ORIGINAL ARTICLE TMPRSS2:ERG blocks neuroendocrine and luminal cell differentiation to maintain prostate cancer proliferation Z Mounir 1 , F Lin 1 , VG Lin 1 , JM Korn 1 , Y Yu 1 , R Valdez 1 , OH Aina 2 , G Buchwalter 3,4 , AB Jaffe 1 , M Korpal 1,5 , P Zhu 1,5 , M Brown 3,4 , RD Cardiff 2 , JL Rocnik 1,6 , Y Yang 1 and R Pagliarini 1 The biological outcome of TMPRSS2:ERG chromosomal translocations in prostate cancer (PC) remains poorly understood. To address this, we compared the transcriptional effects of TMPRSS2:ERG expression in a transgenic mouse model with those of ERG knockdown in a TMPRSS2:ERG-positive PC cell line. This reveals that ERG represses the expression of a previously unreported set of androgen receptor (AR)—independent neuronal genes that are indicative of neuroendocrine (NE) cell differentiation—in addition to previously reported AR-regulated luminal genes. Cell sorting and proliferation assays performed after sustained ERG knockdown indicate that ERG drives proliferation and blocks the differentiation of prostate cells to both NE and luminal cell types. Inhibition of ERG expression in TMPRSS2:ERG-positive PC cells through blockade of AR signaling is tracked with increased NE gene expression. We also provide evidence that these NE cells are resistant to pharmacological AR inhibition and can revert to the phenotype of parental cells upon restoration of AR/ERG signaling. Our findings highlight an ERG-regulated mechanism capable of repopulating the parent tumor through the transient generation of an anti-androgen therapy-resistant cell population, suggesting that ERG may have a direct role in preventing resistance to anti-androgen therapy. Oncogene (2015) 34, 3815–3825; doi:10.1038/onc.2014.308; published online 29 September 2014 INTRODUCTION Prostate cancer (PC) is the most common cancer in men and a leading cause of cancer-related deaths in the western world. 1 Highly recurrent gene fusions involving v-ets erythroblastosis virus E26 transformation-specific (ETS) transcription factors to androgen-regulated genes have emerged as key initiators of PC pathogenesis. Among these, intra-chromosomal translocation of the androgen-regulated transmembrane protease serine 2 (TMPRSS2) gene to the ETS family member ERG is the most prevalent fusion, occurring in about 50% of PC cases. 2 Fusion of the ERG coding sequence with the promoter/5′ untranslated sequence of TMPRSS2 causes ectopic ERG expression under the control of androgen signaling in the prostate. 2 ERG can affect a number of pathways in vitro, with repression of androgen receptor (AR) target gene expression being a commonly observed phenotype. 3,4 In a number of published mouse models, ERG expression can initiate prostate hyperplastic phenotypes 3,5 and can also cooperate with PTEN haplo-insufficiency to promote invasive prostatic adenocarcinoma. 5,6 However, the role of ERG in the maintenance of PC proliferation remains unclear. 3,5,7–10 To better understand how ERG affects prostate cell biology, we developed model systems to compare the biological and transcriptional effects of TMPRSS2:ERG modulation in order to find common phenotypes. Our study identifies a role of ERG in suppressing neuroendocrine (NE) cell differentiation, with implications for PC resistance to anti-androgen therapy. RESULTS ERG attenuates both AR signaling and a neuronal gene signature To better understand how the TMPRSS2:ERG gene fusion alters gene expression in PC, we compared the transcriptional effects of ERG knockdown in the TMPRSS2:ERG-positive human PC cell line VCaP (Figure 1a) to those of TMPRSS2:ERG expression in a transgenic mouse model (Figure 1b). In the first model, two independent doxycycline (Dox)-inducible shRNA sequences (ERG shRNA 1 or ERG shRNA 2) were used to deplete ERG from VCaP cells, which led to robust ERG knockdown at the protein level (Figure 1a). ERG knockdown with either shRNA inhibited the proliferation of VCaP cells in short- (Supplementary Figure S1A) and long-term assays (Supplementary Figure S1B), whereas no effect was observed with a non-targeting control (NTC) shRNA. ERG shRNAs had no effect on colony formation in TMPRSS2:ERG- negative 22Rv1 PC cells (Supplementary Figure S1B), indicating that the proliferative effects of these shRNAs in VCaP are due to on-target inhibition of ERG expression. In the second model, a TMPRSS2:ERG transgenic mouse was generated using a human bacterial artificial chromosome expressing ERG exons 4–11 under the control of the hTMPRSS2 promoter, which mimics the most frequently observed ERG fusion event in PC (Supplementary Figure S1C; and Materials and methods). 2 The prostate-specific expression of ERG in these transgenic mice (Figure 1b) was sufficient to induce hyperplastic phenotypes in the mouse prostate, consistent with some previously published ERG trans- genic models (Supplementary Figures S1D). 3,5,8 1 Novartis Institutes for BioMedical Research, Cambridge, MA, USA; 2 Center for Comparative Medicine, University of California, Davis, CA, USA; 3 Department of Medical Oncology, Dana-Farber Cancer Institute and Harvard Medical School, Boston, MA, USA and 4 Center for Functional Cancer Epigenetics, Dana-Farber Cancer Institute, Boston, MA, USA. Correspondence: Dr R Pagliarini, Novartis Institutes of BioMedical Research, 250, Massachusetts Avenue, Cambridge, MA 02139, USA. E-mail: raymond.pagliarini@novartis.com 5 Current address: H3 Biomedicine, Cambridge, MA, USA. 6 Current address: Sanofi, Cambridge, MA, USA. Received 9 February 2014; revised 5 August 2014; accepted 13 August 2014; published online 29 September 2014 Oncogene (2015) 34, 3815 – 3825 © 2015 Macmillan Publishers Limited All rights reserved 0950-9232/15 www.nature.com/onc