Significance of BRCA2 and RB1 Co-loss in Aggressive Prostate Cancer Progression

Pathologic variants of DNA damage response (DDR) genes are prevalent in a subset of men with metastatic castration-resistant prostate cancer (mCRPC; refs. 1–3). DDR is an essential defense and cell survival mechanism (4). Inherited (germline) or somatic genetic abnormalities of DDR pathway components, primarily insertions or deleterious mutations resulting in protein truncations, occur in 20%–25% of men with mCRPC (1–3). Although BRCA2 mutations are known to confer an increased risk of breast and ovarian cancer (5), recent observations have shown that alterations of BRCA2 are more prevalent than previously appreciated in men with prostate cancer and more frequent than alterations in any other DDR gene (6). In one study, BRCA2 alterations were seen in 13.3% of men with metastatic prostate cancer, while another found germline BRCA2 mutations in 5.3% of men with advanced prostate cancer (2, 3). Importantly, in a cohort of 1,302 men with localized and locally advanced prostate cancer, the 67 patients with BRCA2 germline mutations experienced more rapid progression to mCRPC, with 5-year metastasis-free survival rates of approximately 50%–60%, suggesting a more aggressive phenotype (7). A very recent germline sequencing study in a large cohort of men (7,636 unselected patients with prostate cancer and 12,366 male, cancer-free controls) revealed that pathologic variants of BRCA2 were significantly associated with prostate cancer risk (P < 0.001; ref. 8). Deep sequencing of cell-free DNA (cfDNA) from 202 patients with mCRPC treated with abiraterone acetate or enzalutamide after development of CRPC revealed that defects in BRCA2 and ATM were strongly associated with poor clinical outcomes and resistance to these second-generation antiandrogens, independent of other prognostic factors (9). The mechanisms by which loss of BRCA2 might promote aggressive prostate cancer and confer resistance to androgen deprivation therapy (ADT) and androgen signaling pathway inhibitors are not understood.

Previous studies have shown that loss of RB1 is associated with CRPC progression and metastasis (10, 11). Earlier studies showed that disruption of RB1 modulates androgen receptor (AR) activity in prostate cancer that in turn induces castration resistance and resistance to AR-directed therapeutics, and that the tumor-suppressive function of RB1 is distinct from canonical cell-cycle regulation of RB1 (12, 13). Very recently, Abida and colleagues showed that RB1 alteration was significantly associated with poor overall survival for 128 patients with mCRPC treated with first-line next-generation AR signaling inhibitors (ARSi; abiraterone or enzalutamide; P = 0.002; ref. 14). RB1 is located on chromosome 13q in close proximity to BRCA2. Deletion of chromosome 13q is a frequent event in localized prostate cancer and related to clinical aggressiveness (15, 16). Deletion (loss of heterozygosity) of 13q was associated with higher prostate cancer stage and grade and is common in metastatic disease (17). Moreover, in a study of 7,375 prostate cancer cases, 21% of localized cases harbored deletion of 13q; this deletion was associated with advanced tumor stage and early biochemical recurrence (18). Another prostate cancer study demonstrated that BRCA2 germline mutations are often associated with RB1 heterozygous deletion (19).

Herein, we identify a previously uncharacterized prostate cancer subset characterized by concomitant deletions (homozygous and heterozygous) of BRCA2 and RB1. Furthermore, for the first time, we demonstrate that even single copy loss of BRCA2 and RB1 is sufficient to induce an aggressive phenotype in prostate cancer. In this study, we developed a cell line–based model to examine the consequence of codeletion of BRCA2 and RB1 and demonstrated that this alteration is an independent genomic driver of therapy-resistant aggressive prostate cancer rather than the consequence of exposure to therapy. We further show that co-loss of BRCA2 and RB1 may induce an epithelial-to-mesenchymal transition (EMT) mediated by induction of the transcription factors SLUG or SNAIL or transcriptional coactivator PRRX1.

Tumors that harbor DDR defects, particularly BRCA defects, are sensitive to PARP inhibitors (PARPi) through a synthetic lethality mechanism (20). In a phase II clinical trial of olaparib in 49 patients with mCRPC, 16 (∼33%) showed a significant response to therapy (radiologic progression–free survival P < 0.001, overall survival P = 0.05). Of note, 88% of responders to olaparib harbored homologous recombination repair defects, due, in large part, to aberrations of BRCA2 and ATM (1). In our study, we developed a 3-color FISH method for rapid identification of codeletion of BRCA2 and RB1 in human prostate cancer cells and in mCRPC organoids. We show that PARP inhibition significantly attenuates growth of prostate cancer cell lines and organoids derived from human mCRPC that harbor not only homozygous but also heterozygous codeletion of BRCA2 and RB1. We propose that early recognition and intervention using PARPi-based therapy in prostate cancer cases identified as having BRCA2-RB1 codeletion could lead to substantial clinical benefit.

Human prostate cancer cells LNCaP, 22RV1, DU145, PC3, and VCaP were obtained from ATCC. LNCaP-C42 cells were obtained from VitroMed. The LNCaP-Abl cell line was provided by Zoran Culig (Innsbruck Medical University, Innsbruck, Austria), E006AA-T cells were provided by John T. Isaacs (The Johns Hopkins University School of Medicine, Baltimore, MD), PC3M cells were provided by Raymond C. Bergan (Knight Cancer Institute, Oregon Health & Science University, Portland, OR), and the LAPC4 cell line was provided by Charles Sawyers (Memorial Sloan Kettering Cancer Center, New York, NY). These cells were maintained in 10% FBS (LNCaP, LNCaP-C42, LAPC4, VCaP, 22RV1, DU145, PC3, PC3M, and E006AA) or 10% charcoal-stripped serum (LNCaP-Abl) supplemented with 2 mmol/L of l-glutamine and 1× antibiotic/antimycotic (Gemini Bio-Products) at 37°C in 5% CO₂. Human prostate epithelial cell RWPE1 was obtained from ATCC and cultured in keratinocyte serum-free medium (Thermo Fisher Scientific) at 37°C in 5% CO₂. Cells were authenticated by human short tandem repeat profiling at the MSK Integrated Genomics Operation Core. Patient-derived human prostate cancer organoids were cultured as described previously (21).

Lentiviral vectors encoding CRISPR or short hairpin RNA (shRNA) were generated as described previously (22) and transfected to LNCaP cells using LentiBlast (OZ Biosciences). Stable cells were generated using puromycin and/or hygromycin selection. We have designed three separate guide RNAs (gRNA) for human BRCA2 and human RB1 (Supplementary Table S1) and cloned the gRNAs into a LentiCRISPRv2-puromycin or hygromycin backbone respectively; a third-generation lentiviral backbone that constitutively expresses Cas9. Nontargeting scrambled gRNA (scr gRNA) was used as control. We used a similar strategy for generating 22RV1-RB1 cells and LNCaP-RB1 cells.

To generate BRCA2 knockout LNCaP cells by CRISPR/CAS9 methods, we infected parental LNCaP cells with BRCA2 scr gRNA lentivirus, followed by 5 μg/mL puromycin for 5 days. Loss of BRCA2 in the pooled population of LNCaP cells was analyzed by Western blot analysis using BRCA2-specific antibodies and this pooled population of cells was used for the following experiments. For generation of single cell–derived clones, we plated BRCA2 pooled population cells in very low density (500 cells in each 150-mm tissue culture plate in 20 mL of puromycin-supplemented media). After 4 weeks, single cell–derived clones were isolated using PYREX cloning cylinders (Thermo Fisher Scientific # 99-552-21). To determine the genome targeting efficiency of BRCA2 scr gRNA in the pooled population as well as in single cell–derived clones, we performed T7 endonuclease assay using EnGen Mutation detection kit according to manufacturer's protocol (NEB). The primers corresponding to specific gRNA that were used for PCR amplification are listed in Supplementary Table S1. The T7 assay demonstrated a mixed heterozygous population of cells containing wild-type (wt) and mutant BRCA2 DNA (Supplementary Fig. S1B).

To generate BRCA2 knockout RWPE1 cells, we cloned BRCA2 gRNA2 to LentiCRISPRv2-GFP backbone that constitutively expresses Cas9 and GFP. Lentiviral infected cells were selected by FACS sorting for GFP-positive cells (twice) and analyzed by Western blot analysis. To generate BRCA2-RB1 knockout–knockdown LNCaP cells, we first infected parental LNCaP cells with lentivirus containing BRCA2 gRNA or scr gRNA. Pooled populations of the stable cells were established by puromycin selection and analyzed by Western blot analysis and qPCR. We infected BRCA2-knockout or scr LNCaP cells with lentivirus containing RB1 shRNA followed by hygromycin selection. BRCA2-knockout or scr (gRNA) LNCaP cells also infected with lentiviral nontargeting shRNA (scr-shRNA) were used as control. Cells within 4–10 passages after stable selection were used for the following experiments.

siRNA or cDNA constructs were transiently transfected to indicated cells using the TransIT-X2 system (Mirus). A list of CRISPR, cDNA, shRNA, and SMARTpool siRNA constructs is provided in Supplementary Table S1. Efficiency of knockdown and overexpression was verified by qPCR and Western blot analysis.

Bioinformatic analysis of publicly available genomics data from various clinical cohorts was performed using data obtained from cBioPortal (23, 24) and Oncomine (25). The graphs and Kaplan–Meier survival curves were plotted using GraphPad Prism (version 7). Also used in this study were the cohorts described in the following sources: Armenia and colleagues 2018 (26); Baca and colleagues 2013 (27); Barbieri and colleagues 2012 (28); Beltran and colleagues 2016 (29); Grasso and colleagues 2012 (30); Hieronymus and colleagues 2014 (31); Kumar and colleagues 2016 (32); Robinson and colleagues 2015 (3); Setlur and colleagues 2008 (33); Taylor and colleagues 2010 (34); TCGA 2015 (35); TCGA provisional and pan-cancer prostate, TCGA provisional pan-cancer (unpublished data in cBioPortal); and Zehir and colleagues 2017 (36).

Cells were washed with HBSS and lysed in radioimmunoprecipitation assay (RIPA) buffer unless otherwise noted (50 mmol/L TRIS-HCl pH 7.4, 150 mmol/L NaCl, 1 mmol/L EDTA, 1% Triton X-100, 1% sodium deoxycholate, and 0.1% SDS) supplemented with protease and phosphatase inhibitors (Thermo Fisher Scientific). Protein concentrations were measured using the Bradford protein assay. Western blot was performed using specific antibodies (Supplementary Table S1). For BRCA2 Western blot we used Novex Tris-Glycine Mini Gels, WedgeWell format (6% or 4%–20%, Thermo Fisher Scientific).

Total RNA was extracted using the Direct-zol RNA Kit (Zymo Research) and reverse transcribed with qScript cDNA SuperMix (Quantabio). cDNA corresponding to approximately 10 ng of starting RNA was used for one reaction. qPCR was performed with Taqman Gene Expression Assay (Applied Biosystems). All quantifications were normalized to endogenous GAPDH. Probes used for qPCR are listed in Supplementary Table S1.

Total RNA from indicated cells and control LNCaP cells were isolated and analyzed by RNA sequencing by 50 million 2 × 50 bp reads in the MSK Integrated Genomics Operation Core Facility. RNA-sequencing data were analyzed in Partek. Heatmaps and volcano plots were developed using Partek manufacturer's instructions. Pathway analysis from RNA-sequencing data was performed using gene set enrichment analysis (GSEA) and ToppGene (37). The Molecular Signatures Database (MSigDB) is currently the most useful tool to analyze gene set enrichment from the transcriptomic data (38). Liberzon and colleagues developed a collection of “hallmark” gene sets as a part of MSigDB that summarize and represent specific well-defined biological states or processes and display coherent expression (39). These “hallmark pathways” summarize information across multiple gene sets and therefore provide more defined biological space for GSEA analysis (39). We used this hallmark signature to analyze our RNA sequencing and clinical cohort transcriptome data. Sequencing data are deposited to GEO repository under accession number GSE114155.

For the generation of survival curves using 10-gene (upregulated or downregulated from RNA sequencing) signatures, the Z score for each gene in 10-gene signatures was generated on the basis of the mRNA expression data from the Taylor cohort by using only the subset of primary prostate cancer samples. mRNA signature score was obtained by summing the Z scores. This generated a unique value for each sample in the cohort; this score was then divided into low and high based on the median. These mRNA scores were then correlated to clinical outcomes in the Taylor cohort. The Kaplan–Meier survival curves were generated and compared using the log-rank test.

3D organoid assays were performed as described previously (40). Cells were detached using Accutase (Innovative Cell Technologies), collected using 70-μm cell strainers, counted (1 × 10³ cell/well), and resuspended in serum-free PrEGM BulletKit (Lonza, catalog no. CC-3165 and CC-4177) supplemented with 1:50 B-27 supplement (Thermo Fisher Scientific catalog no. 17504044) and mixed with Matrigel Membrane Matrix (Thermo Fisher Scientific CB-40234C) in a 1:1 ratio. The cell and Matrigel mixture were plated on ultra-low attachment plates and allowed to grow for 2 weeks in serum-free PrEGM BulletKit supplemented with 1:50 B-27 medium. Organoids were counted and photographed using GelCount colony counter (Oxford Optronix). Organoid diameters more than 100 μm were counted.

Cells were plated on cover slips and allowed to grow for 48 hours. Cells were washed with HBSS and fixed in 4% paraformaldehyde for 10 minutes. Cells were permeabilized in 0.2% Triton X100 for 20 minutes in room temperature and blocked in blocking solution (2.5% BSA, 2.5% goat, and 2.5% donkey serum in HBSS) for 1 hour at room temperature followed by incubation with indicated primary antibody in blocking solution in 4°C overnight and then secondary antibody for 1 hour at room temperature. For Phalloidin staining, fixed cells were incubated in 1× Alexa Fluor 594 Phalloidin (Thermo Fisher Scientific) at 4°C overnight. Cells were mounted in mounting media containing DAPI and visualized and photographed under a fluorescent microscope (Nikon).

For MTT assay, cells were plated at 2.5 × 10³ per well in 96-well plates in complete media (10% FBS) or media supplemented with 10% charcoal-stripped serum. Cells were either treated with DMSO or with indicated inhibitors. After indicated times, cells were incubated in 0.5 mg/mL MTT (Invitrogen) for 1 hour at 37°C. MTT crystals were dissolved in isopropanol and absorbance was measured in BioTek plate reader at 570 nm and represented graphically. The BrdU assay was performed by BrdU Cell Proliferation Assay Kit according to manufacturer's instructions (BrdU cell proliferation assay kit, Cell Signalling Technology, # 6813). Cells were plated at 2.5 × 10³ per well in 96-well plates in complete media (10% FBS) or media supplemented with 10% charcoal-stripped serum. Cells were either treated with DMSO or with indicated inhibitors. BrdU incorporation in the proliferating cells was measured in BioTek plate reader at 450 nm and represented graphically. For the Crystal Violet cell proliferation assay, cells (in 96-well plate, treated with indicated drugs or cultured in FBS or CSS supplemented medium) were fixed in chilled 100% methanol for 10 minutes followed by staining with crystal violet (Millipore Sigma) for 2 hours and then washed with water. Crystal violet was dissolved in 1% SDS and absorbance was measured in BioTek plate reader at 595 nm and represented graphically.

Control and indicated LNCaP or RWPE1 cells were seeded at a density of 0.5 × 10⁵ cells per 24-well cell culture plate in complete medium. After 48 hours, a scratch was made with a 10 μL pipette tip in a confluent area of the cell culture dish. Photographs of a selected area of each scratch were taken 48 hours after scratching.

Matrigel invasion and Boyden chamber migration assays were performed as described previously (41). Briefly, cells in serum-free media (2.5 × 10³ cells/well for control LNCaP and variants; 1 × 10³ for PC3M and variants) were added in the top of the Matrigel invasion chamber (Thermo Fisher Scientific catalog no. 08-774-122) or Corning migration chamber (Thermo Fisher Scientific catalog no. 07-200-174). Ten percent FBS in the bottom chamber was used as chemoattractant. After indicated times, cells in the bottom chamber were fixed in methanol and stained with crystal violet, photographed, and counted under phase-contrast microscopy.

All cell lines were harvested and fixed in methanol: acetic acid (3:1). FISH analysis was performed on fixed cells and was based on TCGA data (Supplementary Fig. S3D; ref. 35). A 3-color probe was designed to detect loss of BRCA2 (red) and RB1 (orange). Region 13q12 (green) served as the control. The bacterial artificial chromosome (BAC) clones used in the probe-mix were as follows: BRCA2 (RP11-281G19; labeled with red dUTP), RB1 (RP11-305D15; labeled with orange dUTP), and 13q12 (RP11-867N8 and RP11-1031D16; labeled with green dUTP). All RP11 clones were purchased from the Roswell Park Cancer Institute Genomics Shared Resource (Buffalo, NY). Probe labeling, hybridization, posthybridization washing, and fluorescence detection were performed according to standard laboratory procedures. Prior to hybridization on cell lines, the probe was hybridized on normal peripheral blood (male) and locus specificity was confirmed. Slides were scanned using a Zeiss Axioplan 2i epifluorescence microscope (Carl Zeiss Microscopy) equipped with a 1.4-megapixel CCD camera (CV-M4+CL, JAI) controlled by Isis 5.5.9 imaging software (MetaSystems Group Inc).

The entire hybridized area was scanned through a 63× or 100× objective lens to assess quality of hybridization and signal pattern. Following initial scan, for each cell line, a minimum of 100 nuclei were scored and representative cells/regions imaged. A minimum of 25 metaphases were also analyzed and chromosomes counted to infer ploidy. The call for loss was in relation to ploidy; for example, in a near-tetraploid (∼4n) cell line, copy number ≤3 was considered as loss. Three normal lymphoblastoid cell lines (GM06875A, GM07535B, and GM21677), obtained from Corielle Institute (Camden, NJ), were also analyzed and for each cell line, a minimum of 100 nuclei scored to derive the cut-off values (false-positive). The cut-off value for each gene/locus was calculated as the mean of false-positive plus three times the SD and set at 5% for loss (<2 copies) and applicable to diploid cell lines.

Results are reported as mean ± SD or ± SEM, unless otherwise noted. Comparisons between groups were performed using an unpaired two-sided Student t test (P < 0.05 was considered significant), unless noted. P trends were analyzed by one-way ANOVA. Bar graphs were generated using GraphPad Prism software (version 7.0 GraphPad Software, Inc).

We investigated the consequences of BRCA2 deletion via lentiviral CRISPR/Cas9-mediated stable elimination of BRCA2 in LNCaP cells, a hormone-dependent human prostate cancer cell line. All three gRNAs successfully diminished BRCA2 transcript and protein levels in LNCaP cells (Fig. 1A; Supplementary Fig. S1A top and bottom). Furthermore, the T7 endonuclease assay revealed that all three gRNAs induced heterozygous loss of BRCA2 in LNCaP cells (Supplementary Fig. S1B). Previous studies demonstrated that BRCA2 inactivation impairs homologous recombination of DNA double-strand breaks and therefore BRCA2-null cells become sensitive to PARPi (42). As predicted, BRCA2-null LNCaP cells exhibited enhanced sensitivity to various PARPi and cisplatin (Supplementary Fig. S1C). However, our data also showed that BRCA2 knockout LNCaP cells exhibited more sensitivity toward talazoparib (BMN 673) and rucaparib compared with control gRNA (scr) infected cells (Supplementary Fig. S1C). We detected higher expression of FOLH1 in BRCA2 knockout LNCaP cells compared with control cells (Fig. 1A). We observed that elimination of BRCA2 increases phosphorylation of γH2AX in LNCaP cells (Fig. 1B and C, top), a biomarker for defective repair of double-strand breaks, indicating that CRISPR-mediated elimination of BRCA2 may induce a homologous recombination repair defect in LNCaP cells. We also observed an increase in S2056 autophosphorylation of DNA-PKcs in BRCA2 knockout LNCaP cells, indicating hyperactivation of DNA-PKcs (Fig. 1B and C, bottom). Furthermore, BRCA2-null LNCaP cells exhibited androgen-independent growth, as evidenced by enhanced 2D growth in androgen-deprived charcoal-stripped medium compared with control LNCaP (Fig. 1D, Supplementary Fig. S1E and S1F). Also, the BRCA2-null LNCaP cells exhibited relative resistance to enzalutamide (Fig. 1E; Supplementary Fig. S1D and S1F), indicating that these cells became castration resistant. Similarly, RNAi-mediated transient silencing of BRCA2 in LNCaP and LAPC4 (another androgen-dependent human prostate cancer cell line) cells also exhibited resistance to androgen depletion, as evidenced by growth in charcoal-stripped medium or complete media supplemented with enzalutamide (Supplementary Fig. S1G; Fig. 1F) BRCA2-null LNCaP cells also exhibited enhanced prostatosphere formation in 3D Matrigel cultures (organoids) in the androgen-independent condition (Fig. 1G) indicating that BRCA2-null LNCaP cells are more tumorigenic compared with control LNCaP cells.

To investigate the direct effect of the BRCA2-RB1 codeletion on human prostate cancer cells, we introduced a shRNA of RB1 (shRB1; in a lentiviral stable expression vector) into BRCA2-null LNCaP cells, generating BRCA2-RB1 knockdown LNCaP cells (hereafter called LNCaP-BRCA2-RB1; Fig. 2A; Supplementary Fig. S2A). We observed downregulation of BRCA2 protein and mRNA in RB1 knockdown LNCaP cells (Fig. 2A; Supplementary Fig. S2A). Interestingly, we have also observed that loss of BRCA2 attenuates RB1 protein expression in all BRCA2 knockout LNCaP cells (Fig. 2A; Supplementary Fig. S2B). Similarly, CRISPR-mediated knockout of RB1 also inhibits BRCA2 expression in LNCaP cells (Supplementary Fig. S2C), indicating a possible feed-forward loop between BRCA2 and RB1 expression in prostate cancer cells. We observed induction of E2F-1 in RB1 and/or BRCA2 knockdown/knockout cells (Fig. 2A). Furthermore, BRCA2-RB1 knockout/knockdown LNCaP cells exhibit relative resistance to the CDK4/6 inhibitor palbociclib as determined by MTT assay (Fig. 2B). Our data suggest that depletion of RB1 and/or BRCA2 in LNCaP cells is sufficient to induce canonical downstream pathway suppression by RB1.

LNCaP-BRCA2-RB1 cells exhibited elongated morphology (Fig. 2C). Immunofluorescence staining using phalloidin shows the remodeling of actin filaments in LNCaP-BRCA2-RB1 cells, further supporting the changes of cellular morphology upon co-loss of BRCA2 and RB1 (Fig. 2C). LNCaP-BRCA2-RB1 cells also exhibited enhanced wound migration and invasion through Matrigel (Fig. 2C; Supplementary Fig. S2D). Knockdown/knockout of either RB1 or BRCA2 alone induced an intermediate invasive phenotype (Fig. 2C; Supplementary Fig. S2D).

We observed increased phosphorylation of γH2AX in LNCaP-BRCA2-RB1 cells compared with BRCA2 or RB1 knockout/knockdown LNCaP cells (Fig. 2D; Supplementary Fig. S2E). Furthermore, we observed a very modest increase of S2056 autophosphorylation of DNA-PKcs in LNCaP-BRCA2-RB1 cells compared with BRCA2 knockout LNCaP cells (Fig. 2D; Supplementary Fig. S2F). RB1 loss alone only caused a modest increase of phosphorylation of γH2AX but not S2056 autophosphorylation of DNA-PKcs compared with control LNCaP cells (Fig. 2D; Supplementary Fig. S2F). Treatment with the PARPi olaparib and talazoparib caused more cell growth inhibition in LNCaP-BRCA2-RB1 cells than on BRCA2-null LNCaP cells (Fig. 2E). We were unable to detect any inhibitory effect of olaparib or talazoparib on RB1 knockdown cells compared with control LNCaP cells. These data suggested that co-loss of BRCA2 and RB1 increases sensitivity to PARPi in prostate cancer cells compared with BRCA2 loss alone. In contrast, RB1 loss alone was not associated with sensitivity of prostate cancer cells to PARPi (Fig. 2E).

To further confirm the effect of co-loss of BRCA2 and RB1 on the invasive phenotype of prostate cancer cells, we knocked out RB1 in 22RV1 cells that harbor oncogenic mutation of BRCA2 (T3033Nfs*11; Fig. 5B). RB1 knockout 22RV1 cells exhibit higher Matrigel invasion compared with control 22RV1 cells (Supplementary Fig. S2G).

To understand the molecular consequence of BRCA2-RB1 loss, we performed RNA sequencing on the LNCaP-BRCA2-RB1 cells. Interestingly, we observed a gradation of changes in gene expression in these cells compared with knockdown of either BRCA2 or RB1 alone, which provided further evidence of an additive effect of BRCA2-RB1 co-loss in LNCaP cells (Fig. 2F; Supplementary Table S2). Pathway analysis of upregulated genes in LNCaP-BRCA2-RB1 cells showed that the gene signature is prostate cancer–specific (Fig. 2G, top and bottom; Supplementary Table S3). Using single-sample GSEA (ssGSEA; ref. 43), we developed a 10-gene signature from the 10 mRNAs most upregulated and most downregulated (Supplementary Table S2) by co-loss of BRCA2 and RB1 in LNCaP cells. Both 10-gene signatures strongly predicted early relapse in localized prostate cancer in the Taylor cohort (Fig. 2H). In addition, we performed GSEA on the upregulated transcriptome of LNCaP-BRCA2-RB1 cells (Supplementary Fig. S2H; Supplementary Table S3) and observed that induction of several essential molecular pathways, including regulation of cell differentiation and transcription, were enriched upon co-loss of BRCA2 and RB1 in LNCaP cells. However, we are unable to detect any correlation between previously published RB1 signatures (12, 44) and our LNCaP cell–derived BRCA2-RB1 signature (Supplementary Figs. S2I and S2J).

Our observations prompted us to investigate the molecular mechanism by which the invasive phenotype resulting from co-loss of BRCA2 and RB1 in LNCaP cells occurs. We performed the “hallmark pathways” analysis using GSEA in the upregulated transcriptome of LNCaP-BRCA2-RB1 cells (Fig. 2G, top). We observed increased expression of several EMT and dedifferentiation-related signaling pathways (mTORC1, Hedgehog, TNFα-NFκB, TFGβ), including enrichment of the hallmark EMT signaling pathway (Supplementary Fig. S3A; Supplementary Table S4). We observed decreased expression of E-cadherin and increased expression of the mesenchymal marker vimentin (both translational and transcriptional) in the double knocked down cells compared with control LNCaP cells (Fig. 3A; Supplementary Fig. S3B). Our immunofluorescence staining also showed loss of cell membrane E-cadherin and β–catenin and gain of vimentin in the LNCaP-BRCA2-RB1 cells (Fig. 3B). This observation is consistent with the elongated morphology and actin cytoskeleton remodeling of LNCaP-BRCA2-RB1 cells (Fig. 2C). Moreover, these findings further supported the observation that LNCaP-BRCA2-RB1 cells undergo an EMT-like transformation, while knockdown of BRCA2 or RB1 alone induce a partial EMT-like phenotype (Fig. 3A and B). However, we did not find any changes in expression of AR or the neuroendocrine marker NSE in double knockout/knockdown LNCaP cells compared with control cells (Fig. 3A).

We overexpressed BRCA2 and RB1 in highly aggressive mesenchymal-like PC3M cells that exhibit low endogenous BRCA2 and RB1. Overexpression of BRCA2 and RB1 inhibits vimentin and N-cadherin expression in PC3M cells; however, NSE expression remains unchanged (Fig. 3C). Interestingly, we also observed that overexpression of either of the genes (BRCA2 or RB1) autoregulates the expression of the other in PC3M cells (Fig. 3C), further indicating the feed-forward loop between BRCA2 and RB1 in prostate cancer. BRCA2 and RB1 also exhibit diminished Boyden chamber migration and Matrigel invasion in overexpressed PC3M cells compared with control cells (Supplementary Fig. S3D).

To further validate whether loss of BRCA2 and RB1 is sufficient to induce EMT in prostate cancer cells, we used the immortalized benign human prostate cells RWPE1. RWPE1 cells express significantly lower RB1 protein compared with parental LNCaP cells due to their expression of a single copy of human papilloma virus 18 (HPV 18; Fig. 3D; ref. 45). We used CRISPR to knockout BRCA2 from RWPE1 cells (Fig. 3D) and observed that BRCA2 from knockout RWPE1 cells exhibit elongated morphology and remodeling of actin filament (Fig. 3E). We also observed enhanced wound migration in BRCA2 knockout RWPE1 cells (Fig. 3E). Our immunofluorescence staining also showed loss of cell membrane E-cadherin and β catenin and gain of vimentin in the BRCA2-knockout RWPE1 cells (Fig. 3F). As predicted, BRCA2-null RWPE1 cells also exhibited enhanced sensitivity to PARPi olaparib (Fig. 3G).

We analyzed the transcriptome that is enriched in the BRCA2-RB1 codeleted TCGA provisional prostate cancer cohort and performed GSEA hallmark pathway analyses (Supplementary Table S5). We observed that EMT is one of the common pathways enriched in the BRCA2-RB1–null cell line and TCGA cohort (Supplementary Fig. S3C). More importantly, our analysis of the Setlur prostate cancer cohort (lethal vs. indolent) using Oncomine suite and GSEA also demonstrated enrichment [P = 0.015, q (P_adj value based on FDR) = 0.039, normalized enrichment score (NES) = 1.764] of the EMT pathway (Supplementary Fig. S3F; Supplementary Table S6), indicating the clinical significance of EMT in lethal prostate cancer.

To determine which transcriptional factors were involved in EMT transformation, we analyzed the expression of previously demonstrated EMT-related transcription factors by qPCR (Fig. 3H). We observed upregulation of EMT transcription factors SLUG (SNAI2) and SNAIL (SNAI1) and transcriptional coactivator PRRX1 in LNCaP-BRCA2-RB1 compared with LNCaP cells (Fig. 3H). Relative SLUG expression was significantly (>100-fold) higher compared with other EMT transcription factors in LNCaP-BRCA2-RB1 cells (Fig. 3I). Previously SLUG had been demonstrated as an androgen-regulated transcription factor that facilitates castration resistance in prostate cancer (46). We observed that treatment with androgen (R1881) significantly increased SLUG, but not SNAIL or PRRX1 mRNA in LNCaP-BRCA2-RB1 cells (Fig. 3I). We showed that siRNA-mediated knockdown of SLUG, SNAIL, or PRRX1 inhibits invasiveness compared with control siRNA–transfected LNCaP-BRCA2-RB1 cells or control (scr) LNCaP cells (Fig. 3J).

We analyzed BRCA2 status in a pan-cancer dataset derived from cBioPortal for Cancer Genomics (23, 24) where BRCA2 is frequently altered (BRCA2 alteration frequency >5% of cases; number of cases >50; Supplementary Fig. S4A). We observed more frequent homozygous deletions of BRCA2 in prostate cancer (localized and mCRPC) than in other cancers (whereas other cancers exhibit frequent mutational events; Supplementary Fig. S4A). In the Armenia and colleagues' prostate cancer dataset, which contains both primary (localized) and mCRPC cases (26), we observed BRCA2 alterations in approximately 10% of mCRPC cases compared with only approximately 2.5% in primary cases (P = 2.91e-06; summarized in Supplementary Table S7). BRCA2 alterations are more common than other major DDR pathway components, and are enriched in mCRPC relative to localized disease, suggesting it is associated with, if not a driver of, aggressive disease (Fig. 4A; Supplementary Table S7). Note that the Armenia cohort was not designed to determine germline mutations of DDR pathway components.

Further in-depth analysis of the BRCA2 status in multiple independent publicly available and published prostate cancer datasets (from cBioPortal) revealed that a significant fraction of localized as well as metastatic cases exhibit deletion (homozygous and heterozygous) of BRCA2, which had not been previously described (Supplementary Fig. S4B). Our analysis also revealed that BRCA2 alterations (homozygous or heterozygous deletions, as well as mutations, denoted as BRCA2 alterations throughout this study) were significantly enriched (P = 0.0216) in this combined mCRPC dataset (n = 444) compared with a primary (localized) dataset (n = 925; Supplementary Fig. S4B). While the TCGA provisional cohort was not designed to look at clinical outcomes (overall survival), in the available data, BRCA2 deletion is significantly associated with shorter disease/progression-free survival (5 years; P_trend = 0.0059; Fig. 4B). Interestingly, we are unable to detect any difference in disease progression between patients with homozygous and heterozygous BRCA2 deletions (Fig. 4B). Our observation suggests that even heterozygous loss of BRCA2 may be associated with a more aggressive form of prostate cancer.

Homozygous and even heterozygous deletion of BRCA2 significantly reduced BRCA2 protein levels as determined by reverse-phase protein array (RPPA; P_trend = 0.0083; Fig. 4C). We were unable to detect any difference in BRCA2 protein expression between heterozygous and homozygous cases (Fig. 4C). However, in the same TCGA prostate cancer cohort, we did not detect a relationship between BRCA2 deletion (either homozygous or heterozygous) and BRCA2 mRNA expression (Supplementary Fig. S4C). Heterozygous deletion of BRCA2 is sufficient to reduce protein level but not mRNA level, indicating that single copy loss may lead to haploinsufficiency of BRCA2 protein expression. Decreased BRCA2 protein expression is significantly correlated with shorter disease-free survival (Supplementary Fig. S4D). Taken together, for the first time we demonstrate the potential clinical significance of heterozygous deletion of BRCA2 in primary prostate cancer through loss of BRCA2 protein expression.

A prior sequencing study revealed that codeletion (heterozygous and homozygous) of RB1 and BRCA2 is present in a significant fraction of primary prostate cancers [∼25% in TCGA provisional cohort (Fig. 4D, top); ref. 35]. Interestingly, in the MSK-IMPACT prostate cancer cohort (36), we observed that BRCA2 homozygous deletion, not mutation, is enriched in metastatic cases and cooccurs with homozygous RB1 deletion (Supplementary Fig. S4E). In the TCGA and Taylor prostate cancer datasets, patients with primary prostate cancer who have BRCA2-RB1 codeletion have significantly shorter disease/progression-free survival compared with patients with deletion of neither or of RB1 alone (Fig. 4D bottom; Supplementary Fig. S4H, bottom), while deletion of BRCA2 without RB1 is rare (Fig. 4D, top; Supplementary Fig. S4H, top). Also, BRCA2 copy number and RB1 copy number are correlated in both primary prostate cancer (TCGA) and mCRPC (Kumar) cohorts (Supplementary Fig. S4F). However, note that unlike BRCA2, RB1 mRNA expression is significantly associated with RB1 genomic deletion (heterozygous and homozygous) in primary (TCGA) and mCRPC (Kumar) cohorts (Supplementary Fig. S4G).

Codeletion of BRCA2-RB1 is significantly enriched in high Gleason grade prostate cancer as well as in metastases (Fig. 4E; Supplementary Table S8). However, deletion of RB1 alone is not significantly associated with stage or progression to metastasis (Fig. 4E). The details of the codeletion and P values of each stage are summarized in Supplementary Table S8. We also observed that approximately 10% of low-grade (Gleason 6) cases harbor genomic codeletion of BRCA2 and RB1 (Supplementary Fig. S4I). We established the mRNA expression of the genes that are upregulated due to codeletion of BRCA2 and RB1 in Gleason 6 disease in TCGA provisional prostate cancer cohort (Supplementary Fig. S4I; Supplementary Table S9). To further assess the importance of the BRCA2-RB1 codeletion in low-grade primary prostate cancer, we compared the BRCA2-RB1 loss Gleason 6 gene signature from TCGA to the metastatic prostate cancer signature using Oncomine suite (25). In the Taylor cohort, we observed enrichment of this BRCA2-RB1 loss Gleason 6 gene signature in metastatic prostate cancer (P = 2.00E-20, odds 3.7; Supplementary Fig. S4I).

We extended our study to matched (localized and metastatic) prostate cancer samples in the Kumar and colleagues cohort to further assess the direct association between codeletion of both genes and metastatic progression. Supplementary Table S10 displays the 12 patients with mCRPC in the Kumar and colleagues cohort that had matched localized and metastatic samples. All 8 patients (66.7%) who had codeletion of BRCA2 and RB1 in their localized tumors retained their BRCA2-RB1 codeletion in all of their metastatic tumors (Supplementary Table S10), indicating that this codeletion may be critical to metastatic progression. Interestingly, for the one patient (06-081) who had an RB1 deletion alone in his localized prostate tumor, the RB1 deletion was not seen in all his metastatic tumors. These data suggest that codeletion of BRCA2 and RB1 in primary disease is likely a driver to mCRPC.

In an analysis of the Armenia and colleagues' dataset, which contains both primary and mCRPC cases, we found that BRCA2-RB1 co-loss in early prostate cancer appeared to be significantly associated with increased fraction of genome altered (Fig. 4F). Fraction of genome altered is a biomarker associated with genomic instability and also appears to be associated with prostate tumor aggressiveness (47), suggesting that BRCA2-RB1–null tumors are likely aggressive in nature.

Copy number segment analysis of primary and mCRPC samples from the Armenia and colleagues' dataset indicated frequent deletion of the BRCA2-RB1 region of chromosome 13q (Fig. 4G). We also observed copy number loss of other genes located in the BRCA2-RB1 region in patients who harbored the codeletion of BRCA2 and RB1 (Fig. 4H, top). To further assess the nature of this deletion, we analyzed the mRNA of all the protein-coding genes on chromosome 13q (Fig. 4H, bottom; Supplementary Table S11). We observed that the mRNA expression of chromosome 13q genes between BRCA2 and RB1 was lower in BRCA2-RB1–deleted patients compared with wild-type patients in the TCGA 2015 cohort (Fig. 4H). More in-depth analysis in the TCGA pan-cancer prostate cohort (extended TCGA 2015 cohort) showed that the mRNA expression of genes located downstream of BRCA2 was significantly lower than for genes located upstream of BRCA2 in patients who harbored a codeletion of BRCA2 and RB1 (Fig. 4I). These data indicate an interstitial deletion of the BRCA2-RB1 region in prostate cancer rather than deletion of the entire chromosome 13q arm.

We observed an association between the loss of mRNA expression of BRCA2-RB1 region genes in the mCRPC cohorts compared with primary (localized) prostate cancer. Loss of expression of these genes was seen (to a greater degree) in mCRPC compared with primary cases in the Grasso (P = 2.12E-6, OR 4.4) and Taylor (P = 2.47E-20, OR 12.2) cohorts (Grasso: primary n = 59, mCRPC n = 35; Taylor primary n = 131, mCRPC n = 19; Fig. 4J; Supplementary Fig. S4J). Note that in the Grasso cohort, the mCRPC specimens were isolated by rapid autopsy from metastatic sites (30). Taken together, these data suggest that an interstitial deletion of the BRCA2-RB1 region of chromosome 13q may be associated with castration resistance and metastasis.

To further confirm that in prostate cancer BRCA2 is frequently deleted with RB1 rather than alone, we developed a 3-color FISH probe to apply to human cells. We validated our probes on human peripheral blood and immortalized prostate cells (RWPE-1), in which almost every cell exhibits two copies of BRCA2 and RB1 (Fig. 5A and C; Supplementary Fig. 5A; Supplementary Table S12). Human CRPC cell lines E006AA, DU145, PC3, and PC3M exhibited uniform heterozygous codeletion of BRCA2 and RB1 (Fig. 5A; Supplementary Fig. S5A; Supplementary Table S12). Heterozygous codeletion of BRCA2 and RB1 is associated with high fraction of genome altered in PC3 and DU145 cells, but not in 22RV1 and MDA PC2B cells (absence of codeletion) or in LNCaP cells (partial codeletion) in The Cancer Cell Line Encyclopedia (ref. 48; Fig. 5B). The detailed analysis of the BRCA2-RB1 copy number and ploidy of individual prostate cancer cell lines is shown in Supplementary Table S12. Most importantly, we were able to detect heterozygous codeletion of BRCA2 and RB1 in VCaP cells (not noted in sequencing study), which also display a high fraction of genome altered (Fig. 5A and B; Supplementary Fig. S5A).

We found that approximately 60% of parental LNCaP cells harbor loss of one or more copies of RB1, including approximately 10% with codeletion of BRCA2 (Fig. 5A–C; Supplementary Table S12). We observed heterogeneity in chromosome number (ploidy, 2–10 copies of chromosome/cell) in LNCaP cells indicating the heterogeneous nature of the parental LNCaP cell line (Fig. 5D; Supplementary Table S12). Previous studies have identified a castration-resistant low-PSA subpopulation among parental LNCaP cells (49). This is consistent with our current observation and suggests the clonal expansion of a subpopulation of LNCaP cells in the castrate environment as demonstrated previously (49). Interestingly, the LNCaP-derived hormone-independent LNCaP-Abl cell line (able to grow in androgen-independent culture condition) exhibits uniform co-loss of 1 of 4 copies of BRCA2 and RB1, further indicating this codeletion is directly associated with ADT resistance and also may indicate a clonal expansion of castration-resistant BRCA2-RB1–deleted population from parental LNCaP cells (Fig. 5A, C, and D; Supplementary Table S12).

In these cell lines, the protein and mRNAs of both genes were consistently decreased (Fig. 5E; Supplementary Fig. S5B). Although the castration-resistant LNCaP subclone C42 exhibits uniform heterozygous deletion of RB1 only, attenuation of BRCA2 protein and mRNA is observed as well. This indicates that an additional mechanism of loss of BRCA2 in RB1-deleted cells may lead to the castration-resistant phenotype (Fig. 5A and E; Supplementary Fig. S5B; Supplementary Table S12).

Our immunoblot analysis showed that the human CRPC cell lines DU145, PC3, and the PC3-derivative PC3M that exhibited uniform heterozygous codeletion of BRCA2 and RB1 (Fig. 5A; Supplementary Fig. S5A; Supplementary Table S12) also exhibit the EMT-like phenotype, including upregulation of vimentin and loss of E-cadherin expression (Fig. 5F). However, LAPC4, 22RV1 (mutant BRCA2 but wild-type RB1), and LNCaP (RB1 partial deletion but wild-type BRCA2) exhibit more epithelial-like characteristics (refs. 34, 48; Fig. 5F). Codeletion of BRCA2-RB1 in LNCaP-Abl cells is also associated with upregulation of vimentin protein expression, which is consistent with our current observations (Fig. 5F).

Codeletion of BRCA2-RB1 in the LNCaP-Abl cell line is consistently associated with sensitivity to various PARPi (rucaparib and talazoparib) and platinum drugs compared with parental LNCaP cells (Fig. 5G). Note that although parental LNCaP cells harbor several defects in various DDR genes (Supplementary Fig. S5C), the LNCaP subline LNCaP-Abl exhibits more PARPi-mediated cell growth inhibition compared with parental LNCaP cells (Fig. 5G). Although the COSMIC cancer cell line dataset showed that LNCaP cells harbor a deletion-frameshift mutation of BRCA2 (p.D946fs*14), sequencing studies from The Cancer Cell Line Encyclopedia (Fig. 5B; ref. 48) and the Taylor prostate dataset (34) were unable to detect such BRCA2 mutation in parental LNCaP cells. A prior study also showed that LNCaP cells express a wild-type BRCA2 transcript (50). These data suggest that heterozygous codeletion of BRCA2 and RB1 in LNCaP-Abl cells is sufficient to reduce the mRNA expression of both genes and therefore induce sensitivity to PARPi (Fig. 5G). Similarly, PC3M cells, which also harbor genomic codeletion of BRCA2 and RB1, show sensitivity to various PARPi or platinum drugs (Supplementary Fig. S5D, bottom). In contrast, we have observed that the 22RV1 cell line, which harbors a T3033Nfs*11 mutation in BRCA2, showed sensitivity to cisplatin and modest sensitivity to talazoparib, but not to other PARPi (Supplementary Fig. S5D, top). Taken together, these results indicate that co-loss of BRCA2-RB1 is a cell line–independent event and is frequently associated with castration resistance and leads to heightened sensitivity to PARPi.

3D organoid cultures of human cancers have shown extreme promise in cancer research (21, 51–53). Organoids can potentially be used as avatars for human cancer to study the molecular mechanisms of candidate genes and the effect of drugs. Earlier prostate organoids (MSK-PCa 1–7) were successfully developed from patients with CRPC. These organoids successfully retain the genetic characteristics of patients and grow in vitro as well as in immunodeficient mice (21). We tested the BRCA2-RB1 status by 3-color FISH in three mCRPC organoids that were originally isolated from metastatic sites from castration-resistant tumors (21). As a control, we also analyzed a benign prostate organoid by FISH. We observed that organoid MSK-PCa1 and MSK-PCa3 exhibited heterozygous co-deletion (∼100% of cells) of BRCA2 and RB1; however, MSK-PCa2 largely (94%) exhibited heterozygous deletion of RB1 only (Fig. 6A and B; Supplementary Fig. S6A; Supplementary Table S13). The copy number segment analysis of the prostate cancer organoids matched the FISH analysis, showing codeletion of BRCA2 and RB1 in MSK-PCa1 and MSK-PCa3, and deletion of RB1 only in MSK-PCa2. Heterozygous deletion of BRCA2 and RB1 is consistent with loss of their protein expression (Fig. 6C) as identified in our previous observation in TCGA prostate cancer cohort (Fig. 4C). We did observe upregulation of BRCA2 protein expression in the MSK-PCa2 organoid, which may be due to an extra copy of chromosome 13 (Supplementary Fig. S6A; Supplementary Table S13) rather than due to transcriptional activity of BRCA2. MSK-PCa1 and MSK-PCa3 also showed upregulation of mesenchymal markers N-cadherin and vimentin (the latter only in MSK-PCa1), indicating the EMT-like phenotype of these cells (Fig. 6D). However, MSK-PCa2 exhibited more epithelial morphology (Fig. 6D). We also observed higher SNAIL and PRRX1 mRNA expression in the MSK-PCa1 organoid (Supplementary Fig. S6C).

We observed growth reduction of MSK-PCa1 and MSK-PCa3 compared with the benign organoid when treated with the PARPi olaparib and talazoparib (Fig. 6E). However, PARPi did not have an inhibitory effect on the growth of the MSK-PCa2 organoid. Interestingly, none of the organoids harbored any other known mutation in DDR genes (Supplementary Fig. S6B), indicating that coheterozygous deletion of BRCA2 and RB1 is sufficient to sensitize cells to PARPi treatment–mediated growth inhibition.

We observed BRCA2-RB1 deletion (heterozygous and homozygous) in approximately 30% of all cancer types determined from TCGA pan-cancer cohort (without prostate cancer n = 10,820; Supplementary Fig. S6D, top; Supplementary Table S14). We observed that deletion of either BRCA2 or RB1 or codeletion is associated with shorter overall survival (P_trend < 0.0001; Supplementary Fig. S6D, bottom), indicating that loss of BRCA2 or RB1 alone may also play an important role in disease progression in the pan-cancer scenario.

Mutations in BRCA2, a DDR and cancer susceptibility gene, have been known to confer elevated risk of breast and ovarian cancer, and BRCA2 mutations are prevalent in men with advanced prostate cancer (1). In fact, pathologic variants in DDR pathway genes are prevalent in a substantial subset of men who develop mCRPC. Compared with other DDR pathway components, alterations of BRCA2 have been observed in a higher proportion of men with mCRPC and they are associated with a worse prognosis, especially when the mutations involve loss of gene function (2, 3). In the PROREPAIR-B mCRPC cohort (of 419 patients, 68 had germline DDR mutations, including 14 with BRCA2 mutations), BRCA2 germline mutations were reported to have a deleterious impact on mCRPC outcome (54). Furthermore, integrative whole genome and transcriptome analysis in 101 samples from patients with mCRPC identified deletion (homozygous and heterozygous) of BRCA2 in a significant number of men (55). Several other studies also measure DDR and BRCA2 alteration and confirm a high prevalence. Herein, we observed that BRCA2 and its chromosomal neighbor RB1 are frequently codeleted in primary prostate cancer and this codeletion is enriched in mCRPC. Our current study focuses on understanding the molecular consequences of co-loss of BRCA2 and RB1 in prostate cancer. We observed that CRISPR-mediated knockout of BRCA2 induces FOLH1 expression in LNCaP cells. Recently Paschelis and colleagues reported higher membranous FOLH1/PSMA expression in patients with mCRPC who harbor deleterious aberration of BRCA2 [P < 0.001; median H-score: 300 (165–300; ref. 56]. Our data indicate that CRISPR-mediated knockout of BRCA2 LNCaP cells may exhibit molecular features (upregulation of FOLH1) similar to those observed in patients with mCRPC who harbor aberration of BRCA2. Furthermore, using CRISPR and shRNA-based knockout/knockdown approach, we observed that co-loss of BRCA2 and RB1 in LNCaP cells induces an EMT-like invasive phenotype compared with loss of either gene alone. For the first time, we demonstrate that human prostate cancer cells exhibit a distinct phenotype upon co-loss of BRCA2-RB1 that may lead to aggressive disease.

Biallelic inactivation (germline mutation plus somatic loss of heterozygosity) of BRCA2 or other DDR genes is often associated with prostate cancer progression. However, the role of heterozygous somatic deletion of BRCA2 in prostate cancer is poorly understood. In our current study, we show for the first time that heterozygous loss of BRCA2 significantly reduces BRCA2 protein levels in primary prostate cancer (TCGA cohort). Similarly, in the same cohort, we observed that heterozygous deletion of RB1 also significantly attenuates RB1 mRNA expression. Heterozygous deletions are very difficult to analyze by copy number variation algorithms; therefore, we validated the heterozygous deletion of BRCA2 and RB1 in mCRPC-derived organoids using FISH. We observed depleted BRCA2 and RB1 protein in the organoids that harbor heterozygous codeletion of a BRCA2 and RB1 allele. Our data indicate that heterozygous loss leads to haploinsufficiency of BRCA2 and RB1 protein, which may result in “BRCA2-RB1ness” in prostate cancer. Very recently, Rodrigues and colleagues demonstrated heterogenicity of RB1 expression in advanced mCRPC (57). In the same study, the authors showed that hemizygous/heterozygous loss of one copy of RB1 in mCRPC was associated with absence of RB1 protein expression, further indicating the presence of haploinsufficiency of RB1 in mCRPC (57). Therefore, heterozygous loss of BRCA2-RB1 is likely associated with prostate cancer progression even in primary disease. However, this speculation requires further testing in a larger cohort to confirm the prognostic significance of heterozygous loss of BRCA2 and RB1 in prostate cancer.

Herein, we showed that PARP inhibition reduces growth of prostate cancer cells that harbor homozygous and, importantly, heterozygous co-loss of BRCA2 and RB1. PARPi have demonstrated promise in treating cancers with DDR deficiencies (20, 58). Our data showed that the castration-resistant prostate cancer cell line LNCaP-Abl and the mCRPC organoids MSK-PCa1 and MSK-PCa3 that harbor heterozygous codeletion of BRCA2-RB1 (on the basis of 3-color FISH), undergo significant growth inhibition when treated with olaparib or talazoparib. Very recently, Jonsson and colleagues showed that PARPi exhibit significant clinical benefit (P = 2.2 × 10^–5) to patients harboring biallelic or even heterozygous loss of BRCA (BRCA1 or 2) in BRCA-associated cancer including prostate cancer (59). Therefore, we hypothesized that patients with mCRPC and potentially high-risk localized disease who are identified as having tumors with codeletion (heterozygous/homozygous) of BRCA2-RB1 could significantly benefit from PARPi-based therapy.

Recent findings in genetically engineered mice showed that conditional elimination of BRCA2 in the mouse prostate failed to induce oncogenic transformation (60). BRCA2 is located on chromosome 13q in close proximity to the RB transcriptional corepressor 1 (RB1) gene, and loss of heterozygosity of the BRCA2 and RB1 locus is observed in approximately 30% of sporadic breast tumors (61). RB1 is a gatekeeper gene, whose inactivation is often demonstrated as an important rate-limiting step in tumor initiation (62). We found that BRCA2 is frequently codeleted with RB1 and more so in advanced than primary prostate cancer. Codeletion was associated with higher Gleason grade and metastases, while deletion of RB1 alone was not associated with disease progression. Interestingly, we observed that knockout/knockdown of RB1 partially attenuates BRCA2 expression in prostate cancer cells. Furthermore, we also observed that BRCA2 knockout LNCaP cells exhibit partial loss of RB1 expression, indicating a possible positive feed-forward loop between BRCA2 and RB1 in prostate cancer cells. Previous studies have showed that depletion of RB1 in osteosarcoma cells displayed spontaneous DNA damage evidence by increased γH2AX foci and elevated reactive oxygen species (63) that may lead to loss of BRCA2 expression. However, multiple molecular mechanisms may be involved in this feed-forward regulation of BRCA2 and RB1 and this could be a field of future study. BRCA2-RB1 codeletion was associated with increased genomic instability, suggesting that such tumors may be particularly aggressive. Although our data do not suggest that BRCA2-RB1–null prostate cancer cells are the cell of origin of prostate cancer, detection of this aberration early has importance in that it identifies an aggressive form of the disease.

For over two decades, deletion of chromosome 13q has been known to be a frequent event in prostate cancer (17, 64). Recently, Kluth and colleagues showed that the 13q deletions are heterozygous and associated with high Gleason grade (P < 0.0001) and early biochemical recurrence (P < 0.0001; ref. 18). We observed an interstitial deletion in the BRCA2-RB1 region of chromosome 13q evidenced by the loss of mRNA expression of the BRCA2-RB1 region genes compared with the genes outside this region. Although we were not able to determine the clinical significance of BRCA2-RB1 interstitial deletion in primary prostate cancer (unpublished observation), we do observe significant attenuation of the mRNA expression of these genes (in the BRCA2-RB1 region) in mCRPC cases. We showed that coelimination of BRCA2 and RB1 is sufficient to induce enzalutamide resistance in human prostate cancer cells. Recently, the antiandrogens enzalutamide, apalutamide, and darolutamide were approved by the FDA for nonmetastatic CRPC (65–67). Identification of the interstitial deletion of the BRCA2-RB1 region of chromosome 13q may be a biomarker of resistance to antiandrogen therapy.

Previous studies have shown that progression to invasive carcinoma requires loss of epithelial adhesion and polarity and acquisition of an invasive phenotype (68, 69). In some cases, tumor cells hijack a developmental program of gene expression, EMT, to gain invasive capacity (68–70). EMT is often associated with metastasis of various cancer types (70), but the role of EMT in prostate cancer progression is not well characterized. Furthermore, murine models have shown that conditional knockout of BRCA1 or BRCA2 is sufficient to induce an EMT-like phenotype in mouse mammary tumors (71, 72). Moreover, EMT is also often associated with CRPC (31, 36). Previous studies found that BRCA2 germline–mutated prostate cancers are highly invasive (73), and BRCA2 germline mutations are often associated with RB1 heterozygous deletion in prostate cancer (19). In transcriptome analysis of BRCA2-RB1–deleted cases in TCGA cohort, we identified significant enrichment of the hallmark EMT signature, which is consistent with our experimental observation in BRCA2-RB1–null LNCaP cells. To further understand the clinical significance of EMT in prostate cancer, we analyzed the 363 primary prostate cancer samples from the Swedish Watchful Waiting cohort (including 277 indolent and 86 lethal cases; 5-year follow-up) in Oncomine suite (33). We observed that enrichment of the hallmark EMT signature was associated with lethal disease. Our cell line–based observation and analysis in TCGA cohort indicate that codeletion of BRCA2 and RB1 may induce an EMT-like phenotype in primary tumors and leads to extravasation of tumor cells and metastatic progression to lethal disease.

We also showed that coelimination of BRCA2-RB1 induced SLUG (SNAI2), SNAIL, and PRRX1 expression. Although knockdown of all 3 genes reduces Matrigel invasion in BRCA2-RB1–null prostate cancer cells (Fig. 3J), loss of SLUG exhibits a more profound anti-invasive effect compared with loss of SNAIL or PRRX1. Previous studies have shown that SLUG is crucial for TGFβ-induced EMT in prostate cancer and SLUG-positive cells are enriched in advanced prostate cancer (74). Elevated expression of SLUG was also observed in CRPC tissue (46). SLUG had been demonstrated as an androgen-regulated transcription factor that facilitates castration resistance in prostate cancer (46), and here we showed higher SLUG expression in LNCaP-BRCA2-RB1 cells compared with control cells when exposed to androgen. BRCA2 negatively regulates SLUG expression and EMT in an experimental model of ovarian cancer, which is also consistent with our current observation (75). Therefore, we suggest SLUG might play an important role in BRCA2-RB1 loss–mediated EMT transformation of prostate cancer cells and may be a potential driver of aggressive disease. However, considering the complexity of the process, we believe that additional mechanisms may also be involved in dedifferentiation and EMT in BRCA2-RB1–null prostate cancer cells, a subject for future investigation.

Previous studies have shown that the fraction of the genome altered is a biomarker associated with prostate cancer progression (76). In our current study, we observed that co-loss of BRCA2 and RB1 is associated with a higher fraction of genome altered compared with RB1 loss alone. Loss of BRCA1 in a genetically engineered mouse model was sufficient to induce higher fraction of genome altered (77). Earlier studies also showed that BRCA2-mutated tumors have high copy number burden (78). Therefore, we postulate that codeletion of BRCA2 and RB1 may be an early event associated with, rather than a consequence of, a high fraction of genome altered.

To validate and allow for rapid identification of the codeletion of BRCA2 and RB1, we developed a 3-color FISH-based method. We observed that human prostate cancer cells that harbor single copy deletion of BRCA2 also exhibit heterozygous deletion of RB1. Most importantly, our FISH analysis successfully enabled us to determine the copy number of BRCA2, RB1, and chromosome 13q at the single-cell level and enabled us to demonstrate the heterogeneity of prostate cancer. In addition, our FISH data are consistent with the findings in various independent prostate cancer clinical cohorts that BRCA2 is almost uniformly codeleted (heterozygous) with RB1. We extended our FISH analysis to mCRPC-derived organoids and showed that our FISH method was able to detect codeletion of BRCA2 and RB1 in organoids, consistent with copy number segment analysis. We also observed that the organoids that harbor the codeletion (MSK-PCa1 and 3) were enzalutamide-resistant, whereas MSK-PCa2, which harbors RB1 loss alone, exhibits enzalutamide sensitivity (21). The data from the organoids are consistent with our cell line study and further suggest that patients with prostate cancer who harbor BRCA2-RB1 codeletion in a significant fraction of their primary tumor cells may not benefit from ADT. Our 3-color FISH method is rapid and consistent with genome sequencing data and therefore could be used for early screening of the BRCA2-RB1 codeletion to identify patients at high risk of this form of aggressive prostate cancer, providing an opportunity for early intervention and targeted treatment.

P.W. Kantoff is a paid consultant/advisory board member of Bavarian Nordic Immunotherapeutics, SEER Biosciences, DRGT, Progenity, and OncoCellMDX, reports ownership interest (including stocks, patents, etc.) in Context Therapeutics LLC, DRGT, Placon, and SEER Biosciences, and reports other remuneration from/relationships with Merck, Roche, Context Therapeutics LLC, Tarveda Therapeutics, and Adena Health. No potential conflicts of interest were disclosed by the other authors.

Conception and design: G. Chakraborty, P.W. Kantoff

Development of methodology: G. Chakraborty, Y.Z. Mazzu, Y. Chen, G.J. Nanjangud

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): G. Chakraborty, Y.Z. Mazzu, M.O. Atiq, K. Komura, L. Jehane, R. Hirani, Y. Yoshikawa, G.J. Nanjangud

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): G. Chakraborty, J. Armenia, Y.Z. Mazzu, S. Nandakumar, K. Komura, K. Chadalavada, W. Abida, G.J. Nanjangud, P.W. Kantoff

Writing, review, and/or revision of the manuscript: G. Chakraborty, J. Armenia, Y.Z. Mazzu, S. Nandakumar, K.H. Stopsack, M.O. Atiq, L. Jehane, R. Hirani, Y. Yoshikawa, N.A. Khan, Y. Chen, W. Abida, L.A. Mucci, G.J. Nanjangud, P.W. Kantoff

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): G. Chakraborty, S. Nandakumar, G.-S.M. Lee, P.W. Kantoff

Study supervision: G. Chakraborty, G.-S.M. Lee, P.W. Kantoff

We thank Ralph Garippa, Qiu Xiang, Hsiu Yu Liu, Yunsi Refermat, and staff of the MSK RNAi Core for CRISPR design and T7 assay, staff of the MSK Integrated Genomics Operation Core for RNA-sequencing services, Cindy Lee of the MSK Human Oncology and Pathogenesis Program for organoid culture, Hui Zhao of cBioPortal for gene ontology submission, Michael McGregor and Amy Plofker (MSK) for editing, and Rileen Sinha (Brigham & Women's Hospital/Broad Institute/Harvard Medical School, Boston, MA) for suggestions. This research is funded, in part, through NIH/NCI Cancer Center Support Grant P30 CA008748 to Memorial Sloan Kettering Cancer Center. G. Chakraborty and K.H. Stopsack were supported by a Prostate Cancer Foundation Young Investigator Award. P.W. Kantoff. and L.A. Mucci were supported by a Prostate Cancer Foundation Challenge Award and by NCI grant 5P01CA228696-02.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.