Clinical Chemistry 53:4
645– 656 (2007)
Proteomics and
Protein Markers
Preanalytic Influence of Sample Handling on
SELDI-TOF Serum Protein Profiles
John F. Timms,1,2* Elif Arslan-Low,1 Aleksandra Gentry-Maharaj,1 Zhiyuan Luo,3
Davy T’Jampens,4 Vladimir N. Podust,4 Jeremy Ford,1 Eric T. Fung,4 Alex Gammerman,3
Ian Jacobs,1 and Usha Menon1
Background: High-throughput proteomic methods for
disease biomarker discovery in human serum are promising, but concerns exist regarding reproducibility of
results and variability introduced by sample handling.
This study investigated the influence of different preanalytic handling methods on surface-enhanced laser
desorption/ionization time-of-flight mass spectrometry
(SELDI-TOF MS) protein profiles of prefractionated
serum. We investigated whether older collections with
longer sample transit times yield useful protein profiles,
and sought to establish the most feasible collection
methods for future clinical proteomic studies.
Methods: To examine the effect of tube type, clotting
time, transport/incubation time, temperature, and storage method on protein profiles, we used 6 different
handling methods to collect sera from 25 healthy volunteers. We used a high-throughput, prefractionation strategy to generate anion-exchange fractions and examined
their protein profiles on CM10, IMAC30-Cu, and H50
arrays by using surface-enhanced laser desorption/ionization time-of-flight mass spectrometry.
Results: Prolonged transport and incubation at room
temperature generated low mass peaks, resulting in
distinctions among the protocols. The most and least
stringent methods gave the lowest overall peak variances, indicating that proteolysis in the latter may have
been nearly complete. For samples transported on ice
there was little effect of clotting time, storage method, or
transit time. Certain proteins (TTR, ApoCI, and transferrin) were unaffected by handling, but others (ITIH4
and hemoglobin ␤) displayed significant variability.
Conclusions: Changes in preanalytical handling variables affect profiles of serum proteins, including proposed disease biomarkers. Proteomic analysis of samples from serum banks collected using less stringent
protocols is applicable if all samples are handled
identically.
© 2007 American Association for Clinical Chemistry
Analysis of serum proteomes holds great promise for
identifying novel cancer markers for screening, diagnosis,
and prognosis (1–3 ). Most patients find venipuncture
tolerable, and it is standard practice to monitor disease
progression or response to therapy by collecting serial
blood samples. Recently, various proteomics-based approaches coupled with advanced bioinformatics have
been used to identify putative disease biomarkers in
patient serum and plasma, and several studies have
identified biomarkers that improve the positive predictive
value of disease detection (4, 5 ). The cancer antigen
CA125, a diagnostic and prognostic marker for ovarian
cancer, is increased in only 85% of patients with ovarian
cancer and 50% of those with early stage disease (6 ), and
may also be increased in benign disease. Research using
new technologies is focused on developing a marker that
will outperform CA125 or can be used in combination to
increase the performance of the current test.
Several different methods based on mass spectrometry
(MS)5 have been applied in the search for cancer biomarkers [reviewed in (1, 7 )]. Surface-enhanced laser desorp-
1
Translational Research Laboratory, Institute of Women’s Health, University College London, London, United Kingdom.
2
Cancer Proteomics Group, Ludwig Institute for Cancer Research, London
Branch, London, United Kingdom.
3
Department of Computer Science, Royal Holloway College, University of
London, London, United Kingdom.
4
Ciphergen Biosystems, Inc., Fremont, CA.
* Address correspondence to this author at: Translational Research Laboratory, Institute of Women’s Health, University College London, Huntley
Street, London, WC1E 6DH, UK. Fax 44-207-6796334; e-mail jtimms@
ludwig.ucl.ac.uk.
Received September 15, 2007; accepted January 15, 2007.
Previously published online at DOI: 10.1373/clinchem.2006.080101
5
Nonstandard abbreviations: MS, mass spectrometry; SELDI-TOF MS,
surface-enhanced laser desorption/ionization time-of-flight mass spectrometry; UKCTOCS, United Kingdom Collaborative Trial of Ovarian Cancer
Screening; RT, room temperature; IMAC, immobilized metal affinity capture;
PCA, principal components analysis; ITIH4, inter-␣-trypsin inhibitor heavy
chain H4; ApoCI, apolipoprotein C-I; TTR, transthyretin.
645
646
Tims et al.: Sample Handling Effects on Serum Proteomic Profiles
tion/ionization time-of-flight (SELDI-TOF) MS (8, 9 ) has
been used extensively for serum profiling. In this method,
high-throughput mass profiling with laser desorption/
ionization MS instrumentation is performed on sample
proteins bound selectively to chip surfaces with different
chemical properties. Spectral patterns are then compared
across samples to find discriminating masses or changes
in peak intensities. Initial enthusiasm about these new
technologies has been somewhat tempered by questions
on the robustness of class discriminating algorithms and
method reproducibility (7, 10, 11 ). Increasing evidence
that sample collection and processing can affect protein
profiles and the ability to differentiate between disease
and control samples has cast further doubt on the validity
of some studies (12, 13 ). Transit time, storage conditions,
clotting time, and tube type can all affect serum profiles,
irrespective of true biological variation (14 –17 ). It is likely
that such introduced differences are primarily driven by
proteolysis, although other variables may contribute,
such as agglutination or differential adhesion of serum
polypeptides to tube walls. These findings have raised
concerns about using samples for case-control studies
from older collections, where samples were collected and
transported for different times at ambient temperatures.
Many of these collections are unique, with samples predating cancer diagnosis. One such collection is the United
Kingdom Collaborative Trial of Ovarian Cancer Screening
(UKCTOCS), in which 202 638 postmenopausal women
from 13 centers in the UK were randomized to screening
vs controls; this serum bank will eventually have 500 000
samples, including serial samples from 50 000 women (see
www.ukctocs.org.uk). The collection protocol for this trial
allows blood samples to stand on the clot for 24 –56 h
before processing. Thus, if proteomic technologies are
used for biomarker discovery from such collections, it is
imperative to compare samples collected and processed
using these less stringent protocols with those collected in
accordance with protocols that involve immediate transport on ice.
We examined the impact of diverse serum handling
protocols on protein profiles observed by SELDI-TOF MS.
We also sought to determine the least variable and most
clinically feasible handling method for prospective serum
collections for proteomic studies.
Materials and Methods
sample collection and handling
This study was approved by the local ethics committee
and written informed consent was obtained from all
donors. Serum samples were collected at Barts and the
London NHS Trust on 2 consecutive days from 25
healthy, postmenopausal women randomized to the
CA125 arm of UKCTOCS. Protocol 1 (Green; GN) used
the existing UKCTOCS protocol (see www.ukctocs.
org.uk) with samples collected in Greiner gel tubes,
allowed to clot, centrifuged at room temperature (RT),
divided into aliquots, and placed in straws that were heat
sealed and stored at ⫺80 °C. Time from venipuncture to
centrifugation was 30 h for each sample. Additional
samples from the same volunteers were collected in
Becton Dickinson red-top tubes, allowed to clot at RT for
60 min, subjected to transport/storage on wet ice for 2 h
before centrifugation, transferred to straws, and stored at
⫺80 °C (protocol 2a; Yellow; YE) (14 ). A 3rd protocol used
a 5 min clotting time at RT, followed by transport/storage
on wet ice for 3 h before centrifugation, transfer to straws,
and storage (protocol 2b; Gray; GY). Three variants of
protocol 2b were used to prepare cryovials for storage
instead of straws (protocol 2c; Cryovial; CR), with transport/storage on wet ice for 6 h instead of 3 h (protocol 2d;
Orange; OR) and with transport/storage for 3 h at RT
instead of on wet ice (protocol 2e; White; WH). Handling
protocols are detailed in Table S1 of the Data Supplement
that accompanies this article at (http://www.clinchem.
org/content/vol53/issue4). Although 25 ⫻ 6 protocols
would generate 150 individual serum samples for analysis, because of insufficient material only 13 protocol 2c
samples were available.
sample preparation
Sample preparation details are provided as Supplemental Data. Briefly, samples were thawed, randomized,
and triplicate 25 ␮L aliquots placed into 96-well plates.
After denaturation in urea and dilution, the samples at
pH 9 were put in filter plates containing rehydrated
QHyperD威 F resin (Pall Corporation) and incubated with
shaking. Unbound material was removed on a vacuum
manifold as fraction 1 (FR1), and proteins were eluted in
a step-wise fashion by decreasing pH (FR2, pH 7; FR3, pH
5; FR4, pH 4; and FR5, pH 3) with a final organic solvent
elution (FR6). The 6 fractions were applied to CM10
(weak cation-exchange) and IMAC30-Cu (immobilized
metal affinity capture) arrays in 96-sample bioprocessors
(Ciphergen Biosystems). FR6 samples were also applied
to H50 (hydrophobic) arrays. Chip preparation, sample
application, and matrix application were performed according to the manufacturer’s instructions. All liquid
handling steps were performed on an Aquarius workstation (Tecan).
seldi-tof ms data acquisition and processing
Details of SELDI-TOF MS data acquisition and processing
are provided as Supplemental Data. Briefly, spectra were
acquired on an externally calibrated ProteinChip威 System
Series 4000 instrument, using 2 laser intensities for acquisition of low (2.5–20 kDa) and high (20 –200 kDa) mass
range data. Spectra were processed (baseline subtraction,
deionzing, normalization, spectral alignment, and peak
detection) with CiphergenExpress software, as described
in Supplemental Data. Peak numbers were recorded for
the different handling methods and fraction types. We
examined the differences between the handling methods
by principal components analysis (PCA), hierarchical
647
Clinical Chemistry 53, No. 4, 2007
clustering, and examination of P value by use of mean
peak intensities from triplicate samples. Median variances
were also calculated as descriptors of trends for different
collection/handling methods.
role of the sponsors
peak identification
Ciphergen Biosystems participated in the study design,
donation of reagents, interpretation of results, and writing
of this article, but were not involved in donor selection.
Approval of the paper by Ciphergen Biosystems before its
submission was not required.
Proteins were enriched by liquid chromatography and
ultrafiltration followed by SDS-PAGE. Peptides ⬍5 kDa
were identified by direct sequencing by tandem mass
spectrometry (see below). Proteins ⬎5 kDa were purified
by SDS-PAGE and stained using a Colloidal Blue Staining
Kit (Invitrogen). Selected bands were excised, and one
quarter of each was extracted using 50% formic acid/25%
acetonitrile/15% isopropanol/10% water and reanalyzed
by SELDI-TOF MS to confirm matching of the stained
band with the peak of interest. The remainder was in-gel
digested with trypsin and analyzed by tandem MS using
a Q-STAR威 XL equipped with a PCI-1000 ProteinChip
Interface (Ciphergen Biosystems). MS/MS spectra were
submitted to the database mining tool Mascot (Version
2.1.2; Matrix Science) and searched against the updated
SwissProt or NCBInr databases with the following search
parameters: trypsin, allowing up to 2 missed cleavages (or
semitrypsin if the trypsin search was not successful);
peptide tolerance ⫾50 ppm; MS/MS tolerance ⫾ 0.3 Da;
peptide charge ⫹1. Peak identifications were also confirmed using data from previous publications (3, 18 –31 ).
All are abundant, ubiquitously expressed serum proteins
or proteolytic fragments.
Serum samples were processed 6 different ways (see
Table S1 in the online Data Supplement) to assess the
effects of time and temperature before storage, clotting
time, and storage tube type on SELDI-TOF MS profiles. To
improve coverage, samples were prefractionated using
strong anion-exchange chromatography, and fractions
further spotted onto different SELDI arrays (CM10,
IMAC30-Cu2⫹ and H50) generating 13 different conditions for SELDI-TOF MS analysis. All steps were automated using robotics to improve reproducibility and
throughput. The number of peaks (average of triplicate
samples) in each fraction type was first compared across
the different handling methods (Table 1). Based on previous work, and after preliminary analysis, only fractions
FR1, 4 and 6 were considered since the others gave either
low peak numbers (FR2) or similar patterns of peaks to
those in FR4 (FR3 and FR5). For most fractions, the CM10
surface gave more peaks than the IMAC30 surface, with
FR1 (anion exchange flow-through) on the CM10 chip
surface yielding the most peaks overall. The organic
elution (FR6) gave the highest number of peaks on all chip
Results
Table 1. Numbers of spectral peaks by fraction type and collection method.
Mass range
Protocol
FR1 CM10
FR1 IMAC30
FR4 CM10
FR4 IMAC30
FR6 CM10
FR6 IMAC30
FR6 H50
Total
2.5–20 kDa
20–200 kDa
2.5–200 kDa
ALL
ALL
ALL
117
56
173
71
57
128
64
48
112
52
51
103
82
48
130
71
70
141
77
67
144
534
397
931
2.5–20 kDa
CR
GN
GY
OR
WH
YE
CR
GN
GY
OR
WH
YE
CR
GN
GY
OR
WH
YE
70
80
64
65
69
74
39
44
36
39
39
40
109
124
100
104
108
114
23
45
24
29
30
27
34
34
29
30
29
29
57
79
53
59
59
56
30
39
28
35
43
39
32
34
34
34
32
33
62
73
62
69
75
72
25
31
24
27
27
28
35
33
32
29
32
32
60
64
56
56
59
60
48
51
48
54
54
56
41
43
45
46
46
44
89
94
93
100
100
100
49
45
48
46
48
42
45
41
39
40
39
39
94
86
87
86
87
81
60
61
58
58
56
58
42
40
39
42
40
42
102
101
97
100
96
100
305
352
294
314
327
324
268
269
254
260
257
259
573
621
548
574
584
583
20–200 kDa
2.5–200 kDa
Peaks were picked in CiphergenExpress software using the criteria outlined in Materials and Methods. Peak numbers by fraction type, molecular weight range, and
collection protocol are shown.
648
Tims et al.: Sample Handling Effects on Serum Proteomic Profiles
types. The individual collection methods gave similar
numbers of peaks in each fraction and on each chip type,
although the least stringent protocol 1 (GN) gave an
appreciably higher number in the low mass range, while
protocols 2b (GY) and 2c (CR) gave the lowest peak
numbers (Table 1).
We next conducted PCA to assess how samples and
handling methods grouped together. Protocol 1 (GN) was
the most distinctive method, with most volunteer samples
grouping together and away from samples collected using
the other methods (Fig. 1). This was true for all fraction/
chip surface combinations, but only in the low (2.5–
20 kDa) mass range. A likely explanation for this separation is the extended transport/storage time used in this
protocol. In support of this, partial separation was observed between protocols 2b (GY; 3 h on ice) and 2e
(WH; 3 h at RT), suggesting that temperature before
centrifugation is a major factor influencing spectral patterns (data not shown). Importantly, there was no differences among the protocols, using either mass range, when
samples were transported on ice for 3 vs 6 h (GY vs OR),
when samples were clotted for 60 min vs 5 min (YE vs
GY), or when samples were strawed vs aliquoted into
cryovials for storage (GY vs CR) (data not shown). Protocols were also compared based on P values. For this, the
preprocessed 160 most frequent peaks were selected and
a median intensity value (n ⫽ 3) obtained for each study
Fig. 1. Protocol comparison by principle components analysis.
PCA was performed using Ciphergen Express software to compare samples collected using the different protocols. In the
example shown, peaks from FR1 CM10
2.5–20 kDa were used for the analysis. The
circles denote the clustering of most protocol 1 (GN) samples. A similar clustering of
protocol 1 samples was found for all fraction/chip surface combinations, but only in
the low (2.5–20 kDa) mass range. Note that
the colors used in the Fig. differ from those
used for protocol labeling.
participant and protocol. To test the null-hypothesis that
there is no difference between protocols, P values were
calculated for all 160 peaks using the Wilcoxon sign test.
The Pmin (minimum P value of all 160 values) was then
used to find a measure of agreement between pairs of
protocols by calculating the corresponding “conservative” P value according to the formula: min (n*Pmin;1),
where n ⫽ 160 and the term “min” means that the
minimum of 2 numbers Pmin and 1 is taken; the word
“conservative” refers to the fact that the probability of the
P value not exceeding epsilon is at most epsilon, for any
epsilon between 0 and 1. To create a summary table to
characterize the combination of all 13 fraction/chip types,
the smallest P value in the fractions for each protocol pair
was taken and adjusted according to the formula using
n ⫽ 13 (Table 2). Using this approach, protocols 2b (GY),
2c (CR), and 2d (OR) were most similar and protocol 1
(GN) most dissimilar, in agreement with the PCA. However, there was no significant difference between protocols 1 and 2c, (P ⫽ 0.276), although protocol 2c had the
smaller number of 13 samples, making it difficult to make
reliable conclusions. This may also explain why all other
protocols showed a strong agreement with protocol 2c.
Notably, there was a significant difference between protocols 2a (YE) and 2b (GY), suggesting that clotting time
does influence the protein profiles, a finding which was
not apparent from the PCA.
649
Clinical Chemistry 53, No. 4, 2007
Table 2. Protocol comparison based on Pmin values.
Protocol
1 (GN)
2a (YE)
2b (GY)
2c (CR)
2d (OR)
2e (WH)
1 (GN)
2a (YE)
2b (GY)
2c (CR)
2d (OR)
2e (WH)
1
0.000135
0.00027
0.276123
6.74E-05
6.74E-05
0.000135
1
0.000576
0.276123
0.000144
0.043129
0.00027
0.000576
1
0.723633
0.265808
0.000288
0.276123
0.276123
0.723633
1
0.507813
0.276123
6.74E-05
0.000144
0.265808
0.507813
1
7.21E-05
6.74E-05
0.043129
0.000288
0.276123
7.21E-05
1
Wilcoxon sign test P values were calculated for pairs of protocols by using the median intensity values (n⫽3) for the preprocessed 160 most frequent peaks. The
minimum P value (Pmin) was then used to find a measure of agreement between pairs of protocols by calculating the corresponding “conservative” P value according
to the formula: min (n ⫻ Pmin;1), where n⫽160. To create the summary table shown, to characterize the combination of all 13 fraction/chip types, the smallest P value
in the 13 fractions/chip types for each protocol pair was taken and adjusted according to the formula, using n⫽13.
We next analyzed peak variances to assess the general
stability of the handling methods. Data were calibrated
internally based on known peaks, and median intensities
and SDs taken for each peak across the 25 study participants. These values were used to calculate a coefficient of
variance (SD/median intensity) for each protocol and
fraction type. Median variance values were also calculated and compared across protocols to give an overall
measure of variability. Protocol 2a (YE) gave the lowest
median variances in both mass ranges, followed by protocols 1 (GN) and 2e (WH), suggesting that these were the
most stable methods (Fig. 2A and B). This was corroborated by the observation that these methods gave the
highest numbers of peaks with a median variance ⬍1.0
(Fig. 2C).
Selection of peaks for identification was based on
altered intensity and variance across protocols, with emphasis on differences between protocols 1 (GN) and 2a
(YE); considered the least and most stringent protocols,
respectively. Identifications were in agreement with previous studies, where available (Table 3). Examples of two
peaks that displayed increased intensity in protocol 1
samples, but not those of protocol 2a, are shown in Fig. 3.
Peak 4286 Da (Fig. 3A) was identified as the 4281.78 Da
fragment of inter␣-trypsin inhibitor heavy chain H4
[ITIH4) (see (22 )]. Two other ITIH4 fragments (3157.58 Da
[see (27 )] and 3955.48 Da [see (31 )] and their methionineoxidised forms were also identified, with all displaying
increased intensity in protocol 1 samples (Table 3). An
8144 Da peak appeared to be a superposition of two
peaks. In most protocols the 8144 Da peak represented an
8141.59 Da form of platelet factor 4 (30 ), with an alternatively cleaved signal sequence (data not shown). In protocol 1, the up-regulated peak 8126 Da corresponded to a
C-terminal-truncated fragment of C3a anaphylatoxin
[8126.52 Da; see (3, 23 )] (Fig. 3C). A C1 inhibitor Cterminal fragment (4152.87 Da), an albumin N-terminal
fragment (3156.59 Da), and peaks corresponding to neutrophil defensins 1, 2, and 3 were also noticeably increased in protocol 1 samples. Other identifications
were apolipoprotein C-I (ApoCI; 6330.59 Da) and its
SPA adduct, a truncated form of ApoCII (8204.17 Da),
albumin dimer (138 kDa), transferrin (79 kDa), and trans-
thyretin (13.9 kDa), which were relatively stable across the
different handling methods. In contrast, hemoglobin ␣
(15126.36 Da) and ␤ (15867.28 Da), and fibrinogen ␣
fragments 3262.47 Da and 5904.22 Da and their modified
forms, displayed lower intensities in protocols 1 (GN) and
2e (WH) (Table 3). Several peaks, including those representing the major form of platelet factor 4 [7765.10 Da; see
(30 )], showed altered intensity in protocol 2b (GY) vs 2a
(YE) and 1 (GN), but not the other protocols, suggesting
that their final serum concentration is affected by clotting
time (highlighted in bold in Table 3).
Discussion
One of the main objectives of this study was to select an
optimal protocol for serum collection and handling for a
large case-control study, which would be feasible for both
clinical collection and protein profiling. Variance analysis
showed that protocol 2a (YE) gave the lowest overall
variance when all peaks were considered, and is therefore
the preferred serum collection/handling method. This
method, while clinically feasible, requires rapid transit of
samples on ice to the laboratory for processing and
freezing, and imposes additional logistic and resource
burdens on clinical studies. No obvious effects on overall
peak number, median intensity, or variance were apparent between the various protocols (GY, YE, CR, OR)
where samples were placed on ice. This indicates that
transport times of up to 6 h, and different storage methods
(straws vs cryovials), have little effect on serum protein
profiles as long as samples are on ice. However, the
altered intensity of some peaks (e.g., platelet factor 4 at
7765.18 Da) did show that altering the clotting time has
some effects. Perhaps not surprisingly, PCA showed the
greatest separation for protocol 1 over other methods.
Separation was representative for all 2.5–20 kDa spectra
on all arrays, and was not apparent in the 20 –200 kDa
range, in accordance with increased peak numbers and
intensities in the lower mass range. There was minor
separation attributable to transport/storage at RT vs ice
(GY vs WH). Thus, although strict control on sample
handling protocols has been proposed to reduce variability, these data show that sample collection protocols can
be more flexible with regard to clotting times and trans-
650
Tims et al.: Sample Handling Effects on Serum Proteomic Profiles
Fig. 2. Variance analysis.
(A), median peak variances (standard deviation/median peak intensity) by protocol
and fraction type. Numbers in italics are
greater than an arbitrary cutoff value of 0.6.
(B), graphical representation of median
peak variances for the low (left) and high
(right) mass ranges. (C), number (#) and
percentage (%) of peaks where median
peak variance (SD/median peak intensity)
was ⬍1.0.
port; transport for 3 h at RT or up to 6 h on ice can be used
in clinical studies with limited impact on variability. The
critical issue is that all individual samples must be treated
exactly the same.
The stringent protocol with 1 h clotting and 3 h
transit/storage on ice (YE) has been shown to give the
most reproducible results in a serum profiling analysis
performed using automated magnetic bead-based prefractionation and MALDI-TOF MS (14 ). However, our
study showed that despite transit/storage at RT for 30 h,
protocol 1 (GN) also gave a relatively low overall variance. This finding is critical, because it establishes that
samples collected in older studies with longer transit
times at RT can be used in case-control studies for novel
biomarkers as long as all samples were handled similarly.
Many of these biobanks are unique because they are
associated with long follow-up, and contain samples
stored many years before disease diagnosis.
The greater number of peaks (often with lower variances and higher peak intensities) in the low mass range
for protocol 1 samples suggests that proteolysis in these
samples has gone to completion. A similar trend was also
apparent in protocol 2e (WH) samples, which were incubated for 3 h at RT before storage. These data are in
agreement with a previous SELDI-TOF MS study showing increases in certain peaks with time between venipuncture and sample processing, with some overlap with
the peaks identified here (15 ). In particular, the increased
intensities of the ITIH4, C3a, and C1 inhibitor fragments
and their modified forms in protocol 1 samples provides
F1 CM1020–
200 kDa
F1 CM102.5–
20 kDa
3193.2
3264.5
3423.0
3489.1
3955.8
3969.9
4072.7
4137.6
4286.0
4677.0
5074.4
5492.2
5764.6
5905.3
5914.0
6114.0
6637.8
6843.7
8144.0
8344.0
15158.0
19
21
26
29
33
34
35
37
39
43
47
54
55
57
58
60
63
64
75
76
109
20 26573.0
40 79510.0
43 102333.0
50 138323.0
2770.5
2851.3
2870.9
2954.7
3075.2
3141.3
3159.0
MW (int
cal)
6
8
9
11
14
17
18
Cluster
Fraction type
index
Albumin dimer
Fibrinogen ␣ frag,
5904.22 Da
5904.22 Met-ox
5904.22 SPA adduct
ApoCI, 6630.59 Da
6630.59 Da,
SPA adduct
C3a truncated,
8126.52 Da
8126.52 Da,
SPA adduct
Hb alpha,
15126.36 Da
Transferrin, 3⫹ ion
Transferrin
ITIH4 fragment,
4281.78 Da
Neutrophil defensin 3,
3486.1 Da
ITIH4 fragment,
3955.46 Da
3955.46 Da Met-ox
Fibrinogen ␣ frag,
3262.47 Da
ITIH4 fragment,
3157.58 Da
5904.22 Da, 2⫹ ion
Identification
(20, 29 )
(3, 22 )
(3, 23 )
(3, 29 )
(25 )
(22 )
(31 )
(21, 24 )
(31 )
(27 )
Refs
CR
0.05
3.18
0.06
0.22
0.15
0.13
0.82
7.96
1.29
13.96
1.62
0.22
0.12
0.21
0.06
13.38
0.36
0.17
0.25
0.42
0.80
0.29
0.11
1.51
2.42
0.15
0.13
1.01
1.83
0.04
0.19
0.61
GN
0.25
2.93
0.05
0.23
0.06
0.40
5.07
1.93
0.31
10.63
1.15
0.98
0.42
0.05
0.47
3.01
1.13
0.79
1.71
1.66
4.16
0.08
0.32
0.02
0.01
0.02
0.51
6.27
0.44
0.52
0.54
2.78
0.04
2.65
0.05
0.23
0.31
0.11
0.76
6.80
0.98
14.97
1.61
0.12
0.11
0.17
0.04
10.06
0.60
0.08
0.46
0.97
0.88
0.25
0.12
1.65
2.11
0.15
0.05
0.77
1.49
0.05
0.14
1.10
GY
0.04
2.67
0.05
0.23
0.33
0.15
0.55
8.39
1.49
14.56
1.59
0.11
0.11
0.18
0.05
13.96
0.47
0.07
0.49
0.84
0.76
0.21
0.06
1.84
2.30
0.22
0.06
1.20
1.93
0.06
0.20
0.89
OR
Median
0.06
2.81
0.05
0.23
0.10
0.13
1.11
8.16
1.36
12.53
1.34
0.23
0.17
0.20
0.11
14.38
0.30
0.20
0.77
0.50
0.86
0.12
0.17
1.67
2.72
0.18
0.10
2.02
1.78
0.14
0.22
1.15
WH
0.07
2.66
0.05
0.24
0.22
0.16
1.25
7.87
1.30
12.14
1.41
0.28
0.18
0.19
0.12
12.43
0.43
0.22
0.61
0.84
0.92
0.13
0.10
1.65
2.45
0.20
0.11
1.98
1.67
0.12
0.24
1.13
YE
2.06
0.38
0.34
0.56
0.52
0.79
1.52
0.48
0.58
0.24
0.28
2.36
2.42
0.43
4.18
0.54
1.78
1.13
4.01
3.43
1.88
0.47
0.85
0.93
0.77
1.66
2.11
2.63
0.56
6.12
1.61
1.70
CR
1.13
0.22
0.14
0.21
1.99
0.48
0.52
1.55
1.38
0.32
0.37
0.92
0.71
2.20
0.78
1.39
0.78
0.62
0.40
0.75
0.48
1.16
0.91
28.3
70.3
5.14
0.52
0.54
1.16
0.62
0.65
0.50
GN
Table 3. Selected peaks displaying differential intensity and variance across handling protocols.
4.55
0.38
0.17
0.28
0.54
0.76
1.28
0.62
0.58
0.35
0.38
2.15
1.17
0.66
2.76
0.52
4.20
1.82
1.05
3.74
2.53
0.59
0.98
0.62
0.64
1.12
3.05
2.05
0.50
3.01
1.56
0.88
GY
1.19
0.29
0.23
0.26
0.49
0.60
0.81
0.34
0.46
0.28
0.30
1.36
1.48
0.79
1.63
0.48
4.06
1.43
0.78
3.58
2.55
0.68
1.09
0.72
0.72
0.75
2.44
1.10
0.47
2.32
0.87
1.03
OR
SD/Median
WH
1.04
0.32
0.23
0.30
2.05
0.50
0.61
0.26
0.54
0.28
0.30
1.10
1.08
0.73
1.46
0.44
3.89
0.66
0.67
3.57
1.18
0.87
0.69
0.54
0.50
0.97
2.54
1.23
0.56
1.77
0.94
0.86
YE
1.57
0.19
0.14
0.16
0.40
0.41
0.57
0.30
0.53
0.20
0.20
1.31
0.95
0.56
1.55
0.50
1.43
0.61
0.87
1.64
1.14
0.59
1.03
0.53
0.56
0.53
1.68
0.99
0.53
1.63
1.13
0.66
Clinical Chemistry 53, No. 4, 2007
651
F4 CM10 2.5–
20 kDa
F1 IMAC20–
200 kDa
F1 IMAC 2.5–
20 kDa
3394.8
3956.1
3975.8
4284.1
4297.5
5767.6
5905.5
5921.4
5970.1
6113.8
6850.8
7159.7
8131.5
8159.1
11436.0
15913.0
16
19
20
24
25
30
31
32
33
34
38
39
46
47
56
68
36
7045.8
69 16073.0
1 20037.0
24 26560.0
27 30461.0
31 39771.0
41 79249.0
42 89545.0
44 101944.0
45 106330.0
51 137052.0
19
3817.1
26
4965.7
27
5070.4
29
5755.2
35
6955.1
2863.6
2950.9
3156.0
MW (int
cal)
10
11
12
Cluster
Fraction type
index
Transthyretin, cysteinylated,
2⫹ ion
Transthyretin,
glutathionylated, 2⫹ ion
Transferrin ion
Albumin dimer
Transferrin, 2⫹ ion
Transferrin
Transferrin, 3⫹ ion
Hb beta,
15867.27 Da
C3a truncated, 8126.52
Da
5904.22 Da SPA adduct
ApoCI, SPA adduct
Fibrinogen ␣ frag,
5904.22 Da
5904.22 Da Met-ox
ITIH4 fragment,
3955.46 Da
3955.46, Met-ox
ITIH4 fragment,
4281.78 Da
4281.78, Met-ox
ITIH4 fragment,
3157.58 Da
Identification
(29 )
(29 )
(3, 22 )
(3, 23 )
(25 )
(22 )
(31 )
(27 )
Refs
CR
0.70
0.13
0.03
0.06
0.05
0.29
2.05
0.16
0.06
0.03
0.21
1.13
0.17
0.31
0.41
0.78
0.33
0.22
0.19
3.64
0.39
0.53
0.74
0.30
0.53
0.31
0.19
6.02
0.40
0.51
0.25
0.62
1.58
1.37
0.84
GN
0.74
0.04
0.10
0.19
0.06
0.37
2.79
0.18
0.05
0.02
0.21
1.01
1.77
0.06
0.13
1.07
1.85
0.10
0.08
0.80
0.20
0.13
0.52
0.64
3.11
0.91
0.57
1.12
3.14
1.58
1.30
4.86
5.62
0.35
2.47
0.68
0.21
0.03
0.05
0.08
0.35
2.37
0.16
0.05
0.03
0.24
1.11
0.13
0.24
0.40
0.98
0.31
0.21
0.33
3.14
0.53
0.63
0.76
0.16
0.34
0.50
0.09
5.77
0.59
0.50
0.26
0.54
1.39
1.10
0.62
GY
0.68
0.22
0.03
0.05
0.07
0.37
2.70
0.16
0.05
0.03
0.22
0.99
0.13
0.27
0.39
0.89
0.36
0.15
0.35
3.23
0.46
0.63
0.92
0.17
0.34
0.42
0.11
6.80
0.41
0.63
0.27
0.63
1.33
1.22
0.78
OR
Median
Table 3. Continued.
WH
0.76
0.07
0.03
0.06
0.07
0.40
3.17
0.18
0.05
0.02
0.21
1.01
0.53
0.41
0.51
0.95
0.81
0.19
0.13
4.22
0.59
0.93
0.84
0.23
0.62
0.37
0.19
8.30
0.32
0.73
0.37
0.78
2.27
1.26
1.02
YE
0.78
0.19
0.04
0.05
0.07
0.38
2.86
0.17
0.05
0.02
0.21
1.05
0.28
0.42
0.82
0.96
0.74
0.17
0.31
4.00
0.60
0.86
0.99
0.19
0.33
0.27
0.11
7.11
0.43
0.67
0.45
0.58
1.44
1.33
1.00
CR
0.32
0.40
1.10
0.97
0.72
0.37
0.38
0.30
0.20
0.47
0.31
0.42
1.11
0.67
0.42
0.29
2.52
0.41
0.40
0.66
1.01
0.83
0.46
0.51
3.15
1.99
0.95
0.80
4.50
1.59
2.33
4.62
0.80
0.41
1.04
GN
0.16
1.12
0.78
0.98
0.36
0.23
0.22
0.32
0.28
0.50
0.32
0.29
0.54
2.07
1.06
0.21
0.63
0.70
0.83
0.94
0.64
1.97
0.79
0.68
0.69
1.76
0.80
1.30
0.63
2.04
0.89
0.87
0.52
1.33
0.53
0.24
0.56
1.23
0.65
0.32
0.16
0.25
0.21
0.25
0.33
0.26
0.41
1.32
1.04
0.35
0.35
0.80
3.21
0.49
0.86
0.67
0.66
0.40
0.73
0.85
1.09
1.78
0.74
0.91
1.58
1.73
1.40
1.41
0.69
1.34
GY
0.28
0.57
0.93
0.45
0.52
0.20
0.21
0.29
0.27
0.23
0.34
0.40
1.25
0.64
0.38
0.26
1.43
0.35
0.58
0.71
0.85
0.92
0.36
0.96
0.79
2.31
1.02
0.89
5.69
4.44
1.34
3.85
0.51
0.54
0.44
OR
SD/Median
WH
0.18
1.54
0.74
0.87
0.38
0.24
0.26
0.23
0.26
0.36
0.29
0.32
0.85
0.41
0.50
0.22
0.58
0.49
1.35
0.60
0.62
0.54
0.40
1.19
0.79
2.16
1.79
0.57
6.63
3.01
2.11
3.34
1.24
0.49
1.09
YE
0.18
0.32
0.58
0.60
0.22
0.20
0.22
0.27
0.18
0.25
0.25
0.22
0.87
0.34
0.28
0.24
0.58
0.30
0.30
0.40
0.48
0.41
0.29
0.71
1.43
4.42
1.52
0.46
6.90
5.50
0.62
5.35
0.48
0.40
0.46
652
Tims et al.: Sample Handling Effects on Serum Proteomic Profiles
7568.1
13795.0
13915.0
14091.0
14289.0
2864.5
3156.5
3370.5
3815.1
3881.5
4125.6
4801.6
5754.9
6951.3
7042.9
7146.8
7760.2
13747.0
13888.0
14072.0
14275.0
15884.0
3371.4
3442.0
3473.6
3489.6
4051.4
4153.3
7931.8
8207.2
15148.0
38
54
55
56
57
7
10
12
14
15
16
19
20
27
28
29
30
40
41
42
43
47
14
15
16
17
19
20
43
44
72
F6 CM10 2.5–
20 kDa
F4 IMAC 2.5–
20 kDa
MW (int
cal)
Cluster
Fraction type
index
Identification
C1 inhibitor C-term frag,
4152.87 Da
Hb beta, 15867.27 Da, 2⫹
ion
ApoCII, truncated, 8204.17
Hb alpha, 15126.36 Da
Neutrophil defensin 3,
3486.1 Da
Platelet factor 4,
7765.18 Da
Transthyretin
Transthyretin, cysteinylated
Transthyretin,
glutathionylated
Transthyretin SPA adduct
Hb beta, 15867.27 Da
Neutrophil defensin 2,
3371.01 Da
Neutrophil defensin 1,
3442.09 Da
Transthyretin, cysteinylated,
2⫹ ion
Transthyretin,
glutathionylated, 2⫹ ion
Albumin, N-term frag,
3156.59 Da
Neutrophil defensin 2,
3371.01 Da
Hb alpha, 2⫹ion
Transthyretin
Transthyretin, cysteinylated
Transthyretin,
glutathionylated
Transthyretin SPA adduct
(26, 28 )
(24 )
(21, 24 )
(3, 22 )
(21, 24 )
(26, 28 )
(26, 28 )
(26, 28 )
(30 )
(24 )
(3 )
(26, 28 )
(26, 28 )
(26, 28 )
Refs
CR
0.36
0.39
0.20
0.21
0.82
0.51
0.28
0.69
1.19
0.48
1.43
1.44
4.60
3.91
0.51
1.49
1.63
3.04
0.48
0.76
0.36
0.45
2.37
0.50
0.35
1.62
2.01
0.23
0.55
1.82
1.61
GN
0.37
0.08
0.05
1.04
2.13
0.00
2.21
6.38
1.12
0.12
6.33
1.26
4.72
4.09
0.63
4.48
1.78
2.49
0.99
1.89
1.53
0.16
2.54
2.94
0.42
3.97
4.70
0.06
0.52
1.95
1.90
0.38
0.22
0.16
0.12
0.59
0.43
0.24
0.55
1.13
0.41
1.42
1.42
4.40
3.90
0.54
1.47
1.68
2.31
0.25
1.02
0.57
0.39
2.32
0.42
0.39
1.86
1.73
0.16
0.69
1.87
1.70
GY
0.37
0.32
0.14
0.19
0.73
0.45
0.23
0.47
1.19
0.34
1.32
1.43
4.68
3.95
0.54
1.22
1.71
2.73
0.17
0.94
0.33
0.44
2.30
0.44
0.45
1.76
1.80
0.25
0.64
2.06
2.00
OR
Median
Table 3. Continued.
WH
0.39
0.11
0.07
0.45
0.76
0.17
0.57
1.45
1.25
0.15
2.35
1.51
4.59
4.27
0.52
4.06
1.72
2.43
1.37
0.91
0.54
0.55
2.45
0.74
0.48
1.87
2.28
0.07
0.68
2.08
2.08
YE
0.34
0.18
0.12
0.30
0.70
0.28
0.34
0.80
1.18
0.24
1.78
1.47
4.65
4.11
0.52
3.72
1.70
2.74
1.04
0.94
0.58
0.91
2.35
0.21
0.45
1.52
2.17
0.14
0.67
2.11
2.09
CR
0.43
0.60
0.56
1.08
0.78
0.71
0.67
0.96
0.29
0.44
0.66
0.19
0.19
0.23
0.70
0.83
0.26
0.22
0.98
0.99
1.33
0.55
0.20
0.90
0.32
0.84
0.42
0.51
0.33
0.36
0.46
GN
0.43
0.57
0.58
0.59
0.83
127
0.98
0.70
0.24
0.58
0.43
0.24
0.17
0.20
0.23
0.43
0.15
0.31
0.77
0.30
0.44
1.24
0.17
0.45
0.24
0.58
0.43
1.31
0.20
0.18
0.26
0.46
0.86
0.56
1.10
1.05
1.07
1.12
0.83
0.34
0.44
0.67
0.32
0.31
0.32
0.46
0.51
0.30
0.36
1.95
0.75
0.82
0.59
0.28
1.49
0.30
0.50
0.48
0.91
0.29
0.34
0.35
GY
0.44
0.70
0.69
0.59
0.70
1.16
0.84
0.97
0.24
0.48
0.71
0.24
0.23
0.23
0.36
0.73
0.18
0.34
3.54
0.83
1.36
0.59
0.25
1.10
0.27
0.46
0.37
0.51
0.29
0.28
0.30
OR
SD/Median
WH
0.42
0.93
0.85
0.59
0.74
1.41
0.71
0.89
0.23
1.35
0.49
0.19
0.17
0.19
0.32
0.55
0.17
0.26
0.53
0.68
1.19
0.55
0.18
0.84
0.22
0.48
0.36
1.74
0.19
0.20
0.27
YE
0.45
0.64
0.45
0.68
0.76
1.75
1.26
0.75
0.22
0.37
0.52
0.16
0.13
0.16
0.21
0.46
0.13
0.20
0.66
0.63
0.76
0.30
0.18
2.63
0.25
0.50
0.28
0.68
0.18
0.15
0.21
Clinical Chemistry 53, No. 4, 2007
653
654
8143.5
15853.0
82171.0
83506.0
32526.0
33453.0
42
68
51
52
27
28
Albumin, 2⫹ ion
4170.6
7759.6
23
40
F6 H50 20–
200 kDa
3483.5
21
F6 IMAC20–
200 kDa
3440.4
20
Internally calibrated mass, peak identification, references, median peak intensity, and median variance (SD/median intensity) across different handling protocols are shown. Median intensity values in bold indicate
peaks that change due to altered clotting time. Median variance values shaded gray and in italics are ⬎1.0.
0.60
0.61
0.30
0.19
1.03
0.44
0.54
1.20
0.56
0.23
1.18
0.42
1.44
0.91
0.53
0.34
0.58
0.60
0.90
1.16
0.54
0.28
0.92
0.48
0.67
0.62
0.82
0.33
4.84
1.02
1.24
0.51
0.32
0.27
0.80
0.84
0.08
0.14
0.04
0.05
0.03
0.32
0.11
0.08
0.03
0.04
0.02
0.25
0.03
0.20
0.05
0.05
0.04
0.28
0.03
0.21
0.04
0.05
0.03
0.27
0.11
0.06
0.01
0.02
0.01
0.16
0.04
0.29
0.06
0.06
0.04
0.32
(3, 22 )
Hb beta, 15867.27 Da
0.82
0.36
0.75
0.56
0.80
0.62
0.91
0.50
0.80
0.54
0.80
0.63
0.26
0.52
0.36
0.48
0.27
0.20
0.28
0.22
0.80
0.41
0.31
0.19
(30 )
1.82
1.00
1.26
2.51
0.71
1.50
0.28
0.46
0.11
0.08
2.93
0.16
(21, 24 )
0.66
0.65
0.63
0.57
0.86
0.58
0.89
1.26
0.49
0.48
5.11
0.59
(21, 24 )
3371.4
19
YE
0.70
0.48
0.61
0.75
0.53
0.59
1.33
1.59
0.99
0.86
4.40
0.85
(21, 24 )
1.12
WH
OR
0.71
0.58
GY
GN
0.48
0.53
CR
YE
0.66
0.63
WH
OR
GY
0.46
SD/Median
Median
0.47
2.89
GN
CR
0.44
(3 )
Refs
Identification
Albumin, N-term frag,
3156.59 Da
Neutrophil defensin 2,
3371.01 Da
Neutrophil defensin 1,
3442.09 Da
Neutrophil defensin 3,
3486.1 Da
4152.87 Da Met-ox
Platelet factor 4, 7765.18
Da
3153.6
16
F6 IMAC 2.5–
20 kDa
MW (int
cal)
Cluster
Fraction type
index
Table 3. Continued.
0.39
Tims et al.: Sample Handling Effects on Serum Proteomic Profiles
evidence that these degradation products are generated as
a result of increased proteolysis due to extended transport/storage. Conversely, full-length hemoglobin ␣ and ␤
displayed decreased intensities in protocols 1 (GN) and 2e
(WH), suggesting that they may be subject to degradation.
Similarly, the decreases in fibrinogen ␣ fragments 3262.47
and 5904.22 Da were consistent with further degradation
to smaller undetected forms. It is harder to explain the
increased levels of the neutrophil defensins in the protocol 1 samples. These disulfide bond-containing molecules
are resistant to proteolysis, so an indirect mechanism
must account for their altered intensities. Several other
protein forms did not change significantly with collection
method, revealing them to be relatively stable serum
markers.
It has been suggested that many candidate disease
biomarkers identified in SELDI-TOF MS profiling experiments are abundant acute-phase reactants, and are thus
secondary effects of the diseased state (7 ). For example,
serum transthyretin (TTR) is a known marker for nutritional status and the inflammatory acute-phase response.
In ovarian cancer, TTR was identified as a potential early
diagnostic marker, with decreased TTR levels reported in
the sera of ovarian cancer patients compared with controls, without differences in its microheterogeneity
(20, 32, 33 ). Notably, TTR and its modified forms were
unaffected by the different handling conditions used here,
and displayed relatively low variances across this healthy
cohort. Thus, it would appear that TTR is relatively stable,
making it a more robust disease biomarker. Similarly,
SELDI-TOF MS was previously used to detect ApoCI and
transferrin as classifiers of ovarian, colorectal, and other
cancers (20, 34 ), and our data show that they are also
stable under the conditions tested.
In a recent study, the putative acute-phase protein
ITIH4 was shown to be extensively proteolytically processed, and its fragmentation patterns associated with
different disease conditions (27, 31 ). Fragmentation was
generally consistent with cleavages by endoproteases,
followed by exoproteases, and the observed fragments
were reported to change little under different assay conditions or processing procedures. An up-regulated cleavage fragment of ITIH4 was also shown to enable differentiation of patients with ovarian cancer from healthy
controls or patients with benign pelvic masses (32 ). Our
data provide evidence that ITIH4 is relatively unstable,
with the generation of fragments increasing in serum
maintained at RT for prolonged periods. Hemoglobin ␤
has also been identified as a putative ovarian cancer
biomarker (3, 20 ), but appears from our study to be
relatively unstable in serum. With this in mind, ITIH4
fragments or hemoglobin ␤ may not make robust disease
biomarkers unless strict precautions are taken with sample handling. Future work will involve additional MS/
MS-based identification of the unknown peaks that are
discriminatory for the different collection methods. This
will allow the assessment of their usefulness as potential
655
Clinical Chemistry 53, No. 4, 2007
Fig. 3. Example spectra.
(A), example spectra showing peak 4286
Da (inter␣-trypsin inhibitor heavy chain H4
(ITIH4) fragment, 4281.78 Da) of FR1
CM10 across protocols. Median intensities and standard deviations for all samples are also shown. The peak displayed
a highest intensity in protocol 1 (GN) and
may be generated as a result of increased
proteolysis due to the extended transport/storage at RT used in this protocol.
(B), example showing peak intensities for
8144 Da of FR1 CM10, also increased in
protocol 1. (C), the 8144 Da peak is a
superposition of 2 peaks representing a
C-terminal-truncated form of C3a anaphylatoxin (8126.52 Da), abundant in protocol 1 (GN) samples, and an N-terminal
variant of platelet factor 4 (8141.59 Da;
PF4). The spectra shown have not been
normalized.
disease biomarkers where they have been identified in
other studies.
Our work establishes that the proteomic analysis of
samples from established serum banks, where samples
were not collected in accordance with more stringent
protocols, can be used for proteomic biomarker studies.
The key factor is that all samples in the collection should
have been handled in a similar manner. Cases and controls should be matched for transport time and an assessment made when proteolysis in these samples reaches
completion. Biomarker discovery using a proteomic approach in such case-control sets, such as UKCTOCS, will
involve stable biomarkers rather than labile proteins. For
future studies, the key variable during specimen collec-
656
Tims et al.: Sample Handling Effects on Serum Proteomic Profiles
tion will be transport on ice, and it does not seem to
matter if transport times are then 3 or 6 h.
Financial support from The Eve Appeal and Ciphergen
Biosystems Inc. is gratefully acknowledged.
References
1. Wulfkuhle JD, Liotta LA, Petricoin EF. Proteomic applications for
the early detection of cancer. Nat Rev Cancer 2003;3:267–75.
2. Xiao Z, Prieto D, Conrads TP, Veenstra TD, Issaq HJ. Proteomic
patterns: their potential for disease diagnosis. Mol Cell Endocrinol
2005;230:95–106.
3. Engwegen JY, Gast MC, Schellens JH, Beijnen JH. Clinical proteomics: searching for better tumour markers with SELDI-TOF
mass spectrometry. Trends Pharmacol Sci 2006;27:251–9.
4. Ludwig JA, Weinstein JN. Biomarkers in cancer staging, prognosis
and treatment selection. Nat Rev Cancer 2005;5:845–56.
5. Chatterjee SK, Zetter BR. Cancer biomarkers: knowing the present
and predicting the future. Future Oncol 2005;1:37–50.
6. Brioschi PA, Irion O, Bischof P, Bader M, Forni M, Krauer F. Serum
CA 125 in epithelial ovarian cancer: a longitudinal study. Br J
Obstet Gynaecol 1987;94:196 –201.
7. Diamandis EP. Mass spectrometry as a diagnostic and a cancer
biomarker discovery tool: opportunities and potential limitations.
Mol Cell Proteomics 2004;3:367–78.
8. Merchant M, Weinberger SR. Recent advancements in surfaceenhanced laser desorption/ionization-time of flight-mass spectrometry. Electrophoresis 2000;21:1164 –77.
9. Petricoin EF, Liotta LA. SELDI-TOF-based serum proteomic pattern
diagnostics for early detection of cancer. Curr Opin Biotechnol
2004;15:24 –30.
10. Baggerly KA, Morris JS, Edmonson SR, Coombes KR. Signal in
noise: evaluating reported reproducibility of serum proteomic
tests for ovarian cancer. J Natl Cancer Inst 2005;97:307–9.
11. Diamandis EP. Serum proteomic profiling by matrix-assisted laser
desorption-ionization time-of-flight mass spectrometry for cancer
diagnosis: next steps. Cancer Res 2006;66:5540 –1.
12. Coombes KR, Morris JS, Hu J, Edmonson SR, Baggerly KA. Serum
proteomics profiling–a young technology begins to mature. Nat
Biotechnol 2005;23:291–2.
13. Karsan A, Eigl BJ, Flibotte S, Gelmon K, Switzer P, Hassell P, et al.
Analytical and preanalytical biases in serum proteomic pattern
analysis for breast cancer diagnosis. Clin Chem 2005;51:
1525– 8.
14. Villanueva J, Philip J, Chaparro CA, Li Y, Toledo-Crow R, DeNoyer
L, et al. Correcting common errors in identifying cancer-specific
serum peptide signatures. J Proteome Res 2005;4:1060 –72.
15. Banks RE, Stanley AJ, Cairns DA, Barrett JH, Clarke P, Thompson
D, Selby PJ. Influences of blood sample processing on lowmolecular-weight proteome identified by surface-enhanced laser
desorption/ionization mass spectrometry. Clin Chem 2005;51:
1637– 49.
16. Baumann S, Ceglarek U, Fiedler GM, Lembcke J, Leichtle A, Thiery
J. Standardized approach to proteome profiling of human serum
based on magnetic bead separation and matrix-assisted laser
desorption/ionization time-of-flight mass spectrometry. Clin Chem
2005;51:973– 80.
17. Findeisen P, Sismanidis D, Riedl M, Costina V, Neumaier M.
Preanalytical impact of sample handling on proteome profiling
experiments with matrix-assisted laser desorption/ionization
time-of-flight mass spectrometry. Clin Chem 2005;51:2409 –11.
18. Ruetschi U, Zetterberg H, Podust VN, Gottfries J, Li S, Hviid
Simonsen A, et al. Identification of CSF biomarkers for frontotemporal dementia using SELDI-TOF. Exp Neurol 2005;196:273– 81.
19. Fung ET, Yip TT, Lomas L, Wang Z, Yip C, Meng XY, et al.
Classification of cancer types by measuring variants of host
response proteins using SELDI serum assays. Int J Cancer
2005;115:783–9.
20. Kozak KR, Su F, Whitelegge JP, Faull K, Reddy S, Farias-Eisner R.
Characterization of serum biomarkers for detection of early stage
ovarian cancer. Proteomics 2005;5:4589 –96.
21. Albrethsen J, Bogebo R, Gammeltoft S, Olsen J, Winther B, Raskov
H. Upregulated expression of human neutrophil peptides 1, 2 and
3 (HNP 1–3) in colon cancer serum and tumours: a biomarker
study. BMC Cancer 2005;5:8.
22. Koomen JM, Shih LN, Coombes KR, Li D, Xiao LC, Fidler IJ, et al.
Plasma protein profiling for diagnosis of pancreatic cancer reveals
the presence of host response proteins. Clin Cancer Res 2005;
11:1110 – 8.
23. Li J, Orlandi R, White CN, Rosenzweig J, Zhao J, Seregni E, et al.
Independent validation of candidate breast cancer serum biomarkers identified by mass spectrometry. Clin Chem 2005;51:2229 –
35.
24. Melle C, Ernst G, Schimmel B, Bleul A, Thieme H, Kaufmann R, et
al. Discovery and identification of alpha-defensins as low abundant, tumor-derived serum markers in colorectal cancer. Gastroenterology 2005;129:66 –73.
25. Nomura F, Tomonaga T, Sogawa K, Ohashi T, Nezu M, Sunaga M,
et al. Identification of novel and downregulated biomarkers for
alcoholism by surface enhanced laser desorption/ionization-mass
spectrometry. Proteomics 2004;4:1187–94.
26. Schweigert FJ, Wirth K, Raila J. Characterization of the microheterogeneity of transthyretin in plasma and urine using SELDITOF-MS immunoassay. Proteome Sci 2004;2:5.
27. Song J, Patel M, Rosenzweig CN, Chan-Li Y, Sokoll LJ, Fung ET, et
al. Quantification of fragments of human serum inter-alpha-trypsin
inhibitor heavy chain 4 by a surface-enhanced laser desorption/
ionization-based immunoassay. Clin Chem 2006;52:1045–53.
28. Terazaki H, Ando Y, Suhr O, Ohlsson PI, Obayashi K, Yamashita T,
et al. Post-translational modification of transthyretin in plasma.
Biochem Biophys Res Commun 1998;249:26 –30.
29. Ward DG, Suggett N, Cheng Y, Wei W, Johnson H, Billingham LJ,
et al. Identification of serum biomarkers for colon cancer by
proteomic analysis. Br J Cancer 2006;94:1898 –905.
30. Vermeulen R, Lan Q, Zhang L, Gunn L, McCarthy D, Woodbury RL,
et al. Decreased levels of CXC-chemokines in serum of benzeneexposed workers identified by array-based proteomics. Proc Natl
Acad Sci USA 2005;102:17041– 6.
31. Villanueva J, Shaffer DR, Philip J, Chaparro CA, Erdjument-Bromage H, Olshen AB, et al. Differential exoprotease activities
confer tumor-specific serum peptidome patterns. J Clin Invest
2006;116:271– 84.
32. Zhang Z, Bast RC, Jr., Yu Y, Li J, Sokoll LJ, Rai AJ, et al. Three
biomarkers identified from serum proteomic analysis for the
detection of early stage ovarian cancer. Cancer Res 2004;64:
5882–90.
33. Gericke B, Raila J, Sehouli J, Haebel S, Konsgen D, Mustea A, et
al. Microheterogeneity of transthyretin in serum and ascitic fluid of
ovarian cancer patients. BMC Cancer 2005;5:133.
34. Engwegen JY, Helgason HH, Cats A, Harris N, Bonfrer JM,
Schellens JH, et al. Identification of serum proteins discriminating
colorectal cancer patients and healthy controls using surfaceenhanced laser desorption ionisation-time of flight mass spectrometry. World J Gastroenterol 2006;12:1536 – 44.