Reduction and
Alkylation Reveals Massive Unspecific Side Effects by Iodine-containing
Reagents. Molecular & cellular proteomics: MCP. 2017;16(7):1173–1187.
52. Burkhart JM, Schumbrutzki C, Wortelkamp S, et al. Systematic and
quantitative comparison of digest efficiency and specificity reveals
the impact of trypsin quality on MS-based proteomics. J Proteomics.
2012;75(4):1454–1462.
EXPERT REVIEW OF PROTEOMICS 531
53. Burkhart JM, Vaudel M, Zahedi RP, et al. iTRAQ protein quantification:
a quality-controlled workflow. Proteomics. 2011;11(6):1125–1134.
54. Bittremieux W, Tabb DL, Impens F, et al. Quality control in mass
spectrometry-based proteomics. Mass Spectrom Rev. 2017. Epub
ahead of print.
55. Percy AJ, Mohammed Y, Yang J, et al. A standardized kit for auto-
mated quantitative assessment of candidate protein biomarkers in
human plasma. Bioanalysis. 2015;7(23):2991–3004.
56. Zolg DP, Wilhelm M, Yu P, et al. PROCAL: A set of 40 peptide
standards for retention time indexing, column performance mon-
itoring and collision energy calibration. Proteomics. 2017;17(21).
57. Pichler P, Mazanek M, Dusberger F, et al. SIMPATIQCO: a server-
based software suite which facilitates monitoring the time course
of LC-MS performance metrics on Orbitrap instruments. J Proteome
Res. 2012;11(11):5540–5547.
58. Bereman MS. Tools for monitoring system suitability in LC MS/MS
centric proteomic experiments. Proteomics. 2015;15(5–6):891–902.
59. Sandin M, Teleman J, Malmstrom J, et al. Data processing methods
and quality control strategies for label-free LC-MS protein quanti-
fication. Biochim Biophys Acta. 2014;1844(1Pt A):29–41.
60. Mohammed Y, Percy AJ, Chambers AG, et al. Qualis-SIS: automated
standard curve generation and quality assessment for multiplexed
targeted quantitative proteomic experiments with labeled stan-
dards. J Proteome Res. 2015;14(2):1137–1146.
61. Aittokallio T. Dealing with missing values in large-scale studies: micro-
array data imputation and beyond. Brief Bioinform. 2010;11(2):253–264.
62. Lazar C, Gatto L, Ferro M, et al. Accounting for the Multiple Natures of
Missing Values in Label-Free Quantitative Proteomics Data Sets to
Compare Imputation Strategies. J Proteome Res. 2016;15(4):1116–1125.
63. Tyanova S, Temu T, Sinitcyn P, et al. The Perseus computational
platform for comprehensive analysis of (prote)omics data. Nature
methods. 2016;13(9):731–740.
64. Carr SA, Abbatiello SE, Ackermann BL, et al. Targeted peptide
measurements in biology and medicine: best practices for mass
spectrometry-based assay development using a fit-for-purpose
approach. Molecular & cellular proteomics: MCP. 2014;13(3):907–917.
65. Ebhardt HA, Root A, Sander C, et al. Applications of targeted
proteomics in systems biology and translational medicine.
Proteomics. 2015;15(18):3193–3208.
66. Anderson L, Hunter CL. Quantitative mass spectrometric multiple
reaction monitoring assays for major plasma proteins. Molecular &
cellular proteomics: MCP. 2006;5(4):573–588.
67. Ta A, Se A, Schilling B, et al. Multi-site assessment of the precision
and reproducibility of multiple reaction monitoring-based measure-
ments of proteins in plasma. Nat Biotechnol. 2009;27(7):633–641.
68. Peterson AC, Russell JD, Bailey DJ, et al. Parallel reaction monitoring
for high resolution and high mass accuracy quantitative, targeted
proteomics. Molecular & cellular proteomics: MCP. 2012;11
(11):1475–1488.
69. Cox J, Mann M. MaxQuant enables high peptide identification
rates, individualized p.p.b.-range mass accuracies and proteome-
wide protein quantification. Nat Biotechnol. 2008;26(12):1367–1372.
70. Sturm M, Bertsch A, Gropl C, et al. OpenMS - an open-source
software framework for mass spectrometry. BMC bioinformatics.
2008;9:163.
71. Rost HL, Sachsenberg T, Aiche S, et al. OpenMS: a flexible open-
source software platform for mass spectrometry data analysis.
Nature methods. 2016;13(9):741–748.
72. Chalkley RJ, Clauser KR. Modification site localization scoring: stra-
tegies and performance. Molecular & cellular proteomics: MCP.
2012;11(5):3–14.
73. Beck F, Lewandrowski U, Wiltfang M, et al. The good, the bad, the
ugly: validating the mass spectrometric analysis of modified pep-
tides. Proteomics. 2011;11(6):1099–1109.
74. Mikesh LM, Ueberheide B, Chi A, et al. The utility of ETD mass
spectrometry in proteomic analysis. Biochim Biophys Acta.
2006;1764(12):1811–1822.
75. Gillet LC, Navarro P, Tate S, et al. Targeted data extraction of the
MS/MS spectra generated by data-independent acquisition: a new
532 R. MNATSAKANYAN ET AL.
concept for consistent and accurate proteome analysis. Molecular &
cellular proteomics: MCP. 2012;11(6):O111 016717.
76. Simburger JM, Dettmer K, Oefner PJ, et al. Optimizing the SWATH-MS-
workflow for label-free proteomics. J Proteomics. 2016;145:137–140.
77. Distler U, Kuharev J, Navarro P, et al. Label-free quantification in ion
mobility-enhanced data-independent acquisition proteomics. Nat
Protoc. 2016;11(4):795–812.
78. Percy AJ, Chambers AG, Yang J, et al. Comparison of standard- and
nano-flow liquid chromatography platforms for MRM-based quan-
titation of putative plasma biomarker proteins. Anal Bioanal Chem.
2012;404(4):1089–1101.
79. Ross PL, Huang YN, Marchese JN, et al. Multiplexed protein quantitation
in Saccharomyces cerevisiae using amine-reactive isobaric tagging
reagents. Molecular & cellular proteomics: MCP. 2004;3(12):1154–1169.
80. Treitz C, Cassidy L, Hockendorf A, et al. Quantitative proteome
analysis of Caenorhabditis elegans upon exposure to nematicidal
Bacillus thuringiensis. J Proteomics. 2015;113:337–350.
81. Thompson A, Schafer J, Kuhn K, et al. Tandem mass tags: a novel
quantification strategy for comparative analysis of complex protein
mixtures by MS/MS. Anal Chem. 2003;75(8):1895–1904.
82. Erickson BK, Jedrychowski MP, McAlister GC, et al. Evaluating multi-
plexed quantitative phosphopeptide analysis on a hybrid quadru-
pole mass filter/linear ion trap/Orbitrap mass spectrometer. Anal
Chem. 2015;87(2):1241–1249.
83. Venne AS, Solari, FA, Faden F, et al. An improved workflow for
quantitative N-terminal charge-based fractional diagonal chroma-
tography (ChaFRADIC) to study proteolytic events in Arabidopsis
thaliana. Proteomics. 2015;15(14):2458–69.
84. Hoffert JD, Pisitkun T, Saeed F, et al. Dynamics of the G protein-
coupled vasopressin V2 receptor signaling network revealed by
quantitative phosphoproteomics. Molecular & cellular proteomics:
MCP. 2012;11(2):M111 014613.
85. Parker BL, Palmisano G, Edwards AV, et al. Quantitative N-linked
glycoproteomics of myocardial ischemia and reperfusion injury
reveals early remodeling in the extracellular environment.
Molecular & cellular proteomics: MCP. 2011;10(8):M110 006833.
86. Prudova A, Auf Dem Keller U, Butler GS, et al. Multiplex
N-terminome analysis of MMP-2 and MMP-9 substrate degradomes
by iTRAQ-TAILS quantitative proteomics. Molecular & cellular pro-
teomics: MCP. 2010;9(5):894–911.
87. Melo-Braga MN, Verano-Braga T, Leon IR, et al. Modulation of protein
phosphorylation, N-glycosylation and Lys-acetylation in grape (Vitis
vinifera) mesocarp and exocarp owing to Lobesia botrana infection.
Molecular & cellular proteomics: MCP. 2012;11(10):945–956.
88. Huang H, Haar Petersen M, Ibanez-Vea M, et al. Simultaneous
Enrichment of Cysteine-containing Peptides and Phosphopeptides
Using a Cysteine-specific Phosphonate Adaptable Tag (CysPAT) in
Combination with titanium dioxide (TiO2) Chromatography. Mol
Cell Proteomics. 2016;15(10):3282–3296.
89. Erickson BK, Jedrychowski MP, McAlister GC, et al. Evaluating multi-
plexed quantitative phosphopeptide analysis on a hybrid quadru-
pole mass filter/linear ion trap/orbitrap mass spectrometer. Anal
Chem. 2015;87(2):1241–1249.
90. Vaudel M, Burkhart JM, Radau S, et al. Integral quantification accu-
racy estimation for reporter ion-based quantitative proteomics
(iQuARI). J Proteome Res. 2012;11(10):5072–5080.
91. Karp NA, Huber W, Sadowski PG, et al. Addressing accuracy and
precision issues in iTRAQ quantitation. Molecular & cellular proteo-
mics: MCP. 2010;9(9):1885–1897.
92. Gillette MA, Carr SA. Quantitative analysis of peptides and proteins
in biomedicine by targeted mass spectrometry. Nature methods.
2013;10(1):28–34.
93. Rose CM, Merrill AE, Bailey DJ, et al. Neutron encoded labeling for
peptide identification. Anal Chem. 2013;85(10):5129–5137.
94. Potts GK, Voigt EA, Bailey DJ, et al. Neucode Labels for Multiplexed,
Absolute Protein Quantification. Anal Chem. 2016;88(6):3295–3303.
95. Boesch AW, Zhao Y, Landman AS, et al. A multiplexed assay to
detect antimicrobial peptides in biological fluids and cell secre-
tions. J Immunol Methods. 2013;397(1–2):71–76.
96. Treindl F, Ruprecht B, Beiter Y, et al. A bead-based western for
high-throughput cellular signal transduction analyses. Nat
Commun. 2016;7:12852.
97. Shi T, Song E, Nie S, et al. Advances in targeted proteomics and
applications to biomedical research. Proteomics. 2016;16(15-
16):2160–2182.
98. Bourmaud A, Gallien S, Domon B. Parallel reaction monitoring
using quadrupole-Orbitrap mass spectrometer: principle and appli-
cations. Proteomics. 2016;16(15-16):2146–2159.
99. Domon B, Gallien S. Recent advances in targeted proteomics for
clinical applications. Proteomics Clin Appl. 2015;9(3–4):423–431.
100. Mohammed Y, Domanski D, Jackson AM, et al. PeptidePicker: a scien-
tific workflow with web interface for selecting appropriate peptides
for targeted proteomics experiments. J Proteomics. 2014;106:151–161.
101. Se A, Dr M, Schilling B, et al. Design, implementation and multisite
evaluation of a system suitability protocol for the quantitative
assessment of instrument performance in liquid chromatography-
multiple reaction monitoring-MS (LC-MRM-MS). Molecular & cellular
proteomics: MCP. 2013;12(9):2623–2639.
• A standardization approach for MRM measurements is demon-
strated to yield precise inter- and intra-lab results at 11 locations.
102. Whiteaker JR, Halusa GN, Hoofnagle AN, et al. CPTAC Assay Portal: a
repository of targeted proteomic assays. Nature methods. 2014;11
(7):703–704.
103. Beck F, Geiger J, Gambaryan S, et al. Temporal quantitative phos-
phoproteomics of ADP stimulation reveals novel central nodes in
platelet activation and inhibition. Blood. 2017;129(2):e1–e12.
• Combination of TiO2 enrichment and PRM for targeted valida-
tion of phosphopeptide candidates obtained from discovery
phosphoproteomics. In total 23 phosphopeptides were ana-
lyzed in 7 different conditions using platelets obtained from
5 healthy donors, which would correspond to a total of 800
Western blots.
104. Beck F, Geiger J, Gambaryan S, et al. Time-resolved characterization
of cAMP/PKA-dependent signaling reveals that platelet inhibition is
a concerted process involving multiple signaling pathways. Blood.
2014;123(5):e1–e10.
105. Anderson NL, Anderson NG, Haines LR, et al. Mass spectrometric
quantitation of peptides and proteins using Stable Isotope
Standards and Capture by Anti-Peptide Antibodies (SISCAPA). J
Proteome Res. 2004;3(2):235–244.
106. Razavi M, Frick LE, LaMarr WA, et al. High-throughput SISCAPA
quantitation of peptides from human plasma digests by ultrafast,
liquid chromatography-free mass spectrometry. J Proteome Res.
2012;11(12):5642–5649.
107. Lin D, Alborn WE, Slebos RJ, et al. Comparison of protein immunopre-
cipitation-multiple reaction monitoring with ELISA for assay of biomar-
ker candidates in plasma. J Proteome Res. 2013;12(12):5996–6003.
108. Popp R, Basik M, Spatz A, et al. How iMALDI can improve clinical
diagnostics. Analyst. 2018;143(10):2197–2203.
109. Li H, Popp R, Frohlich B, et al. Peptide and Protein Quantification
Using Automated Immuno-MALDI (iMALDI). Journal of visualized
experiments: JoVE. 2017 126.
110. Jiang J, Parker CE, Hoadley KA, et al. Development of an immuno
tandem mass spectrometry (iMALDI) assay for EGFR diagnosis.
Proteomics Clin Appl. 2007;1(12):1651–1659.
111. Popp R, Li H, LeBlanc A, et al. Immuno-Matrix-Assisted Laser
Desorption/Ionization Assays for Quantifying AKT1 and AKT2 in
Breast and Colorectal Cancer Cell Lines and Tumors. Anal Chem.
2017;89(19):10592–10600.
112. Hoeppe S, Schreiber TD, Planatscher H, et al. Targeting peptide
termini, a novel immunoaffinity approach to reduce complexity in
mass spectrometric protein identification. Molecular & cellular pro-
teomics: MCP. 2011;10(2):M110 002857.
113. Marth JD, Grewal PK. Mammalian glycosylation in immunity. Nat
Rev Immunol. 2008;8(11):874–887.
114. Varinou L, Ramsauer K, Karaghiosoff M, et al. Phosphorylation of the
Stat1 transactivation domain is required for full-fledged IFN-gamma-
dependent innate immunity. Immunity. 2003;19(6):793–802.

115. Kaneko Y, Nimmerjahn F, Ravetch JV. Anti-inflammatory activity of
immunoglobulin G resulting from Fc sialylation. Science. 2006;313
(5787):670–673.
116. Viatour P, Merville MP, Bours V, et al. Phosphorylation of NF-kappaB
and IkappaB proteins: implications in cancer and inflammation.
Trends Biochem Sci. 2005;30(1):43–52.
117. Hakomori S. Tumor malignancy defined by aberrant glycosylation and
sphingo(glyco)lipid metabolism. Cancer Res. 1996;56(23):5309–5318.
118. Hwang RF, Yokoi K, Bucana CD, et al. Inhibition of platelet-derived
growth factor receptor phosphorylation by STI571 (Gleevec)
reduces growth and metastasis of human pancreatic carcinoma in
an orthotopic nude mouse model. Clinical cancer research: an
official journal of the American Association for Cancer Research.
2003;9(17):6534–6544.
119. Tepper K, Biernat J, Kumar S, et al. Oligomer formation of tau
protein hyperphosphorylated in cells. J Biol Chem. 2014;289
(49):34389–34407.
120. Grundke-Iqbal I, Iqbal K, Tung YC, et al. Abnormal phosphorylation
of the microtubule-associated protein tau (tau) in Alzheimer cytos-
keletal pathology. Proc Natl Acad Sci U S A. 1986;83(13):4913–4917.
121. Wang JZ, Grundke-Iqbal I, Iqbal K. Glycosylation of microtubule-
associated protein tau: an abnormal posttranslational modification
in Alzheimer’s disease. Nat Med. 1996;2(8):871–875.
122. Scheuner D, Mierde DV, Song B, et al. Control of mRNA translation
preserves endoplasmic reticulum function in beta cells and main-
tains glucose homeostasis. Nat Med. 2005;11(7):757–764.
123. Ortsäter H, Grankvist N, Honkanen RE, et al. Protein phosphatases
in pancreatic islets. J Endocrinol. 2014;221(3):R121–R144.
124. Meyerovitch J, Backer JM, Kahn CR. Hepatic phosphotyrosine phos-
phatase activity and its alterations in diabetic rats. J Clin Invest.
1989;84(3):976–983.
125. Brownlee M. Advanced protein glycosylation in diabetes and aging.
Annu Rev Med. 1995;46:223–234.
126. Blume-Jensen P, Hunter T. Oncogenic kinase signalling. Nature.
2001;411(6835):355–365.
127. Zanivan S, Meves A, Behrendt K, et al. In vivo SILAC-based proteo-
mics reveals phosphoproteome changes during mouse skin carci-
nogenesis. Cell Rep. 2013;3(2):552–566.
128. Huelsemann MF, Patz M, Beckmann L, et al. Hypoxia-induced p38
MAPK activation reduces Mcl-1 expression and facilitates sensitivity
towards BH3 mimetics in chronic lymphocytic leukemia. Leukemia.
2015;29(4):981–984.
129. Hakomori S. Glycosylation defining cancer malignancy: new wine in
an old bottle. Proc Natl Acad Sci U S A. 2002;99(16):10231–10233.
130. Decker RS, Decker ML, Kulikovskaya I, et al. Myosin-binding protein
C phosphorylation, myofibril structure, and contractile function
during low-flow ischemia. Circulation. 2005;111(7):906–912.
131. El-Armouche A, Pohlmann L, Schlossarek S, et al. Decreased phos-
phorylation levels of cardiac myosin-binding protein-C in human and
experimental heart failure. J Mol Cell Cardiol. 2007;43(2):223–229.
132. Schoutsen B, Blom JJ, Verdouw PD, et al. Calcium transport and
phospholamban in sarcoplasmic reticulum of ischemic myocar-
dium. J Mol Cell Cardiol. 1989;21(7):719–727.
133. Nc T, Ab H, Jr G, et al. Quantitative analysis of cell signaling and
drug action via mass spectrometry-based systems level phospho-
proteomics. Proteomics. 2009;9(6):1469–1487.
134. Kumar A, Baycin-Hizal D, Shiloach J, et al. Coupling enrichment and
proteomics methods for understanding and treating disease.
Proteomics Clin Appl. 2015;9(1–2):33–47.
135. Liu T, Qian WJ, Gritsenko MA, et al. Human plasma N-glycoproteome
analysis by immunoaffinity subtraction, hydrazide chemistry, and
mass spectrometry. J Proteome Res. 2005;4(6):2070–2080.
136. Jung K, Cho W, Regnier FE. Glycoproteomics of plasma based on
narrow selectivity lectin affinity chromatography. J Proteome Res.
2009;8(2):643–650.
137. Wang L, Li F, Sun W, et al. Concanavalin A-captured glycoproteins
in healthy human urine. Molecular & cellular proteomics: MCP.
2006;5(3):560–562.
EXPERT REVIEW OF PROTEOMICS 533
138. Lewandrowski U, Moebius J, Walter U, et al. Elucidation of
N-glycosylation sites on human platelet proteins: a glycoproteomic
approach. Molecular & cellular proteomics: MCP. 2006;5(2):226–233.
139. Block TM, Comunale MA, Lowman M, et al. Use of targeted glyco-
proteomics to identify serum glycoproteins that correlate with liver
cancer in woodchucks and humans. Proc Natl Acad Sci U S A.
2005;102(3):779–784.
140. Zhao J, Simeone DM, Heidt D, et al. Comparative serum glycopro-
teomics using lectin selected sialic acid glycoproteins with mass
spectrometric analysis: application to pancreatic cancer serum. J
Proteome Res. 2006;5(7):1792–1802.
141. Alley WR Jr., Vasseur JA, Goetz JA, et al. N-linked glycan structures
and their expressions change in the blood sera of ovarian cancer
patients. J Proteome Res. 2012;11(4):2282–2300.
142. Sonneveld ME, de Haas M, Koeleman C, et al. Patients with IgG1-
anti-red blood cell autoantibodies show aberrant Fc-glycosylation.
Sci Rep. 2017;7(1):8187.
143. Jin C, Kenny DT, Skoog EC, et al. Structural diversity of human
gastric mucin glycans. Molecular & cellular proteomics: MCP.
2017;16(5):743–758.
144. Hinneburg H, Korac P, Schirmeister F, et al. Unlocking Cancer
Glycomes from Histopathological Formalin-fixed and Paraffin-
embedded (FFPE) Tissue Microdissections. Molecular & cellular pro-
teomics: MCP. 2017;16(4):524–536.
145. Pagel O, Loroch S, Sickmann A, et al. Current strategies and find-
ings in clinically relevant post-translational modification-specific
proteomics. Expert Rev Proteomics. 2015;12(3):235–253.
146. von Stechow L, Francavilla C, Olsen JV. Recent findings and tech-
nological advances in phosphoproteomics for cells and tissues.
Expert Rev Proteomics. 2015;12(5):469–487.
147. Engholm-Keller K, Larsen MR. Technologies and challenges in large-
scale phosphoproteomics. Proteomics. 2013;13(6):910–931.
148. Riley NM, Coon JJ. Phosphoproteomics in the Age of Rapid and
Deep Proteome Profiling. Anal Chem. 2016;88(1):74–94.
149. Engholm-Keller K, Birck P, Storling J, et al. TiSH–a robust and
sensitive global phosphoproteomics strategy employing a combi-
nation of TiO2, SIMAC, and HILIC. J Proteomics. 2012;75(18):5749–
5761.
150. Geiger T, Cox J, Ostasiewicz P, et al. Super-SILAC mix for quantitative
proteomics of human tumor tissue. Nature methods. 2010;7(5):383–385.
151. Schweppe DK, Rigas JR, Gerber SA. Quantitative phosphoproteomic
profiling of human non-small cell lung cancer tumors. J Proteomics.
2013;91:286–296.
152. Dazert E, Colombi M, Boldanova T, et al. Quantitative proteomics and
phosphoproteomics on serial tumor biopsies from a sorafenib-treated
HCC patient. Proc Natl Acad Sci U S A. 2016;113(5):1381–1386.
153. Kennedy JJ, Yan P, Zhao L, et al. Immobilized Metal Affinity
Chromatography Coupled to Multiple Reaction Monitoring
Enables Reproducible Quantification of Phospho-signaling.
Molecular & cellular proteomics: MCP. 2016;15(2):726–739.
154. de Graaf EL, Kaplon J, Mohammed S, et al. Signal Transduction
Reaction Monitoring Deciphers Site-Specific PI3K-mTOR/MAPK
Pathway Dynamics in Oncogene-Induced Senescence. J Proteome
Res. 2015;14(7):2906–2914.
155. Loroch S, Zahedi RP, Sickmann A. Highly sensitive phosphoproteo-
mics by tailoring solid-phase extraction to electrostatic repulsion-
hydrophilic interaction chromatography. Anal Chem. 2015;87
(3):1596–1604.
156. Alpert AJ. Electrostatic repulsion hydrophilic interaction chromato-
graphy for isocratic separation of charged solutes and selective
isolation of phosphopeptides. Anal Chem. 2008;80(1):62–76.
157. Di Conza G, Trusso Cafarello S, Loroch S, et al. The mTOR and PP2A
Pathways Regulate PHD2 Phosphorylation to Fine-Tune HIF1alpha
Levels and Colorectal Cancer Cell Survival under Hypoxia. Cell Rep.
2017;18(7):1699–1712.
•• Absolute quantitation of phosphorylation levels in colorectal
cancer biopsies using a tailored ERLIC-PRM approach and
requiring only 32 µg of total tissue protein.
534 R. MNATSAKANYAN ET AL.
158. Chien KY, Liu HC, Goshe MB. Development and application of a
phosphoproteomic method using electrostatic repulsion-hydrophi-
lic interaction chromatography (ERLIC), IMAC, and LC-MS/MS ana-
lysis to study Marek’s Disease Virus infection. J Proteome Res.
2011;10(9):4041–4053.
159. Hao P, Ren Y, Dutta B, et al. Comparative evaluation of electrostatic
repulsion-hydrophilic interaction chromatography (ERLIC) and
high-pH reversed phase (Hp-RP) chromatography in profiling of
rat kidney proteome. J Proteomics. 2013;82:254–262.
160. Chen X, Wu D, Zhao Y, et al. Increasing phosphoproteome cover-
age and identification of phosphorylation motifs through combina-
tion of different HPLC fractionation methods. J Chromatogr B Analyt
Technol Biomed Life Sci. 2011;879(1):25–34.
161. Mertins P, Yang F, Liu T, et al. Ischemia in tumors induces early and
sustained phosphorylation changes in stress kinase pathways but
does not affect global protein levels. Molecular & cellular proteo-
mics: MCP. 2014;13(7):1690–1704.
162. Gajadhar AS, Johnson H, Slebos RJ, et al. Phosphotyrosine signaling
analysis in human tumors is confounded by systemic ischemia-
driven artifacts and intra-specimen heterogeneity. Cancer Res.
2015;75(7):1495–1503.
163. Petrera A, Lai ZW, Schilling O. Carboxyterminal Protein Processing
in Health and Disease: key Actors and Emerging Technologies. J
Proteome Res. 2014;13(11):4497–4504.
164. Mossmann D, Vögtle FN, Taskin Asli A, et al. Amyloid-β Peptide
Induces Mitochondrial Dysfunction by Inhibition of Preprotein
Maturation. Cell Metab. 2014;20(4):662–669.
165. McCarty SM, Percival SL. Proteases and Delayed Wound Healing.
Advances in wound care. 2013;2(8):438–447.
166. Csiszar A, Kutay B, Wirth S, et al. Interleukin-like epithelial-to-
mesenchymal transition inducer activity is controlled by proteolytic
processing and plasminogen-urokinase plasminogen activator
receptor system-regulated secretion during breast cancer progres-
sion. Breast cancer research: BCR. 2014;16(5):433.
167. Gevaert K, Goethals M, Martens L, et al. Exploring proteomes and
analyzing protein processing by mass spectrometric identification
of sorted N-terminal peptides. Nat Biotechnol. 2003;21(5):566–569.
168. Kleifeld O, Doucet A, Auf Dem Keller U, et al. Isotopic labeling of
terminal amines in complex samples identifies protein N-termini
and protease cleavage products. Nat Biotech. 2010;28(3):281–288.
169. Venne AS, Vogtle FN, Meisinger C, et al. Novel highly sensitive,
specific, and straightforward strategy for comprehensive N-terminal
proteomics reveals unknown substrates of the mitochondrial pepti-
dase Icp55. J Proteome Res. 2013;12(9):3823–3830.
170. Mahrus S, Trinidad JC, Barkan DT, et al. Global sequencing of
proteolytic cleavage sites in apoptosis by specific labeling of pro-
tein N termini. Cell. 2008;134(5):866–876.
171. Dix MM, Simon GM, Cravatt BF. Global identification of caspase
substrates using PROTOMAP (protein topography and migration
analysis platform). Methods in molecular biology. 2014;1133:61–70.
172. Segura V, Garin-Muga A, Guruceaga E, et al. Progress and pitfalls in
finding the ‘missing proteins’ from the human proteome map.
Expert Rev Proteomics. 2017;14(1):9–14.
173. Baker MS, Ahn SB, Mohamedali A, et al. Accelerating the search for
the missing proteins in the human proteome. Nat Commun.
2017;8:14271.
174. Lange PF, Huesgen PF, Nguyen K, et al. Annotating N Termini for
the Human Proteome Project: N Termini and Nα-Acetylation Status
Differentiate Stable Cleaved Protein Species from Degradation
Remnants in the Human Erythrocyte Proteome. J Proteome Res.
2014;13(4):2028–2044.
175. Eckhard U, Marino G, Abbey SR, et al. The Human Dental Pulp
Proteome and N-Terminome: Levering the Unexplored Potential
of Semitryptic Peptides Enriched by TAILS to Identify Missing
Proteins in the Human Proteome Project in Underexplored
Tissues. J Proteome Res. 2015;14(9):3568–3582.
176. Klein T, Fung S-Y, Renner F, et al. The paracaspase MALT1 cleaves
HOIL1 reducing linear ubiquitination by LUBAC to dampen lym-
phocyte NF-κB signalling. 2015;6:8777.
177. Wildes D, Wells JA. Sampling the N-terminal proteome of human
blood. Proc Natl Acad Sci U S A. 2010;107(10):4561–4566.
178. Wiita AP, Hsu GW, Lu CM, et al. Circulating proteolytic signatures of
chemotherapy-induced cell death in humans discovered by
N-terminal labeling. Proc Natl Acad Sci U S A. 2014;111(21):7594–7599..
• N-terminomics study identifying potential proteolytic markers
of chemotherapy-induced cell death, which were also vali-
dated by MRM.
179. Solari FA, Mattheij NJ, Burkhart JM, et al. Combined Quantification
of the Global Proteome, Phosphoproteome, and Proteolytic
Cleavage to Characterize Altered Platelet Functions in the Human
Scott Syndrome. Molecular & cellular proteomics: MCP. 2016;15
(10):3154–3169.
180. Shema G, Nguyen MTN, Solari FA, et al. Simple, scalable, and
ultrasensitive tip-based identification of protease substrates. Mol
Cell Proteomics. 2018;17(4):826–834.
181. Chung HS, Wang SB, Venkatraman V, et al. Cysteine oxidative
posttranslational modifications: emerging regulation in the cardio-
vascular system. Circ Res. 2013;112(2):382–392.
182. Kim HJ, Ha S, Lee HY, et al. ROSics: chemistry and proteomics of
cysteine modifications in redox biology. Mass Spectrom Rev.
2015;34(2):184–208.
183. Gould N, Doulias PT, Tenopoulou M, et al. Regulation of protein
function and signaling by reversible cysteine S-nitrosylation. J Biol
Chem. 2013;288(37):26473–26479.
184. Waszczak C, Akter S, Jacques S, et al. Oxidative post-translational
modifications of cysteine residues in plant signal transduction. J
Exp Bot. 2015;66(10):2923–2934.
185. Nietzel T, Mostertz J, Hochgrafe F, et al. Redox regulation of mito-
chondrial proteins and proteomes by cysteine thiol switches.
Mitochondrion. 2016;33:72–83.
186. Sun J, Protein ME. S-nitrosylation and cardioprotection. Circ Res.
2010;106(2):285–296.
187. Watanabe Y, Cohen RA, Matsui R. Redox Regulation of Ischemic
Angiogenesis- Another Aspect of Reactive Oxygen Species. Circ J.
2016;80(6):1278–1284.
188. Wang SB, Foster DB, Rucker J, et al. Redox regulation of mitochon-
drial ATP synthase: implications for cardiac resynchronization ther-
apy. Circ Res. 2011;109(7):750–757.
189. Yuan K, Liu Y, Chen HN, et al. Thiol-based redox proteomics in
cancer research. Proteomics. 2015;15(2–3):287–299.
190. Murphy E, Kohr M, Menazza S, et al. Signaling by S-nitrosylation in
the heart. J Mol Cell Cardiol. 2014;73:18–25.
191. Butterfield DA, Dalle-Donne I. Redox proteomics: from protein
modifications to cellular dysfunction and disease. Mass Spectrom
Rev. 2014;33(1):1–6.
192. Raju K, Doulias PT, Evans P, et al. Regulation of brain glutamate
metabolism by nitric oxide and S-nitrosylation. Sci Signal. 2015;8
(384):ra68.
193. Duan J, Gaffrey MJ, Qian WJ. Quantitative proteomic characteriza-
tion of redox-dependent post-translational modifications on pro-
tein cysteines. Mol Biosyst. 2017;13(5):816–829.
194. McDonagh B. Detection of ROS Induced Proteomic Signatures by
Mass Spectrometry. Front Physiol. 2017;8:470.
195. Nagumo K, Tanaka M, Chuang VT, et al. Cys34-cysteinylated human
serum albumin is a sensitive plasma marker in oxidative stress-
related chronic diseases. PLoS One. 2014;9(1):e85216.
196. Delobel J, Prudent M, Crettaz D, et al. Cysteine redox proteomics of
the hemoglobin-depleted cytosolic fraction of stored red blood
cells. Proteomics Clin Appl. 2016;10(8):883–893.
197. Sonego G, Abonnenc M, Tissot JD, et al. Redox Proteomics and
Platelet Activation: Understanding the Redox Proteome to Improve
Platelet Quality for Transfusion. Int J Mol Sci. 2017;18:2.
198. Wojdyla K, Williamson J, Roepstorff P, et al. SNO/SOH TMT strategy
for combinatorial analysis of reversible cysteine oxidations. J
Proteomics. 2015;113:415–434.
199. Guo J, Gaffrey MJ, Su D, et al. Resin-assisted enrichment of thiols as
a general strategy for proteomic profiling of cysteine-based rever-
sible modifications. Nat Protoc. 2014;9(1):64–75.

200. García-Santamarina S, Boronat S, Domènech A, et al. Monitoring in
vivo reversible cysteine oxidation in proteins using ICAT and mass
spectrometry. Nat Protoc. 2014;9(5):1131–1145.
201. Pan KT, Chen YY, Pu TH, et al. Mass spectrometry-based quantita-
tive proteomics for dissecting multiplexed redox cysteine modifica-
tions in nitric oxide-protected cardiomyocyte under hypoxia.
Antioxid Redox Signal. 2014;20(9):1365–1381.
202. Forrester MT, Thompson JW, Foster MW, et al. Proteomic analysis of
S-nitrosylation and denitrosylation by resin-assisted capture. Nat
Biotechnol. 2009;27(6):557–559.
203. Wolhuter K, Whitwell HJ, Switzer CH, et al. Evidence against Stable
Protein S-Nitrosylation as a Widespread Mechanism of Post-transla-
tional Regulation. Mol Cell. 2018;69(3):438–450 e435.
204. Chouchani ET, James AM, Methner C, et al. Identification and
quantification of protein S-nitrosation by nitrite in the mouse
heart during ischemia. J Biol Chem. 2017;292(35):14486–14495.
205. Ibanez-Vea M, Huang H, Martinez De Morentin X, et al. Characterization
of Macrophage Endogenous S-Nitrosoproteome Using a Cysteine-
Specific Phosphonate Adaptable Tag in Combination with TiO2
Chromatography. J Proteome Res. 2018;17(3):1172–1182.
206. Yang J, Gupta V, Tallman KA, et al. Global, in situ, site-specific
analysis of protein S-sulfenylation. Nat Protoc. 2015;10(7):1022–1037.
207. Poole LB, Karplus PA, Claiborne A. Protein sulfenic acids in redox
signaling. Annu Rev Pharmacol Toxicol. 2004;44:325–347.
208. Yang J, Gupta V, Carroll KS, et al. Site-specific mapping and quan-
tification of protein S-sulphenylation in cells. Nat Commun.
2014;5:4776.
209. Li R, Klockenbusch C, Lin L, et al. Quantitative Protein Sulfenic Acid
Analysis Identifies Platelet Releasate-Induced Activation of Integrin
beta2 on Monocytes via NADPH Oxidase. J Proteome Res. 2016;15
(12):4221–4233.
210. Bechtel TJ, Weerapana E. From structure to redox: the diverse func-
tional roles of disulfides and implications in disease. Proteomics.
2017;17:6.
211. Kang DH, Lee DJ, Lee KW, et al. Peroxiredoxin II is an essential
antioxidant enzyme that prevents the oxidative inactivation of
VEGF receptor-2 in vascular endothelial cells. Mol Cell. 2011;44
(4):545–558.
212. Lu S, Fan SB, Yang B, et al. Mapping native disulfide bonds at a
proteome scale. Nat Methods. 2015;12(4):329–331.
213. Murdoch CE, Shuler M, Haeussler DJ, et al. Glutaredoxin-1 up-
regulation induces soluble vascular endothelial growth factor
receptor 1, attenuating post-ischemia limb revascularization. J Biol
Chem. 2014;289(12):8633–8644.
214. Duan J, Kodali VK, Gaffrey MJ, et al. Quantitative Profiling of Protein
S-Glutathionylation Reveals Redox-Dependent Regulation of
Macrophage Function during Nanoparticle-Induced Oxidative
Stress. ACS Nano. 2016;10(1):524–538.
215. McGarry DJ, Chen W, Chakravarty P, et al. Proteome-wide identifi-
cation and quantification of S-glutathionylation targets in mouse
liver. Biochem J. 2015;469(1):25–32.
216. Chen R, Mias GI, Li-Pook-Than J, et al. Personal omics profiling
reveals dynamic molecular and medical phenotypes. Cell.
2012;148(6):1293–1307.
217. Parker CE, Borchers CH. Mass spectrometry based biomarker dis-
covery, verification, and validation–quality assurance and control of
protein biomarker assays. Mol Oncol. 2014;8(4):840–858..
• Review article about biomarker discovery with MS.
218. Geyer PE, Holdt LM, Teupser D, et al. Revisiting biomarker discovery
by plasma proteomics. Mol Syst Biol. 2017;13(9):942.
219. Hernandez B, Parnell A, Pennington SR. Why have so few proteomic
biomarkers “survived” validation? (Sample size and independent vali-
dation considerations). Proteomics. 2014;14(13-14):1587–1592.
EXPERT REVIEW OF PROTEOMICS 535
220. Christin C, Hoefsloot HC, Smilde AK, et al. A critical assessment of
feature selection methods for biomarker discovery in clinical pro-
teomics. Molecular & cellular proteomics: MCP. 2013;12(1):263–276.
221. Skates SJ, Gillette MA, LaBaer J, et al. Statistical design for biospeci-
men cohort size in proteomics-based biomarker discovery and
verification studies. J Proteome Res. 2013;12(12):5383–5394.
222. Diz AP, Carvajal-Rodriguez A, Skibinski DO. Multiple hypothesis
testing in proteomics: a strategy for experimental work. Molecular
& cellular proteomics: MCP. 2011;10(3):M110 004374.
223. Loroch S, Trabold K, Gambaryan S, et al. Alterations of the platelet
proteome in type I Glanzmann thrombasthenia caused by different
homozygous delG frameshift mutations in ITGA2B. Thromb
Haemost. 2017;117(3):556–569.
224. Kammers K, Cole RN, Tiengwe C, et al. Detecting Significant
Changes in Protein Abundance. EuPA open proteomics.
2015;7:11–19.
225. Datta S, Satten GA, Benos DJ, et al. An empirical bayes adjustment
to increase the sensitivity of detecting differentially expressed
genes in microarray experiments. Bioinformatics (Oxford, England).
2004;20(2):235–242.
226. Wettenhall JM, Smyth GK. limmaGUI: a graphical user interface for
linear modeling of microarray data. Bioinformatics (Oxford,
England). 2004;20(18):3705–3706.
227. Ting L, Cowley MJ, Hoon SL, et al. Normalization and statistical
analysis of quantitative proteomics data generated by meta-
bolic labeling. Molecular & cellular proteomics: MCP. 2009;8
(10):2227–2242.
228. Ritchie ME, Phipson B, Wu D, et al. limma powers differential
expression analyses for RNA-sequencing and microarray studies.
Nucleic Acids Res. 2015;43(7):e47.
229. Boja ES, Rodriguez H. Proteogenomic convergence for under-
standing cancer pathways and networks. Clin Proteomics.
2014;11(1):22.
230. Alfaro JA, Sinha A, Kislinger T, et al. Onco-proteogenomics: cancer
proteomics joins forces with genomics. Nature methods. 2014;11
(11):1107–1113.
231. Schwamborn K, Caprioli RM. Molecular imaging by mass spectro-
metry–looking beyond classical histology. Nat Rev Cancer. 2010;10
(9):639–646.
232. Domanski D, Murphy LC, Borchers CH. Assay development for the
determination of phosphorylation stoichiometry using multiple
reaction monitoring methods with and without phosphatase treat-
ment: application to breast cancer signaling pathways. Anal Chem.
2010;82(13):5610–5620.
233. The M, Tasnim A, Kall L. How to talk about protein-level false
discovery rates in shotgun proteomics. Proteomics. 2016;16
(18):2461–2469.
234. Zahedi RP. Joining forces: studying multiple post-translational
modifications to understand dynamic disease mechanisms. Expert
Rev Proteomics. 2016;13(12):1055–1057.
235. Diehl HC, Beine B, Elm J, et al. The challenge of on-tissue diges-
tion for MALDI MSI- a comparison of different protocols to
improve imaging experiments. Anal Bioanal Chem. 2015;407
(8):2223–2243.
236. Kopczynski D, Coman C, Zahedi RP, et al. Multi-OMICS: a critical
technical perspective on integrative lipidomics approaches.
Biochim Biophys Acta. 2017;1862(8):808–811.
237. Conrads TP, Petricoin EF 3rd. The Obama Administration’s Cancer
Moonshot: a Call for Proteomics. Clinical cancer research: an official
journal of the American Association for Cancer Research. 2016;22
(18):4556–4558.
238. Aelion CM, Airhihenbuwa CO, Alemagno S, et al. The US Cancer
Moonshot initiative. Lancet. Oncol. 2016;17(5):e178–180.