Modern Pathology (2009) 22, 1564–1574 1564 & 2009 USCAP, Inc. All rights reserved 0893-3952/09 $32.00

Nuclear expression of the RNA-binding RBM3 is associated with an improved clinical outcome in breast cancer

Annika Jo¨gi1,2, Donal J Brennan3, Lisa Ryde´n4, Kristina Magnusson5,Ma˚rten Ferno¨ 6, Olle Sta˚l7, Signe Borgquist6, Mathias Uhlen8,Go¨ran Landberg1, Sven Pa˚hlman1,2, Fredrik Ponte´n4 and Karin Jirstro¨m1,2

1Department of Laboratory Medicine, Center for Molecular Pathology, Malmo¨ University Hospital, Lund University, Malmo¨, Sweden; 2CREATE Health Center for Translational Cancer Research, Lund University, Lund, Sweden; 3UCD School of Biomolecular and Biomedical Science, UCD Conway Institute, University College Dublin, Belfield, Dublin, Ireland; 4Division of Surgery, Department of Clinical Sciences, Lund University, Lund, Sweden; 5Department of Genetics and Pathology, Rudbeck Laboratory, Uppsala University, Uppsala, Sweden; 6Division of Oncology, Department of Clinical Sciences, Lund University Hospital, Lund, Sweden; 7Department of Biomedicine and Surgery, Division of Oncology, Faculty of Health Sciences, Linko¨ping University, Linko¨ping, Sweden and 8Department of Biotechnology, AlbaNova University Center, Royal Institute of Technology, Stockholm, Sweden

Single-strand RNA-binding (RBPs) are involved in many aspects of RNA metabolism and in the regulation of transcription. The RBP RBM3 was recently suggested to be a proto-oncogene in ; however, such a role has not been corroborated by previous studies in the colon or other tumor types, and the prognostic implications of tumor-specific RBM3 expression remain unclear. Mono-specific against RBM3 were generated. specificity was confirmed using siRNA gene silencing, western blotting and immunohistochemistry on a panel of breast cancer cell lines. Using tissue microarrays and IHC, RBM3 protein expression was examined in 48 normal tissues and in 20 common cancers. Additional analysis in two independent breast cancer cohorts (n ¼ 1016) with long-term follow-up was also carried out. RBM3 was upregulated in cancer compared to normal tissues. The nuclear expression of RBM3 in breast cancer was associated with low grade (Po0.001), small tumors (Po0.001), estrogen receptor (ER) positivity (Po0.001) and Ki-67 negativity (Po0.001) in both the breast cancer cohorts. An increased nuclear expression of RBM3 was associated with a prolonged overall and recurrence-free survival. The prognostic value was particularly pronounced in hormone receptor-positive tumors and remained significant in multivariate interaction analysis after controlling for tamoxifen treatment (HR: 0.49, 95% CI: 0.30–0.79, P ¼ 0.004). These data strongly indicate that nuclear RBM3 is an independent favorable prognostic factor in breast cancer, and seems to have a specific role in ER-positive tumors. Modern Pathology (2009) 22, 1564–1574; doi:10.1038/modpathol.2009.124; published online 4 September 2009

Keywords: RBM3; breast cancer; prognosis; estrogen receptor; antibody-based proteomics

Single-strand RNA-binding proteins (RBPs) are in the regulation of gene transcription.1,2 Reversible involved in many aspects of RNA metabolism and binding of RBPs to RNA is accomplished through specific sequences within the proteins, such as the ‘RNA-recognition motif’ (RRM), ‘RNA-binding mo- Correspondence: Professor K Jirstro¨m, Department of Laboratory tif’ (RBM), ‘ribonucleoprotein motif’ (RNP), the Medicine, Center for Molecular Pathology, Malmo¨ University arginine-rich motif (ARM), the cold-shock domain Hospital, Lund University, SE-205 02 Malmo¨, Sweden. (CSD) and many others.2 RBM proteins have E-mail: [email protected] Received 7 April 2009; revised 21 July 2009; accepted 22 July been recognized as a specific subgroup of RRM/ 2009; published online 4 September 2009 RBM/RNP containing RBPs. Each of the 10

www.modernpathology.org RBM3 and prognosis in breast cancer AJo¨gi et al 1565 proteins designated as RBM proteins contain Materials and methods between 1 and 4 copies of the RRM consensus sequence.1 The functional importance of the Breast Cancer Cohorts RRM domain is suggested by its evolutionary Consecutive cohort (cohort I) conservation across species and by its presence in A total of 512 consecutive cases of primary breast virtually every organelle of the cell in which RNA is cancer were diagnosed between 1988 and 1992 at 1 present. the Department of Pathology, Malmo¨ University Initially identified in a human fetal brain tissue Hospital. The median age at diagnosis was 64.2 3 cDNA library, the RBM3 gene maps to Xp11.23, and years (range: 27–96), and the median follow-up time is one of three X--related RBM from diagnosis until the first breast cancer event was (RBMX, RBM3 and RBM10). The RBM3 gene encodes 106 months (range: 0–207). This cohort is described alternatively spliced transcripts, with the longest in detail elsewhere.15,16 Ethical permission was reading frame encoding for a 157 amino-acid protein obtained from the local ethics committee at Lund containing one RRM domain and a glycine-rich University (Dnr 613/02). region.3 The RBM3 protein binds to both DNA and RNA.4 Although RBM proteins have been proposed to be Premenopausal randomized tamoxifen trial (cohort II) a novel family of apoptosis regulators,1,5–7 the exact A total of 564 premenopausal women with stage II function of RBM3 has yet to be fully elucidated. (pT2N0, pT1N1M0 and pT2N1M0) invasive breast RBM3 transcripts have been found in various human cancer were enrolled in a randomized trial of 2 years 3 tamoxifen treatment vs control between 1984 and tissues, and in vitro, RBM3 is one of the earliest 17 proteins synthesized in response to mild hypother- 1991. The median follow-up for patients without a 8 breast cancer event was 13.9 years. The study design mia. A correlation between the - 18,19 related RBM genes (RBMX, RBM3 and RBM10) and is described in detail elsewhere. Tissue micro- the proapoptotic Bax gene has been shown in breast arrays were constructed from 500 available tumor cancer.5 Downregulation of RBM3 is associated with specimens. Ethical permission was obtained from melanoma progression;9 however, a recent report the ethics committees at Lund and Linko¨ping described RBM3 as a proto-oncogene that protects Universities. against mitotic catastrophe in colorectal cancer cell lines.10 Taken together, the role of RBM3 in the context of tumor initiation and progression in Tissue Microarray Construction various cancer forms remains unknown and, to date, HPA-screening tissue microarrays were constructed no studies have addressed the role of tumor-specific as described previously.13 Before tissue microarray RBM3 protein expression in relation to disease construction from the extended cohorts, all breast outcome. cancer cases were histopathologically re-evaluated Using an antibody-based proteomics strategy, a on hematoxylin & eosin-stained slides. Areas repre- comprehensive atlas of human protein expression sentative of cancer were then marked and tissue patterns has been generated through the Human microarrays were constructed as described pre- Protein Atlas program (http://www.proteinatlas. 20 11,12 viously using either a manual tissue arrayer org). To date, more than 3000 antibodies (MTA-1, Beecher, WI, USA) (cohort I) or an (corresponding to more than 2600 different automated tissue arrayer (ATA-27, Beecher) (cohort human proteins) have been screened in tissue II). Two 1.0-mm cores (cohort I) or 0.6-mm cores microarrays, representing 48 types of normal tissues, (cohort II) were extracted from each donor block and 216 human tumor samples representing the 20 most assembled in a recipient block as described pre- common forms of human cancer and 47 cell viously.19 lines,13,14 In addition to generating a map of protein expression patterns, this atlas can also be used as a platform for the discovery of new diagnostic, 12 RBM3 Antibody Generation and prognostic and predictive cancer . Immunohistochemistry One such candidate is RBM3, which, on the basis of its differential expression among the breast A mono-specific Protein Epitope Signature Tags cancer samples in the protein atlas was further antibody was generated using a recombinant protein evaluated in tissue microarrays constructed from fragment corresponding to the human RBM3 gene two independent breast cancer cohorts––one con- (HPA 3624) as described previously.21 Sections of secutive cohort (n ¼ 512) and one cohort containing 4-mm thickness were dried, deparaffinized and tumors from a randomized tamoxifen trial con- rehydrated. After heat-mediated antigen retrieval at ducted in premenopausal women (n ¼ 500). The pH 9.9 (Target Retrieval Solution, Dako, Glostrup, latter cohort enabled assessment of the prognostic Denmark), IHC was performed in Techmate 500 impact of RBM3 expression and also its potential (Dako) using the anti-RBM3 (HPA 3624) antibody at value as a predictor of response to tamoxifen a dilution of 1:100. For all other antibodies, heat- treatment. mediated antigen retrieval was performed using

Modern Pathology (2009) 22, 1564–1574 RBM3 and prognosis in breast cancer 1566 AJo¨gi et al

microwave treatment for 2 Â 5 min in a citrate buffer Statistics before being processed either in the Ventana Bench- mark system (Ventana Medical Systems, AZ, USA) Chi-square test was used for comparison of RBM3 using pre-diluted antibodies to estrogen receptor expression and relevant clinico-patholgical charac- (ER) (Anti-ER, clone 6F11), progesterone receptor teristics. Kaplan–Meier analysis and log-rank test (PR) (Anti-PgR, clone 16) and Her2 (Pathway CB- were used to illustrate differences in both recurrence- USA, 760–2694) or in the Techmate 500 system free and overall survival according to the RBM3 (Dako) for Ki-67 (1:200, M7240; Dako). DAB (3,30- expression. Recurrence-free survival considered loco- diaminobenzidine) was used as a chromogenic regional and distant metastasis or breast cancer- substrate and slides were counterstained using specific death as the primary event, whereas con- hematoxylin. tralateral events were excluded. Cox regression Nuclear and cytoplasmic RBM3 expressions were proportional hazards models were used to estimate assessed using the same scoring system as that used the impact of RBM3 expression on recurrence-free in the web-based annotation of the HPA.22 Nuclear and overall survival in both univariate and multi- and cytoplasmic staining fractions were assigned a variate analysis, after adjusting for tumor size, age at score from 0 (0–1%), 1 (2–25%), 2 (26–75) to 3 diagnosis, ER, HER2, lymph-node status and (475%) and the staining intensities within the Nottingham Histological Grade. All calculations were respective subcellular locations were noted as performed using SPSS version 15.0 (SPSS, Chicago, 0 ¼ negative, 1 ¼ weak, 2 ¼ moderate and 3 ¼ strong. IL, USA). All statistical tests were two-sided and a ER, PR, HER2 and Ki-67 were assessed as described P-value o0.05 was considered statistically previously.15 ER and PR negativity was defined as significant. o10% positively staining nuclei, according to the current clinical guidelines in Sweden. Results RBM3 Protein Expression in Normal and Tumor Tissues Cell Culture and Gene Silencing Antibody-based screening of RBM3 in 48 different The human breast cancer cell lines, MCF-7, MDA- human tissues showed moderate nuclear and/or MB-231, T47D, and the human immortalized breast cytoplasmic expression in the breast and epididy- epithelial cell line MCF-10A were maintained under mal epithelium, lung macrophages, skeletal muscle, standard conditions. Transfection with siRNA parathyroid, placenta, skin, testes and tonsil against RBM3 (Applied Biosystems, Carlsbad, CA, (Figure 1). In a normal breast, scattered expression USA) or control siRNA (Applied Biosystems), was of RBM3 was observed in luminal rather than in performed with Lipofectamine 2000 (Invitrogen, myoepithelial cells (Figure 1c). Analysis in 20 Carlsbad, CA, USA) with a final concentration of different cancer types showed increased expression 50 nM siRNA. Cells were harvested for protein or of RBM3 in breast cancer (10 of 12) (Figure 1). RNA recovery 48 and 72 h post transfection initia- Increased RBM3 expression was also evident in tion. IHC was performed on cells after fixation in malignant carcinoid (3 of 4), testicular cancer (6 of paraformaldehyde (4%), dehydration and paraffin 12) and in non-melanoma skin cancer (9 of 11) embedding. (Figure 1).

Expression of RBM3 in Human Breast Cancer and PCR and Western Blotting Breast Cancer Cell Lines RNA isolation (RNeasy, Qiagen), complementary RBM3 expression was examined in a panel of DNA generation (Reverse Transcriptase Kit, Applied cell lines—MCF-7, T47D, MDA-MB-231 and MCF- Biosystems) and quantitative real-time PCR (qPCR) 10A. Western blot analysis using the anti-RBM3 (SYBR green PCR mastermix, Applied Biosystems) antibody recognized a single band at 17 kDa were performed as described previously.23,24 For (Figure 2a). The differential expression of RBM3 immunoblotting, cells were lysed in RIPA buffer was evident, with highest levels observed in MCF-7 (10 mM Tris-HCl pH 7.2, 160 mM NaCl, 1% Triton-X, cells (Figure 2a). IHC performed on the cell lines 1% sodium deoxycholate, 0.1% SDS, 1 nM EDTA, confirmed the western blot findings (Figure 2a). 1 mM EGTA) and supplemented with protease Western blot analysis of protein extracts collected inhibitor cocktail (Roche Applied Science, Burgess from six invasive ER-positive breast tumors Hill, West Sussex, UK). For each sample 60 mgof showed that RBM3 was expressed in four samples protein was separated by 10% SDS-PAGE and (Figure 2b). blotted onto PVDF membranes (Millipore, Billerica, Antibody specificity was confirmed using MA, USA). RBM3 was detected by HPA 3624 diluted siRNA-mediated gene silencing of RBM3 performed 1:250 in 5% non-fat milk PBST. on MCF-7 and MDA-MB-231 cell lines. RBM3

Modern Pathology (2009) 22, 1564–1574 RBM3 and prognosis in breast cancer AJo¨gi et al 1567

Figure 1 Expression pattern of RBM3 protein in normal and cancer tissues. A summary view showing levels of RBM3 protein expression in a panel of normal human tissues (a). Mild-to-moderate positivity (yellow) was observed in the skin (b), breast epithelium (c) and parathyroid (d). Lung macrophages, skeletal muscle, placenta, testes and tonsil also showed low levels of RBM3 protein expression. A summary view showing expression of RBM3 protein in 216 cancers from 20 different tumor types (e). The high frequency of moderate expression (yellow) in breast cancer must be noted. RBM3 expression was also evident in malignant carcinoid (f), non-melanoma skin cancer (g) and testicular cancer (h). Other tumor types showed lower levels of RBM3 protein expression.

gene expression was successfully reduced after Nuclear RBM3 Expression in Breast Cancer is transfection with two different siRNA sequences Associated with Favorable Clinico-Pathological (siRNA 58 and 59) as shown by qPCR at 72 h post Parameters transfection (Supplementary Figure 1). Gene silen- cing was also confirmed by western blot analysis RBM3 protein expression was then examined in two (Figure 2c). Significantly, IHC staining for RBM3 independent extended breast cancer cohorts. RBM3 of cells transfected with siRNA against RBM3 was present in various fractions and intensities in and control siRNA showed specific loss of signal tumor cell nuclei, whereas cytoplasmic staining, in the cells transfected with siRNA against RBM3 when present, was evident in the majority of tumor compared with control RNA or no siRNA (Figure cells in varying intensities (Figure 2e). After anti- 2d). The loss of RBM3 signal with siRNA- body optimization and immunohistochemical pro- mediated knockdown corroborates the specificity cessing, RBM3 protein expression was evaluable in of the antibody in both IHC and immunoblot 241 (47%) cases in cohort I and in 311 (61%) cases applications. in cohort II. To evaluate our study for any potential

Modern Pathology (2009) 22, 1564–1574 RBM3 and prognosis in breast cancer 1568 AJo¨gi et al

Modern Pathology (2009) 22, 1564–1574 RBM3 and prognosis in breast cancer AJo¨gi et al 1569 selection bias, baseline clinicopathological charac- levels of RBM3, using a cutoff of 75% positive teristics from the evaluated or ‘RBM3-known’ cases nuclei, was associated with an improved OS in and from the unevaluated or ‘RBM3-unknown’ cases cohort I (P ¼ 0.001) and in cohort II (P ¼ 0.006) in both are presented in Supplementary Table 1. (Figure 3). This shows that there were significantly more ER- Cox regression univariate analysis showed that and PR-positive tumors in the ‘RBM3-known’ cases increased RBM3 nuclear expression (475%) was in cohort I, otherwise no significant differences were associated with an improved recurrence-free survi- evident between the ‘RBM3-known’ and ‘RBM3- val (RR ¼ 0.58, 95% CI: 0.36–0.95, P ¼ 0.03) and OS unknown’ tumors in both cohorts, indicating the (RR ¼ 0.55, 95% CI: 0.38–0.78, P ¼ 0.001) in cohort I absence of any selection bias. (Table 2). Multivariate analysis within the same In cohort I, 150 tumors (62%) expressed high cohort showed that increased expression of nuclear levels of RBM3 (475% positive nuclei), whereas in RBM3 was an independent predictor of an improved cohort II only 121 tumors (39%) expressed high recurrence-free survival (RR ¼ 0.52, 95% CI: 0.27– levels of RBM3. This could be attributed to the fact 0.99, P ¼ 0.05) and overall survival (RR ¼ 0.56, 95% that cohort II consisted of exclusively premenopau- CI: 0.34–0.92, P ¼ 0.02) after controlling for grade, sal patients, and also contained a larger proportion nodal status, tumor size, ER, PR and Her2 status of high-grade, ER-negative tumors. Despite these (Table 2). Given the previous observation of a differences, the nuclear expression of RBM3 was possible relationship between RBM3 expression associated with a number of favorable clinicopatho- and ER positivity (Table 1), the relationship between logical parameters in both breast cancer cohorts, RBM3 and outcome in ER-positive tumors was such as small tumor size, low grade, Ki-67 negativity examined and RBM3 was identified as an indepen- as well as ER and PR positivity (Table 1). No dent predictor of improved overall survival in ER- association was evident between nuclear RBM3 positive patients (RR ¼ 0.49, 95% CI: 0.29–0.83, expression and Her2 or nodal status. In both cohorts, P ¼ 0.008) (Table 2). RBM3 had no impact on a significant correlation between nuclear and cyto- outcome in hormone receptor-negative patients plasmic staining was evident (R ¼ 0.154, P ¼ 0.016 (data not shown). for cohort I and R ¼ 0.214, Po0.001 for cohort II), Increased RBM3 nuclear expression (475%) was but apart from a positive correlation with ER also associated with an improved recurrence-free positivity in both cohorts (P ¼ 0.04 and P ¼ 0.02, survival (RR ¼ 0.70, 95% CI: 0.50–0.99, P ¼ 0.04) and respectively) and PR positivity in cohort II OS (RR ¼ 0.61, 95% CI: 0.42–0.87, P ¼ 0.006) in (P ¼ 0.01), cytoplasmic staining was not associated cohort II. Cox multivariate analysis showed that with any relevant clinicopathological parameters. RBM3 was not an independent predictor of recur- rence-free or overall survival in cohort II (Table 3). However, a subgroup analysis of ER-positive pa- RBM3 Expression is an Independent Predictor of tients using Cox multivariate regression analysis Disease Outcome in ER-Positive Breast Cancer adjusted for treatment bias revealed that increased RBM3 nuclear expression (475%) was associated Having shown that RBM3 nuclear expression was with an improvement in recurrence-free survival associated with a less aggressive phenotype, the (RR ¼ 0.56, 95% CI: 0.36–0.90, P ¼ 0.02) and OS relationship between RBM3 protein expression (RR ¼ 0.49, 95% CI: 0.30–0.79, P ¼ 0.004) (Table 3). and disease outcome in two independent breast Given these findings, the relationship between cancer cohorts was examined. Nuclear, but not cyto- RBM3 expression and tamoxifen response in cohort plasmic, expression was associated with impro- II was investigated. As recurrence-free survival ved survival and analysis of different thresholds was the end point of the original trial, this was showed an optimal cutoff at 75% positivity in used in this analysis. Kaplan–Meier analysis that both cohorts (Figure 3). RBM3 protein expression compared 2 years adjuvant tamoxifen to no systemic data were thus dichotomized into o75% vs 475% treatment (control), stratified according to RBM3 positive nuclear staining and this dichotomized expression, showed no relationship between RBM3 variable was used for further statistical analyses. expression and tamoxifen response in premenopau- Kaplan–Meier analysis showed that increased sal breast cancer (Figures 3e and f).

Figure 2 RBM3 is expressed in human breast-derived cell lines and breast tumors. Western blotting in a panel of human cell lines, breast cancer derived MCF-7, T47D and MDA-MB-231, and the immortalized breast epithelial cell line MCF-10A recognized the 17-kDa RBM3 protein (a, upper panel). Western blot was confirmed using IHC on MCF-7, MDA-MB-231 and T47D cell lines. RBM3 expression was confined to the nuclei in MCF-7 and MDA-MB-231 cells, whereas cytoplasmic expression was also evident in T47D cells (panel a, lower panel). Western blot of protein extracts from six breast tumors detected with anti-RBM3 (lower panel). In all, 60 mg protein was separated by SDS-PAGE and blotted onto PVDF membranes. Below each panel, actin staining of the same membranes is shown (b). Antibody specificity was confirmed using siRNA-mediated gene silencing. Two different commercially available (Applied Biochemistry) siRNAs against RBM3 and control siRNA was transfected into MCF-7 and MDA-MB-231 cells, and the RBM3 protein expression was assessed at 72 h using western blotting (c). IHC staining of siRNA-transfected cells (d). MCF-7 and MDA-MB-231 cells transfected with control siRNA or siRNA against RBM3 (siRNA58 and siRNA59). IHC was performed analogous to that on tissue microarrays. Examples of tumors with negative, low, moderate and high levels of RBM3 expression (e).

Modern Pathology (2009) 22, 1564–1574 RBM3 and prognosis in breast cancer 1570 AJo¨gi et al

Table 1 Relationship between RBM3 and clinicopathological parameters in two breast cancer cohorts

RBM3 nuclear fraction Cohort I n ¼ 512 (241) Cohort II n ¼ 500 (311) n (%) 0–1% 2–25% 26–75% 475% 0–1% 2–25% 26–75% 475% 19 (7.9) 26 (10.8) 46 (19.1) 150 (62.2) P-value 54 (17.4) 58 (18.6) 78 (25.1) 121 (38.9) P-value

Tumor size r20 mm 7 (36.8) 14 (53.9) 26 (56.5) 106 (70.7) 0.001 7 (13.0) 18 (31.0) 31 (39.7) 51 (42.2) 0.001 420 mm 12 (63.2) 12 (46.1) 20 (43.5) 44 (29.3) 47 (87.0) 40 (69.0) 47 (60.3) 69 (57.0) Missing 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 1 (0.8)

Grade I 1 (5.3) 4 (15.4) 7 (15.2) 40 (26.6) o0.001 2 (3.7) 2 (3.5) 10 (12.8) 16 (13.2) o0.001 II 3 (15.8) 14 (53.8) 19 (41.3) 73 (48.7) 10 (18.5) 14 (24.1) 37 (47.4) 66 (54.6) III 15 (78.9) 17 (65.4) 20 (43.5) 37 (24.7) 41 (75.9) 39 (67.2) 31 (39.7) 37 (30.6) Missing 0 (0) 0 (0) 0 (0) 0 (0) 1 (1.9) 3 (5.2) 0 (0) 2 (1.6)

Node status Negative 10 (52.6) 14 (53.8) 23 (50.0) 91 (60.7) 0.30 19 (35.2) 19 (32.8) 18 (23.1) 29 (24.0) 0.10 Positive 7 (36.8) 9 (34.6) 19 (41.3) 45 (30.0) 35 (64.8) 39 (67.2) 60 (76.9) 91 (75.2) Missing 2 (10.5) 3 (11.5) 4 (8.7) 14 (9.3) 0 (0) 0 (0) 0 (0) 1 (0.8)

ER status Negative 9 (47.4) 3 (11.5) 5 (10.9) 8 (5.3) o0.001 40 (74.1) 28 (48.3) 17 (21.8) 19 (15.7) o0.001 Positive 10 (52.6) 23 (88.5) 41 (89.1) 138 (92.0) 13 (24.1) 28 (48.3) 58 (74.4) 97 (80.2) Missing 0 (0) 0 (0) 0 (0) 4 (2.7) 1 (1.8) 2 (3.4) 3 (3.8) 5 (4.1)

PR status Negative 12 (63.2) 6 (23.1) 8 (17.4) 34 (22.7) 0.015 38 (70.4) 29 (50.0) 22 (28.2) 23 (19.0) o0.001 Positive 5 (26.3) 14 (53.8) 30 (65.2) 81 (54.0) 13 (24.1) 26 (44.8) 52 (66.7) 89 (73.6) Missing 2 (10.5) 6 (23.1) 8 (17.4) 35 (23.3) 3 (5.5) 3 (5.2) 4 (5.1) 9 (7.4)

HER2 (IHC) 0 13 (6.8) 15 (57.8) 19 (41.3) 76 (51.0) 0.88 36 (66.7) 31 (53.4) 43 (55.1) 60 (49.6) 0.38 1+ 1 (5.3) 4 (15.4) 13 (28.3) 40 (26.7) 6 (11.1) 7 (12.1) 14 (17.9) 17 (14.0) 2+ 3 (15.8) 4 (15.4) 6 (13.0) 10 (6.7) 1 (1.8) 4 (6.9) 7 (9.0) 9 (7.4) 3+ 2 (10.5) 1 (3.8) 510.9) 11 (7.3) 8 (14.8) 9 (15.5) 8 (10.3) 17 (14.1) Missing 0 (0) 2 (7.7) 3 (6.5) 13 (8.7) 3 (7.6) 7 (12.1) 6 (7.7) 18 (14.9)

Ki67 (%) 0–10 5 (26.3) 9 (34.6) 12 (26.1) 67 (44.6) o0.001 13 (24.1) 14 (24.1) 37 (47.4) 51 (42.2) o0.001 11–25 2 (10.5) 8 (30.8) 19 (41.3) 52 (34.7) 14 (25.9) 18 (31.0) 20 (25.6) 34 (28.1) 425 12 (53.2) 9 (34.6) 15 (32.6) 25 (16.7) 23 (42.6) 22 (37.9) 18 (23.1) 16 (13.2) Missing 0 (0) 0 (0) 0 (0) 6 (4.0) 4 (7.4) 4 (6.9) 3 (3.9) 20 (16.5)

ER, estrogen receptor; PR, progesterone receptor; IHC, immunohistochemistry.

Discussion was an independent predictor of recurrence-free and overall survival in ER-positive patients after con- This is the first study to evaluate tumor-specific trolling for other established prognostic parameters expression of the RBP, RBM3. Using an antibody- and tamoxifen treatment (Table 3). Increased levels based proteomics approach to explore RBM3 protein of nuclear RBM3 expression were not predictive of expression in normal tissues and tumors, we tamoxifen response in premenopausal breast cancer showed that RBM3 is seldom expressed in normal (Figure 3). These findings support the role of RBM3 tissues, but is expressed in a number of common as an important prognostic factor in breast cancer. cancers (Figure 1). Immunohistochemistry per- RBM3 is a cold-shock protein8 and the role of formed on breast cancer tissue microarrays showed these proteins in tumor progression is still not fully that nuclear RBM3 expression was associated with understood. Cold shock has been suggested to be an small, low-grade, ER-positive tumors in two inde- important mediator of a novel type of mitotic cell pendent breast cancer cohorts (Table 1). Increased death known as caspase-independent mitotic nuclear expression of RBM3 protein was associated death;25 however, the role of RBM3 in this process with an improved overall and recurrence-free remains unknown. CSD protein A has also been survival in both cohorts, and a subset analysis of shown to inhibit angiogenesis and lymphangiogen- ER-positive patients showed that increased levels of esis in a Lewis lung carcinoma xenograft mouse RBM3 was an independent predictor of overall model.26 RBM3 may influence tumor progression by survival in cohort I (Table 2). In cohort II, RBM3 controlling global protein expression within the cell

Modern Pathology (2009) 22, 1564–1574 RBM3 and prognosis in breast cancer AJo¨gi et al 1571

a b 1.0 1.0

0.8 0.8

0.6 0.6 > 75% > 75%

0.4 0.4 1-25% Overall survival Overall Overall survival Overall

0-1% 0-75% 0.2 0.2 26-75%

0.0 P < 0.001 0.0 P = 0.001

051015 20 051015 20 Time (years) Time (years) c d 1.0 1.0

0.8 0.8 > 75% > 75% 0.6 0.6 0-1% 1-25% 0.4 0.4 0-75% 26-75% Overall survival Overall Overall survival Overall

0.2 0.2

0.0 P = 0.018 0.0 P = 0.006

051015 20 051015 20 Time (years) Time (years) e f 1.0 1.0

0.8 0.8 Tamoxifen

Tamoxifen 0.6 0.6

Control

0.4 0.4

Control

Recurrence-free survival 0.2 Recurrence-free survival 0.2

0.0 P = 0.200 0.0 P = 0.113

051015 20 051015 20 Time (years) Time (years) Figure 3 Nuclear expression of RBM3 is associated with an improved OS, but is not a predictor of tamoxifen response. Kaplan–Meier analysis of overall survival in cohort I (a and b) and in cohort II (c and d) showing 75% positive nuclei as the optimal threshold for RBM3. Kaplan–Meier analysis of recurrence-free survival after 2 years tamoxifen compared with no adjuvant treatment (control) in premenopausal breast cancer identify no benefit for tamoxifen in tumors expressing high (e) or low (f) levels of RBM3. This analysis was confined to ER-positive patients in cohort II who had participated in a randomized control trial.

Modern Pathology (2009) 22, 1564–1574 RBM3 and prognosis in breast cancer 1572 AJo¨gi et al

Table 2 Cox univariate and multivariate analysis of recurrence-free and overall survival in cohort I

Recurrence-free survival Overall survival

RR (95% CI) P-value RR (95% CI) P-value

All patients (n ¼ 241) RBM3 Univariate Univariate Low 1.00 1.00 High 0.58 (0.36–0.95) 0.03 0.55 (0.38–0.78) 0.001

RBM3 Multivariatea Multivariatea Low 1.00 1.00 High 0.52 (0.27–0.99) 0.05 0.56 (0.34–0.92) 0.02

ER- positive patients (n ¼ 210) RBM3 Univariate Univariate Low 1.00 1.00 High 0.63 (0.37–1.08) 0.095 0.54 (0.37–0.80) 0.002

RBM3 Multivariatea Multivariatea Low 1.00 1.00 High 0.54 (0.27–1.11) 0.092 0.49 (0.29–0.83) 0.008

aMultivariate analysis included adjustment for grade, age, nodal status, ER status, PR status, Her2 and tumor size.

Table 3 Cox regression analysis of recurrence-free and overall survival in different patient strata in cohort II

Recurrence-free survival Overall survival

RR (95% CI) P-value RR (95% CI) P-value

All patients (n ¼ 300) RBM3 Univariate Univariate Low 1.00 1.00 High 0.70 (0.50–0.99) 0.04 0.61 (0.42–0.87) 0.006

RBM3 Multivariate Multivariate Low 1.00 1.00 High 0.78 (0.51–1.19) 0.25 0.68 (0.44–1.06) 0.09

RBM3 Multivariate+treatment adjusted Multivariate+treatment adjusted Low 1.00 1.00 High 0.79 (0.51–1.20) 0.27 0.69 (0.44–1.07) 0.10

ER positive (n ¼ 196) RBM3 Univariate Univariate Low 1.00 1.00 High 0.55 (0.36–0.85) 0.006 0.52 (0.33–0.81) 0.003

RBM3 Multivariate Multivariate Low 1.00 1.00 High 0.57 (0.36–0.90) 0.02 0.49 (0.30–0.79) 0.004

RBM3 Multivariate+treatment adjusted Multivariate+treatment adjusted Low 1.00 1.00 High 0.56 (0.36–0.90) 0.02 0.49 (0.30–0.79) 0.004

Multivariate analysis included adjustment for grade, age, nodal status, ER status, PR status, Her2 and tumor size.

by altering microRNA levels.27 RBPs have also been associated with favorable clinicopathological vari- described as a novel family of apoptosis modula- ables and an improved prognosis in two indepen- tors,1 and a correlation between the X chromosome- dent breast cancer cohorts. related RBM genes (RBMX, RBM3 and RBM10) and In vitro studies of the role of RBM3 in tumor the pro-apoptotic Bax gene has been shown in breast progression have been conflicting. Baldi et al9 cancer.5 Our data add to these findings by the showed that RBM3 was downregulated in metastatic observation that increased RBM3 expression was melanoma cell lines, whereas Sureban et al10

Modern Pathology (2009) 22, 1564–1574 RBM3 and prognosis in breast cancer AJo¨gi et al 1573 described RBM3 as a proto-oncogene that prevents Taken together, our data show that high nuclear mitotic catastrophe in colorectal cancer cell lines. In expression of RBM3 is clearly associated with a the study by Sureban et al,10 immunohistochemistry significantly improved survival, particularly in ER- was performed in a small number of colorectal positive tumors. Although some studies may contra- cancer specimens. It is not clear which subcellular dict these findings,10 the majority of studies on location was evaluated, and no relationship with members of the extended RBM family suggest a prognosis was described. protective role in tumor progression. RBM8A and Our data emphasize the fact that antibody-based RBM8B interact with the tumor suppressor gene approaches, using immunohistochemistry-based OVCA1,32 and RBMY is downregulated during analysis of clinically annotated tumor specimens, testicular cancer progression.33,34 Additional studies are indispensable in the validation of putative on well-annotated tumor samples are warranted to cancer biomarkers. Such an approach allows for confirm the utility of RBM3 as a prognostic the assessment of candidate expression in biomarker in various cancer forms. different subcellular locations and their prognostic impact. We identified nuclear RBM3 as an indepen- dent prognostic marker (Tables 2 and 3); however, Acknowledgements the ratio of nuclear to cytoplasmic staining may also affect disease outcome, eg, as previously shown for We thank Elise Nilsson for excellent assistance with survivin in breast cancer using automated image construction of the tissue microarrays. This work analysis.28 It is noteworthy that, as visualized in was supported by grants from Swegene, the Knut Figures 3a and b, the survival trends for the and Alice Wallenberg Foundation, the Swedish intermediate categories, ie, tumors with 1–25% Foundation for Strategic Research, the Swedish nuclear positivity, were inconsistent between the Cancer Society, Gunnar Nilsson’s Cancer Founda- two cohorts. However, the number of tumors in this tion, the Crafoord Foundation and the Research group was rather small, particularly in cohort I funds of Malmo¨ University Hospital. The UCD (10.8%). In cohort II, survival curves for RBM3- Conway Institute is funded by the Programme for negative and high tumors are clearly separated at 5 Third Level Institutions (PRTLI), as administered by years of follow-up (Figure 3c), and then converging. the Higher Education Authority (HEA) of Ireland. In this study, we used the same scoring system as in the protein atlas for the extended analysis, but further studies are warranted to determine optimal Disclosure/conflict of interest cutoff values. The authors declare no conflict of interest. Breast cancer has revolutionized the application of biomarkers. Assessment of hormone receptor status and Her2 is compulsory for all newly diagnosed breast cancers. Nevertheless, additional References biomarkers are required to identify prognostic 1 Sutherland LC, Rintala-Maki ND, White RD, et al. RNA subgroups and to tailor individual treatment binding motif (RBM) proteins: a novel family of 29 regimens. The role of adjuvant chemotherapy in apoptosis modulators? J Cell Biochem 2005;94:5–24. node-negative patients remains controversial. At 2 Burd CG, Dreyfuss G. Conserved structures and present, the majority of premenopausal women diversity of functions of RNA-binding proteins. with node-negative disease receive adjuvant Science 1994;265:615–621. chemotherapy; however, the absolute survival 3 Derry JM, Kerns JA, Francke U. RBM3, a novel human benefit from treatment is only 3% at 5 years and gene in Xp11.23 with a putative RNA-binding domain. 5% at 10 years.30 It has been estimated that 70–80% Hum Mol Genet 1995;4:2307–2311. of these patients, who are inherently at low risk, 4 Wright CF, Oswald BW, Dellis S. Vaccinia virus late transcription is activated in vitro by cellular hetero- would have survived without adjuvant therapy and geneous nuclear ribonucleoproteins. J Biol Chem thus avoided potentially toxic side effects; however, 2001;276:40680–40686. there is a lack of accurate prognostic markers to 5 Martinez-Arribas F, Agudo D, Pollan M, et al. Positive identify those patients who would benefit from correlation between the expression of X-chromosome chemotherapy.31 RBM genes (RBMX, RBM3, RBM10) and the proapop- Thus, an urgent need exists to identify markers, totic Bax gene in human breast cancer. J Cell Biochem which would allow for the selection of a ‘good 2006;97:1275–1282. prognosis’ subgroup that may not require adjuvant 6 Martin-Garabato E, Martinez-Arribas F, Pollan M, et al. systemic chemotherapy. The findings from cohort II The small variant of the apoptosis-associated indicated that the prognostic value of RBM3 seemed X-chromosome RBM10 gene is co-expressed with caspase-3 in breast cancer. Cancer Genomics Proteo- to be independent of tamoxifen treatment and o2% mics 2008;5:169–173. of the patients received adjuvant chemotherapy. In 7 Kita H, Carmichael J, Swartz J, et al. Modulation of light of this, RBM3 may be a promising marker polyglutamine-induced cell death by genes identified for selecting patients who will not benefit from by expression profiling. Hum Mol Genet 2002;11: adjuvant treatment. 2279–2287.

Modern Pathology (2009) 22, 1564–1574 RBM3 and prognosis in breast cancer 1574 AJo¨gi et al

8 Danno S, Nishiyama H, Higashitsuji H, et al. Increased mono-specific antibodies for tissue profiling. Proteo- transcript level of RBM3, a member of the glycine-rich mics 2005;5:4327–4337. RNA-binding protein family, in human cells in 22 Bjo¨rling E, Lindskog C, Oksvold P, et al. A web-based response to cold stress. Biochem Biophys Res Commun tool for in silico biomarker discovery based on tissue- 1997;236:804–807. specific protein profiles in normal and cancer tissues. 9 Baldi A, Battista T, De Luca A, et al. Identification of Mol Cell Proteomics 2007;7:825–844. genes down-regulated during melanoma progression: a 23 Lofstedt T, Jogi A, Sigvardsson M, et al. Induction of cDNA array study. Exp Dermatol 2003;12:213–218. ID2 expression by hypoxia-inducible factor-1: a role in 10 Sureban SM, Ramalingam S, Natarajan G, et al. dedifferentiation of hypoxic neuroblastoma cells. J Translation regulatory factor RBM3 is a proto-onco- Biol Chem 2004;279:39223–39231. gene that prevents mitotic catastrophe. Oncogene 24 Wellmann S, Buhrer C, Moderegger E, et al. Oxygen- 2008;27:4544–4556. regulated expression of the RNA-binding proteins 11 Uhlen M, Bjorling E, Agaton C, et al. A human RBM3 and CIRP by a HIF-1-independent mechanism. protein atlas for normal and cancer tissues based on J Cell Sci 2004;117:1785–1794. antibody proteomics. Mol Cell Proteomics 2005;4: 25 Kitagawa K, Niikura Y. Caspase-independent mitotic 1920–1932. death (CIMD). Cell Cycle 2008;7:1001–1005. 12 Ponten F, Jirstrom K, Uhlen M. The Human Protein 26 Saito Y, Nakagami H, Kurooka M, et al. Cold shock Atlas–a tool for pathology. J Pathol 2008;216:387–393. domain protein A represses angiogenesis and lym- 13 Kampf C, Andersson AC, Wester K, et al. Antibody- phangiogenesis via inhibition of serum response based tissue profiling as a tool for clinical proteomics. element. Oncogene 2008;27:1821–1833. Clinical Proteomics 2004;1:285–299. 27 Dresios J, Aschrafi A, Owens GC, et al. Cold stress- 14 Andersson AC, Stromberg S, Backvall H, et al. induced protein Rbm3 binds 60S ribosomal subunits, Analysis of protein expression in cell microarrays: a alters microRNA levels, and enhances global protein tool for antibody-based proteomics. J Histochem synthesis. Proc Natl Acad Sci USA 2005;102: Cytochem 2006;54:1413–1423. 1865–1870. 15 Borgquist S, Holm C, Stendahl M, et al. Oestrogen 28 Brennan DJ, Rexhepaj E, O’Brien SL, et al. Altered receptors alpha and beta show different associations to cytoplasmic-to-nuclear ratio of survivin is a prognostic clinicopathological parameters and their co-expression indicator in breast cancer. Clin Cancer Res 2008;14: might predict a better response to endocrine treatment 2681–2689. in breast cancer. J Clin Pathol 2008;61:197–203. 29 Brennan DJ, Kelly C, Rexhepaj E, et al. Contribution of 16 Helczynska K, Larsson AM, Holmquist Mengelbier L, DNA and tissue microarray technology to the identi- et al. Hypoxia-inducible factor-2alpha correlates to fication and validation of biomarkers and personalised distant recurrence and poor outcome in invasive breast medicine in breast cancer. Cancer Genomics Proteo- cancer. Cancer Res 2008;68:9212–9220. mics 2007;4:3–16. 17 Ryden L, Jonsson PE, Chebil G, et al. Two years of 30 EBCTCG. Polychemotherapy for early breast cancer: an adjuvant tamoxifen in premenopausal patients with overview of the randomised trials. Lancet 1998;352: breast cancer: a randomised, controlled trial with long- 930–942. term follow-up. Eur J Cancer 2005;41:256–264. 31 Brennan DJ, O’Brien SL, Fagan A, et al. Application of 18 Ryde´n L, Stendahl M, Jonsson H, et al. Tumor-specific DNA microarray technology in determining breast VEGF-A and VEGFR2 in postmenopausal breast cancer cancer prognosis and therapeutic response. Expert patients with long-term follow-up. Implication of a Opin Biol Ther 2005;5:1069–1083. link between VEGF pathway and tamoxifen response. 32 Salicioni AM, Xi M, Vanderveer LA, et al. Identifica- Breast Cancer Res Treat 2005;89:135–143. tion and structural analysis of human RBM8A and 19 Brennan DJ, Jirstrom K, Kronblad A, et al. CA IX is an RBM8B: two highly conserved RNA-binding motif independent prognostic marker in premenopausal proteins that interact with OVCA1, a candidate tumor breast cancer patients with one to three positive lymph suppressor. Genomics 2000;69:54–62. nodes and a putative marker of radiation resistance. 33 Lifschitz-Mercer B, Elliott DJ, Leider-Trejo L, et al. Clin Cancer Res 2006;12:6421–6431. Absence of RBM expression as a marker of intratubular 20 Kononen J, Bubendorf L, Kallioniemi A, et al. Tissue (in situ) germ cell neoplasia of the testis. Hum Pathol microarrays for high-throughput molecular profiling of 2000;31:1116–1120. tumor specimens. Nat Med 1998;4:844–847. 34 Schreiber L, Lifschitz-Mercer B, Paz G, et al. Lack of 21 Nilsson P, Paavilainen L, Larsson K, et al. Towards a RBM expression as a marker for carcinoma in situ of human proteome atlas: high-throughput generation of prepubertal dysgenetic testis. J Androl 2003;24:78–84.

Supplementary Information accompanies the paper on Modern Pathology website (http://www.nature.com/ modpathol)

Modern Pathology (2009) 22, 1564–1574