|Home | About | Journals | Submit | Contact Us | Français|
Characterization of the prostate cancer transcriptome and genome has identified chromosomal rearrangements and copy number gains/losses, including ETS gene fusions, PTEN loss and androgen receptor (AR) amplification, that drive prostate cancer development and progression to lethal, metastatic castrate resistant prostate cancer (CRPC)1. As less is known about the role of mutations2–4, here we sequenced the exomes of 50 lethal, heavily-pretreated metastatic CRPCs obtained at rapid autopsy (including three different foci from the same patient) and 11 treatment naïve, high-grade localized prostate cancers. We identified low overall mutation rates even in heavily treated CRPC (2.00/Mb) and confirmed the monoclonal origin of lethal CRPC. Integrating exome copy number analysis identified disruptions of CHD1, which define a subtype of ETS fusionnegative prostate cancer. Similarly, we demonstrate that ETS2, which is deleted in ~1/3 of CRPCs (commonly through TMPRSS2:ERG fusions), is also deregulated through mutation. Further, we identified recurrent mutations in multiple chromatin/histone modifying genes, including MLL2 (mutated in 8.6% of prostate cancers), and demonstrate interaction of the MLL complex with AR, which is required for AR-mediated signaling. We also identified novel recurrent mutations in the AR collaborating factor FOXA1, which is mutated in 5 of 147 (3.4%) prostate cancers (both untreated localized prostate cancer and CRPC), and showed that mutated FOXA1 represses androgen signaling and increases tumour growth. Proteins that physically interact with AR, such as the ERG gene fusion product, FOXA1, MLL2, UTX, and ASXL1 were found to be mutated in CRPC. In summary, we describe the mutational landscape of a heavily treated metastatic cancer, identify novel mechanisms of AR signaling deregulated in prostate cancer, and prioritize candidates for future study.
Although localized prostate cancer is highly curable, more than 32,000 U.S. men die annually of metastatic disease. Androgen-deprivation therapy results in rapid responses in men with metastatic prostate cancer, however nearly all patients eventually progress (castrate resistant prostate cancer [CRPC]). Although CRPC was thought to be androgen signaling independent, recent evidence demonstrates that androgen signaling is often maintained through varied mechanisms (reviewed in1,5). Gene expression and copy number profiling studies have identified recurrent gene fusions, chromosomal gains and losses, and deregulated pathways in prostate cancer 1. Resequencing studies have characterized the mutational spectrum of prostate cancer3,4,6, and the genomes of seven localized prostate cancers have been reported by Berger et al. 7. More recently, Kumar et al. reported the exomes of xenografts from 16 CRPC cases2.
Here, we sequenced the exomes of 50 lethal heavily pretreated CRPCs (WA2-60) obtained at rapid autopsy8, including three distinct sites in the same patient, and eleven treatment naïve, high-grade localized prostate cancers (T1-T97) (Supplementary Table 1). Sequencing results, including coverage statistics, mutation rates, validation rates, mutational spectrum, confirmation of the monoclonal origin of CRPC, and overlap with mutations observed in previous studies are provided in the Supplementary Results, Supplementary Figures 1–6 & Tables 2–6.
We used exome sequencing data to identify somatic copy number alterations 9 (see Methods and Supplementary Fig. 7 and Supplementary Tables 7–9), and as shown in Supplementary Fig. 8 we identified recurrent aberrations previously associated with prostate cancer development and progression (Supplementary Results). We additionally performed array CGH (aCGH) copy number and gene expression profiling on a matched cohort of benign prostate tissues, localized prostate cancers (3/59 sequenced) and 35 CRPCs (31/35 sequenced) (Supplementary Table 10). Profiles were uploaded into Oncomine (www.oncomine.com) for automated data processing, analysis and visualization, and are available for exploration. aCGH profiles were similar to copy number analysis by exome sequencing and to other prostate cancer profiling studies available in Oncomine (Supplementary Fig. 9). Global gene expression profiles were similar to previous studies (analyses available in Oncomine), with exceptions described in the Supplementary Results and Supplementary Fig. 10. Finally, we performed transcriptome sequencing of 11 prostate cancer cell lines to identify likely somatic variants (see Supplementary Results, Methods and Tables 11–15).
From our exome data, we identified nine genes that were significantly mutated (false discovery rate ≤ 0.10) (Fig. 1, Supplementary Tables 16&4), six of which have been reported as recurrently mutated in prostate cancer: TP53, AR, ZFHX3, RB1, PTEN and APC. Three significantly mutated genes do not have described roles in prostate cancer: MLL2, OR5L1 and CDK12. MLL2 encodes a H3K4-specific histone methyltransferase that is recurrently mutated in multiple cancers and CDK12 was recently identified as significantly mutated in ovarian serous carcinoma10. Additionally, using several approaches, we identified multiple significantly mutated pathways, including WNT signaling, and a PTEN interaction network (Supplementary Figure 11 & Tables 17&18); observations on significantly mutated genes and pathways are provided in the Supplementary Results.
Multiple candidate driver mutations in genes associated with AR signaling, DNA damage response, histone/chromatin modification, the spindle checkpoint, and classical tumour suppressors and oncogenes were also identified (Fig. 1). For example, we identified two deleterious mutations in PRKDC (I1137fs and E640*), which encodes the catalytic subunit of the DNA-dependent protein kinase involved in DNA double strand break repair and recombination, in patient T96, who had an extremely aggressive localized prostate cancer. Additional mutated genes in these pathways are described in the Supplementary Results.
To identify potential CRPC drivers, we considered genes with recurrent high-level gains or losses present in peaks of global copy number change, and compared results to mutated genes (Supplementary Fig. 12), as described in the Supplementary Results. Thus, we were intrigued by the peak of copy number loss on chr 5q21 (Fig 2a and Supplementary Fig. 12) harboring CHD1, which encodes an ATP-dependent chromatin-remodeling enzyme reported as deregulated in 3 of 7 prostate cancer genomes by Berger et al. (one somatic splice-site mutation and two rearrangements)7. As described in the Supplementary Results & Figure 13, across our exome and aCGH copy number analysis, we identified focal deletion/mutation of CHD1 (CHD1−) in 10/119 (8%) prostate cancers, which was significantly associated with ETS gene fusion negative (ETS−)status (two sided Fisher’s exact test, p=0.02). Association of CHD1 gene expression and genomic CHD1− status is shown in Supplementary Fig. 13b.
We next analyzed the association between CHD1− and ETS status in three prostate cancer aCGH and nine expression profiling studies (totaling 835 additional cancers) using Oncomine (Supplementary Results). As shown in Supplementary Figure 14 and Table 19 and Figure 2b, in total, across 13 DNA and RNA based studies, we identified 50 of 954 (5.2%) prostate cancers as CHD1−, 48 of which (96%) were ETS− (p<0.0001, two sided Fisher’s exact test). Together, our integrated analysis identifies CHD1−/ ETS− as a novel prostate cancer subtype.
ETS genes have a central role in prostate cancer, most commonly through fusion to androgen driven genes (i.e. TMPRSS2:ERG), and as the majority of ETS gene fusion positive (ETS+) CRPCs retained marked over-expression of the rearranged ETS gene (ERG, ETV1 or ETV5), our results support active androgen signaling in the majority of these men (Supplementary Fig. 13b). We next further explored the role of ETS genes in prostate cancer by evaluating our dataset for aberrations in additional ETS genes. Importantly, two CRPCs harbored deleterious mutations in ETV3, (P327fs in WA56 and W38* in WA26, both ETS+), which does not have a described role in prostate cancer. In addition, we were intrigued by the mutation of ETS2 (R437C) in WA30 (ETS−), as ETS2 shows a similar DNA binding motif as ERG11 and is located immediately telomeric to ERG (head-to-head orientation) in the commonly deleted region in TMPRSS2:ERG fusions through deletion. We also identified focal deletions extending telomeric from ERG in ERG+ cancers, as well as focal deletion of ETS2 in WA31, which shows outlier under-expression of ETS2 (Supplementary Results, & Fig 15). Additionally, the R437C mutation in ETS2 occurs in the ETS domain at a DNA contacting residue conserved in class I ETS transcription factors11, which include all ETS genes known to be involved in gene fusions in prostate cancer (Fig. 2c). Given observations that prostate cancers with TMPRSS2:ERG fusions through deletion may be more aggressive than those with fusions through insertion, we and others have speculated that the intervening region may harbor tumour suppressors, including ETS212–14. As shown in Fig 2d, we demonstrate that over-expression of wild type ETS2 results in decreased migration, invasion and proliferation in VCaP cells, while the R437C mutation has opposite effects, supporting a potential tumour suppressive role for ETS2.
In addition to MLL2 and CHD1, our integrated analysis identified mutations and copy number aberrations in multiple other genes involved in chromatin/histone modification (Fig. 1), including multiple members of the MLL complex (see Supplementary Results and Fig 3). Besides CHD1, which shows deregulation in both localized prostate cancer and CRPC (Fig. 2 and Supplementary Figs. 13&14), mutations of other chromatin/histone remodeling genes were infrequent in localized prostate cancer and concentrated in a single sample (T97, Fig. 1). Hence, we hypothesized that the mutated chromatin/histone remodelers we identified may mediate AR signaling through interaction. Thus, we immunoprecipitated endogenous AR from VCaP cells (ERG+ CRPC, active AR signaling) and blotted for members of the MLL complex (MLL2, MLL, ASH2L), UTX, ASXL1 and CHD1. FOXA1, a known direct interacting AR cofactor15, and EZH2 (a H3K27 histone methyltransferase over-expressed in CRPC), were evaluated as positive and negative controls, respectively. As shown in Figure 3a, members of the MLL complex, UTX and ASXL1 all interact with AR, while interaction with CHD1 was not observed. Reverse immunoprecipitation confirmed interactions between AR and MLL, MLL2, ASH2L and FOXA1 (Supplementary Fig. S16a).
As the MLL complex is implicated in epigenetic transcriptional activation, we studied its role in AR signaling. RNA interference of MLL or ASH2L using independent siRNAs (Supplementary Fig. 16b) significantly inhibited AR signaling, as assessed by inhibition of R1881 (synthetic androgen) stimulation of KLK3 (PSA) expression, with two siRNAs each against MLL or ASH2L inhibiting KLK3 expression at 24 hours by > 7.5 fold (each p<0.001) (Fig. 3b). Together, our data suggest that aberrations in AR and interacting proteins, including chromatin/histone remodelers, ETS genes (exemplified by ERG, which directly interacts with AR16) and known AR co-regulators including FOXA1 (see below), are common in CRPC (Fig. 3c).
Given the central role of AR signaling in CRPC and the selection for aberrations in AR occurring in CRPC, we were intrigued by the identification of a somatic 2 bp insertion in FOXA1 (S453fs) in the localized prostate cancer sample T12, and 340fs and P358fs indels in DU-145 and LAPC-4 (identified by transcriptome sequencing), respectively, given the well described role of FOXA1 in AR signaling17–22. Thus, we screened 101 localized and 46 CRPCs (including foci from all exome sequenced CRPCs), and identified somatic mutations of FOXA1 in 4 localized prostate cancers and 1 CRPC (total 5 of 147, 3.4%). Importantly, 4 of the 5 mutations, as well as both indels identified in the transcriptome screen, occurred in the C-terminal transactivating domain (Fig. 4a). As described in the Supplementary Results, we demonstrate that stable expression in LNCaP cells of either wild type FOXA1 or observed FOXA1 mutants increase proliferation in the presence of androgen (Fig 4b&c), represses the AR transcriptional program (Fig. 4d), and result in increased soft agar colony (Supplementary Fig 17) and xenograft growth (Fig 4e).
Our integrated, exome-based profiling of the mutational landscape of CRPC is notable for representing a large cohort of heavily pre-treated patients with lethal metastatic disease, which are not commonly studied, and provides insights into the resistance mechanisms that evolve in refractory tumours. Additionally, we identified a diverse series of potentially driving mutations and copy number alterations in both known and novel genes and pathways, including FOXA1. Lastly, our integrative genomics dataset provides a useful resource for the study of lethal prostate cancer, as well as determinants of, or resistance mechanisms to, radiation and chemotherapy.
See Supplementary Methods for source of prostate tissues and cell lines, nucleic acid isolation, exome and transcriptome sequencing and data analysis, mutation validation by Sanger sequencing, aCGH and DNA microarray expression profiling, ETS2 in vitro experiments, AR interactor immunoprecipitation, western blotting, and siRNA experiments, FOXA1 screening and in vitro and in vivo experiments.
The authors thank the patients and families who participated in the Rapid Autopsy Program. The authors thank Chandan Kumar, Jay Shendure, Mark Chaisson, and Ali Mortazavi for assistance with next generation sequencing data analysis, Karen Giles for assistance with manuscript preparation, and Sooryana Varambally, Anastasia Yocum, Terrence Barrette and Matthew Iyer for technical assistance. Supported in part by the NIH S.P.O.R.E. (P50 CA69568) to K.J.P. and A.M.C., the Early Detection Research Network (U01 CA111275 and U01 CA113913) to A.M.C., R01CA132874 and the National Functional Genomics Center (W81XWH-09-2-0014) to A.M.C. A.M.C. and K.J.P. are supported by the Prostate Cancer Foundation and are American Cancer Society Clinical Research Professors and A. Alfred Taubman Scholars. A.M.C. is supported by the Doris Duke Foundation. D. R. is supported by a DOD Postdoctoral Award (W81XWH-11-1-0339). J.R.P. is supported by a DOD Predoctoral Award (PC094290). S.A.T. and J.C.B. were supported by Young Investigator Awards from the Prostate Cancer Foundation.
Conflict of interest
The University of Michigan has been issued a patent on the detection of ETS gene fusions in prostate cancer, on which S.A.T., R.M., D.R.R. and A.M.C. are listed as co-inventors. The University of Michigan licensed the diagnostic field of use to Gen-Probe, Inc. S.A.T. has served as a consultant to Compendia Biosciences and has received honoraria from Ventana/Roche. A.M.C. has served as a consultant for Gen-Probe, Inc. and Ventana/Roche. D.R.R. and A.M.C. are co-founders of Compendia Biosciences, which licensed Oncomine from the University of Michigan. M.A. is an employee of Compendia Biosciences. The remaining authors declare no conflicts of interest.
Statement of WorkS.A.T., K.J.P., and A.M.C. conceived the study. K.J.P. established the rapid autopsy program and K.J.P., R.M., J.S., L.P.K. and S.A.T. carried out rapid autopsies and assisted in tissue procurement and analysis. Y.W., D.R., X.C., N.P., and X.J. isolated DNA and RNA and carried out whole exome and transcriptome sequencing. X.J. and X.C. performed gene expression and aCGH. C.G., M.Q., L.S., R.J.L., G.A.R., F.V., B.J.R., and S.A.T. carried out bioinformatics and biostatistical analysis of sequencing data. Y.W., S.M.D., D.R., and S.Y.C. carried out Sanger sequencing based validation. R.J.L., M.A., D.R.R., X.C., X.J., and S.A.T. analyzed gene expression profiling and aCGH data. A.P.K. and J.R.P. carried out studies on AR interactions and function. I.A.A. carried out ETS2 studies and Y.W., B.A., D.R., and J.C.B. carried out FOXA1 studies. S.A.T., C.S.G. and A.M.C. wrote the manuscript, which was reviewed by all authors. Copy number and gene expression data is available from GEO (GSE35988).