High nuclear RBM3 expression is associated with an improved prognosis in colorectal cancer

www.clinical.proteomics-journal.com

REPRINT

Now indexed by

MEDLINE

ProteomicsClinical Applications

RESEARCH ARTICLE

High nuclear RBM3 expression is associated with an

improved prognosis in colorectal cancer

Barbara Hjelm1, Donal J. Brennan2, Nooreldin Zendehrokh3, Jakob Eberhard4, Bjorn Nodin3,Alexander Gaber 3, Fredrik Ponten5, Henrik Johannesson6, Kristina Smaragdi3,Christian Frantz7, Sophia Hober1, Louis B. Johnson7, Sven Pahlman3,8, Karin Jirstrom3

and Mathias Uhlen1,9

1 Department of Proteomics, AlbaNova University Center, Royal Institute of Technology, Stockholm, Sweden2 UCD School of Biomolecular and Biomedical Science, UCD Conway Institute, University College Dublin, Belfield,

Dublin, Ireland3 Center for Molecular Pathology, Department of Laboratory Medicine, Lund University, Skane University Hospital,

Malmo, Sweden4 Division of Oncology, Department of Clinical Sciences, Lund University, Skane University Hospital, Lund, Sweden5 Department of Genetics and Pathology, Rudbeck Laboratory, Uppsala University, Uppsala, Sweden6 Atlas Antibodies AB, AlbaNova University Center, Stockholm, Sweden7 Division of Surgery, Department of Clinical Sciences, Colorectal Unit, Lund University, Malmo University Hospital,

Malmo, Sweden8 CREATE Health Center for Translational Cancer Research, Lund University, Lund, Sweden9 Science for Life Laboratory, Royal Institute of Technology, Stockholm, Sweden

Received: March 25, 2011

Revised: June 15, 2011

Accepted: August 30, 2011

Purpose: In this study, we investigated the prognostic impact of human RBM3 expression in

colorectal cancer using tissue microarray-based immunohistochemical analysis.

Experimental design: One polyclonal antibody and four monoclonal anti-RBM3 antibodies

were generated and epitope mapped using two different methods. Bacterial display revealed

five distinct epitopes for the polyclonal antibody, while the four mouse monoclonal antibodies

were found to bind to three of the five epitopes. A peptide suspension bead array assay

confirmed the five epitopes of the polyclonal antibody, while only one of the monoclonal

antibodies could be mapped using this approach. Antibody specificity was confirmed by

Western blotting and immunohistochemistry, including siRNA-mediated knock-down. Two

of the antibodies (polyclonal and monoclonal) were subsequently used to analyze RBM3

expression in tumor samples from two independent colorectal cancer cohorts, one conse-

cutive cohort (n 5 270) and one prospectively collected cohort of patients with cancer of the

sigmoid colon (n 5 305). RBM3-expression was detected, with high correlation between both

antibodies (R 5 0.81, po0.001).

Results: In both cohorts, tumors with high nuclear RBM3 staining had significantly

prolonged the overall survival. This was also confirmed in multivariate analysis, adjusted for

established prognostic factors.

Conclusion and clinical relevance: These data demonstrate that high tumor-specific nuclear

expression of RBM3 is an independent predictor of good prognosis in colorectal cancer.

Keywords:

Antibodies / Colorectal cancer / Epitope mapping / Prognosis / RNA-binding

protein

Abbreviations: CRC, colorectal cancer; FACS, fluorescence

activated cell sorting; HPA, Human Protein Atlas; IHC, immuno-

histochemistry; PrEST, protein epitope signature tag; TMA,

tissue microarray

Correspondence: Professor Mathias Uhlen, Science for Life

Laboratory, Royal Institute of Technology, Stockholm, Sweden

E-mail: [email protected]

Fax: 146-8-5537-8481

& 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.clinical.proteomics-journal.com

624 Proteomics Clin. Appl. 2011, 5, 624–635DOI 10.1002/prca.201100020

1 Introduction

Based on an initial discovery within the Human Protein

Atlas (HPA) Program [1, 2], nuclear expression of the

human RNA-binding protein RBM3 was demonstrated to be

associated with favorable clinicopathological parameters and

an independent marker of good prognosis in two breast

cancer cohorts [3]. Recently, we have shown that increased

nuclear expression of RBM3 is also associated with a

favorable prognosis in epithelial ovarian cancer and that

down-regulation of RBM3 confers reduced cisplatin sensi-

tivity in ovarian cancer cells [4]. The value of RBM3 as a

prognostic biomarker in colorectal cancer has not yet been

investigated, but RBM3 has been suggested to act as an

oncogene that protects against mitotic catastrophe in color-

ectal cancer cell lines [5]. Taken together, these reports

suggest that RBM3 might influence the outcome of patients

with many different types of cancers, but its function in the

context of tumor initiation and progression is still not fully

understood.

RNA-binding proteins with RNA-binding motifs (RBM)

are involved in many aspects of RNA processing and regula-

tion of gene transcription [6, 7]. RBM3, one of three X-chro-

mosome-related RBM-genes (RBMX, RBM3 and RBM10),

was initially identified in a human fetal brain tissue cDNA

library and maps to Xp11.23 [8]. The RBM3 gene encodes

alternatively spliced transcripts, with the longest reading

frame encoding for a 157 amino acid protein containing one

RRM domain and a glycine-rich region [8]. The RBM3 protein

has been shown to bind to both DNA and RNA [9], however

its exact function still remains to be elucidated.

RBM3 transcripts have been found in various human

tissues [9] and in vitro, RBM3 is one of the earliest proteins

synthesized in response to cold shock [10]. RBM proteins

have been proposed to represent a novel family of apoptosis

regulators [7, 11] and a correlation between expression of the

X-chromosome-related RBM-genes (RBMX, RBM3 and

RBM10) and the proapoptotic Bax gene has been demon-

strated in breast cancer [11]. However, RBM3 has also been

proposed to suppress cell death in a similar fashion to the

X-linked inhibitor of apoptosis [12]. Furthermore, down-

regulation of RBM3 has been demonstrated in gene

expression studies of an in vitro model of melanoma

progression [13].

Due to its potential important role as a cancer biomarker,

we decided to generate both polyclonal and monoclonal

antibodies to the RBM3 protein in rabbit and mice,

respectively. In order to develop these antibodies for various

diagnostic assays it is important to map their binding

regions, in part to allow the development of paired antibody

assays, such as sandwich-based ELISAs [14] or padlock

assays [15]. Another important reason to define the epitopes

is to ensure antibodies with non-overlapping epitopes,

which means low probability of obtaining identical cross-

reactivity staining toward unrelated proteins. This makes it

likely that common staining patterns in various immuno-

based assays are due to correct specificity toward the

intended target. No previous information regarding the

protein target is needed, which makes paired antibodies

extremely useful for validation of antibodies in many

applications, including common immunological methods

such as Western blot, immunohistochemistry and immu-

nofluorescence.

The most common strategy for epitope mapping is to use

overlapping synthetic peptides in an array format, such as

ELISA [16] or peptide array [17]. An alternative strategy was

recently described [18], based on a bacterial cell display

method, in which the protein target is fragmented using

gene technology methods and random fragments are

displayed on the cell surface and binding regions are

assayed based on binding of the displayed fragments to the

antibody. An additional advantage of this method is that,

theoretically, conformational epitopes could be assayed if

the conformation of the protein target is folded on the

surface of the cell.

Here, we have used both bacterial display and peptide

arrays to explore the linear and conformational epitopes of

the antibodies toward RBM3. We show that the bacterial

display can be efficiently used to map the various binding

regions of the polyclonal antibody, while the peptide scan-

ning method allows further fine mapping of the binding

region. The polyclonal antibody and one of the monoclonal

antibodies were then selected to examine the prognostic

impact of tumor-specific RBM3 expression in two indepen-

dent colorectal cancer (CRC) cohorts. The results demon-

strate a high correlation between the two antibodies and that

nuclear expression of RBM3 is an independent predictor of

a prolonged overall survival in CRC patients.

2 Material and methods

2.1 Patients

2.1.1 Cohort I

A consecutive cohort of 270 patients, 137 (50.7%) women

and 133 (49.3%) men, surgically treated for CRC between

January 1st 1990 and December 31st 1991. Forty-two

(15.6%) patients had Stage I disease, 118 (43.1%) Stage II,

70 (25.9%) Stage III, and 40 (14.8%) Stage IV disease.

Tumor location was in the colon in 217 (80.4%) cases, in the

rectum in 52 (19.3%) cases and missing in one case. Median

age at diagnosis was 73.36 years (range 37.59–93.51) and

after a median follow-up of 3.73 years (range 0–18.53), 32

(11.9%) patients were alive and 237 (87.8%) dead. Infor-

mation on vital status was obtained from the population

register. Information about cause-specific survival was not

available for this cohort and follow-up data were missing for

one patient. Data on neoadjuvant radiation therapy for rectal

cancer petients or adjuvant chemotherapy were not available

for this cohort.

Proteomics Clin. Appl. 2011, 5, 624–635 625


2.1.2 Cohort II

The original patient cohort consisted of 339 retrospectively

identified cases from a prospective database of patients who

underwent resection for cancer of the sigmoid colon at

Malmo University Hospital between January 1st 1993 and

December 31st 2003. Tumors from 309 cases could be

retrieved from the archives and, after histopathological re-

evaluation, four additional cases diagnosed as in situ carci-

noma were excluded. The remaining cohort of 305 patients

consisted of 157 (51.5%) men and 148 (48.5%) women.

Fortyseven (15.4%) patients had Stage I, 129 (42.3%) Stage

II, 84 (27.5%) Stage III and 45 (14.8%) Stage IV disease.

Information about vital status and cause of death was

obtained from the Swedish Cause of Death Registry. Median

age at diagnosis was 74 years (39–97) and after a median

follow-up time of 5.35 (range 0–15.80) years, 193 patients

(63%) had died and 93 (30%) of them died from CRC.

Information about treatment and recurrence (local, regional

or distant) was obtained from patient records and/or

pathology reports. Adjuvant chemotherapy had been given

to 38 patients (27 Stage III and 11 Stage IV) and palliative

treatment to 28 patients (14 Stage III and 14 Stage IV). No

patients with Stage II disease had received adjuvant

chemotherapy. All patients were surgically treated at Malmo

University Hospital. Ethical permission has been obtained

from the Ethics Committee at Lund University (ref. nos.

470-06, 447-07 and 35/08), whereby informed consent was

deemed not to be required other than by the opt-out method.

2.2 Tissue microarray construction

Prior to tissue microarray-construction (TMA), all cases

were histopathologically re-evaluated on Haematoxylin &

Eosin-stained slides. Areas representative of cancer were

then marked and TMAs constructed as described previously

[19]. In brief, two 1.0 mm cores were taken from each tumor

and mounted in a new recipient block using a semi-auto-

mated arraying device (TMArrayer, Pathology Devices,

Westminster, MD, USA)

2.3 Generation of recombinant antigen and

antibodies

A 134 amino acid-long fragment of the human protein

RBM3, called Protein Epitope Signature Tag (PrEST) was

selected with the in-house developed bioinformatics tool

PRESTIGE [20] on the basis of low sequence homology to

other human proteins. cDNA corresponding to this region

was generated by RT-PCR using a human total RNA pool as

template [21]. The fragment was cloned into an expression

vector and sequence-verified prior transformation to

Escherichia coli. The expressed recombinant protein fusions

were IMAC-purified under denaturing conditions and vali-

dated on mass spectrometry before immunization of rabbit.

The sera from the immunized animal was purified by a two-

step immunoaffinity protocol as previously described [21] to

yield the polyclonal antibody HPA003624. The monoclonal

antibodies were developed as described elsewhere [4].

2.4 Protein array

The planar array analysis of the affinity purified antibody was

performed as previously described [22]. Totally, 192 different

PrESTs were diluted to 40mg/mL in 0.1 M urea and 1" PBS

(pH 7.4) and 50mL of each PrEST was transferred to a 96-well

spotting plate and subsequently spotted and immobilized in

duplicates onto epoxide slides (Corning Life Sciences, Acton,

MA, USA). After washing and blocking, slides were incu-

bated with affinity purified antibody diluted 1:1000. Slides

were washed once before addition and incubation of

secondary goat anti-rabbit antibody (Invitrogen, Carlsbad, CA,

USA). After a final wash, slides were dried and scanned using

a G2565BA array scanner (Agilent, La Jolla, CA, USA).

2.5 Epitope mapping using bacterial display

A previously developed protocol was used for epitope

mapping [18]. The gene fragment used for antigen produc-

tion was amplified by PCR and the DNA product was

sheared by sonication to random fragments of size

50–350 bp. These were cloned into a staphylococcal display

vector and transformed to Staphylococcus carnosus. Cell

aliquots corresponding to a 10-fold coverage of the library

size were incubated with 0.35 ng mAb and 3.5 ng pAb

respectively, in reaction volumes of 70mL. Cells were washed

and incubated with Alexa Fluors 488 goat anti-rabbit or

anti-mouse antibodies (Invitrogen) and washed again before

analysis with fluorescent activated cell sorting (FACS). Cells

expressing peptides recognized by the antibodies were

enriched in a first round of sorting and in a second analysis

binding cells were sorted out and sequenced by dye-termi-

nator cycle sequencing. Finally, the sequences were aligned

back to the RBM3 sequence.

2.6 Western blot

Antibodies were analyzed by running approximately 15 mg of

total protein lysate from the RT-4 cell line and the U-251MG

cell line on precast 10–20% CriterionTM SDS-PAGE gradient

gels (Bio-Rad Laboratories, Hercules, CA, USA). The

proteins were separated under reducing conditions, followed

by electroblotting to PVDF membranes using Criterion

GelTM Blotting Sandwiches (Bio-Rad Laboratories), all

according to the manufacturer’s recommendations. The

SDS-PAGE gel was stained using GelCodes Blue Stain

Reagent (Pierce, Rockford, IL, USA) and the membranes

626 B. Hjelm et al. Proteomics Clin. Appl. 2011, 5, 624–635


were blocked (5% dry milk, 0.5% Tween20, 1" TBS; 0.1 M

Tris-HCl, 0.5 M NaCl) for 1 h at RT prior to addition of

antibodies. After incubation for 1 h with the primary anti-

bodies, diluted 1/250 in blocking buffer, the membranes

were washed 4" 5 min in 1" TBS with 0.05% Tween20.

The secondary HRP-conjugated swine anti-rabbit or anti-

mouse antibody (DakoCytomation, Glostrup, Denmark) was

diluted 1/3000 in blocking buffer and incubated for 1 h

before a final round of washing to remove unbound mate-

rial. Chemiluminescence detection was carried out using a

Chemidoc CCD-camera system (Bio-Rad Laboratories) with

SuperSignals West Dura Extended Duration Substrate

(Pierce) according to the manufacturer’s protocol.

2.7 Peptide mapping

Peptide mapping was performed as described previously

[23]. Twentyfive biotinylated synthetic peptides (Sigma-

Aldrich, St Louis, MO, USA) were designed to be 15 amino

acids long with a ten amino acids overlap to cover the

PrEST-sequence. All peptides were dissolved in DMSO and

diluted to 50 mM in 100 mL PBS (pH 7.4) supplemented with

1 mg/mL BSA (PBS-B). Totally, 50mL of each peptide mix

was incubated with 105 neutravidin-coated beads in a total

volume of 150mL PBS-B for 60 min in RT. Beads were

washed and a bead mixture containing all 25 bead IDs was

prepared. Monoclonal and polyclonal antibodies were dilu-

ted to 50 ng/mL and mixed with around 1250 beads per

ID. Antibodies were subsequently incubated with 25mL

R-Phycoerythrine-labeled anti-rabbit or anti-mouse IgG

antibody (5 mg/mL, Jackson Immunoresearch, West Grove,

PA, USA) and analyzed using LX200 instrumentation with

Luminex IS 2.3 software (Luminex, Austin, TX, USA).

2.8 Analysis of staining patterns

For assessment of nuclear RBM3 expression, both the fraction

of positive cells and staining intensity were taken into account

using a modification of the previously applied semi-

quantitative scoring system [4]. Nuclear fraction (NF) was

categorized into four groups, namely 0 (0–1%), 1 (2–25%), 2

(26–75) and 3 (475%) and nuclear staining intensity (NI)

denoted as 0–2, whereby 0 5 negative, 1 5 intermediate and

2 5 moderate-strong intensity. A combined nuclear score (NS)

of NF"NI, which had a range of 0–9, was then constructed.

Cytoplasmic staining intensity was denoted as 0 5 negative,

1 5 intermediate and 2 5 moderate-strong, and the fraction of

positive cells not taken into account.

2.9 Cell lines and reagents

The human colorectal cancer cancer cell line SW480 was

maintained in RPMI-1640 supplemented with glutamine,

10% FBS and 1% penicillin/streptomycin in a humidified

incubator of 5% CO2 at 371C.

2.10 siRNA knockdown of RBM3 gene expression

Transfection with siRNA against RBM3 (Applied Biosys-

tems, Carlsbad, CA, USA) or control siRNA (Applied

Biosystems) was performed with Lipofectamine 2000 (Invi-

trogen) with a final concentration of 50 nM siRNA. All

siRNA experiments were performed using two independent

RNA oligonucleotides (]58 and ]59) targeting RBM3.

2.11 PCR and Western blotting

Total RNA isolation (RNeasy, QIAgen, Hilden, Germany),

cDNA synthesis (Reverse Transcriptase kit, Applied

Biosystems, Warrington, UK) and quantitative real-time

PCR (QPCR) analysis with SYBR Green PCR master mix

(Applied Biosystems) were performed as described

previously [24, 25]. Quantifications of expression levels were

calculated by using the comparative Ct method, normal-

ization according to house keeping genes; HMBS (forward

primer: 50-GGC AAT GCG GCT GCA A-30, reverse primer:

50-GGG TAC CCA CGC GAA TCA C-30), SDHA (forward

primer: 50-TGG GAA CAA GAG GGC ATC TG-30, reverse

primer 50-CCA CCA CTG CAT CAA ATT CAT G-30) and

UBC (forward primer: 50-ATT TGG GTC GCG GTT CTT

G-30, reverse primer: 50-TGC CTT GAC ATT CTC GAT

GGT-30). For RBM3 amplification, forward primer with

sequence 50-CTT CAG CAG TTT CGG ACC TA-30 and

reverse primer with sequence 50-ACC ATC CAG AGA CTC

TCC GT-30. All primers were designed using Primer Express

(Applied Biosystems).

For immunoblotting, cells were lysed in ice-cold lysis

buffer (150 mM NaCl, 50 mM Tris-HCL pH 7.5, 1%

Triton X-100, 50 mM NaF, 1 mM Na3VO4, 1 mM PMSF) and

supplemented with protease inhibitor cocktail Complete

Mini (Roche, Basel, Switzerland). For Western

blotting, 20–50 mg of protein were separated on 12% SDS-

PAGE gels and transferred onto nitrocellulose membranes

(Hybond ECL, Amersham Pharmacia Biotech, Buck-

inghamshire, UK). The membranes were probed with

primary antibodies followed by HRP-conjugated secondary

antibodies (Amersham Life Science, Alesbury, UK) and

visualized using the Enhanced ChemiLuminescence detec-

tion system (ECL) and ECL films (Amersham Pharmacia

Biotech). RBM3 was detected by the polyclonal RBM3

antibody (1:150, HPA003624) and by the mouse monoclonal

anti-RBM3 antibody (1B5, Atlas Antibodies AB,

Stockholm, Sweden) diluted 1:500 in blocking solution (5%

BSA, 1" PBS, 0.1% Tween20). Membranes were stripped

and re-probed with an alpha-tubulin antibody (CalBiochem,

San Diego, CA, USA) at a dilution of 1:1000, to provide a

loading control.



2.12 Cell pellet arrays

Cell lines were fixed in 4% formalin and processed in

gradient alcohols. Cell pellets were cleared in xylene and

washed multiple times in molten paraffin. Once processed,

cell lines were arrayed in duplicate 1.0 mm cores using a

manual tissue arrayer (Beecher, WI, USA) and Immuno-

histochemistry (IHC) was performed on 5 mm sections

using the HPA003624 antibody diluted 1:250 and the 1B5

antibody diluted 1:1000.

2.13 Statistical analysis

Pearson correlation coefficient was used to compare nuclear

scores for both antibodies. Chi-square test and Pearson

correlation test were used for comparison of RBM3 expres-

sion and relevant patient- and tumor characteristics. Overall

survival (OS) was assessed by calculating the risk of death

from all causes. Recurrence was defined as local, regional or

distant recurrence or death from CRC and risk of recurrent

disease was referred to as recurrence-free survival (RFS) in

Cohort II. The Kaplan–Meier method and log rank test were

used to estimate OS in different strata. A Cox proportional

hazards model was used for estimation of hazard ratios

(HRs) in both univariate- and multivariate analysis, adjusted

for age, gender, stage and differentiation grade. All statis-

tical tests were two-sided and p-values o0.05 considered

significant. Calculations were performed with the statistical

package SPSS 17.0 (SPSS, IL, USA).

3 Results

3.1 Generation and analysis of antigen and

polyclonal antibody

A 134 amino acid-long Protein Epitope Signature Tag

(PrEST) was selected as antigen by analyzing the human

RBM3 gene using the software package PRESTIGE [20]. The

antigen was expressed in E. coli, purified, MS verified and

used for immunization of rabbit (data not shown). The

rabbit serum was affinity purified against the recombinant

antigen to generate polyclonal antibodies. Specificity of the

polyclonal antibody was analyzed on a microarray spotted

with 192 different human antigen fragments [22], which

showed binding to the corresponding antigen and no cross-

reactivity to others (data not shown). Western blot analysis

using extracts from two human cell lines subsequently

demonstrated a single band corresponding in size to the

expected molecular weight of the protein target (17.2 kDa) in

both cell lysates from the two cell lines RT-4 and U-251MG

(Fig. 1A).

3.2 Generation of mouse monoclonal antibodies

and Western blotting

Mouse monoclonal antibodies were generated using the

same recombinant antigen fragment of 134 amino acids.

Four hybridoma clones were selected for production of

antigen-specific monoclonal antibodies. These monoclonal

antibodies were also used for Western blot analysis and the

results (Fig. 1B–E) show a major band corresponding in size

to the expected molecular weight (17.2 kDa) of the protein

target for three of the four monoclonal antibodies. The

antibody 9B11 does not seem to be functional in the

Western blot analysis (Fig. 1B). Some faint background

bands of larger molecular sizes could be observed for the

monoclonal antibody 6F11.

3.3 Epitope mapping of the polyclonal antibody

using bacterial surface display

Epitope mapping of the polyclonal antibody (HPA003624)

directed toward the 134 amino acid antigen was performed.

A RBM3 library consisting of 105 clones was constructed

according to an earlier developed protocol [18]. The primary

cell sorting using the antibody revealed a few clones with

binding to the epitope (data not shown) and these were

gated out and used for another round of sorting. The second

sorting shows enrichment of many binders and different

populations of clones were sorted out by gating as indicated

by the five colors in Fig. 2A The isolated clones were

sequenced and aligned back to the original antigen

sequence to deduce the consensus epitope. The overall

results show binders to at least five separate regions of the

antigen as shown by the consensus regions in Fig. 2A. The

antibodies with the strongest apparent binding (blue)

showed exclusive binding to the N-terminal region of the

fragment (ADEQALEDHFSSF), while the antibodies with

medium binding (red, yellow and green) showed binding to

separate epitopes primarily in the middle of the antigen

fragment. Red epitopes determined to TNPEHAS and

Figure 1. Western Blot characterization of polyclonal antibody

HPA003624 (A), monoclonal antibody 9B11 (B), monoclonal

antibody 7G3 (C), monoclonal antibody 6F11 (D) and monoclonal

antibody 1B5 (E). 1: RT-4; 2: U-251MG sp.



Figure 2. Epitope mapping of polyclonal antibody HPA003624 and the monoclonal antibodies 9B11, 7G3, 6F11 and 1B5 towards RBM3

using bacterial surface display. (A) FACS dot plot showing the second sorting of the bacterial displayed RBM3 library incubated with

antibody HPA003624. The different sorted populations of binding cells are indicated with different colors and acquired sequences are

shown to the right of the dot plot in corresponding color. On top, consensus epitopes concluded as the minimal sequence needed for

binding of the antibody. (B) Epitope mapping of four monoclonal antibodies. FACS dot plots showing the second sorting of the bacterial

displayed RBM3 library incubated with separate monoclonal antibodies. The sorted populations of binding cells are indicated with

different colors and acquired sequences are shown to the right of the dot plots in corresponding color. On top, consensus epitopes

concluded as the minimal sequence needed for binding of the respective antibody. (C) Epitope mapping of polyclonal and monoclonal

antibodies using synthetic peptides, Intensity plot showing binding profile of the different antibodies to 15-mer peptides coupled to

beads. The peptides were designed with a lateral shift of 5 amino acids covering the whole PrEST-sequence. Peptide IDs are shown on the

x-axis and mean fluorescence intensity on the y-axis. (D) Consensus epitopes for polyclonal and monoclonal antibodies obtained from

both bacterial surface display and suspension bead array. A scale for the protein sequence is found at the top and the bottom. Black bars

indicate the epitopes discovered by suspension bead array and colored bars show epitopes discovered by bacterial surface display.



DYNGRNQGGYDRYSG. Yellow epitopes mapped to

sequence EHASVAMRAMNGES and RGGGFGAHGRG

and green epitope mapped to the sequence GFGAHGRG.

The antibodies with the weakest apparent binding (purple)

showed binding to the sequence DQGYGSGRYYD in the

C-terminal part of the antigen.

3.4 Epitope mapping of the monoclonal antibodies

using bacterial surface display

Epitope mapping of the monoclonal antibody was

performed using the same RBM3 staphylococcal library

consisting of 105 clones. The library was incubated sepa-

rately with each monoclonal antibody and secondary

reagents for analysis in a flow cytometer. An enrichment of

binding clones was observed in the second flow cytometric

analysis and these populations were gated as shown in Fig.

2B and collected. The clones were sequenced and after

alignment back to the original antigen sequence consensus

epitopes were concluded. The results in Fig. 2B show that

the monoclonal 9B11 bound to two different populations of

clones, which span the same N-terminal region of the

antigen (AAADEQ). The epitopes of the monoclonal anti-

bodies 7G3 (SGRYYD) and 1B5 (GSGRYYD) overlap to a

region around amino acids 90–100 while the epitope for

monoclonal 6F11 (HGRGRSYSRG) is shifted slightly

N-terminally around amino acids 80–90.

3.5 Epitope mapping of polyclonal and monoclonal

antibodies using synthetic peptides

Fifteen amino acid-long synthetic peptides were synthesized

spanning the whole antigen with a lateral shift of five amino

acids. The peptides were coupled to color-coded beads in

order to evaluate binding to the each antibody in a

suspension bead array assay using a flow sorting instru-

ment. The results (Fig. 2C) show that the polyclonal anti-

body bound to five separate regions (DEQALEDHFSSFGPI;

TFTNP; GTRGGGFGAH; GDQGYGSGRY and RNQGGY-

DRYSGGNY) across the antigen (red). These epitopes

overlap with the epitopes identified by bacterial display as

indicated in Fig. 2D. Interestingly, only the monoclonal

6F11 showed a detectable binding in the peptide array

(Fig. 2C). The three other monoclonal antibodies showed no

binding to the peptides and it is tempting to speculate that

this is probably due to the fact that the epitopes have a

conformational component and that these conformations

are not formed when using the 15 amino acid synthetic

peptide. It was reassuring that the consensus sequence

(GFGAHGRGRS) of the linear epitope of the monoclonal

antibody 6F11 is overlapping with the consensus sequence

from the bacterial display (Fig. 2D) again showing the

reliability of the two independent methods for epitope

mapping. Since the monoclonal antibody 6F11 gave some

weak bands of larger sizes on the Western blots (Fig. 1D),

the polyclonal antibody (HPA003624) and one of the

monoclonal antibodies (1B5) was selected for further

studies.

3.6 Validation of specificity using siRNA assays

The specificity of the two RBM3 antibodies was further

confirmed by siRNA-mediated knockdown of RBM3 in

SW480 cells. Quantitative PCR confirmed successful gene

silencing with an 85–90% reduction in RBM3 expression

(Fig. 3A). IHC performed on formalin-fixed, paraffin-

embedded siRNA-transfected SW480 cells also revealed a

marked decrease in immunoreactivity in the RBM3 knock-

down cells compared to controls for both antibodies

(Fig. 3C). This could also be confirmed by Western blotting

(Fig. 3B) using both the polyclonal antibody and the

monoclonal 1B5 antibody showing down-regulation of the

RBM3 band following siRNA treatment.

3.7 Correlation between the polyclonal and the

monoclonal antibodies in the tumor TMAs

Tumors from two independent patient cohorts (n 5 270 and

n 5 305 respectively) assembled in tissue microarrays

(TMAs) were analyzed. The tissue microarrays were

immunohistochemically stained with the polyclonal anti-

body and the monoclonal 1B5 antibody. Tumors were

grouped into negative 5 0 (combined NS (0–1), inter-

mediate 5 1 (combined NS 2–3) and strong 5 2 (combined

NS43) as described previously [4]. Examples of tumors

stained negative, intermediate and high are shown in

Fig. 4a. Both antibodies revealed a differential RBM3 tumor-

specific expression, particularly in the nuclei, but also in the

cytoplasm. In Cohort I, the staining could be assessed with

both antibodies in 256 (95%) cases and the corresponding

number was 271 (89%) in Cohort II. There was an excellent

correlation between the nuclear scores for both antibodies

(R 5 0.81, po0.001).

3.8 Relationship to clinicopathological parameters

and survival analysis

RBM3 expression did not correlate with age at diagnosis,

gender, disease stage or differentiation grade in either

cohort, with similar findings for both antibodies (data not

shown). In Cohort I, Kaplan–Meier analysis based on both

antibodies demonstrated a non-significant stepwise trend

toward a prolonged overall survival as RBM3 protein

expression increased (Fig. 3B). Dichotomisation of RBM3

NS, comparing negative to any expression, resulted in a

significant association between increased RBM3 expres-

sion and prolonged overall survival in tumors for whom



expression data from both antibodies were available (n 5 255)

(Fig. 3B). In Cohort II, the stepwise association between

RBM expression, as assessed by both antibodies, and

prolonged overall survival, could be confirmed. This analysis

was also restricted to tumors for whom the data from both

antibodies were available (n 5 271) (Fig. 3B). Univariate Cox

regression analysis demonstrated the association between

RBM3 and improved overall survival (Table 1). Multivariate

Cox regression analysis controlling for age, gender, disease

stage and differentiation grade confirmed that RBM3 as

assessed by both antibodies was an independent predictor of

prolonged overall survival in both cohorts (Table 1). In

Cohort II, the prognostic value of RBM3 for RFS was also

evaluated in Stage I–III patients, whereby it was demon-

strated that RBM3 expression as assessed by the polyclonal

antibody was significantly associated with a prolonged RFS

in both univariate analysis (HR 5 0.60; 95% CI 0.36–0.99,

po0.047) and multivariate analysis (HR 5 0.55; 95% CI

0.33–0.92, p 5 0.024). No significant association to RFS was

however seen for the monoclonal antibody, neither in

univariate (HR 5 0.84; 95% CI 0.51–1.39, p 5 0.506) nor in

multivariate analysis (HR 5 0.71; 95%CI 0.48–1.05,

p 5 0.088). When using OS as endpoint, however, RBM3

was a significant prognostic factor in both uni-and multi-

variate analysis in Stage I–III patients, with consistent

findings for both antibodies (data not shown). The prog-

nostic significance of RBM3 was not altered by adjustment

for adjuvant chemotherapy in the multivariate analysis,

neither in the full cohort nor in subgroup analysis of Stage

I–III patients (data not shown).

In Cohort I, tumor location (colon versus rectum) was

not prognostic but RBM3 remained an independent prog-

nostic factor when tumor location was included in the

multivariate analysis (data not shown). In this cohort,

Figure 3. Specificity of the RBM3 antibodies HPA003624 and 1B5 tested in SW 480 colorectal cancer cells. RBM3 mRNA levels after

transfection with siRNA against RBM3 were determined by qPCR (A). RBM3 protein expression was substantially decreased, as detected

by both antibodies, in si-RBM3-RNA transfected cells compared with controls, demonstrated by (B) Western blot and (C) immunocy-

tochemistry.



however, RBM3 expression was only a significant prognostic

factor in tumors located in the colon (n 5 208) and not rectal

(n 5 47) cancers (data not shown).

4 Discussion

In this study, we investigated the prognostic value of RBM3

in colorectal cancer as assessed by immunohistochemistry

in tumors from two independent patient cohorts, one

retrospectively collected consecutive colorectal cancer cohort

and one prospectively collected cohort with cancers of the

sigmoid colon. Using two different RBM3 antibodies, both

of which had undergone epitope mapping, we demonstrated

that both antibodies could produce highly similar results. In

both CRC cohorts, RBM3 was an independent predictor of a

prolonged overall survival. These findings are in line with

recent findings in breast cancer and epithelial ovarian

cancer [3, 4]. In this study, we focused on the overall survival

as primary endpoint since this information was available for

Figure 4. (A) immunohistochemical images representing examples of tumors where staining of RBM3 was denoted as (a) negative

(nuclear score 5 0–1), (b) intermediate (nuclear score 5 2–3) and (c) strong (nuclear score 43), using the polyclonal HPA003624 antibody

(top row) and the monoclonal 1B5 antibody (bottom row). (B) Kaplan Meier analysis of overall survival according to immunohisto-

chemical RBM3 staining with the HPA003624 and 1B5 antibodies, respectively, in Cohort I (a–d) and Cohort II (e–h). Strata were defined as

negative, intermediate and strong expression (a, c, e, g) and negative versus positive expression (b, d, f, h).



both cohorts, but similar associations, although only

significant for the polyclonal antibody, were seen for RFS in

patients with Stage I–III disease in Cohort II. However, as

data on recurrence had been collected retrospectively and

their accuracy depend on the availability of information in

the patient records, these results should be interpreted with

some caution. Future studies investigating the impact of

RBM3 expression on RFS should preferably be performed in

cohorts where this information has been recorded prospec-

tively. In light of recently published data indicating that

RBM3 predicts response to platinum-based chemotherapy

in epithelial ovarian cancer [4], it will also be of interest to

investigate the impact of RBM3 as a predictor of response to

adjuvant chemotherapy in controlled treatment trials

including CRC patients with metastatic disease

It is noteworthy that the results are contradictory to the

previously published in vitro data proposing that increased

RBM3 expression is associated with a more aggressive

phenotype in colorectal cancer cell lines [5]. In the study by

Sureban et al [5], tissue-based analyses were performed on a

limited number (n 5 15) of cases of colorectal cancer with no

prognostic information, and no stratification based on

cytoplasmic or nuclear localization of RBM3 was done. In

addition, the observed increase in expression of RBM3 was

based on a comparison between tumors of different stages

and adjacent normal colon, and the actual stage-dependent

differences in RBM3 mRNA and protein appear to be

minimal when tumors of different stages were compared

with each other [5]. However, since the previous study

suggests that high RBM3 expression would be associated with

an unfavorable outcome for CRC patients, while our study

presented here suggests in contrast a favorable outcome, we

have made an extensive effort to validate and epitope map

several independent antibodies to support our conclusions.

Based on systematic validation and epitope mapping

approaches, it was possible to select antibodies with differ-

ent epitopes validated both in Western blots and immuno-

histochemistry. Paired monoclonal antibodies directed

Table 1. Cox uni- and multivariate analysis of overall survival according to RBM3 expression assessed by two different antibodies in twoindependent patient cohorts

Cohort I Cohort II

HR (95%CI) p-value (n) HR (95%CI) p-value (n)

HPA003624 Univariate UnivariateRBM3 negative 1.00 72 1.00 144RBM3 positive 0.74(0.55–0.99) 0.040 183 0.52 (0.38–0.71) o0.001 127

Multivariate MultivariateRBM3 negative 1.00 72 1.00 144RBM3 positive 0.73 (0.55–0.98) 0.039 183 0.61 (0.44–0.83) 0.002 127

1B5 Univariate UnivariateRBM3 negative 65 1.00 129RBM3 positive 0.73(0.54–0.98) 0.035 190 0.57 (0.42–0.77) o0.001 142

Multivariate MultivariateRBM3 negative 1.00 65 1.00 129RBM3 positive 0.67(0.49–0.91) 0.010 190 0.55 (0.40–0.75) o0.001 142

Multivariate analysis included adjustment for age (4/o5 75 years), gender, stage (I–II versus III–IV) and differentiation grade (high-intermediate versus low).

Clinical Relevance

Colorectal cancer is one of the most common forms

of cancer worldwide, and there is an obvious need

for novel biomarkers for improved prognostication

and treatment stratification of colorectal cancer

patients. In this study, we demonstrate that nuclear

expression of the RNA-binding protein RBM3 is a

good prognostic marker in colorectal cancer, also

when adjusted for established clinicopathological

parameters. For this purpose, tissue microarrays

with tumor specimens from two independent

patient cohorts representing 575 tumors have been

stained immunohistochemically with two well-

characterized antibodies (using Western blot and

epitope mapping), whereby both antibodies showed

an excellent correlation with each other and highly

similar associations to outcome. We would like to

emphasize the need for these validated affinity

reagents, preferably also with non-overlapping

epitopes, in the quest of biomarkers in personalized

medicine.



toward separate and non-overlapping protein epitopes were

found (1B5 and 6F11) and one of them (1B5) was used

together with a polyclonal antibody in all assays. Both anti-

bodies were shown to be functional with respect to down-

regulation of the target protein using an siRNA-mediated

knock-down assay.

In summary, our results suggest that increased levels of

nuclear expression of the RNA-binding protein RBM3

confer a prolonged overall survival of colorectal cancer

patients. The generation of RBM3 specific antibodies with

known epitopes will allow for easier functional validation of

the role of RBM3 in multiple tumor types. While the

approach employed in this study does not allow for any

mechanistic insight into the association between RBM3

expression and a favorable prognosis in vivo, our data

emphasize the usefulness of IHC-based tissue micro arrays

for validation of cancer biomarkers of potential clinical

relevance. Nevertheless, further validation on tumors from

large patient cohorts is warranted in order to confirm the

prognostic and/or treatment predictive value of RBM3 in

CRC. It would also be of interest to extend the functional

studies of RBM3 to further understand the molecular

mechanisms behind the beneficial prognostic effect of its

tumor-specific expression, which has now been observed in

several cancer forms.

The authors acknowledge the entire staff of the HumanProtein Atlas project. This work was supported by grants fromthe Knut and Alice Wallenberg Foundation.

The authors have declared no conflict of interest.

5 References

[1] Ponten, F., Jirstrom, K., Uhlen, M., The Human Protein Atlas

– a tool for pathology. J. Pathol. 2008, 216, 387–393.

[2] Uhlen, M., Oksvold, P., Fagerberg, L., Lundberg, E. et al.,

Towards a knowledge-based Human Protein Atlas. Nat.

Biotechnol. 2010, 28, 1248–1250.

[3] Jogi, A., Brennan, D. J., Ryden, L., Magnusson, K. et al.,

Nuclear expression of the RNA-binding protein RBM3 is

associated with an improved clinical outcome in breast

cancer. Mod. Pathol. 2009, 22, 1564–1574.

[4] Ehlen, A., Brennan, D. J., Nodin, B., O’Connor, D. P.

et al., Expression of the RNA-binding protein RBM3 is

associated with a favourable prognosis and cisplatin

sensitivity in epithelial ovarian cancer. J. Transl. Med. 2010,

8, 78.

[5] Sureban, S. M., Ramalingam, S., Natarajan, G., May, R.

et al., Translation regulatory factor RBM3 is a proto-onco-

gene that prevents mitotic catastrophe. Oncogene 2008, 27,

4544–4556.

[6] Burd, C. G., Dreyfuss, G., Conserved structures and diversity

of functions of RNA-binding proteins. Science 1994, 265,

615–621.

[7] Sutherland, L. C., Rintala-Maki, N. D., White, R. D., Morin,

C. D., RNA binding motif (RBM) proteins: a novel family of

apoptosis modulators? J. Cell Biochem. 2005, 94, 5–24.

[8] Derry, J. M., Kerns, J. A., Francke, U., RBM3, a novel human

gene in Xp11.23 with a putative RNA-binding domain. Hum.

Mol. Genet. 1995, 4, 2307–2311.

[9] Wright, C. F., Oswald, B. W., Dellis, S., Vaccinia virus late

transcription is activated in vitro by cellular heterogeneous

nuclear ribonucleoproteins. J. Biol. Chem. 2001, 276,

40680–40686.

[10] Danno, S., Nishiyama, H., Higashitsuji, H., Yokoi, H. et al.,

Increased transcript level of RBM3, a member of the

glycine-rich RNA-binding protein family, in human cells in

response to cold stress. Biochem. Biophys. Res. Commun.

1997, 236, 804–807.

[11] Martinez-Arribas, F., Agudo, D., Pollan, M., Gomez-Esquer,

F. et al., Positive correlation between the expression of

X-chromosome RBM genes (RBMX, RBM3, RBM10) and the

proapoptotic Bax gene in human breast cancer. J. Cell

Biochem. 2006, 97, 1275–1282.

[12] Kita, H., Carmichael, J., Swartz, J., Muro, S. et al., Modu-

lation of polyglutamine-induced cell death by genes iden-

tified by expression profiling. Hum. Mol. Genet. 2002, 11,

2279–2287.

[13] Baldi, A., Battista, T., De Luca, A., Santini, D. et al., Identi-

fication of genes down-regulated during melanoma

progression: a cDNA array study. Exp. Dermatol. 2003, 12,

213–218.

[14] Mendoza, L. G., McQuary, P., Mongan, A., Gangadharan, R.

et al., High-throughput microarray-based enzyme-linked

immunosorbent assay (ELISA). Biotechniques 1999, 27,

778–780, 782–776, 788.

[15] Schweitzer, B., Wiltshire, S., Lambert, J., O’Malley, S. et al.,

Inaugural article: immunoassays with rolling circle

DNA amplification: a versatile platform for ultrasensitive

antigen detection. Proc. Natl. Acad. Sci. USA 2000, 97,

10113–10119.

[16] Geysen, H. M., Meloen, R. H., Barteling, S. J., Use of peptide

synthesis to probe viral antigens for epitopes to a resolution

of a single amino acid. Proc. Natl. Acad. Sci. USA 1984, 81,

3998–4002.

[17] Poetz, O., Ostendorp, R., Brocks, B., Schwenk, J. M. et al.,

Protein microarrays for antibody profiling: specificity and

affinity determination on a chip. Proteomics 2005, 5,

2402–2411.

[18] Rockberg, J., Lofblom, J., Hjelm, B., Uhlen, M., Stahl, S.,

Epitope mapping of antibodies using bacterial surface

display. Nat. Methods 2008, 5, 1039–1045.

[19] Kononen, J., Bubendorf, L., Kallioniemi, A., Barlund, M.

et al., Tissue microarrays for high-throughput molecular

profiling of tumor specimens. Nat. Med. 1998, 4, 844–847.

[20] Berglund, L., Bjorling, E., Jonasson, K., Rockberg, J. et al., A

whole-genome bioinformatics approach to selection of

antigens for systematic antibody generation. Proteomics

2008, 8, 2832–2839.

[21] Agaton, C., Falk, R., Hoiden Guthenberg, I., Gostring, L.

et al., Selective enrichment of monospecific polyclonal



antibodies for antibody-based proteomics efforts. J. Chro-

matogr. A 2004, 1043, 33–40.

[22] Nilsson, P., Paavilainen, L., Larsson, K., Odling, J. et al.,

Towards a human proteome atlas: high-throughput

generation of mono-specific antibodies for tissue profiling.

Proteomics 2005, 5, 4327–4337.

[23] Larsson, K., Eriksson, C., Schwenk, J. M., Berglund, L. et al.,

Characterization of PrEST-based antibodies towards human

Cytokeratin-17. J. Immunol. Methods 2009, 342, 20–32.

[24] Lofstedt, T., Jogi, A., Sigvardsson, M., Gradin, K. et al.,

Induction of ID2 expression by hypoxia-inducible factor-1: a

role in dedifferentiation of hypoxic neuroblastoma cells.

J. Biol. Chem. 2004, 279, 39223–39231.

[25] Wellmann, S., Buhrer, C., Moderegger, E., Zelmer, A. et al.,

Oxygen-regulated expression of the RNA-binding proteins

RBM3 and CIRP by a HIF-1-independent mechanism. J. Cell

Sci. 2004, 117, 1785–1794.



High nuclear RBM3 expression is associated with an improved prognosis in colorectal cancer

Documents

Transcript of High nuclear RBM3 expression is associated with an improved prognosis in colorectal cancer