Recurrent SKIL-activating rearrangements in ETS-negative prostate cancer

Oncotarget6235 www.impactjournals.com/oncotarget www.impactjournals.com/oncotarget/ Oncotarget, Vol. 6, No.8 Recurrent SKIL-activating rearrangements in ETS-negative prostate cancer Matti Annala1,2,*, Kati Kivinummi1,2,*, Joonas Tuominen1,3, Serdar Karakurt1,3, Kirsi Granberg1,2, Leena Latonen1,3, Antti Ylipää1,2, Liisa Sjöblom1,3, Pekka Ruusuvuori1,2, Outi Saramäki1,3, Kirsi M. Kaukoniemi1,3, Olli Yli-Harja2, Robert L. Vessella4, Teuvo L.J. Tammela5, Wei Zhang6, Tapio Visakorpi1,3 and Matti Nykter1,2 1 Institute of Biosciences and Medical Technology - BioMediTech, University of Tampere, Tampere, Finland 2 Institute of Biosciences and Medical Technology - BioMediTech, Tampere University of Technology, Tampere, Finland 3 Fimlab Laboratories, Tampere University Hospital, Tampere, Finland 4 Department of Urology, University of Washington, Seattle, WA, USA 5 Department of Urology, Tampere University Hospital and Medical School, University of Tampere, Tampere, Finland 6 Department of Pathology, University of Texas M.D. Anderson Cancer Center, Houston, TX, USA * These authors contributed equally to this work Correspondence to: Matti Nykter, email: [email protected] Correspondence to: Tapio Visakorpi, email: [email protected] Keywords: prostate cancer, sequencing, fusion gene, SKIL Received: October 27, 2014 Accepted: January 15, 2015 Published: January 31, 2015 This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. ABSTRACT Prostate cancer is the third most common cause of male cancer death in developed countries, and one of the most comprehensively characterized human cancers. Roughly 60% of prostate cancers harbor gene fusions that juxtapose ETSfamily transcription factors with androgen regulated promoters. A second subtype, characterized by SPINK1 overexpression, accounts for 15% of prostate cancers. Here we report the discovery of a new prostate cancer subtype characterized by rearrangements juxtaposing the SMAD inhibitor SKIL with androgen regulated promoters, leading to increased SKIL expression. SKIL fusions were found in 6 of 540 (1.1%) prostate cancers and 1 of 27 (3.7%) cell lines and xenografts. 6 of 7 SKIL-positive cancers were negative for ETS overexpression, suggesting mutual exclusivity with ETS fusions. SKIL knockdown led to growth arrest in PC-3 and LNCaP cell line models of prostate cancer, and its overexpression led to increased invasiveness in RWPE-1 cells. The role of SKIL as a prostate cancer oncogene lends support to recent studies on the role of TGF-β signaling as a rate-limiting step in prostate cancer progression. Our findings highlight SKIL as an oncogene and potential therapeutic target in 1-2% of prostate cancers, amounting to an estimated 10,000 cancer diagnoses per year worldwide. INTRODUCTION Prostate cancer is diagnosed in over 900,000 men worldwide every year, making it the second most common cancer among men [1]. The standard-of-care for localized prostate cancer is radical prostatectomy or radiation therapy, whereas advanced tumors are treated with systemic therapies that inhibit androgen signaling [2]. More specific drug targets and driver mutations have been sought through extensive genomic characterization efforts [3,4]. We now know that genomic rearrangement plays a major role in the onset of prostate cancer, with 60% of tumors harboring chromosomal rearrangements that juxtapose androgen regulated promoters with the ETS family transcription factors ERG, ETV1, ETV4 or FLI1 [5,6]. However, the products of these fusion genes have proven difficult to target with small molecule inhibitors. Oncotarget6236 www.impactjournals.com/oncotarget More recently, overexpression of the trypsin inhibitor SPINK1 was found to define a second prostate cancer subtype mutually exclusive with ETS overexpression [7]. Monoclonal SPINK1 antibodies have shown efficacy in preclinical models [8], suggesting that SPINK1 inhibition may prove a beneficial treatment strategy in the 15% of prostate cancers positive for SPINK1 overexpression. Recent studies have identified other alterations mutually exclusive with ETS fusions, including mutations in the SPOP gene [4], and deletions of the chromatin remodeling gene CHD1 [9]. Despite these discoveries, a significant fraction of prostate cancers do not harbor any of the above alterations. In addition to ETS fusions and associated events, genomic characterization studies have identified nonsynonymous mutations in TP53, MED12, and PTEN [3,4,10], and gross deletions of the tumor suppressor genes PTEN, RB1 and TP53 [3,4,10]. More recently, a number of studies have highlighted the role of attenuated TGF-β signaling in prostate cancer progression [11–13]. SMAD4, a critical component of the TGF-β signaling cascade, is inactivated in a subset of advanced prostate cancers through promoter hypermethylation [14] or somatic mutation [10], and its expression is reduced in metastatic prostate cancer [11]. Mouse studies have shown that TGF-β signaling inhibits progression of PTEN-null tumors, and that SMAD4 deletion can overcome this inhibition [11]. In pancreatic adenocarcinoma, biallelic inactivation of SMAD4 is observed in 50% of tumors [15]. In this study, we performed transcriptome and lowcoverage whole genome sequencing on 28 untreated and 13 castration resistant prostate cancers, and identified a new prostate cancer subtype characterized by activating rearrangements of the SMAD inhibitor SKIL. RESULTS Sample acquisition and sequencing Fresh-frozen tissue from 12 benign prostatic hyperplasias (BPH), 28 untreated prostate cancers (PC), and 13 castration resistant prostate cancers (CRPC) was acquired from the Tampere University Hospital (Tampere, Finland). All samples contained a minimum of 70% cancerous or hyperplastic epithelial cells. PC samples were obtained by radical prostatectomy and locally recurrent CRPCs by transurethral resection of the prostate (Supplementary Table 1). Mean age at diagnosis was 60.8 years (range: 47.4-71.8) and mean PSA at diagnosis was 10.8 ng/ml (range: 3.5-48.1). Libraries were prepared for paired-end analysis on the Illumina HiSeq 2000. On average, we obtained 150 million paired end reads per sample from the low coverage whole genome sequencing, and 110 million paired end reads from the whole transcriptome sequencing (Supplementary Table 2). Discovery of recurrent SKIL-activating rearrangements Fusion events involving an ETS family transcription factor or SPINK1 overexpression were identified in 32 of 41 tumors based on transcriptome sequencing (Figure 1a, Supplementary Table 3). No SPOP mutations were identified in our cohort. We identified a novel TMPRSS2SKIL fusion gene in one CRPC sample (Figure 1AB, Supplementary Figure 1) and validated it using Sanger sequencing (Figure 1B) and fluorescence in situ hybridization (Figure 1C). The fusion merged the first three exons of TMPRSS2 with full length SKIL and led to SKIL overexpression due to the androgen regulated TMPRSS2 promoter (Figure 1D). SKIL encodes a SKIlike protein that inhibits TGF-β signaling by binding to and disrupting the heteromeric SMAD complex [16]. To search for more positive cases, we screened 76 additional tumors (Supplementary Table 1) and 22 LuCaP xenografts with qRT-PCR, and identified SKIL overexpression in one xenograft and one clinical sample. Transcriptome sequencing of these samples revealed a SLC45A3-SKIL fusion in LuCaP-77 and a MIPEP-SKIL fusion in the clinical sample, confirming SKIL as a recurrent 3’ fusion partner in prostate cancer (Figure 1E-F). Analysis of transcriptome sequencing data from the Cancer Genome Atlas (TCGA) prostate adenocarcinoma project revealed additional SKIL-activating rearrangements in 4 of 423 samples, with concomitant SKIL overexpression (Figure 2A-B). In the Taylor et al. dataset, two ETS-negative samples (PCA0015 and PCA0056) exhibited outlier overexpression of SKIL, but were excluded from further analysis due to lack of sequencing data [3]. Interestingly, all 5 SKIL-positive clinical samples with clinical information (two TCGA samples lacked clinical data) contained a Gleason grade 5 component or represented metastatic prostate cancer, suggesting that SKIL-activating alterations may associate with high-grade prostate cancer. None of the seven SKIL rearrangements disrupted the protein coding sequence of SKIL, suggesting that fulllength SKIL protein is necessary for oncogenic function. 6 of 7 SKIL rearrangements involved an androgen regulated promoter, indicating selection towards juxtapositions with highly active promoters. 6 of 7 fusion positive samples were negative for ETS overexpression, suggesting mutual exclusivity between SKIL and ETS rearrangements (p = 0.047, Fisher’s exact test). Sample TCGA-YL-A8SJ overexpressed both SKIL and ETV1 (Figure 2A), although we found no reads supporting an ETV1 rearrangement in either the transcriptome or exome sequencing data for this sample. In sample TCGA-HC-7211, the rearrangement between ACPP and SKIL had an unexpected structure, Oncotarget6237 www.impactjournals.com/oncotarget Figure 1: Recurrent SKIL-activating rearrangements in prostate cancer. (A) Matrix showing mutually exclusive overexpression of ERG, ETV1, ETV4, SPINK1, and SKIL in a transcriptome sequencing cohort of 41 prostate cancers. Red rectangles indicate overexpression, and black inner rectangles indicate fusion events. (B) Structure of the TMPRSS2-SKIL fusion gene identified in sample CRPC_348. Black lines indicate exon-exon junctions with transcriptome sequencing evidence. Fusion transcript was validated with Sanger sequencing from cDNA. (C) Fluorescence in situ hybridization validates the fusion at genomic level. One example of a fusion positive cell is shown. (D) SKIL expression in the transcriptome sequencing cohort of 41 prostate cancers and 12 BPHs. SKIL is strongly overexpressed in the TMPRSS2-SKIL positive sample. (E) SKIL expression was measured using qRT-PCR in a validation cohort of 76 prostatectomy samples. Sample PC_11423 exhibited SKIL overexpression and was found to contain a MIPEP-SKIL rearrangement by transcriptome sequencing. (F) SKIL expression was measured using qRT-PCR in LuCaP xenografts and cell line models of prostate cancer. Xenograft LuCaP-77 was found to contain an SLC45A3-SKIL rearrangement by transcriptome sequencing. Oncotarget6238 www.impactjournals.com/oncotarget with the first exon and promoter of ACPP placed downstream of SKIL in antisense orientation (Figure 2B). Despite the non-canonical structure, the rearrangement led to strong overexpression of full-length SKIL (Figure 2A), possibly due to chromatin remodeling induced by the androgen regulated ACPP promoter. The antisense promoter provoked expression of a spliced antisense transcript composed of cryptic exons located in SKIL introns (Supplementary Figure 2). Expression of the antisense transcript did not appear to interfere with SKIL splicing, as sense transcripts were normally spliced. Level of nuclear SKIL protein is elevated in a fusion positive sample To determine whether fusion positive clinical samples also overexpressed SKIL at the protein level, we used a monoclonal antibody to perform immunohistochemistry on the fusion positive TURP sample CRPC_348 and 8 negative controls. The TURP sample exhibited strong nuclear and modest cytoplasmic staining for SKIL, while negative controls showed no staining or only weak cytoplasmic staining for SKIL (Figure 3A). We also showed the overexpression of SKIL at the RNA level in CRPC_348 using RNA in situ hybridization (Figure 3B). Figure 2: SKIL-activating rearrangements in the TCGA prostate adenocarcinoma sequencing cohort. (A) Barplot showing expression of ERG, ETV1, ETV4, SPINK1 and SKIL in TCGA samples. Four samples exhibited significant SKIL overexpression and were found to harbor SKIL-activating rearrangements. (B) Structures of the ACPP-SKIL, SLC45A3-SKIL, MIPOL1-SKIL and HMGN2P46-SKIL rearrangements. Black lines indicate exon-exon junctions with transcriptome sequencing evidence. Oncotarget6239 www.impactjournals.com/oncotarget SKIL regulates proliferation and invasiveness of prostate cancer cells To better understand the biological role of SKIL in prostate cancer cells, we knocked SKIL down in PC-3 cells using two different siRNA (Figure 4A), and observed reduced cell growth (Figure 4B), invasiveness (Figure 4c), and colony formation (Figure 4D) relative to scrambled siRNA. The effect on cell growth was replicated in LNCaP cells (Figure 4E-F). To show that SKIL expression is truly androgen dependent in cells where SKIL is fused with androgen regulated promoters, we extracted LuCaP-77 xenograft tissue (with an SLC45A3-SKIL fusion) from castrate and non-castrate mice 1, 3 and 7 days postcastration. We then used qRT-PCR to quantify SKIL and KLK3 (PSA) expression at each timepoint. Expression values were normalized against TATA-box binding protein (TBP) and compared against non-castrate control mice. We observed a strong reduction in the expression of both KLK3 and SKIL on day 7, indicating androgen regulated expression (Figure 4G). Next, we created a SKIL overexpression model by transfecting immortalized prostate epithelial cells (RWPE1) with a pCI-Neo vector expressing hemagglutinin (HA) tagged SKIL [17]. Control RWPE-1 cells were transfected with an empty pCI-Neo vector (EV). In comparison to control cells, SKIL-transfected RWPE-1 cells exhibited higher expression of SKIL at both RNA and protein levels (Figure 5A-B), and greater invasive potential in a matrigel invasion assay (n = 3, p = 0.044, unpaired two-tailed t-test) (Figure 5C). SKIL overexpression had no effect on the growth of RWPE-1 cells (data not shown). Other genomic alterations in the sequencing cohort A previous study has shown that combined deletion of SMAD4 and PTEN in mouse prostates leads to aggressive prostate cancer with 100% penetrance [11]. Therefore, we set out to check whether SKIL-rearranged cancers harbored concomitant PTEN deletions. PTEN expression was not aberrantly low in any of the 7 SKILpositive cases, and only 1 of 4 SKIL-positive cancers in the TCGA cohort showed evidence of PTEN deletion, suggesting that SKIL rearrangements do not require combined PTEN loss. Genes associated with ERG overexpression, such as COL2A1, ALOX15, CRISP3, B3GNT6 and TDRD1, were not overexpressed in SKILrearranged tumors. Figure 3: Immunohistochemistry and RNA in situ hybridization of SKIL in a SKIL-rearranged tumor. (A) Anti-SKIL staining of paraffin-embedded sections from the SKIL-rearranged TURP sample CRPC_348 and two representative prostatectomies negative for SKIL rearrangement showing no staining or weak cytoplasmic staining. (B) RNA in situ hybridization of CRPC_348 and two representative prostatectomies negative for SKIL rearrangement, with probes targeting SKIL mRNA. Nuclei were stained with hematoxylin. Oncotarget6240 www.impactjournals.com/oncotarget Expression analysis of the TGF-β and BMP pathways (that both converge on SMAD4) revealed significantly reduced expression of TGF-β and BMP ligands in both untreated and castration resistant prostate cancer (Figure 6, Supplementary Figure 3). Other genomic alterations in SKIL-positive tumors included TP53 mutation in CRPC_348, hemizygous TP53 deletion in TCGA-KK-A8IL and TCGA-YL-A8SJ, hemizygous PTEN deletion in TCGA-YL-A8SJ, hemizygous NKX31 deletion in TCGA-HC-7211 and TCGA-YL-A8SJ, and an MLL3 frameshift deletion mutation in CRPC_348 (Supplementary Table 4). In addition to SKIL rearrangements, we found various other genomic alterations in our sequencing cohort (Figure 7, Supplementary Figure 4, Supplementary Table 5). Since the original sequencing did not include paired normal tissues, we used targeted sequencing on paired blood samples to filter out germline variants. The tumor suppressor TP53 was nonsynonymously mutated in 2 PCs and 5 CRPCs, with additional lossof-heterozygosity in 3 CRPCs. PTEN was disrupted by a stopgain mutation in one CRPC sample and deleted in eight other tumors. The AR negative tumor PC_6864 carried a KRAS p.G12R mutation that is known to cause constitutive KRAS activation in cancers of the colon, pancreas and lungs, but is less common in prostate cancer [18]. Sample CRPC_489 harbored two distinct AKT1 mutations, one of which (p.E17K) has been associated with dysregulated tissue growth in the Proteus syndrome [19]. We also identified somatic mutations that altered the forkhead domain and the N-terminal transactivation domain of the AR cofactor FOXA1. Four samples were positive for the HOXB13 p.G84E germline variant that has been associated with prostate cancer susceptibility [20], including one homozygous sample. One AR-negative CRPC sample had acquired a DOT1L-HES6 fusion. We and others have shown that HES6 overexpression is sufficient to induce completely androgen independent growth in prostate cancer cells [21,22]. In both PC and CRPC, we identified frequent alterations in chromatin modifiers including nonsynonymous mutations in CHD4, MLL3, HDAC5, KDM5B and MBD6, and a homozygous deletion of KDM6A in one sample (Figure 7). DISCUSSION SKIL (also known as SnoN) is a 684 amino acid Figure 4: In vitro knockdown experiments on SKIL. (A) SKIL expression was silenced in PC3 cells using two siRNAs, resulting in (B) reduced growth (n = 4), (C) invasion (n = 4) and (D) colony formation (n = 2). (E) SKIL expression was silenced in LNCaP cells using two siRNAs, resulting in (F) reduced growth (n = 4). (G) qRT-PCR time series of SKIL and PSA expression in castrate and non-castrate mice carrying LuCaP-77 xenografts (n = 2). Error bars, s.e.m. with first-order error propagation; *P<0.05; **P<0.01; ***P<0.001, unpaired two-tailed t-test. Oncotarget6241 www.impactjournals.com/oncotarget Figure 6: Genomic and transcriptomic changes in the context of the TGF-β signaling pathway. Genes are shown as boxes with two halves: the left half shows the percentage of untreated prostate cancers with two-fold upregulation (red) or downregulation (blue) relative to BPH, and the right half shows the same for castration resistant prostate cancers. Arrows indicate interactions between proteins or genes, the interaction type is written next to the arrow. Figure 5: In vitro overexpression experiments on SKIL. (A) Anti-SKIL western blot showing increased SKIL protein in SKILtransfected RWPE-1 cells. MCF-7 cells transfected with SKIL or scrambled siRNA are used to validate the band. (B) qRT-PCR quantification of SKIL in RWPE-1 cells transfected with SKIL or empty vector. (C) Matrigen invasion assay on RWPE-1 cells transfected with SKIL or empty vector (n = 3). Error bars, s.e.m. with first-order error propagation; *P<0.05; **P<0.01; ***P<0.001, unpaired two-tailed t-test. Oncotarget6242 www.impactjournals.com/oncotarget nuclear protein that is ubiquitously expressed in human tissues [23,24] and shares its domain structure with SKI, a protein that was originally discovered through its similarity with the transforming component of the Sloan-Kettering Virus [25]. Overexpression of either SKI or SKIL in chicken embryo fibroblasts is sufficient to induce oncogenic transformation [26], and SKIL expression is elevated in many human cancers, including cancers of the skin, breast, colon and blood [27]. The 3q26 locus is amplified in several cancer types, and SKIL (along with TLOC1) has been highlighted as the most potent oncogene in this region [28]. Both SKI and SKIL contain an 80 aa SAND-like domain that can bind with the MH2 domain of SMAD4 [29]. SMAD4 (coSmad) is an irreplaceable part of the heteromeric SMAD complexes that act as downstream mediators of TGF-β signaling. These heteromeric complexes are formed when SMAD4 binds with one or more receptor SMADs such as SMAD2 and SMAD3 [29]. The complex then translocates to the nucleus and activates transcription of TGF-β responsive genes. Binding of SKI/SKIL with the SMAD4 MH2 domain inhibits this transcriptional activation by preventing SMAD complex formation [29] or by recruitment of the nuclear co-repressor NCOR1 [16]. In addition to its binding with SMAD4, SKIL can also bind with the MH2 domains found in receptor SMADs through a domain located close to its N-terminal [16]. Neither SKI nor SKIL has been shown to directly bind DNA, despite both proteins containing a Dachshund homology domain that shares features with the forkhead/winged-helix family of DNA binding proteins [27]. In addition to the inhibitory role of SKIL on TGF-β signaling, a recent study has proposed that SKIL may play a role in regulating epithelial-to-mesenchymal transition (EMT) by inducing expression of SNAI2 (SLUG), a master regulator of EMT [28]. The same study also found that SMAD4 knockdown increased the invasiveness of human mammary epithelial cells, while SKIL overexpression had no effect on cell growth, in agreement with our findings [28]. Another recent study has proposed that SKIL can interact with and promote the activity of estrogen receptor α in the nuclei of breast carcinoma cells. The interaction occurs via two highly conserved nuclear receptor binding LxxLL-like motifs in SKIL [30]. This Figure 7: Characterization of genomic and transcriptomic changes in our prostate cancer sequencing cohort. Expression and copy number changes are shown in blue and red. Point mutations and indels affecting protein coding sequences are shown in green. White squares indicate missing data. Oncotarget6243 www.impactjournals.com/oncotarget finding is intriguing as it suggests a potential interaction between SKIL and androgen receptor, as some LxxLL motifs can bind with the ligand binding domain of AR [31]. The SKIL rearrangements reported in this paper occurred in both untreated and castration resistant prostate cancers, and involved the 5’ partner genes TMPRSS2, SLC45A3, MIPOL1, ACPP, MIPEP and HMGN2P46. The androgen regulated genes TMPRSS2 and SLC45A3 are the most common 5’ partners involved in ETS fusions in prostate cancer [6]. MIPOL1 is another androgen regulated gene that is involved in MIPOL1-ETS rearrangements in prostate cancers [32]. Expression of ACPP is androgen regulated and highly prostate specific among normal tissue types [33]. MIPEP is not generally considered androgen regulated or highly expressed in the prostate, but we observed high expression of both MIPEP and SKIL in the MIPEP-SKIL positive sample. HMGN2P46 is a pseudogene that is strongly expressed in AR-expressing prostate cancers but not expressed in AR-negative cancers in our cohort, suggesting an androgen regulated promoter. We conclude that six SKIL rearrangements involved an androgen regulated promoter, and one involved an otherwise highly active promoter. The discovery of SKIL-activating rearrangements in both untreated and castration resistant prostate cancers suggests that SKIL rearrangements may represent an early event in prostate tumorigenesis. This hypothesis is supported by the fact that SKIL rearrangements appear to be mutually exclusive with ETS fusions, which represent a known early event in prostate cancer progression [34]. The complex structure of two rearrangements highlights the fact that rearrangements can affect gene expression in a clinically significant manner without disrupting the transcribed portion of a gene. As an example of this complexity, one SKIL-activating rearrangement in our study juxtaposed an active promoter to a position downstream of SKIL in opposite orientation, and still resulted in strongly elevated expression of full length SKIL transcript. Based on the cohorts studied in this manuscript, we estimate that SKIL-activating rearrangements are found in 1-2% of diagnosed prostate cancers. Due to the high incidence of prostate cancer, this fraction translates to an estimated 10,000 diagnoses and 3,000 deaths caused by SKIL-positive prostate cancers per year worldwide [1]. Whether SKIL-positive prostate cancers differ in their clinical course from other prostate cancers remains to be evaluated in a larger study. Since 6 of 7 SKIL rearrangements involved androgen regulated promoters, we expect that existing treatment modalities based on androgen ablation will be effective at treating SKILrearranged cancers. Nonetheless, SKIL provides an intriguing new molecular target for personalized therapy, and highlights the role of TGF-β signaling in prostate cancer progression. MATERIALS AND METHODS Sequencing cohort Fresh-frozen tissue specimens from 12 benign prostate hyperplasias, 28 untreated prostate cancers, and 13 castration resistant prostate cancers were acquired from Tampere University Hospital (Tampere, Finland). Untreated prostate cancer samples were obtained by radical prostatectomy and locally recurrent CRPCs by transurethral resection of the prostate. Samples were snap-frozen and stored in liquid nitrogen. Histological evaluation and Gleason grading were performed by a pathologist based on hematoxylin/eosin-stained slides. All samples contained a minimum of 70% cancerous or hyperplastic cells. The use of clinical material was approved by the ethical committee of the Tampere University Hospital and the National Authority for Medicolegal Affairs. Written informed consent was obtained from the subjects. Validation cohort 76 additional hormonally untreated PC prostatectomy samples were acquired from the Tampere University Hospital (Tampere, Finland). Samples were snap-frozen and stored in liquid nitrogen. Histological evaluation and Gleason grading were performed by a pathologist based on hematoxylin/eosin-stained slides. Samples were confirmed to contain a minimum of 70% cancerous or hyperplastic cells by hematoxylin-eosin staining. Mean age at diagnosis was 62.1 years (range: 47.4-71.8), mean PSA at diagnosis was 11.8 (range: 3.1551.5). The use of clinical material was approved by the ethical committee of the Tampere University Hospital and the National Authority for Medicolegal Affairs. Written informed consent was obtained from the subjects. Cell lines and xenografts Prostate cancer cell lines PC-3, LNCaP, DU145, 22Rv1 and immortalized prostate epithelial cell line RWPE-1 were obtained from American Type Cell Collection (Manassas, VA, USA). LAPC-4 cell line was kindly provided by Dr. Charles Sawyers (University of California at Los Angeles, Los Angeles, CA, USA), VCaP and DuCaP by Dr. Jack Schalken (Radboud University Nijmegen Medical Center, Nijmegen, the Netherlands), and EP156T by Dr. Varda Rotter (Weizmann Institute of Science, Rehovot, Israel). All cell lines were cultured under recommended conditions. 22 previously established LuCaP-series xenografts were provided by R.L.V. These xenografts have been derived from primary and metastatic Oncotarget6250 www.impactjournals.com/oncotarget 38. Kim D, Pertea G, Trapnell C, Pimentel H, Kelley R, Salzberg S. TopHat2: accurate alignment of transcriptomes in the presence of insertions, deletions and gene fusions. Genome Biol. 2013;14(4):R36. 39. Venkatachalan SP, Bushman JD, Mercado JL, Sancar F, Christopherson KR, Boileau AJ. Optimized expression vector for ion channel studies in Xenopus oocytes and mammalian cells using alfalfa mosaic virus. Pflugers Arch. 2007;454(1):155–63. 40. Langmead B, Salzberg SL. Fast gapped-read alignment with Bowtie 2. Nat. Methods 2012;9(4):357–9. 41. The 1000 Genomes Project Consortium. A map of human genome variation from population-scale sequencing. Nature 2010;467(7319):1061–73. 42. Wang K, Li M, Hakonarson H. ANNOVAR: functional annotation of genetic variants from high-throughput sequencing data. Nucleic Acids Res. 2010;38(16):e164. 43. Bamford S, Dawson E, Forbes S, Clements J, Pettett R, Dogan A, Flanagan A, Teague J, Futreal PA, Stratton MR, Wooster R. The COSMIC (Catalogue of Somatic Mutations in Cancer) database and website. Br. J. Cancer 2004;91(2):355–8.