Hindawi Publishing Corporation
International Journal of Proteomics
Volume 2013, Article ID 756039, 13 pages
Issuesand Applications in Label-FreeQuantitativeMass
Xianyin Lai,1LianshuiWang,2and Frank A. Witzmann1
1Department of Cellular & Integrative Physiology, Biotechnology Research & Training Center, Indiana University School of Medicine,
Indianapolis, IN 46202, USA
2School of Informatics and Computing, Indiana University, Bloomington, IN 47408, USA
Correspondence should be addressed to Xianyin Lai; firstname.lastname@example.org
Received 2 October 2012; Revised 17 October 2012; Accepted 31 October 2012
Academic Editor: Bomie Han
Copyright © 2013 Xianyin Lai et al. is is an open access article distributed under the Creative Commons Attribution License,
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
To address the challenges associated with differential expression proteomics, label-free mass spectrometric protein quanti�cation
methods have been developed as alternatives to array-based, gel-based, and stable isotope tag or label-based approaches. In
this paper, we focus on the issues associated with label-free methods that rely on quantitation based on peptide ion peak area
measurement. ese issues include chromatographic alignment, peptide quali�cation for quantitation, and normalization. In
platform, that overcome these difficulties and enable comprehensive, accurate, and reproducible protein quantitation in highly
complex protein mixtures from experiments with many sample groups. As examples of the utility of this approach, we present a
variety of cases where the platform was applied successfully to assess differential protein expression or abundance in body �uids,
in vitro nanotoxicology models, tissue proteomics in genetic knock-in mice, and cell membrane proteomics.
Protein quanti�cation for differential expression analy-
sis or expression pro�ling represents the most challeng-
ing aspect in proteomics technology. is task is typi-
cally carried out through array-based , two-dimensional-
electrophoretic (2-DE-) based  or mass-spectrometry-
(MS-) based approaches [3, 4]. MS-based approaches are
normally referred to as “bottom-up” rather than “top-down,”
because the top-down approach has not yet reached its
full potential. In bottom-up quantitative approaches, com-
plex protein mixtures are digested enzymatically, peptides
from each protein are separated by liquid chromatogra-
phy (LC) and detected by MS, and protein quanti�cation
is completed at the peptide level and then combined to
calculate a summarized value for the protein from which
they come. Early in the evolution of quantitative MS-based
proteomic technology, stable isotope labeling methods were
developed [5–8]. Following that premise, many new label-
based methods have arisen. However, all of these suffer from
several limitations: (i) additional sample processing steps
in the experimental work�ow, (ii) high cost of the labeling
reagents, (iii) variable labeling efficiency, and (iv) difficulty
in analyzing low-abundance peptides in multiple samples,
especially when numerous experimental groups are studied
Following the development of label-based approaches,
label-free approaches emerged to overcome the drawbacks
associated with label-based approaches mentioned above.
To illustrate the principles of label-free quantitative mass
example. Along with the elution (retention time) of a peptide
from an LC column (Figure 1(a)), the peptide peak height
(intensity) and mass-to-charge (m/z) are recorded in a mass
spectrum (Figure 1(b)), the peptide ion may be chosen as
a precursor ion at some speci�c time point according to
speci�c parameter settings to generate an MS/MS spectrum
where intensity and m/z of the product ions in the MS/MS
spectrum are recorded (Figure 1(c)). e peptide ion may
be chosen as a precursor ion multiple times (Figures 2(a)–
2(d)). From the LC chromatogram, MS spectrum, and the
MS/MS spectrum, all information associated with peptide
2 International Journal of Proteomics
51.552 52.553 53.5
T: ITMS + c ESI full ms [300–20000]
St 0 10P1 #4036 RT: 51.45 AV: 1 NL: 7.82E3
837.92 1080.6 1219.061364.421548.351736.261888.96
20100203 St 0 10P1 #4037 RT: 51.46 AV: 1 NL: 1.31E2
T: ITMS + c ESI d w full ms2 786.09@cid35 [205–20000]
time points, forming a peptide peak. e scan time points for (b) and (c) are labeled in red. (b) At scan #4036, a full MS scan is performed,
and all the peptide ions including ion m/z 786.09 are recorded. (c) At scan #4037, an MS/MS scan is performed, and the ion m/z 786.09 is
chosen as a precursor ion to generate product ions, providing peptide fragmentation information for peptide identi�cation.
abundance, such as the peptide peak intensity (height or
area of a peak), the peptide precursor ion peak height, and
the peak height of product ions, can be extracted. Using
such information individually or combinatorially, numer-
ous label-free methods have been developed, including two
extensively applied but fundamentally different strategies:
quantitation based on spectral counting  and peptide ion
peak area .
Spectral counting estimates protein abundance by count-
protein. In the example shown in Figures 1 and 2, ten
they were acquired either before or aer the actual peptide
elution peak. Data acquisition in this manner is most typical
when the dynamic exclusion mode is applied to identify
substantially more peptides. In this case, it is erroneous
to hypothesize that peptide abundance is correlated with
the number of spectra. Although spectral counting has
been applied to study differential protein composition in
complex biological samples, when low-end MS and limited
LC separation (such as one-dimensional LC) are applied,
protein quanti�cation using spectral counting is challenging
International Journal of Proteomics3
51.5 52 52.55353.5
400 600800100012001400 160018002000
20100203 St 0 10P1 #4039 RT: 51.48 AV: 1 NL: 1.08E2
T: ITMS + c ESI d w full ms2 785.56@cid35 [205–20000]
400600800 10001200 140016001800 2000
20100203 St 0 10P1 #4243 RT: 53.61 AV: 1 NL: 7.24E2
T: ITMS + c ESI d w full ms2 786@cid35 [205–20000]
F 2: Multiple MS/MS scans of a single peptide. (a) e peptide is eluted from an LC column. e scan time points for the last MS/MS
scan before the peak and the �rst two MS/MS scans aer the peak are label in red. (b) At scan �4039, an MS/MS scan of the precursor ion
m/z 786.09 is performed almost one-half minute before the apex of the peptide peak. (c) At scan �4243, the �rst repeat MS/MS scan of the
precursor ion m/z 786.09 appears two minutes later and is performed almost one-and-a-half minutes aer the apex of the peptide peak. (d)
All the MS/MS scans of the peptide are listed. Most of them appear far behind the peptide peak.
dramatically and adversely affects spectral acquisition, and
(ii) coeluting peptides compete for MS/MS analysis and
Peak area has been applied extensively in the quanti�ca-
tion of small molecule compounds [12–14] and is the most
reliable measurement for quanti�cation. �ecause peptides
perform similarly to small molecule compounds in liquid
chromatography-tandem mass spectrometry (LC-MS/MS)
able method for peptide and protein quanti�cation. �umer-
ous soware packages for label-free quanti�cation of LC-
MS/MS-derived data based on peptide peak area have been
developed and applied in proteomics research. However,
peptide and protein quanti�cation is not as straightforward
4International Journal of Proteomics
05 1015 2025 3035 40 4555
Relative abundanceRT: 0–59.98
1 2 3 4 5 6 7 8 9 10
Intensity (peak area)
449.49 (RT: 37.39)
1 2 3 4 5 6 7 8 9 10
Intensity (peak area)
786.24 (RT: 51.98)
1 2 3 4 5 6 7 8 9 10
Intensity (peak area)
433.23 (RT: 54.82)
F 3: A typical LC-MS chromatogram of three standard peptides. (a) 4 pmol of Angiotensin III, Fibrinopeptide B, and Angiotensin I
were mixed and analyzed by LC-MS/MS 10 times. (b–d) Intensity of Angiotensin I, Fibrinopeptide B, and Angiotensin I at mean m/z 449.49,
show that identical amounts of peptides generate totally different peak area, and the fold difference is as large as 7.45 (6.63𝐸𝐸?????.??𝐸𝐸??6).
as quanti�cation of small molecule compounds, because
in a single run as compared to small molecule analysis,
where the focus is only on one or a few small molecule
compounds. To accurately quantify peptides and proteins,
several issues must be addressed. is paper will present
these issues, describe ways in which to address them, and
provide examples where comprehensive quantitation has
been carried out successfully.
2.1. LC-MS Alignment. To extract the peptide peak area,
two basic parameters, m/z and retention time, must be
determined. Typically, the m/z value is measured repro-
ducibly in low resolution mass spectrometers such as the
LTQ and extremely reproducibly in high resolution mass
spectrometers such as LTQ-Orbitrap. In general, we have
consistently observed that the retention time of each peptide
in LC normally varies by roughly 3min . When dynamic
exclusion is enabled, the MS/MS scan of a peptide does not
normally occur at the peptide elution peak. For example, a
typical LC-MS chromatogram of three standard peptides is
shown in Figure 3(a). e m/z value variations of three stan-
dard peptides from 10 injections are 0.36, 0.89, and 0.26Da
for Angiotensin I, Fibrinopeptide B, and Angiotensin I at
mean m/z 449.49, 786.24, and 433.23Da, respectively. eir
retention time 38.76, 53.26, and 56.08min, respectively. As
shown in Figures 1 and 2, the MS/MS spectra appear at least
0.5min before the apex of the elution peak and 1.6min aer
of peptide elution peaks can be a signi�cant issue in certain
sary. For instance, when high resolution mass spectrometry
is applied, peptides are completely separated in the m/z
dimension, forming individual peaks. If the m/z, retention
time, and ID of each peptide peak are extracted from an
individual sample, the comparison of multiple samples does
not necessarily require alignment. If comparison of multiple
when low resolution mass spectrometry is applied, peptides
cannot be separated completely in the m/z dimension, and
they rarely form individual peaks. e apex of the peptide
elution peak is thus difficult to determine. Using MS/MS
from individual samples is doable, but alignment of multiple
samples enables more accurate estimation of peptide elution
retention time, leading to more accurate quanti�cation.
A second issue relates to when it is best to perform the
alignment. Should it be performed before or aer peptide
identi�cation� In the development of label-free quanti�ca-
tion techniques, researchers initially followed a work�ow
similar to that used in 2-DE image analysis, for example,
as two-dimensional images . In 2-DE gel image anal-
ysis, information regarding protein identity is unavailable.
International Journal of Proteomics5
Consequently, using resolved proteins as landmarks to align
images is the best and only option, although it is likely
that protein “spots” arising from dissimilar proteins may
be aligned as the same protein. However, in an LC-MS/MS
analysis, information regarding peptide identity can be
obtained easily so that every identi�ed peptide can serve
as a landmark, making thousands of landmarks available.
erefore, to improve quantitative accuracy, alignment aer
peptide identi�cation should be carried out.
e third issue regarding alignment relates to how the
alignment should be carried out. Currently, �ve approaches
typically are used to align the retention time of peptides
across multiple injections. e most popular alignment
is similar to that used in 2-DE image analysis, such as
OpenMS/TOPP , MapQuant , Msight , msIn-
spect , and MZmine . Enhancement of the 2-DE
mode alignment is accomplished by adding peptide ID
information to improve alignment accuracy, in programs
such as SuperHirn  and PEPPeR . A third approach
is to align the “Base Peak” chromatograms, as performed
by SIEVE from ermo Scienti�c, and a fourth approach is
an alignment using MS/MS scan times (identi�ed peptides),
such as that found in IDEAL-Q .
We have developed an alternative approach to those
dimensional alignment to determine peptide retention
time using a clustering method . Approaches 1 and 2
perform well only for high resolution data. However, for low
resolution data, these approaches are incapable of generating
analyzable patterns. SIEVE was designed for both high and
low resolution data, aligning the Base Peak chromatogram
via Recursive Base-Peak Framing to generate a unique
“frame” for each group of peaks within a speci�ed m/z and
retention time range. However, when low-resolution data are
analyzed, SIEVE performs poorly for the following reasons.
First, this approach ignores the fact that the retention time
variation of each peptide is not consistent with that of the
Base Peak. Second, it is difficult for SIEVE to align two
dissimilar distributions and align to a void. Finally, the
“frame” (a rectangular region containing x and y coordinates
in the m/z versus retention time plane) de�ned by SIEVE
is too subjective. Because SIEVE generates “frames” with
only a one-size time-width, its application in complex
when the same sample is injected twice, speci�c peptide
peaks will have variable widths. IDEAL-Q uses a linear
regression model to determine the retention time of each
peptide. However, it ignores the fact that most peptides have
multiple MS/MS spectra and appear at different time points,
generating one of the six elution patterns observed by Lai
et al. . Only IdentiQuantXL considers all these issues
and performs an individual, three-dimensional (m/z, RT,
and MS/MS ID) alignment to determine peptide retention
time using a clustering method, providing the most accurate
retention time determination and widest application.
In IdentiQuantXL, all MS/MS scan times of a peptide
are collected, a scan time distribution pattern is generated, a
cluster is applied when the scan time range is over 3min, and
the weighted mean scan time is determined as the peptide
retention time. In this method, the alignment is processed
using the MS/MS scan times to ensure that only the same
peptides are aligned across injections. Because most peptides
have multiple MS/MS spectra and appear at different time
points, the clustering approach is better able to determine
peptide peak retention time accurately.
���� �������� �������� ��� ������� �������������� When
peptides in a complex sample are analyzed by LC-MS/MS,
their behavior in each component of the analytical platform
is diverse, a phenomenon that largely has been ignored.
First, peptides are unlikely to be identi�ed in every injection.
Because of proteomic sample complexity, tens of thousands
dant peptides are likely to be selected for MS/MS analysis in
each injection. Frequently, an individual peptide’s intensity
may be higher or lower than other coeluted peptides in
various injections, leading to selection for fragmentation and
subsequent identi�cation of different peptides in different
of whole rat kidney lysates were analyzed, generating 7,361
peptides , 25.6% of the peptides (1,883) were identi�ed in
1 injection and only 0.8% of the peptides (61) were identi�ed
in all 35 injections. If a peptide is not analyzed by MS/MS
in an injection, that does not mean it is absent from this
in all other injections in that experiment. However, if the
once-identi�ed peptide is a misidenti�ed peptide, extraction
of this peptide from all other injections leads to erroneous
quanti�cation. e use of peptides with low identi�cation
frequency for quantitation constitutes a risk of error, but
no clear cutoff exists to exclude such peptides. e higher
a peptide’s identi�cation frequency, the greater con�dence
one can have that the peptide actually exists in all the
Second, peptides have multiple elution patterns across
single chromatographic condition, for example, one speci�c
column with speci�c mobile phases and gradient, cannot be
optimal for each of the thousands of peptides in a single
sample injection. Six elution patterns have been summarized
: (i) the peptide is completely eluted at one time point and
very consistent in different sample injections analyzed; (ii) it
is completely eluted at one time point but not consistently
in all injections; (iii) its abundance is very high in some
injections and cannot be bound completely to the column
matrix, so some amount of the peptide is eluted before
the main peak; (iv) it is bound to the column matrix too
tightly, eluting mainly at �rst, with a remainder eluting at
higher organic concentration, aer the main peak; (v) it has
an overall elution pattern combining patterns (iii) and (iv);
and (vi) the peptide has poor chromatographic performance
and is eluted at indistinct time points across all injections.
Consequently, some peptides having an elution pattern like
(vi) do not form a peak, precluding quantitation.
ird, the peak area coefficients of variation (CVs) of
peptides from two injections of one sample are highly vari-
able. Due to the fact that chromatographic conditions cannot
6 International Journal of Proteomics
be optimal for all of the thousands of peptides detected
in a single injection, some peptides have a very small CV,
while other peptides have a very large CV. For instance, a
comparison of the intensities of 4pmol of the mass stan-
dards Angiotensin III, Fibrinopeptide B, and Angiotensin
I revealed that their CVs were 15.0%, 11.4%, and 63.4%,
respectively  (Figures 3(b)–3(d)).
Finally, when multiple peptides are used to calculate a
protein’s abundance in a comparative study, individual pep-
tides oen exhibit fold changes that are different from other
peptides from the same protein. Several explanations for this
phenomenon are possible: (i) some peptides have greater
variation than others under the same chromatographic con-
ditions; (ii) posttranslational modi�cation (PTM) variably
affects the relative abundance of unmodi�ed peptides; (iii)
peptide sharing among diverse proteins causes inconsistent
effects on some peptides; (iv) carry-over of some peptides
causes random abundance changes, and (v) differential reg-
ulation of isoforms, misidenti�cation, and misquanti�cation
also may occur [9, 25].
From the discussion above, a fundamental concept
emerged that not every peptide can be used to quantify a
corresponding protein. �nquanti�able peptides are caused
by limitations in mass spectrometry, uniform (not individu-
ally optimized) chromatographic conditions, PTMs, peptide
sharing, and so forth. ese issues normally cannot be
addressed prior to data generation, and, currently, there are
no proper means to correct them. One practical way is to
eliminate the unquanti�able peptides before they are used
for protein quanti�cation. Lai et al. have developed multiple
�lters (peptide frequency, retention time, intensity CV, and
correlation) to speci�cally address this issue and enhance the
accuracy of protein quanti�cation .
2.3. Normalization. e aim of normalization is to remove
sion data obtained from laser or optical microarrays and
then applied to label-free quanti�cation in high-throughput
proteomics . Numerous normalization algorithms have
been developed and applied. Among them, global normal-
ization (central tendency), linear regression, local regression,
and quantile techniques are the commonly used methods for
When global normalization is applied, it is assumed that
every peptide produces the same peak area at the same
abundance. However, this is not the case in proteomic
analysis. For example, when three standard peptides were
injected in identical amounts (Figures 3(b)–3(d)), LC-MS
generated totally different peak areas, and the fold difference
was as large as 7.45 (6.63𝐸𝐸 ? ????.??𝐸𝐸 ? ?6).
Linear regression normalization is applied to remove
. However, in a typical experiment where multiple sam-
ples are compared, a blank is normally introduced between
samples to wash the column and eliminate carry-over.
Furthermore, even if minor carry-over occurs, it typically
involves just a few peptides, not all of them .
Local regression normalization is used to eliminate sys-
tematic bias generated from the effects of ion suppression
on measured peptide abundances, or on measured peptide
abundances approaching detector saturation or background
. Again, these situations only happen to a few peptides,
not every peptide. It is difficult to determine which peptides
are suitable for this normalization.
Quantile regression normalization is employed under
the assumption that the distribution of peptide abundances
in different samples is expected to be similar and can be
accounted for by adjusting these distributions . However,
in an actual experiment, peptide abundances in samples
from different groups or even the same group differ because
of biological variability or experimental manipulation or
Normalization using singly and multiply spiked internal
standards has been reported [24, 28]. is strategy is not
new and has been applied in pharmaceutical analyses for
many years. Isotopic labeled standards, the multiple reaction
monitoring (MRM) technique, and highly optimized LC
conditions are applied in this strategy to obtain speci�c
internal standard peaks. However, tens of thousands of
peptides may exist in a single sample, and their peak area
response with their abundance may be signi�cantly different.
Several internal standards are incapable of representing all
the peptides and are thus of marginal value.
e systematic biases considered by some researchers
sample loaded, and measurement errors [26, 29]. However,
these biases are not compelling enough to warrant normal-
ization. In proteomics experiments, proteins are denatured,
reduced, and alkylated. Protein degradation rarely occurs
during careful sample preparation when sample preparation
is conducted by skilled technicians. Each sample is injected
by autosampler with a high degree of accuracy and precision.
When an appropriate normalization method is not available,
the process of normalization is unnecessary and should be
avoided, as it may introduce new errors in quanti�cation.
Alternatively, use of sample quality control (QC) is
a better way to obtain accurate results. QC samples and
experimental samples should be analyzed in the same batch.
If the QC samples indicate the instrument and running
conditions are optimal, the data obtained from experimental
samples can be considered valid and the systematic biases
As an example of this concept, in an LFQMS analysis
where none of the normalization techniques was applied
, suitable precision was achieved, based on the acceptance
criteria of assay performance by which biomarker assays are
evaluated, CV ± 25% acting as the default value (±30% at the
lower limit of quantitation, LLOQ) . In a repeatability
test of a simple 3 peptide sample, two quanti�able peptides
(Angiotensin III and Fibrinopeptide B) had an intensity CV
≤ 15%. In the repeatability test of a more complex sample
(kidney tissue), 772 (76.4%), 99 (9.8%), 102 (10.1%), and 38
(3.8%) out of 1,011 proteins had a CV ≤ 15%, ≤20% and
>15%, ≤30% and >20%, and ≤50% and >30%, respectively,
indicating that 96.2% of proteins had CVs ≤ 30%. ese data
indicate that elimination of unquanti�able peptides is more
International Journal of Proteomics7
vital to accurate quantitation than normalization. In current
LC-MS technology, no ideal normalization techniques exist.
�sing inappropriate or even �awed normalization will not
improve the analysis and may introduce additional errors,
thus it is better that no normalization is applied. However,
�ltration of unquanti�able peptides is absolutely necessary
for an accurate analysis.
e true test of the utility of an LFQMS platform lies in the
comparative analysis of differential expression in complex
across several experimental groups. In the following exam-
ples, we demonstrate the applicability of the IdentiQuantXL
platform in overcoming many of the limitations of bottom-
prehensive and accurate analysis of protein abundance and
differential expression in a range of experimental situations.
3.1. Aqueous Humor Proteomics. In the initial test of a
beta version of IdentiQuantXL, we analyzed the protein
composition of aqueous humor (AH) to investigate the role
it might play in Fuchs endothelial corneal dystrophy (FECD)
. AH is the biologic �uid in the anterior chamber of
the eye that protects and supplies nutrients to the cornea,
lens, and trabecular meshwork. A balance between produc-
tion and drainage of AH is critical to maintaining normal
intraocular pressure, and its protein composition has been
shown to change dramatically in various ocular conditions.
Albumin and immunoglobulin G (IgG) depleted samples
were obtained from male and female patients with and
without FECD, and both depleted AH and albumin-bound
proteins were analyzed. We identi�ed �4 nonredundant AH
proteins, most of which were identi�ed in previous AH
proteomic studies of patients with cataracts, in the albumin-
depleted fraction. e levels of �ve of these were signi�cantly
lower (afamin, complement C3, histidine-rich glycopro-
tein, immunoglobulin heavy [IgH], and protein family with
sequence similarity 3, member C [FAM3C]), while the levels
of one, suprabasin, was signi�cantly higher in patients with
FECD compared to controls (𝑃𝑃 ? ?????. We also identi�ed
34 proteins in the albumin-bound fraction, four of which
were signi�cantly elevated in patients with FECD including
a hemoglobin fragment, immunoglobulin kappa (IgK) and
immunoglobulin lambda (IgL) (𝑃𝑃 ? ?????. Additionally, we
were unable to detect any signi�cant differences in protein
levels based on gender. Because FECD is a progressive
disorder, regression analyses were performed to determine
any signi�cant correlations with age, and of interest retinol-
binding protein 3 was signi�cantly correlated with age in
patients with FECD (𝑃𝑃 ? ?????, whereas no proteins in
the control group correlated with age. is was the �rst
demonstrating the utility of the LFQMS approach.
In a related study , we investigated the effect of
implantation of a glaucoma shunt device on inappropri-
ate accumulation of plasma-derived proteins in the AH.
We identi�ed 135 nonredundant proteins in the albumin-
depleted fraction, 13 of which differed signi�cantly between
shunted and control groups. ose proteins play a role in
oxidative stress, apoptosis, in�ammation, and/or immunity.
Many of the identi�ed proteins were novel proteins not
previously associated with glaucoma. All but complement
C4 were known plasma proteins and the elevated levels of
these proteins in the aqueous humor suggested that the
glaucoma shunt device caused either a breach in blood-
aqueous barrier or chronic trauma, increasing in�ux of
oxidative, apoptotic, and in�ammatory proteins that could
potentially cause corneal endothelial damage.
3.2. Nanoparticle Exposure in Biological Systems. To assess
the effect of functionalized carbon nanotube (f-CNT) and
silver nanoparticle (AgNP) exposures, LFQMS was used to
generate differential protein expression pro�les in a novel in
vitro intestinal cell model that features Caco-2 and HT29-
MTX cells in coculture. ese adenocarcinoma cell lines
are derived from intestinal absorptive and mucus-secreting
goblet cell types, respectively. eir combination and unique
utilization represent a physiologically relevant GI model
system characterized by tight junctions, polarity, and mucus
secretion covering the entire monolayer. In a related study,
we used LFQMS to compare the composition of proteins in
coronas that form around f-CNT and AgNP in cell culture
media supplemented with fetal bovine serum. As described
below, the results of both of these various studies, in which
the IdentiQuantXL platform was the key analytical tool,
provide excellent examples of the utility of this platform in
3.2.1. Nanotoxicoproteomics of Functionalized Carbon Nan-
otube Exposure. To assess the biological effects of low level,
water dispersible, functionalized carbon nanotube (f-CNT)
exposure in the Caco-2/HT29-MTX coculture model (75%
Caco-2, 25% HT29-MTX), cell protein expression was quan-
ti�ed and compared  using the IdentiQuantXL platform.
lated single-walled carbon nanotubes (SWCNT-COOH),
carboxylated multiwalled carbon nanotubes (MWCNT-
COOH), and poly(vinylpyrrolidone) (PVP) polymer func-
tionalized multiwalled carbon nanotubes (MWCNT-PVP)
for 48h at 500pg/mL and 10𝜇𝜇g/mL. e identi�cation of
8,081 peptides (from technical replicates of 𝑛𝑛 ? ? samples; 70
injections total) led to 4,743 protein database hits that were
globally identi�ed with >90% con�dence, corresponding to
2,282 unique, nonredundant proteins that were compared
across the dose groups. Among these, 428 proteins were
differentially expressed (𝑃𝑃 ? ?????. As listed in Table 1,
mean CV across the 2,282 proteins was less than 17% in all
samples except SWCNT-COOH 500pg/mL and MWCNT-
PVP 10𝜇𝜇g/mL where considerable differential expression
was observed. Control group CV averaged only 12%. At the
high dose, the extent of differential protein expression was
CNT-speci�c and directly related to CNT colloidal stability.
Surprisingly, cells responded to low level MWCNT-PVP
exposure with 3-fold greater differential expression than at
8 International Journal of Proteomics
T 1: Mean coefficient of variation across 2,282 Caco-2/HT29-
MTX proteins in functionalized carbon nanotube exposures and
the high level. Bioinformatic analysis indicated signi�cant
and CNT-speci�c effects on relevant molecular and cellular
functions and canonical pathways, with little overlap across
CNT type and in the absence of overt toxicity.
3.2.2. Nanotoxicoproteomics of Silver Nanoparticle Exposure.
Analysis of 20nm AgNP-citrate exposed to Caco-2/HT29-
MTX cells for 3 and 24h at 50𝜇𝜇g/mL revealed signi�cant
alterations in cellular protein expression. A total of 4,112
unique proteins, homologs, or splice variants were identi�ed,
quanti�ed and their abundance statistically compared. e
24 in�ections total) corresponded to 13,445 identi�ed pep-
tides and 8,343 protein database hits, many of which were
redundant. As listed in Table 2, mean CV across the 4,112
proteins was less than 20% in AgNP-treated samples and less
than 17.3% in the controls. In fact, the number of proteins
with CV ≤ 15% was 55% and 46% in 3h and 24h controls,
e 3-hour exposure signi�cantly altered the expression
of 156 proteins (130 decreased, 26 increased) while 168 (137
decreased, 31 increased) were differentially expressed aer
24h. Volcano plots in Figures 4(a) and 4(b) broadly illustrate
the proteome response. Consistent with our previous NP
studies, protein expression altered at early time points is
different from that occurring in longer exposures. As the
lular response at these time points emerged. At 3h, proteins
associated with protein synthetic processes, ubiquitination
pathways, and EIF2 signaling pathways were downregulated,
as were those in the glycolytic/gluconeogenic pathway. ese
changes indicate events that result in a decline in global
cellular protein synthesis. At 24h, proteins of the gly-
colytic/gluconeogenic pathway are no longer affected, while
proteins associated with cellular infection are upregulated.
When one compares the number of proteins identi�ed
and quanti�ed in this study to the f-CNT study described in
Section 3.2.1, it is apparent that signi�cantly more proteins
were evaluated in the AgNP experiments. Indeed, two-thirds
of peptides were detected, leading to an 80% increase in
the number of nonredundant proteins that were identi�ed
and quanti�ed (4,112 versus 2,282). e difference lies in
the ratio of Caco-2 to HT29-MTX cells in coculture and
T 2: Mean coefficient of variation across 4,112 Caco-2/HT29-
MTX proteins in AgNP exposures and LFQMS analysis.
the duration of cell culture. In the f-CNT study, the ratio
was 75:25 whereas in the AgNP study we found a ratio of
90:10 to improve the cell monolayer stability and integrity.
Decreasing the percentage of mucus-secreting HT29-MTX
cells signi�cantly reduced the amount of mucus present
when cells were recovered for sample preparation, and this
enabled a proportionally higher yield of cellular proteins in
the sample. A more minor consideration relates to the fact
that f-CNT exposures were carried out for 48h, compared
to 3 and 24h in the AgNP study, a point at which protein
synthesis, growth, and proliferation are comparatively lower.
us, the difference may be a function of longevity.
Similar successful applications of this platform to
nanoparticle exposure in other cell culture systems have
been conducted and can be found in the literature [34–36].
3.2.3. Nanoparticle Protein Corona Composition. e protein
corona (PC) that forms on nanoparticles when they are
exposed to protein-containing biological �uids changes its
bioactivity in cells . Based on our own previous observa-
tions [33, 38] that structurally similar nanoparticles can have
divergent biological effects in cell culture systems, we used
the IdentiQuantXL platform to investigate the composition
(nanoclay tubes, SWCNT-Raw, SWCNT-COOH, MWCNT-
Raw, MWCNT-Pure, MWCNT-COOH, and MWCNT-PVP)
mentioned earlier in this paper .
Proteomic analysis identi�ed and quanti�ed 366 differ-
ent protein components of the various NT coronas. e
PC which formed on the nanoclay tubes consisted of the
fewest number, 82 different proteins, whereas the SWCNT-
COOH corona contained the most, at 181. For reference
purposes, analysis of the 10% FBS-DMEM media alone
revealed 2,507 nonredundant proteins, polypeptides, or pro-
tein fragments/isoforms, and a list of these along with their
peptide sequence and abundance data is provided along with
the published manuscript .
All NT coronas were found to consist of 14 common
proteins, including alpha-1-antiproteinase, alpha-2-HS-
glycoprotein, alpha-S1-casein, apolipoprotein A-I, apolipo-
protein A-II, keratin, type I cytoskeletal 10, keratin, type I
cytoskeletal 15, keratin, type II cytoskeletal 1, keratin, type II
cytoskeletal 5, keratin, type II cytoskeletal 6A, keratin, type
II cytoskeletal 6C, keratin, type II cytoskeletal 75, serum
albumin, and titin. e �ve most abundant coronal proteins
(titin, serum albumin, apolipoprotein A-I, apolipoprotein
A-II, and alpha-S1-casein) exhibited signi�cant differences
across the various NTs, while the relative contributions of
alpha-1-antiproteinase (aka alpha-1-antitrypsin in humans),
alpha-2-HS-glycoprotein, and the 7 keratins to the NT
International Journal of Proteomics9
Protein POF1BProtein POF1B
6 6 6 6 6 6 6 6 6
3 h24 h
F 4: Volcano plots of all 4,112 Caco-2/HT29-MTX cell proteins with Log2 fold-change in expression on the x axis and P value on the
y axis. (a) 20nm AgNP-citrate 12.5𝜇𝜇g/mL, 3h effect. (b) 20nm AgNP-citrate 12.5𝜇𝜇g/mL, 24h effect. Blue ellipses surround signi�cantly
decreased proteins; red ellipses are increased proteins. (c) Venn diagram illustrates that AgNP-mediated effects are unique at the two time
points, and little overlap in biological effect occurs.
coronas were not signi�cantly different. With the exception
serum proteins are commonly found in NP coronas formed
in human plasma/serum. Titin is the 14th most abundant
protein in the FBS-DMEM media whereas albumin is 1st,
alpha-2-HS-glycoprotein 2nd, and alpha-1-antiproteinase
3rd, and Apo-AI is 17th, while alpha-S1-casein and the
keratins (other than keratin 1) are far less abundant in the
culture medium. Importantly, the presence of the latter
proteins in the PC of all NPs suggests a selective enrichment
that is not related to their concentrations in the media. It
should also be mentioned that all of the above proteins
are highly abundant in human plasma according to the
most recent version of the Human Peptide Atlas database
(http://www.peptideatlas.org), with the exception of alpha-
S1-casein, which is not a component of human plasma .
e most abundant PC protein was Xin actin-binding
MWCNT-Pure, MWCNT-PVP, and SWCNT-COOH coro-
nas. XIRP2, aka mXin𝛽𝛽 and myomaxin, is a 382,300Da
protein expressed in cardiac and skeletal muscle where it
interacts with �lamentous actin and 𝛼𝛼-actinin through the
medium. Like titin, this largely abundant coronal protein is
associated with intracellular �lamentous proteins. e ample
presence of XIRP2 in the media and in NT coronas may
be in the form of protein fragments that are more common
to fetal serum and less so in adult human or bovine sera
where they are known to interact with albumin . Other
proteins may also be present in the PC via their association
with bovine serum albumin, as part of the albuminome
10International Journal of Proteomics
Exp. 1 (2 hr LC)
Top 5 pituitary proteins
(of 1,361 hits; 646 unique)
Exp.2 (3 hr LC)
Top 5 pituitary proteins
(of 4,352 hits; 2,416 unique)
RNA helicase DDX1
F 5: Impact of brain saline perfusion (Exp. 2) to reduce blood-protein contamination and expand the LC gradient by 50% (Exp.
2) on protein abundance among the top 5 pituitary proteins identi�ed and quanti�ed in the two experiments. WT: wild type mouse; �I:
𝐿𝐿?𝐿𝐿?𝑊𝑊???𝑊𝑊𝑊𝑊𝑊𝑊?𝑊𝑊???𝑊𝑊𝑊𝑊𝑊𝑊knock-in mouse;∗𝑃𝑃 ? ???????, 𝑞𝑞 ? ????.
[43–45]. For instance, the keratins identi�ed in the PCs may
be there through their interaction with albumin directly, or
indirectly via their known interaction with apolipoproteins,
which also interact with albumin . e results above
represent one of the most comprehensive analyses of both
protein corona constituents and the protein composition of
utility of the LFQMS platform.
3.3. Pituitary Proteomics. As part of an investigation of
the molecular differences between wild-type (WT) and
𝐿𝐿?𝐿𝐿?𝑊𝑊???𝑊𝑊𝑊𝑊𝑊𝑊?𝑊𝑊???𝑊𝑊𝑊𝑊𝑊𝑊knock-in mice (a model of pediatric
conducted a quantitative proteomic analysis of the proteins
samples (from which 20𝜇𝜇g of peptides were injected), a
total of 2,416 nonredundant proteins were identi�ed and
quanti�ed from 7,680 unique peptides representing 4,352
protein database hits. Coefficients of variation across all
2,416 proteins in WT and knock-in were 14.0% and 20.6%,
respectively; 67% of WT and 33% of the knock-in proteins
had CV ≤ 15%. e latter can be explained by the fact that
425 proteins were differentially expressed in the knock-in
mice (𝑃𝑃 ? ???????, corresponding to 𝑞𝑞 ? ????). Within this
upregulated. Of the pituitary hormones detected, prolactin
(PRL) was 17.1-fold lower in the 𝐿𝐿?𝐿𝐿?𝑊𝑊???𝑊𝑊𝑊𝑊𝑊𝑊?𝑊𝑊???𝑊𝑊𝑊𝑊𝑊𝑊animals
when compared to WT. Proopiomelanocortin (POMC, the
precursor peptide containing ACTH) and 𝛼𝛼GSU also were
and 1.6-fold, respectively. LH𝛽𝛽 and TSH𝛽𝛽 were signi�cantly
higher in the 𝐿𝐿?𝐿𝐿?𝑊𝑊???𝑊𝑊𝑊𝑊𝑊𝑊?𝑊𝑊???𝑊𝑊𝑊𝑊𝑊𝑊mice, at 1.8- and 1.7-
fold. GH, the most abundant protein detected, was not
signi�cantly different between 𝐿𝐿?𝐿𝐿?𝑊𝑊???𝑊𝑊𝑊𝑊𝑊𝑊?𝑊𝑊???𝑊𝑊𝑊𝑊𝑊𝑊and WT,
whereas a negative control, ACTA1 (alpha actin), showed no
represents the most comprehensive proteomic analysis of the
pituitary to date. is uniquely extensive dataset provides
a resource for others investigating pituitary physiology and
may serve to suggest biomarkers for future studies.
Not only did this investigation demonstrate the utility
of the IdentiQuant LFQMS platform, but also the opti-
mization of the approach demonstrated the critical impor-
tance of maximal LC conditions and proper sample prepa-
ration to accurate and reproducible protein quantitation.
In an initial analysis of unperfused pituitary lysates from
𝐿𝐿?𝐿𝐿?𝑊𝑊???𝑊𝑊𝑊𝑊𝑊𝑊?𝑊𝑊???𝑊𝑊𝑊𝑊𝑊𝑊and WT mice, a 2h LC gradient was
used. As Figure 5 illustrates, those conditions resulted in the
identi�cation of only 1,361 database hits and the identi�ca-
tion and quanti�cation of only 636 nonredundant proteins.
An examination of the most abundant proteins detected
revealed that hemoglobin alpha was the third-most abun-
dant protein (along with albumin and several other serum
proteins not shown), strongly suggesting substantial blood
International Journal of Proteomics 11
contamination. When new pituitary samples were dissected
from brains perfused with ice cold saline (0.9%) and the LC
gradient was expanded to 3h, database hits increased 3.2-
fold and nonredundant protein quantitation increased 3.7-
fold. e blood-related proteins were not eliminated entirely,
but their relative abundance was signi�cantly reduced (<top
100 proteins). Perfusion and LC gradient expansion also
increased the relative quantitation of the most abundant
pituitary protein, somatotropin (growth hormone), along
with several other hormones and pituitary proteins.
form was recently used to verify the effectiveness and repro-
ducibility of a novel membrane enrichment method .
Using HT29-MTX cells and a single cell disruption step
(in a hypotonic reagent using liquid nitrogen), a single low
speed centrifugation isolation step, and three wash steps
using high speed centrifugation, the author identi�ed and
quanti�ed 2,362 nonredundant proteins from an average
yield of 237𝜇𝜇g membrane proteins per 10 million cells.
Across various experiments, 99% of the quanti�ed proteins
had a CV ≤ 30%, and 2,001 proteins (85%) had a CV
≤ 15%. Western blot and LC-MS/MS results of markers
for cytoplasm, nucleus, mitochondria, and their membranes
indicated that the enriched membrane fraction was highly
pure by the absence of, or presence of trace amounts of,
non-membrane marker proteins. ese results not only
document the effectiveness of the enrichment method, but
identify and quantify proteins from membrane fractions, a
notoriously difficult task [49, 50].
Label-free mass spectrometric protein quanti�cation based
on peptide peak area is a viable alternative to array-based,
gel-based, stable isotope tag or label-based, and spectral
associated with quantitation based on peptide ion peak area
measurement: chromatographic alignment, peptide quali-
�cation for quantitation, and normalization. To overcome
these issues, we incorporated several approaches into a label-
free quantitative mass spectrometry platform, IdentiQuan-
comprehensive, accurate, and reproducible protein identi�-
tein mixtures from experiments with many sample groups.
is work was supported by the following grants: NIEHS
 L. A. Liotta, V. Espina, A. I. Mehta et al., “Protein microarrays:
meeting analytical challenges for clinical applications,” Cancer
Cell, vol. 3, no. 4, pp. 317–325, 2003.
 T. Rabilloud and C. Lelong, “Two-dimensional gel elec-
74, no. 10, pp. 1829–1841, 2011.
 F. Xie, T. Liu, W. J. Qian, V. A. Petyuk, and R. D. Smith,
“Liquid chromatography-mass spectrometry-based quantita-
tive proteomics,” Journal of Biological Chemistry, vol. 286, no.
29, pp. 25443–25449, 2011.
 M. W. Linscheid, R. Ahrends, S. Pieper, and A. Kühn, “Liquid
chromatography-mass spectrometry-based quantitative pro-
teomics,” Methods in Molecular Biology, vol. 564, pp. 189–205,
 J. Ji, A. Chakraborty, M. Geng et al., “Strategy for qualitative
and quantitative analysis in proteomics based on signature
peptides,” Journal of Chromatography B, vol. 745, no. 1, pp.
 S. P. Gygi, B. Rist, S. A. Gerber, F. Turecek, M. H. Gelb, and R.
Aebersold, “Quantitative analysis of complex protein mixtures
no. 10, pp. 994–999, 1999.
 X. Yao, A. Freas, J. Ramirez, P. A. Demirev, and C. Fenselau,
“Proteolytic 18O labeling for comparative proteomics: model
vol. 73, no. 13, pp. 2836–2842, 2001.
R. D. Smith, “Proteome analysis using selective incorporation
of isotopically labeled amino acids,” Journal of the American
Society for Mass Spectrometry, vol. 11, no. 1, pp. 78–82, 2000.
 X. Lai, L. Wang, H. Tang, and F. A. Witzmann, “A novel
alignmentmethod andmultiple�lters forexclusionof unquali-
�ed peptides to enhance label-free quanti�cation using peptide
10, pp. 4799–4812, 2011.
 W. M. Old, K. Meyer-Arendt, L. Aveline-Wolf et al., “Compar-
ison of label-free methods for quantifying human proteins by
shotgun proteomics,” Molecular and Cellular Proteomics, vol. 4,
no. 10, pp. 1487–1502, 2005.
 X. Wang, W. Zhu, K. Pradhan et al., “Feature extraction in the
 D. Tsikas, M. T. Suchy, A. Mitschke, B. Beckmann, and
F. M. Gutzki, “Measurement of nitrite in urine by gas
chromatography-mass spectrometry,” Methods in Molecular
Biology, vol. 844, pp. 277–293, 2012.
“Rapid and robust con�rmation and quanti�cation of 11-nor-
delta9-tetrahydrocannabinol-9-carboxylic acid (thc-cooh) in
urine by column switching lc-ms-ms analysis,” Journal of Mass
Spectrometry, vol. 47, no. 1, pp. 124–130, 2012.
 K. Heinig, T. Wirz, F. Bucheli, V. Monin, and A. Gloge,
“Sensitive determination of a pharmaceutical compound and
its metabolites in human plasma by ultra-high performance
liquid chromatography-tandem mass spectrometry with on-
line solid-phase extraction,” Journal of Pharmaceutical and
Biomedical Analysis, vol. 54, no. 4, pp. 742–749, 2011.
 P. M. Palagi, D. Walther, M. Quadroni et al., “MSight: an image
analysis soware for liquid chromatography-mass spectrome-
try,” Proteomics, vol. 5, no. 9, pp. 2381–2384, 2005.
challenges associated with alcohol research,” Alcohol Research
and Health, vol. 31, no. 3, pp. 251–255, 2008.
 M. Berth, F. M. Moser, M. Kolbe, and J. Bernhardt, “e state
of the art in the analysis of two-dimensional gel electrophoresis
12International Journal of Proteomics
images,” Applied Microbiology and Biotechnology, vol. 76, no. 6,
pp. 1223–1243, 2007.
 O. Kohlbacher, K. Reinert, C. Gröpl et al., “TOPP—the
 K. C. Leptos, D. A. Sarracino, J. D. Jaffe, B. Krastins, and G.
M. Church, “MapQuant: open-source soware for large-scale
protein quanti�cation,” Proteomics, vol. 6, no. 6, pp. 1770–1782,
 D. May, W. Law, M. Fitzgibbon, Q. Fang, and M. McIntosh,
“Soware platform for rapidly creating computational tools
for mass spectrometry-based proteomics,” Journal of Proteome
Research, vol. 8, no. 6, pp. 3212–3217, 2009.
 M. Katajamaa, J. Miettinen, and M. Orešič, “MZmine: toolbox
for processing and visualization of mass spectrometry based
molecular pro�le data,” Bioinformatics, vol. 22, no. 5, pp.
 L. N. Mueller, O. Rinner, A. Schmidt et al., “SuperHirn—a
novel tool for high resolution LC-MS-based peptide/protein
pro�ling,” Proteomics, vol. 7, no. 19, pp. 3470–3480, 2007.
pattern recognition,” Molecular and Cellular Proteomics, vol. 5,
no. 10, pp. 1927–1941, 2006.
 C. C. Tsou, C. F. Tsai, Y. H. Tsui et al., “IDEAL-Q, an
cient peptide alignment approach and spectral data validation,”
Molecular and Cellular Proteomics, vol. 9, no. 1, pp. 131–144,
 F. Erhard and R. Zimmer, “Detecting outlier peptides in
quantitative high-throughput mass spectrometry data,” Journal
of Proteomics, vol. 75, no. 11, pp. 3230–3239, 2012.
 S. J. Callister, R. C. Barry, J. N. Adkins et al., “Normalization
spectrometry and label-free proteomics,” Journal of Proteome
Research, vol. 5, no. 2, pp. 277–286, 2006.
 L. Mayne, Z. Y. Kan, P. S. Chetty, A. Ricciuti, B. T. Walters,
and S. W. Englander, “Many overlapping peptides for protein
hydrogen exchange experiments by the fragment separation-
Mass Spectrometry, vol. 22, no. 11, pp. 1898–1905, 2011.
 B. Mann, M. Madera, Q. Sheng, H. Tang, Y. Mechref, and
M. V. Novotny, “ProteinQuant Suite: a bundle of automated
soware tools for label-free quantitative proteomics,” Rapid
Communications in Mass Spectrometry, vol. 22, no. 23, pp.
 W. Zhu, J. W. Smith, and C. M. Huang, “Mass spectrometry-
based label-free quantitative proteomics,” Journal of Biomed-
icine and Biotechnology, vol. 2010, Article ID 840518, 6 pages,
 C. H. Chau, O. Rixe, H. McLeod, and W. D. Figg, “Validation
of analytic methods for biomarkers used in drug development,”
Clinical Cancer Research, vol. 14, no. 19, pp. 5967–5976, 2008.
 M. R. Richardson, Z. M. Segu, M. O. Price et al., “Alterations in
corneal dystrophy,” Molecular Vision, vol. 16, pp. 2376–2383,
 A. Anshu, M. O. Price, M. R. Richardson et al., “Alterations in
device,” Molecular Vision, vol. 17, pp. 1891–1900, 2011.
 X. Lai, B. L. Blazer-Yost, J. W. Clack et al., “Protein expression
pro�les of intestinal epithelial co-cultures aer low-level expo-
sure to functionalized carbon nanotubes,” International Journal
of Biomedical Nanoscience and Nanotechnology. In press.
 T. Xia, R. F. Hamilton Jr., J. C. Bonner et al., “Inter-laboratory
comparison of in vitro nanotoxicological assays from the niehs
nanogo consortium,” Environmental Health Perspectives. In
 A. K. Vidanapathirana, X. Lai, S. C. Hilderbrand et al., “Multi-
walled carbon nanotube directed gene and protein expression
in cultured human aortic endothelial cells is in�uenced by
 S. C. Tilton, N. J. Karin, A. Tolic et al., “ree human cell types
respond to multi-walled carbon nanotubes and titanium diox-
ide nanobelts with cell-speci�c transcriptomic and proteomic
expression patterns,” Nanotoxicology. In press.
 I. Lynch, T. Cedervall, M. Lundqvist, C. Cabaleiro-Lago, S.
Linse, and K. A. Dawson, “e nanoparticle-protein complex
as a biological entity� a complex �uids and surface science
Science, vol. 134-135, pp. 167–174, 2007.
 B. L. Blazer-Yost, A. Banga, A. Amos et al., “Effect of carbon
and protein expression,” Nanotoxicology, vol. 5, no. 3, pp.
 J. H. Shannahan, J. M. Brown, R. Chen et al., “Comparison of
nanotube-protein corona composition in cell culture media,”
Small. In press.
 T. Farrah, E. W. Deutsch, G. S. Omenn et al., “A high-
con�dence human plasma proteome reference set with esti-
mated concentrations in peptideatlas,” Molecular & Cellular
Proteomics, vol. 10, no. 9, Article ID M110.006353, 2011.
 H. T. Huang, O. M. Brand, M. Mathew et al., “Myomaxin
Is a novel transcriptional target of MEF2A that encodes a
Xin-related 𝛼𝛼-actinin-interacting protein,” Journal of Biological
Chemistry, vol. 281, no. 51, pp. 39370–39379, 2006.
 D. Pacholsky, P. Vakeel, M. Himmel et al., “Xin repeats de�ne a
novel actin-binding motif,” Journal of Cell Science, vol. 117, no.
22, pp. 5257–5268, 2004.
 D. Scumaci, M. Gaspari, M. Saccomanno et al., “Assessment
of an ad hoc procedure for isolation and characterization of
human albuminome,” Analytical Biochemistry, vol. 418, no. 1,
pp. 161–163, 2011.
and its albuminome,” Proteomics, vol. 1, no. 1, pp. 73–88, 2007.
 R. L. Gundry, M. Y. White, J. Nogee, I. Tchernyshyov, and
J. E. Van Eyk, “Assessment of albumin removal from an
immunoaffinity spin column: critical implications for pro-
teomic examination of the albuminome and albumin-depleted
samples,” Proteomics, vol. 9, no. 7, pp. 2021–2028, 2009.
 M. Zhou, D. A. Lucas, K. C. Chan et al., “An investigation into
the human serum ‘interactome’,” Electrophoresis, vol. 25, no. 9,
pp. 1289–1298, 2004.
 K. L. Prince, S. C. Colvin, X. Lai, F. A. Witzmann, and S.
J. Rhodes, “Developmental analysis and in�uence of genetic
background on the lhx3 w227ter mouse model of combined
pituitary hormone de�ciency disease,” Endocrinology. In press.
 X. Lai, “A reproducible method to enrich membrane proteins
with high-purity and high-yield for an lc-ms/ms approach in
quantitative membrane proteomics,” Electrophoresis. In press.
International Journal of Proteomics 13 Download full-text
 J. M. Gilmore and M. P. Washburn, “Advances in shotgun
proteomics and the analysis of membrane proteomes,” Journal
of Proteomics, vol. 73, no. 11, pp. 2078–2091, 2010.
 S. J. Cordwell and T. E. ingholm, “Technologies for plasma
membrane proteomics,” Proteomics, vol. 10, no. 4, pp. 611–627,