iTRAQ identification of candidate serum biomarkers associated with metastatic progression of human prostate cancer.

Abstract

A major challenge in the management of patients with prostate cancer is identifying those individuals at risk of developing metastatic disease, as in most cases the disease will remain indolent. We analyzed pooled serum samples from 4 groups of patients (n = 5 samples/group), collected prospectively and actively monitored for a minimum of 5 yrs. Patients groups were (i) histological diagnosis of benign prostatic hyperplasia with no evidence of cancer 'BPH', (ii) localised cancer with no evidence of progression, 'non-progressing' (iii) localised cancer with evidence of biochemical progression, 'progressing', and (iv) bone metastasis at presentation 'metastatic'. Pooled samples were immuno-depleted of the 14 most highly abundant proteins and analysed using a 4-plex iTRAQ approach. Overall 122 proteins were identified and relatively quantified. Comparisons of progressing versus non-progressing groups identified the significant differential expression of 25 proteins (p<0.001). Comparisons of metastatic versus progressing groups identified the significant differential expression of 23 proteins. Mapping the differentially expressed proteins onto the prostate cancer progression pathway revealed the dysregulated expression of individual proteins, pairs of proteins and 'panels' of proteins to be associated with particular stages of disease development and progression. The median immunostaining intensity of eukaryotic translation elongation factor 1 alpha 1 (eEF1A1), one of the candidates identified, was significantly higher in osteoblasts in close proximity to metastatic tumour cells compared with osteoblasts in control bone (p = 0.0353, Mann Whitney U). Our proteomic approach has identified leads for potentially useful serum biomarkers associated with the metastatic progression of prostate cancer. The panels identified, including eEF1A1 warrant further investigation and validation.

Full text

Page 1

iTRAQ Identification of Candidate Serum Biomarkers
Associated with Metastatic Progression of Human
Prostate Cancer
Ishtiaq Rehman1*., Caroline A. Evans1,2., Adam Glen1, Simon S. Cross3, Colby L. Eaton1, Jenny Down1,
Giancarlo Pesce1, Joshua T. Phillips4, Ow Saw Yen2, George N. Thalmann5, Phillip C. Wright2, Freddie C.
Hamdy6
1Department of Human Metabolism, The Medical School, The Mellanby Centre for Bone Research, University of Sheffield, Sheffield, United Kingdom, 2Biological and
Environmental Systems Group, Department of Chemical and Biological Engineering, ChELSI Institute, University of Sheffield, Sheffield, United Kingdom, 3Academic Unit
of Pathology, Department of Neuroscience, Faculty of Medicine, Dentistry and Health, University of Sheffield, Sheffield, United Kingdom, 4Department of Urology, City
Hospital, Birmingham, United Kingdom, 5Department of Urology, Anna-Seiler-Haus, University of Bern, InselspitalInselspital, Bern, Switzerland, 6Nuffield Department of
Surgical Sciences, University of Oxford, John Radcliffe Hospital, Oxford, United Kingdom
Abstract
A major challenge in the management of patients with prostate cancer is identifying those individuals at risk of developing
metastatic disease, as in most cases the disease will remain indolent. We analyzed pooled serum samples from 4 groups of
patients (n=5 samples/group), collected prospectively and actively monitored for a minimum of 5 yrs. Patients groups were
(i) histological diagnosis of benign prostatic hyperplasia with no evidence of cancer ‘BPH’, (ii) localised cancer with no
evidence of progression, ‘non-progressing’ (iii) localised cancer with evidence of biochemical progression, ‘progressing’, and
(iv) bone metastasis at presentation ‘metastatic’. Pooled samples were immuno-depleted of the 14 most highly abundant
proteins and analysed using a 4-plex iTRAQ approach. Overall 122 proteins were identified and relatively quantified.
Comparisons of progressing versus non-progressing groups identified the significant differential expression of 25 proteins
(p,0.001). Comparisons of metastatic versus progressing groups identified the significant differential expression of 23
proteins. Mapping the differentially expressed proteins onto the prostate cancer progression pathway revealed the
dysregulated expression of individual proteins, pairs of proteins and ‘panels’ of proteins to be associated with particular
stages of disease development and progression. The median immunostaining intensity of eukaryotic translation elongation
factor 1 alpha 1 (eEF1A1), one of the candidates identified, was significantly higher in osteoblasts in close proximity to
metastatic tumour cells compared with osteoblasts in control bone (p=0.0353, Mann Whitney U). Our proteomic approach
has identified leads for potentially useful serum biomarkers associated with the metastatic progression of prostate cancer.
The panels identified, including eEF1A1 warrant further investigation and validation.
Citation: Rehman I, Evans CA, Glen A, Cross SS, Eaton CL, et al. (2012) iTRAQ Identification of Candidate Serum Biomarkers Associated with Metastatic Progression
of Human Prostate Cancer. PLoS ONE 7(2): e30885. doi:10.1371/journal.pone.0030885
Editor: Karl X. Chai, University of Central Florida, United States of America
Received August 26, 2011; Accepted December 27, 2011; Published February 15, 2012
Copyright: ? 2012 Rehman et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This work was supported by grants from PROMET (project no. FP6-LSH-5-2004-018858), and P-MARK (contract no. LSHC-CT-2004-503011), under the
sixth EU Framework programme and the NCRI ProMPT (Prostate cancer Mechanisms of progression and Treatment) collaborative grant to Dr. Hamdy, and grants
from Engineering and Physical Sciences Research (EPSRC) to Dr. Wright: EP/E036252/1 and GR/S84347/01. We are grateful to Dr. Nicholas Hoyle at Roche
Diagnostics, Penzberg, Germany for his support towards consumables and reagents for the project. The funders had no role in study design, data collection and
analysis, decision to publish or preparation of the manuscript.
Competing Interests: We are grateful to Dr. Nicholas Hoyle at Roche Diagnostics, Penzberg, Germany for his support towards consumables and reagents for
the project. This does not alter the authors’ adherence to all the PLoS ONE policies on sharing data and materials.
* E-mail: i.rehman@sheffield.ac.uk
. These authors contributed equally to this work.
Introduction
In Europe and the US, prostate cancer is the second most
common cancer diagnosis and the third most common cause of
cancer-related deaths in men [1,2]. Moreover, the incidence has
increased since the widespread introduction of prostate specific
antigen (PSA) testing [3]. Most patients with prostate cancer are
diagnosed at an early stage, but even with screening over 7% of
cases develop metastatic disease [4]. In men with distant metastasis
the prognosis is poor, with an average survival of 24 to 48 months.
Bone is the most common site for prostate cancer metastasis and is
associated with bone pain, spinal cord compression and marrow
failure [4]. Currently, bone metastatic lesions are determined by
imaging such as isotope bone scanning, however, the identification
of a serum based biomarker(s) for predicting the susceptibility of
patients to develop bone metastasis could enable a more accurate
clinical assessment of the disease and help guide therapy.
The diagnosis of prostate cancer is most commonly made by a
triad of serum prostate specific antigen (PSA) measurements,
digital rectal examination (DRE), and histological assessment of
transrectal ultrasound (TRUS) guided biopsy material [5].
Although PSA is a FDA approved biomarker for prostate cancer
detection, its usefulness is controversial as it has been shown to be
unreliable for disease diagnosis, and in particular for distinguishing
PLoS ONE | www.plosone.org1February 2012 | Volume 7 | Issue 2 | e30885

Page 2

indolent from aggressive forms of the disease [6,7]. Additionally,
PSA is associated with a high degree of false-positive and false-
negative test results, as levels may be elevated in non-cancer
conditions of the prostate, including benign prostatic hyperplasia
(BPH). Thus, additional biomarkers are urgently needed which
could improve the diagnostic specificity of PSA and predict the
likelihood of disease progression.
Blood and its products, such as plasma and serum are ideal fluids
for the identification of cancer biomarkers since they contain
proteinsbothsecretedandshedfromcancercells,combinedwiththe
ease of sampling. However, the variable composition and large
dynamic range of proteins present in plasma (estimated to be 1010
orders of magnitude or more), pose formidable challenges in
identifying clinically relevant biomarkers amongst the background
of abundant proteins such as albumin, immunoglobulin and
transferrin [8]. Of the 22 or so most abundant proteins in plasma,
these constitute more than 99% of the mass of the total plasma
proteins, while the remaining 1% are thought to be composed of the
medium/low abundance proteins and include the biomarker pool
[9]. The large orders of magnitude in protein concentration have
hampered previous mass spectrometry based efforts aimed at
identifying clinically relevant biomarkers, mainly due to a suppres-
sion of ionization of the low abundance proteins by the higher
abundance proteins [10]. However, prior removal of some of the
most highly abundant proteins has been shown to improve the
detection of relatively lower abundant proteins [11,12]. Although
there are many different protein fractionation methodologies based
on differences in molecular weight, shape, charge, pI, hydrophobic-
ityand affinitythroughspecificbiomolecularinteractions,ithasbeen
reported that high abundance protein separation using the antibody
based IgY-12 immunodepletion system is highly reproducible [13].
Amongst the proteomic technologies used for biomarker
identification, ‘isobaric Tags for Relative and Absolute Quantita-
tion’ (iTRAQ) has the advantages of being relatively high
throughput,and cansimultaneouslyprovide information on peptide
quantitation and identification, as previously reported by us and
others [14–16]. Briefly, in a typical workflow samples are reduced,
alkylated and proteolytically digested to generate peptides. The
peptides are labeled with a set of iTRAQ reagents (in a 4 or 8-plex
format), pooled and fractionated by strong cation exchange (SCX).
The fractions are then analyzed by liquid chromatography tandem
mass spectrometry (LC-MS/MS), with the resultant mass spectra
providing sequence information (from the peptide fragments), and
relative quantification (from the reporter group ions).
In an effort to identify novel proteins associated with the metastatic
progression of human prostate cancer, we have performed a 4-plex
iTRAQ analysis using pooled serum samples collected prospectively
from 4 well defined groups of patients who were actively monitored
for at least 5 years, and selected to represent the spectrum of prostatic
disease. Following data analysis, a number of candidates were found
to be significantly differentially expressed in cancer samples
compared with benign samples. One of the candidates identified as
being significantly up-regulated in cancer groups was eukaryotic
translation elongation factor 1 alpha 1 (eEF1A1), and was further
investigated by immunohistochemistry using prostate tissue samples
fromlocalizedandmetastaticcases.The biomarkerleadsidentified in
our ‘discovery’ phase study, including eEF1A1 are discussed in
relation to their significance to prostate cancer progression.
Materials and Methods
Patients and serum collection
Peripheral blood was collected prospectively from patients
attending the Urology clinic at the Royal Hallamshire Hospital at
the time of their initial visit, following written informed consent.
Blood sample collection was approved by the Ethics Committee of
the University of Sheffield. Serum was collected by allowing the
blood to coagulate for 30 min, centrifuged at 1,2006g for 10 min
at 4uC and then stored at 280uC in 100 ml aliquots. All blood
samples were collected prior to the administration of any
treatment. Twenty serum samples were carefully selected to
represent the various stages and grades of prostate disease and
pooled (n=5 patients/group), to form 4 patient groups. All
patients were actively monitored for at least 5 years from the time
of their initial blood sampling. The 4 patient groups were: Group
1: histological diagnosis of benign prostatic hyperplasia ‘BPH’,
with no evidence of cancer by at least 2 independent sets of
prostatic biopsies, and a PSA level below 10 ng/ml (mean age of
61 yrs). Group 2: histological diagnosis of prostate cancer with a
PSA level below 10 ng/ml, and no evidence of a rising PSA
following 5 yrs active monitoring - ‘non-progressing’ group, (mean
age of 67 yrs); Group 3: histological diagnosis of clinically localised
cancer with an initial PSA level below 13 ng/ml, followed by 3
consecutive rises in PSA levels during 5 yrs of active monitoring -
‘progressing’ group, (mean age of 69 yrs); Group 4: patients with a
PSA.19 ng/ml and evidence of bone metastasis from a positive
radionucleotide bone scan - ‘metastatic’ group, (mean age of
73 yrs). The differences in the median ages of the patients were
found not to be statistically significant between the 4 groups
(p=0.146, Kruskal-Wallis test). The disease characteristics of the
20 patients comprising the 4 groups are shown in Table S1.
Patient tissue material
Tissue microarrays (TMAs), comprised of 56 cases of prostatic
adenocarcinoma ranging in single Gleason grades (i.e. grade 2,
n=5; grade 3, n=32; grade 4, n=9; grade 5, n=10), and 40
cases of adjacent non-malignant tissue, and were constructed as
previously described [17,18]. An additional 23 cases of bone
biopsies from patients both with and without prostate cancer
skeletal metastasis were obtained by 8 mm trephine biopsy
performed under general anesthesia. Informed patient consent
and Ethics Committee approval was obtained prior to the study
(South Sheffield Research Ethics Committee, SSREC/02/155
and 00/172).
Cell lines
Human prostate cancer cell lines LNCaP, PC-3, DU145 and
VCaP were purchased from the American Type Culture
Collection (ATCC), (http://www.lgcstandards-atcc.org/). DuCaP
cells were obtained via the PRIMA project consortium (http://
www.primaproject.org/participants.php).
LNCaP-LN3, PC-3M and PC-3M-LN4 cells were a kind gift
from Dr. Curtis Pettaway (University of Texas, M.D. Anderson
Cancer Centre), [19]. The LNCaP-C4-2 and LNCaP-C4-2B cell
lines were obtained from Prof. George Thalmann [20]. The TE-
85 osteosarcoma cells were a kind gift from Prof. J.A. Gallagher,
University of Liverpool. All prostate cancer cell lines were cultured
as previously described and confirmed to be free from Mycoplas-
ma [15].
The LNCaP-Pro5,
IgY-14 affinity depletion of serum samples
Pooled serum samples were depleted of the 14 most common
plasma proteins using the Seppro IgY-14 depletion system [21].
Previous studies have shown that serum pooling followed by
depletion of the most highly abundant proteins is an effective
strategy to reduce the dynamic range of proteins, and thus
enhance the identification of serum biomarkers, as demonstrated
using the quantitative proteomic method of iTRAQ(R) [22].
Serum Biomarkers for Prostate Cancer Metastasis
PLoS ONE | www.plosone.org2 February 2012 | Volume 7 | Issue 2 | e30885

Page 3

Serum samples from 5 patients representing each of the 4 patient
groups were pooled in equal volumes to give a total volume of
200 ml for each group (40 ml of each sample). The pooled serum
samples were shipped on dry ice to Genway (Digilabs Biovision,
Germany), for immuno-depletion using the Seppro Ig-Y14 system.
The flow-through fraction (depleted of albumin, immunoglobulin
IgG, fibrinogen, transferrin, IgA, IgM, haptoglobin, alpha2-
macroglobin, alpha1-acid glycoprotein, alpha1-antitrypsin, Apo
A-I HDL, Apo A-II HDL, complement C3 and LDL (ApoB)), was
used for subsequent iTRAQ analysis.
iTRAQ sample labelling and SCX fractionation
Prior to iTRAQ analysis, samples were concentrated and buffer
exchanged using 5 kDa molecular weight cut-off spin concentra-
tors (Millipore). The samples were buffer exchanged three times
against 500 ml of 1 M triethylammoniumbicarbonate (TEAB)
buffer and concentrated to a volume of approximately 80 ml.
Samples were labelled with the iTRAQ reagents according to the
manufacturers instructions, and as previously described [14,15].
Each sample was labelled with one of the four iTRAQ reagents
(iTRAQ reporter ions of 114.1, 115.1, 116.1, 117.1 mass/charge
ratio). The tag labelling order was BPH- 117; localised non-
progressing cancer-116; progressing cancer-115; metastatic dis-
ease-114). Labelled samples were pooled and fractionated by
strong cation exchange (SCX), using a BioLC HPLC column
(Dionex, Surrey, UK), and analyzed by LC-MS/MS as previously
described [14,15].
Tandem mass spectrometry analysis
Mass spectrometry (MS) was performed using a QStar XL
Hybrid ESI Quadrupole time-of-flight tandem mass spectrometer,
ESI-qQ-TOF-MS/MS (Applied Biosystems, Framingham, MA;
MDS-Sciex, Concord, Ontario, Canada), coupled with an online
capillary liquid chromatography system (Famos, Switchos and
Ultimate from Dionex/LC Packings, Amsterdam, The Nether-
lands). The dried samples were resuspended in 60 l of 3%
acetonitrile and 0.1% formic acid ready for the MS, and 10–15 l
(depending on the peptide concentration as seen in the peak
intensity of the SCX chromatogram) were injected to the nano-
LC-ESI-MS/MS system for each analysis. Initial separation took
place on a PepMap C18RP capillary column (LC Packings) with a
constant flow rate of 0.3 l/min. LC buffers A and B were made up
as 3% acetonitrile, 0.1% formic acid and 97% acetonitrile, 0.1%
formic acid, respectively. The gradient was started as 97% buffer
A and 3% buffer B for 3 minutes, followed by 3 to 25% buffer B
for 120 minutes, 90% buffer B for 7 minutes and finally 97%
buffer A for 7 minutes. Data acquisition in the mass spectrometer
was set to the positive ion mode, with a selected mass range of
350–1800 m/z. Tandem mass spectrometry was performed on
peptides with +2, +3, +4 charge states across a scan range of 65–
2000 m/z.
Protein identification and relative quantification
Protein identification and relative quantification was carried out
as previously described [23,24]. Identification of peptide precursor
and fragments was performed by database searching against the
Swiss-Prot and Trembl Homo sapiens protein database (41070,
71449 ORFs respectively, downloaded from UniProt, May 2010).
Parameters for searching were set up as follows: MS tolerance was
0.4 and MS/MS tolerance were set at: peptide tolerance 0.4 Da,
charge +2, +3 and +4, min peptide length, z-score, max p-value
and AC score were 6, 6, 1026and 6 respectively. Phenyx default
‘turbo’ scoring was enabled with mass tolerance restriction of
0.1 Da for MS and MS/MS and the minimum percentage of the
peptide sequence coverage by b+(b), b2+(b++), y+(y) or y2+(y++)
fragment series was set to the default value of 20%. Target
database search space was restricted to tryptic peptides with a
maximal of 1 miscleavage. Modifications were set as: 4-plex
iTRAQ mass shifts (+144 Da, K and N-term), methylthiol
(+46 Da) and oxidation of methionine (+16 Da). False discovery
rates (FDR) were estimated using a concatenated target-decoy
database as described by Elias and Gygi [25]. Protein changes
were qualified using a t- test algorithm developed in house [23].
Immunohistochemistry for eEF1A1
Immunohistochemistry was performed essentially as previously
described [18]. Sections of bone and prostatic tissues were cut
(4 m) and mounted on superfrost slides (VWR International,
Germany). Slides were incubated with mouse monoclonal anti-
eEF1A1 antibody (Santa-Cruz Biotechnology, CA, USA Cat. No
sc21758), at 0.4 mg/ml in 2% horse serum overnight at 4uC.
Sections were washed twice in PBS-Tween 20 (PBST), and
incubated for 30 min with anti-mouse IgG ImmPRESS HRP
(Vector Laboratories, Cat. NoMP-7402). After further washing in
PBST, localisation of antibody/antigen complex was visualized
using the ImmPACT DAB system (Vector Laboratories, Cat. No
SK-4105). Control sections were incubated with anti-mouse IgG
isotype control (Vector laboratories I-2000), diluted to 0.4 mg/ml
in 2% normal horse serum. eEF1A1 immunostaining was assessed
for both intensity and cellular localization by an experienced
histopathologist (Dr. Simon Cross), who was blinded to the study.
Each case was assigned a staining intensity, ranging from 0–3,
where 0=absent; 1=weak; 2=moderate and 3=Intense staining
as previously described [18].
Western blotting
Western blotting was performed essentially as previously
described [26]. Anti-EF-Tu goat polyclonal IgG primary antibody
was used at a concentration of 1:1000 (Santa Cruz, Cat.
No. sc12990). Secondary antibody was HRP-conjugated rabbit
anti-goat (Abcam), and was used at a concentration of 1:1400
(Abcam). Dual color precision plus molecular weight markers were
used for size estimation (Bio-Rad, Hertfordshire, UK).
Reverse-Transcription PCR and Sequencing
TotalRNAwasextracted fromprostatecancercelllinesusingTRI
reagent (Sigma-Aldrich, UK), according to the manufacturer’s
instructions. The RNA was quantified spectrophotometrically and
2 mg was reverse transcribed into cDNA using the SuperScript III
Reverse Transcriptase kit with 250 ng of random primers, according
to the manufacturer’s instructions (Invitrogen, UK). PCR primers
specific to the eEF1A1 isoform were designed manually, using the
Ensembl cDNA sequence: ENST00000316292
ensembl.org/index.html). The eEF1A1-forward primer sequence
was (59-39): TCCTTCAAGTATGCCTGGGTCT (eEF1A1-F1),
corresponding to nucleotide positions 157–178. The eEF1A1-reverse
primersequencewas: TGGCACAAATGCTACTGTGTCG
(eEF1A2-R1), corresponding to nucleotide positions 555–576, to
giveanexpected PCR productsizeof420 bp.SimilarlyPCRprimers
specific for the eEF1A2 isoform were designed using the Ensembl
cDNA sequence: ENST00000298049. The eEF1A2-forward primer
sequence was: AGGAGGCTGCTCAGTTCACCT (eEF1A2-F3),
corresponding to nucleotide positions 1004–1024; and the eEF1A2-
reverse primer sequence was: CCGCTCTTCTTCTCCACGTTC
(eEF1A2-R3),correspondingto nucleotide positions1317–1336,with
an expected PCR product size of 334 bp. Primers were synthesized
using the commercial facility at Eurofins MWG Operon (http://
www.eurofinsdna.com/products-services/oligonucleotides0.html).
(http://www.
Serum Biomarkers for Prostate Cancer Metastasis
PLoS ONE | www.plosone.org3 February 2012 | Volume 7 | Issue 2 | e30885

Page 4

Reverse transcription PCR was performed by using 1 ml of
cDNA from each of the cell lines, 10 pmol of each forward/
reverse primer, and 0.5 ml of AccuPrime Taq DNA polymerase
(Invitrogen, UK), in 20 ml volumes. Thermocycling was performed
under the following conditions: Initial denaturation at 94uC for
5 minutes; 30 PCR cycles of 94uC for 1 min, 58uC for 1 min, and
72uC for 1 min, and a final extension of 72uC for 7 minutes.
Amplified PCR products (10 ml) were separated on a 2.5% agarose
gel containing ethidium bromide and imaged using the GelDoc
XR+Molecular Imager (Bio-Rad). Band intensities were measured
using the Quantity One software (Bio-Rad).
PCR products were sequenced at the Genetics core facility,
University of Sheffield (http://www.shef.ac.uk/medicine/research/
corefacilities/genetics.html). DNA sequences were visualised using
the Chromas Lite version 2.01 software, freely downloaded from
http://www.technelysium.com.au/chromas_lite.html.
Results
Hierarchical cluster analysis of protein profiles
Analysis of the 4 pooled groups of patients identified 122
proteins with associated information on their relative expression
levels (Table S2). The false discovery rate was 1.4% which is
within the acceptable limit of 5% [25]. For the iTRAQ dataset, 75
unique proteins were identified and relatively quantified with $2
unique peptides, and these data were used for statistical analyses to
determine alterations in protein levels between groups. Hierarchi-
cal clustering was performed to group the data based on the
degree of similarity between the BPH and cancer groups.
Agglomerative clustering using the squared Euclidean distance
between log10value of iTRAQ ratios and smallest intercluster
dissimilarity linkage procedure was performed (Mathematica 7.0.0
for Mac), to generate the dendrogram shown in Figure 1. A key
feature of the dendrogram is separation of the patient groups
according to the stage of their disease. Patients with metastasis
separated the furthest and are thus considered to be the most
different from patients with BPH. The patients with BPH clustered
more closely with patients in the non-progressing group compared
with the progressing and metastatic groups.
Gene ontology annotation
To assess the range of proteins identified, gene ontology (GO)
annotations for biological processes were assigned using the
Protein ANalysis THrough Evolutionary Relationships (PAN-
THER) software, which links protein accession codes to the
corresponding entries in the gene ontology database [14,27]. The
PANTHER analysis revealed the presence of many common
plasma proteins such as those associated with complement
mediated immunity (14%), immunity and defence (27%), prote-
olysis (21%), blood clotting (7%), and protein metabolism and
modification (22%).
For the biological network analysis, the Metacore platform
(GeneGo, Inc., St. Joseph, MI), was employed as previously
described [14], which revealed that many of the differentially
expressed proteins such as C4, C4a, C5, C5b, C9 and C6 mapped
to the classical immune response pathway (Figure S1).
Diagnostic biomarker leads
Differences in protein levels are reported following a t-test analysis
as previously described [23]. The p-values were calculated based on
thenumberofpeptidesusedforthequantificationandthevarianceof
the iTRAQ reporter ratios derived from these peptides. A p-
value#0.01 was used to assign significant changes in protein levels
between sample sets. Importantly, the protein changes reported as
significant differential expression were selected based upon statistical
significance rather than fold change [28]. Some of these differences
are expected to be potentially larger due to the known under
estimation associated with iTRAQ based quantification [24].
During our analysis we were interested in proteins showing both
increased and decreased expression levels, as previous studies have
reported that both significantly up- and down-regulated proteins
may be of clinical relevance [15,29].
The identification of proteins differentially expressed in non-
progressing, progressing and metastatic patient groups relative to
the BPH group were of interest as these could provide leads for
potentially useful diagnostic and prognostic biomarkers (Table S3).
Thus, a comparison between the non-progressing cancer group
versus the BPH group showed a significant differential expression
of 22 proteins; 7 of which were up-regulated and 15 down-
regulated (Table S3a). Similarly, a comparison between the
progressing patient group versus the BPH group identified the
differential expression of 19 proteins; 11 of which showed
significant over-expression and 8 showed down-regulation (Table
S3b). Comparisons of the metastatic patient group versus the BPH
group identified the differential expression of 35 proteins, with 19
proteins showing significant over-expression and 16 showing
significant down-regulation (Table S3c). Additionally, C-reactive
protein (CRP) was found to be elevated 41.1 fold in the serum of
patients with metastatic disease compared to patients with BPH
(Table S2).
Prognostic biomarker leads
Once a diagnosis of cancer has been made the next steps are to
establish the extent (stage) of disease, in an attempt to predict those
patients in which the disease is likely to progress from the patients
in which the disease is likely to remain localized, and to obtain
prognostic information. Currently, pre-treatment PSA levels,
biopsy Gleason grade and clinical staging are used to provide
prognostic information; however, these parameters are associated
with a number of limitations. Thus, a comparison of patients with
progressing versus non-progressing disease identified the signifi-
cant differential expression of 25 proteins; 13 up-regulated and 12
down-regulated (Table S4).
Differential protein levels associated with disease
progression
In addition to the comparisons above, protein differences were
mapped according to the stage of prostate cancer development
Figure 1. Hierarchical cluster analysis of the 4 patient groups
studied. Samples were clustered based on the similarity of their
protein expression profiles observed in log10of the iTRAQ ratios and a
dendrogram generated to indicate the relationship between the
samples. Squared Euclidean distance between clusters (single linkage)
is shown. Varying lengths of the branch points indicate the degree of
similarity; the shorter the branch the higher the degree of similarity.
doi:10.1371/journal.pone.0030885.g001
Serum Biomarkers for Prostate Cancer Metastasis
PLoS ONE | www.plosone.org4February 2012 | Volume 7 | Issue 2 | e30885

Page 5

and progression i.e. as the cancer developed from non-malignant
epithelium and progressed to locally advanced and metastatic
disease (Figure 2). The lists of differences are based on
comparisons between the non-progressing versus BPH group;
progressing versus non-progressing group; and metastatic versus
progressing cancer groups. From Figure 2, it is apparent that
individual proteins, ‘pairs’ of proteins and ‘panels’ of proteins
(defined as $3 proteins), were seen to be differentially expressed at
certain stages of disease development and progression. For
instance, individual proteins such as alpha-2-macroglobulin,
lumican and serum amyloid P-component were seen to be
differentially expressed between the non-progressing versus the
BPH group. Other proteins such as beta-2-glycoprotein 1 and
somatomedin-B (blue font), were both seen to be relatively
decreased as a ‘pair’ in progressing versus non-progressing
samples, while both Apolipoprotein A-1V and Complement
component 4B were seen to be relatively increased in expression
in metastatic samples versus progressing samples (orange font).
Additionally, two ‘panels’, were seen to be altered as the disease
developed and progressed. The first panel comprised of 3 proteins:
afamin, alpha-2-HS-glycoprotein chain B and fibronectin 1
(shown in red font), and was seen to be relatively increased in
expression in the non-progressing versus the BPH group, but
decreased as the cancer progressed and remained relatively low as
the cancer metastasized. The second panel comprised of 7 proteins:
alpha-1-antichymotrypsin; cDNA FLJ55673, highly similar to
complement factor B; cDNA FLJ54228, highly similar to leucine-
rich alpha-2-glycoprotein; cDNA FLJ58564, highly similar to plasma
protease C1 inhibitor; Ceruloplasmin, Complement C5 and
Complement component C9b (green font), and were seen to be
relatively decreased in expression in the non-progressing group
compared with the BPHgroup, and relativelyincreased in expression
as the cancer progressed i.e. were relatively increased in the
progressing group and remained elevated in the metastatic group.
Interestingly, eukaryotic translation elongation factor 1 alpha 1
(eEF1A1), (a.k.a EF-Tu), was seen to show significant increased
expression in non-progressing cancer relative to BPH, and its
expression was further increased with disease progression, and was
maintained during metastasis (Table S2 and Figure 2). eEF1A1
was thetophitfollowing
VETGVLKPGMVVTFAPVNVTTEVK peptide identified in
the serum samples. Comparison of the full length amino acid
sequence of eEF1A1 with its isoform eEF1A2, indicated that the
peptide sequence was unique to eEF1A1. The corresponding
peptide in eEF1A2 differs by a single amino acid where valine is
substituted by isoleucine. Since these amino acids have a 14 Da
difference in molecular mass we could confidently assign the
identified peptide to correspond to the eEF1A1 isoform.
theblastp searchof the
Figure 2. Proteins showing significant differential expression (up-regulated and down-regulated) according to disease
progression. The list of differentially expressed proteins shown are based on comparisons between non-progressing versus BPH; progressing
versus non-progressing and metastatic versus progressing groups. Note the differential expression of proteins either individually (black font), as pairs
(blue, orange and purple fonts), or as a panel ($3 proteins, green and red fonts). (*)=identified as a single high confidence peptide.
doi:10.1371/journal.pone.0030885.g002
Serum Biomarkers for Prostate Cancer Metastasis
PLoS ONE | www.plosone.org5 February 2012 | Volume 7 | Issue 2 | e30885