Abstract
Carriers of germline BRCA2 pathogenic sequence variants have elevated aggressive prostate cancer risk and are candidates for precision oncology treatments. We examined whether BRCA2-deficient (BRCA2d) prostate tumors have distinct genomic alterations compared with BRCA2-intact (BRCA2i) tumors. Among 2536 primary and 899 metastatic prostate tumors from the ICGC, GENIE, and TCGA databases, we identified 138 primary and 85 metastatic BRCA2d tumors. Total tumor mutation burden (TMB) was higher among primary BRCA2d tumors, although pathogenic TMB did not differ by tumor BRCA2 status. Pathogenic and total single nucleotide variant (SNV) frequencies at KMT2D were higher in BRCA2d primary tumors, as was the total SNV frequency at KMT2D in BRCA2d metastatic tumors. Homozygous deletions at NEK3, RB1, and APC were enriched in BRCA2d primary tumors, and RB1 deletions in metastatic BRCA2d tumors as well. TMPRSS2-ETV1 fusions were more common in BRCA2d tumors. These results identify somatic alterations that hallmark etiological and prognostic differences between BRCA2d and BRCA2i prostate tumors.
Similar content being viewed by others
Introduction
Male carriers of BRCA2 germline pathogenic sequence variants (PSV) experience 2.6-fold higher lifetime risk of prostate cancer and a 7.3–8.6-fold higher risk of developing early-onset (<65 years) prostate cancer1,2,3. Germline BRCA2 PSV are associated with higher tumor stage, Gleason grade, and prostate-specific antigen (PSA) levels at diagnosis4,5,6. Intraductal carcinoma of the prostate (IDCP) occurs more commonly in tumors that harbor a germline BRCA2 mutation than in sporadic prostate cancers, and likewise confers a higher risk of mortality7. Prostate cancers with a germline BRCA2 PSV are associated with higher rates of lymph node involvement, metastases, and prostate cancer-specific death for both primary and metastatic cancers4,8,9. Germline or somatic BRCA2 loss occurs in ~13% of metastatic prostate cancers, compared with 3% in primary tumors10,11. The presence of a germline BRCA2 PSV also directs therapeutic management with PARP inhibitors, although PARP inhibitors are not yet uniformly available globally12.
Prostate tumors that are BRCA2-deficient (BRCA2d) have aggressive genomic profiles that may contribute to the worse outcomes observed in this subset. Compared with BRCA2-intact (BRCA2i) tumors that do not contain a PSV, germline BRCA2d prostate tumors have an elevated proportion of the genome altered7. The mean count of copy number alterations in prostate cancers was reported to be 3-fold higher among germline carriers of BRCA2 PSV relative to non-carriers, with gains considerably more common in the region encompassing c-MYC13. Amplifications of the Wnt/β-catenin pathway modulators MED12 and MED12L were also more common among germline BRCA2d tumors7. Additionally, germline BRCA2d prostate tumors have been shown to experience global hypomethylation relative to BRCA2i tumors7.
We hypothesize that BRCA2d tumors represent a unique phenotype in prostate cancer. Identification of tumor genomic aberrations in BRCA2d prostate tumors may provide insight into the mechanisms of BRCA2-associated prostate carcinogenesis and progression, which could have downstream implications for prevention or therapeutics. We assembled multi-omic data including single nucleotide variants (SNVs), copy number alterations (CNAs), and structural variants (SVs) from multiple public databases to create the largest and most comprehensive dataset of BRCA2d prostate tumors to date and compared these with BRCA2i prostate tumors.
Results
BRCA2d prostate tumors
A total of 2536 primary and 899 metastatic prostatic adenocarcinomas from the International Cancer Genome Consortium (ICGC)14, the American Association for Cancer Research Project Genomics Evidence Neoplasia Information Exchange (GENIE)15, and The Cancer Genome Atlas (TCGA) databases16 were identified with available SNV, CNA, or SV) data (Fig. 1). One hundred thirty-eight primary tumors (5.4%) harbored a somatic PSV leading to BRCA2d, while 2398 tumors (94.6%) had no alterations or a non-pathogenic alteration at BRCA2 and were denoted BRCA2i. Fifty-three primary BRCA2d tumors had pathogenic SNVs, 74 had CNAs, and seven had SVs affecting BRCA2, while the remaining four tumors had multiple PSV affecting BRCA2 (Supplementary Table 1). Patients with losses of heterozygosity (LOH; n = 184) or non-pathogenic SNVs (n = 68) at BRCA2 were considered BRCA2i.
Eighty-five metastatic tumors (9.5%) were adjudicated to be BRCA2d, while 814 (90.5%) were BRCA2i. Among the metastatic BRCA2d tumors, 36 harbored pathogenic SNVs, 39 had CNAs, and five had SVs affecting BRCA2, and the remaining five tumors had multiple PSVs at BRCA2 (Supplementary Table 1). There were 40 BRCA2i patients with LOH at BRCA2 and 22 patients with non-pathogenic SNVs. Clinical and pathological characteristics of primary and metastatic tumors by tumor BRCA2 status are shown in Supplementary Table 3.
Single nucleotide variant (SNV) analyses
Among 531 patients with primary tumors from the ICGC database, the median SNV TMB was 0.942 per mb (range: 0.015–6.111) (Fig. 2a). The median SNV TMB in BRCA2d tumors (median = 1.103, range: 0.338–5.152) was significantly higher than among BRCA2i tumors (median = 0.925, range: 0.015–6.111; p = 0.011 from Wilcoxon rank-sum test) (Fig. 2b). The median pathogenic SNV TMB was 0.004/mb (range: 0.000–0.021) (Fig. 2c). However, in contrast to the total SNV TMB, there was no difference in the pathogenic SNV TMB between BRCA2d tumors (median = 0.004, range: 0.000–0.013) and BRCA2i tumors (median = 0.004, range: 0.000–0.021; p = 0.20) (Fig. 2d).
Panels display a the distribution of total TMB in all tumors, b the distribution of total TMB by BRCA2 status, c the distribution of pathogenic TMB in all tumors, and d the distribution of pathogenic TMB by BRCA2 status. TMB is estimated from samples with whole genome sequencing from the ICGC dataset (n = 531). In panels b and d, the lower, middle, and upper lines correspond to the lower quartile, median, and upper quartile of TMB, respectively. P-values are from a two-sided Wilcoxon rank-sum test. BRCA2d BRCA2-deficient; BRCA2i BRCA2-intact.
The most common pathogenic SNVs among BRCA2d and BRCA2i primary tumors from the combined ICGC and GENIE data are shown in Fig. 3a, b. KMT2D had the highest pathogenic SNV frequency among primary BRCA2d tumors (12.5%) with a significantly higher frequency among BRCA2d tumors than BRCA2i tumors (4.6%, FDR-adjusted p [padj] = 0.0004). TP53 had the second highest pathogenic SNV frequency among BRCA2d tumors, but its frequency did not differ between BRCA2d (11.6%) and BRCA2i (15.5%) primary tumors (padj = 0.27) (Fig. 3a, b, Supplementary Table 4). Pathogenic SNVs at PTEN (7.2%), SPOP (6.9%), KMT2C (5.8%), CSMD1 (5.7%), SYNE1 (5.7%), CSMD3 (5.7%), and FOXA1 (5.6%) were also present in ≥5% of primary BRCA2d tumors; however, these genes also did not differ in mutation frequency by tumor BRCA2 status after multiple testing correction. When evaluating total SNV frequency, SNVs at 11 genes were significantly more frequent among primary BRCA2d tumors, while none occurred at higher frequency among BRCA2i tumors (Supplementary Table 5) (padj < 0.05). Of these genes, CSMD3 was the most commonly altered among BRCA2d tumors (67.1%), followed by LRP1B (66.2%), and CSMD1 (64.3%). The total SNV frequency for KMT2D was also significantly higher for BRCA2d (13.2%) than for BRCA2i tumors (5.8%, padj = 0.003). Pathogenic SNVs in oncogenic pathways were not differentially enriched between BRCA2d and BRCA2i tumors. (Supplementary Table 6). Finally, the mean total and pathogenic transitions and transversions per patient did not differ by BRCA2 status (Supplementary Fig. 1).
Among metastatic tumors from GENIE, TP53 had the highest pathogenic SNV frequency in both BRCA2d and BRCA2i tumors (25.9% vs. 35.6%, padj = 0.07) (Fig. 3c, d, Supplementary Table 7). FOXA1 (18.4%), KMT2D (15.9%), APC (15.3%), KMT2C (13.7%), and ZFHX3 (12.5%) had pathogenic SNVs in ≥10% of metastatic BRCA2d tumors. No gene differed in pathogenic SNV frequency by BRCA2 status following FDR correction. The pathogenic SNV frequency at SPOP was 7.4% for BRCA2d tumors and 10.1% for BRCA2i tumors (padj = 0.45). When evaluating total SNV frequency, SNVs were enriched at KMT2D (17.1% vs. 6.9%), RAD51B (8.0% vs. 0.7%), BRIP1 (6.1% vs. 1.0%), BRCA1 (5.9% vs. 0.9%), and RAD50 (3.8% vs. 0.1%) in metastatic BRCA2d tumors relative to metastatic BRCA2i tumors (all padj < 0.05) (Supplementary Table 8). Other genes with an SNV frequency ≥ 10% among metastatic BRCA2d tumors include TP53 (28.2%), FOXA1 (18.4%), APC (16.5%), KMT2C (16.4%), ZFHX3 (14.1%) and GRIN2A (11.1%), but SNV frequency did not significantly differ by BRCA2 status for these genes.
Pathogenic SNV frequency did not significantly differ after multiple testing correction for any single gene between primary and metastatic BRCA2d tumors in GENIE (Supplementary Table 9). Pathogenic SNVs at FOXA1 were nominally less frequent among primary BRCA2d tumors (5.5%) than metastatic BRCA2d tumors (19.4%; padj = 0.03), as were pathogenic SNVs at AR (1.5% vs.10.3%; padj = 0.04).
We evaluated Catalog of Somatic Mutations in Cancer (COSMIC) signature similarity in BRCA2d and BRCA2i primary tumors (Supplementary Table 10)17. The only unique signature identified in BRCA2d tumors was defective homologous recombination DNA damage repair (ID6). In both BRCA2d and BRCA2i, COSMIC signatures were identified related to spontaneous deamination of 5-methylcytosine (clock-like signature; SBS1) and slippage during DNA replication of the replicated DNA strand (ID1 in both BRCA2d and BRCA2i; ID2 in BRCA2i only). We infer these to be signatures associated with prostate cancer, and not specific to BRCA2d prostate tumors. Finally, we identified a signature similar to defective DNA mismatch repair among BRCA2i tumors only.
Copy number alteration (CNA) analyses
In primary tumors from the GENIE and TCGA datasets, homozygous deletions were significantly more frequent at three genes in BRCA2d tumors relative to BRCA2i tumors after multiple testing correction: NEK3 (20.0% vs 3.4%), RB1 (11.6% vs 2.1%), and APC (3.8% vs 0.4%) (Fig. 4a, Supplementary Table 11). Amplifications at MYC (5.8%), NSMCE2 (4.3%), NBN (3.6%), and AR (2.9%) were the most common high-level amplifications among BRCA2d primary tumors, but no gene differed in high-level amplification frequency by BRCA2 status. Frequencies of low-level gains and LOH in primary tumors by BRCA2 status are presented in Supplementary Table 10.
In metastatic tumors, the frequency of homozygous deletions at RB1 was significantly higher among BRCA2d than BRCA2i tumors (17.7% versus 5.2%, padj = 0.0001) (Fig. 4b, Supplementary Table 12). No other single gene difference in homozygous deletion frequency by BRCA2 status after multiple testing correction. The frequency of PTEN homozygous deletions was non-significantly lower among BRCA2d tumors (14.1%) than BRCA2i tumors (20.8%). Homozygous deletions at CHD1 also did not differ by BRCA2 status (5.7% BRCA2d vs 3.1% BRCA2i, padj = 0.28). As was observed in primary tumors, no gene differed in high-level amplification frequency by BRCA2 status in metastatic tumors. Among genes with high-level amplification frequencies ≥ 10% among BRCA2d tumors, amplifications at MYC, TCEB1, RECQL4, AGO2, PRDM14, and PREX2 were non-significantly more common among BRCA2d tumors, while AR was non-significantly more commonly amplified among BRCA2i tumors. Frequencies of low-level gains and LOH in metastatic tumors by BRCA2 status are presented in Supplementary Table 12.
The frequency of homozygous deletions at RB1 was nominally higher among metastatic BRCA2d tumors (16.1%) than primary BRCA2d tumors (4.4%, padj = 0.03) in the GENIE data, as was the frequency of homozygous deletions at PTEN (12.4% vs. 1.5%, padj = 0.01) (Supplementary Table 13). However, these frequencies were not significantly different following FDR correction. Likewise, no single gene differed in high-level amplification frequency between primary and metastatic BRCA2d tumors, although amplifications at AR were nominally more frequent in metastatic tumors (15.4% vs. 4.4%, padj = 0.03).
Structural variant (SV) analysis
The frequencies of TMPRSS2- and ETS-related SVs were estimated in the TCGA and GENIE data. In addition to 848 SVs that affected TMPRSS2 and/or ETS family genes, 33 SVs affected a gene within 2Mbp of TMPRSS2 or an ETS family gene and were assumed to impair function of that gene (Supplementary Table 14). The frequency of re-annotated SVs that were inferred to affect TMPRSS2 or an ETS family gene did not differ by tumor BRCA2 status. The frequency of TMPRSS2-ETS fusions was similar between primary BRCA2d (32.8%) and BRCA2i (31.1%) tumors (p = 0.78) (Table 1), as was the frequency of TMPRSS2-ERG fusions (28.4% vs 29.9%, p-0.62). Similarly, no differences in the frequency of TMPRSS2-ETV1, TMPRSS2-intragenic, or other SVs involving ETS genes were observed between BRCA2d and BRCA2i primary tumors.
In metastatic tumors, TMPRSS2-ETV1 fusions were more common among BRCA2d (3.0%) than BRCA2i tumors (0.1%; p = 0.017). In metastatic tumors, the prevalence of TMPRSS2-ETS fusions did not differ by tumor BRCA2 status, nor did the prevalence of TMPRSS2-ERG, TMPRSS2-intragenic or other SVs involving ETS family genes.
Discussion
There is mounting evidence that the molecular signatures of prostate tumors in men with BRCA2d prostate tumors either from germline or somatic mutation exhibit a different molecular signature relative to BRCA2i. Our results confirm some prior reports of mutational patterns in BRCA2d 18,19, but also identify new mutational patterns in BRCA2d prostate tumors in part because of the increased sample size of our dataset.
We report that the TMB of SNVs in BRCA2d tumors is significantly higher than in BRCA2i tumors, in line with prior findings that the somatic mutation rate is 2.9-fold higher in high-grade BRCA2d prostate tumors20. However, we found no significant difference in pathogenic SNV TMB between BRCA2d and BRCA2i tumors. We also observed that pathogenic and total SNV frequencies at KMT2D were higher in BRCA2d primary tumors than BRCA2i tumors. KMT2D encodes a histone methyltransferase that methylates the Lys-4 position of histone H3, and is a member of the ASCOM protein complex, which has been shown to be a transcriptional regulator of the beta-globin and estrogen receptor genes. KMT2D is associated with activation of PKN1, which stimulates transcription of the AR-regulated kallikrein genes KLK2 and KLK321. KMT2D has been reported to be highly mutated in prostate tumors, and high KMT2D transcription is associated with poor prostate cancer prognosis. KMT2D epigenetically activates PI3K/AKT pathway and epithelial-mesenchymal transition by targeting LIFR and KLF4 and thus serves as a putative therapeutic target for prostate cancer22. Our observation that KMT2D is more commonly mutated in BRCA2d suggests that it plays a role in BRCA2-associated prostate carcinogenesis and may identify therapeutic targets for BRCA2d prostate cancer.
BRCA2 plays a critical role in the regulation of homologous recombination (HR) repair of double-stranded DNA breaks23, and protein partners involved in this process have been described. We identified elevated SNV frequencies in known BRCA2-interacting pathway genes RAD51B, BRIP1, BRCA1, and RAD50 in metastatic BRCA2d tumors relative to BRCA2i tumors. These results confirm the role of these loci in BRCA2-associated pathways involved in prostate carcinogenesis and suggest that they are also involved in BRCA2d prostate tumorigenesis and progression. This observation is consistent with reports of mutation signatures associated with HR defects in BRCA2 prostate tumors that do not have BRCA2 germline PSV24. We did not observe differences in SPOP mutation or CHD1 loss frequency by tumor BRCA2 status, although these genes have been implicated in impaired HR repair25,26,27.
We identified higher homozygous deletion frequencies for NEK3, RB1, and APC in BRCA2d primary tumors, and for RB1 deletions in metastatic BRCA2d tumors. Chromosome 13q14, which contains RB1 (chromosome 13q14.2) and NEK3 (chromosome 13q14.3), is among the most commonly altered loci in prostate tumors, with RB1 and BRCA2 being commonly co-deleted in prostate tumors28,29. In human prostate tumor cell line models, concomitant BRCA2 and RB1 loss induces an epithelial-to-mesenchymal transition, resulting in a more aggressive tumor phenotype30. Alterations at this locus are associated with poor prognosis and unfavorable tumor characteristics31,32. Our observations suggest that not only is this locus important for prostate carcinogenesis and for prediction of poor prognosis, it may also in part explain the difference in prostate tumor aggressiveness and poor prognosis seen in BRCA2d prostate cancer.
CNA in prostate tumors among carriers of BRCA2 PSV have been reported to be 3-fold higher than in tumors without BRCA2 PSV, with copy number gains being more common in the region encompassing c-MYC13. While c-MYC amplifications were among the most common events observed in our dataset, we did not identify a significant excess in BRCA2d compared with BRCA2i. This observation may indicate different effects of BRCA2 PSV compared with BRCA2d tumors, which were not exclusively associated with germline PSV.
APC is a tumor suppressor gene that activates elements of the Wnt/β-catenin signaling pathway, and is mutated in 3-10% of prostate tumors33. APC is involved in phosphorylation of β-catenin and subsequent degradation. Knowledge of this pathway has led to the development of Wnt signaling inhibitors. APC promoter methylation exists at high levels in prostate tumors and is a poor prognosis indicator in prostate cancer34. While we did not have data to address APC promoter methylation, we report 3.8% APC deletions in primary BRCA2d compared with 0.4% in BRCA2i. These data are consistent with elevated levels of APC inactivation in BRCA2d. The lower proportion of APC mutations reported here in BRCA2i tumors compared with other reports in the literature is due to our association of APC CNA only, and not other mutational types or promoter methylation. This result builds upon the findings of Taylor et al. who observed that amplifications of the Wnt/β-catenin pathway modulators MED12 and MED12L were also more common among BRCA2 PSV carriers and that BRCA2d prostate cancers have been shown to experience global hypomethylation relative to sporadic cancers7.
Fusion proteins involving TMPRSS2, particularly those involving ERG and other ETS family members, are common SV alterations in prostate tumors. TMPRSS2-ETS fusions, and TMPRSS2-ERG fusions specifically, were common in both BRCA2d and BRCA2i, but we did not identify a difference in the frequency of these events by BRCA2 status in either primary or metastatic tumors, as has been reported for germline BRCA2d prostate cancer7,35,36. We observed a subset of SVs which affected genes in close proximity to TMPRSS2 or ETS family genes and were assumed to impact the function of the neighboring gene. Given the low frequency of this occurrence, inclusion or exclusion of these SVs inferred to be TMPRSS2-ETS fusions did not substantially affect estimates of SV frequencies. However, misclassification of functional TMPRSS2 and/or ETS family genes may exist if consideration of nearby mutated genes are not considered. This observation is dependent on the interval considered. We were also not able to assess the effects of these SV events on gene function with the data available. We did observe a significantly elevated frequency of TMPRSS2-ETV1 fusions in metastatic BRCA2d tumors compared to BRCA2i. ETV1 is a target of the androgen receptor (AR). ETV1 and AR, interact in prostate tissue to regulate cell invasion37. Decreased ETV1 expression disrupts the ability of both androgen-dependent and androgen-independent prostate cell invasion, independent of TMPRSS2 fusion.
Our analysis represents the largest to date of primary and metastatic BRCA2d prostate tumors and examined signatures across multiple classes of somatic mutation. While we report biologically plausible associations between somatic mutations and BRCA2d and BRCA2i prostate tumors, our analysis is limited in a number of ways. First, while it is well known that germline BRCA2 PSV are associated with more aggressive disease, we were not able to evaluate how somatic mutational events were correlated with clinical traits. We used a surrogate comparison of primary vs. metastatic tumors to compare mutations in these two tumor groups, but this is not an adequate surrogate for the severity analyses that may be required to guide inferences about prognosis or management of disease. We applied multiple tools to ascertain BRCA2-deficiency38,39,40; however, the potential for misclassification of BRCA2d or BRCA2i tumors remains and use of alternative tools may have led to different designations of tumor BRCA2-deficiency41. Such misclassification would likely attenuate differences in alteration patterns between BRCA2d or BRCA2i tumors. Our analysis maximized the sample size available by integrating data across multiple publicly available sources, including TCGA, ICGC, and GENIE. These sources use different platforms and approaches for variant calling. As a result, misclassification of alterations likely differs across the sources, but we do not expect that the extent of misclassification differed for BRCA2d and BRCA2i tumors. In our efforts to account for potential misclassification, we noted that the annotations and platforms used (e.g., panel vs. whole exome/genome sequencing) could in some cases not be harmonized. Therefore, we harmonized only those data that could be combined in any single analysis (Fig. 1).
Our results provide evidence that somatic mutational patterns in prostate tumors may in part explain why BRCA2d tumors have more aggressive characteristics than BRCA2i tumors and focus attention on novel molecular events and pathways that may be used to understand the unique etiology of BRCA2d tumors. We identify Wnt/β-catenin, PI3K, and homologous double-stranded break repair pathways as hallmarks of BRCA2d prostate tumors. These patterns suggest great potential for molecularly targeted screening, monitoring, therapeutic development, and clinical management of these tumors.
Methods
Study sample and data processing
De-identified mutation files for SNVs, CNAs, and SVs and clinical data from the International Cancer Genome Consortium (ICGC) PRAD-CA, PRAD-CN, PRAD-FR, and PRAD-UK projects were downloaded from the ICGC Data Portal (https://dcc.icgc.org/releases/release_28)14. The American Association for Cancer Research (AACR) Project Genomics Evidence Neoplasia Information Exchange (GENIE) de-identified SNV, CNA, and SV alteration and clinical data for prostate adenocarcinomas were retrieved from Sage Bionetworks Synapse platform—release 8 (https://www.synapse.org)15. De-identified SNV, CNA, SV, and clinical data for prostate adenocarcinomas in the TCGA PanCancer Atlas were retrieved from cBioportal16. The ICGC and GENIE data were filtered to retain only primary and metastatic samples. The ICGC SNVs were processed using maftools to obtain a common data format as SNVs from GENIE and TCGA42. GENIE CNAs were obtained in CNA formats that indicated a gene was disrupted in a specific tumor sample. GENIE data were further filtered to retain only samples with available SNV, CNA, and SV profiling unless the sample had an alteration affecting BRCA2. The eight retained panels for BRCA2i samples were: COLU-CCCP-V1, DFCI-ONCOPANEL-1, MSK-IMPACT341, MSK-IMPACT410, MSK-IMPACT468, VICC-01-T5A, VICC-01-T7, VICC-01-DX1). The panels included for BRCA2d samples include DFCI-ONCOPANEL-2, DFCI-ONCOPANEL-3, DFCI-ONCOPANEL-3.1, DUKE-F1-T7, GRCC-MOSC4, GRCC-MOSC3, UHN-555-PROSTATE-V1, UHN-555-V1, and YALE-OCP-V3, in addition to the eight used for BRCA2i samples. This study is compliant with all relevant ethical regulations and was deemed exempt by the institutional review board of Dana-Farber Cancer Institute (Protocol #19-236). The GENIE, ICGC, and TCGA consortia were responsible for obtaining patient informed consent or waivers of consent.
All samples from ICGC, TCGA, and GENIE were systematically reviewed by pathologists to confirm the histopathologic diagnosis and to ensure tumor cellularity met required thresholds defined by dataset and contributing institution. Data regarding patient receipt of neoadjuvant therapy prior to tumor resection were not available for primary tumors in GENIE and ICGC, while 99.6% of patients in TCGA, which contains primary tumors only, did not receive neoadjuvant therapy.
The analytic sample set is summarized in a flow diagram in Fig. 1. A total of 4298 patients with prostate adenocarcinomas were identified across ICGC (n = 703), GENIE (n = 3101), and TCGA (n = 494). Samples with only blood and non-tumor tumor tissue data were excluded from the analyses, as were samples for two female patients. GENIE samples for which BRCA2 alterations were not profiled were excluded from analysis. If multiple primary or metastatic samples were available for a given patient, a single sample was selected using the following hierarchy: (1) the sample where the BRCA2 PSV was detected was selected, (2) the sample with maximal available genomic data types was selected (e.g., a sample with SNV and CNA profiling was prioritized over a sample with SNV profiling only), (3) a sample was randomly selected if neither of the first two criteria was fulfilled. Analyses of SNVs were conducted in a harmonized merged ICGC, GENIE, and TCGA dataset. ICGC CNA and SV could not be adequately harmonized with the corresponding GENIE and TCGA data. Thus, CNA and SV analyses were limited to the GENIE and TCGA data. Sample sizes of the final analytic population by dataset and genomic data type for primary and metastatic tumors are shown in Fig. 1.
Identification of BRCA2-deficient tumors
BRCA2d status was determined by assessing the presence of SNV, CNA, and SV involving BRCA2 in tumors. Deep deletions or amplifications of BRCA2 were assumed to result in impaired BRCA2 function, while tumors with LOH or low-level gains were considered BRCA2i. Likewise, patients with SV affecting BRCA2 were classified as BRCA2d. The potential impact of somatic SNV on BRCA2 function was evaluated by variant type and consequence using OncoKB, SnpEff, and ClinVar38,39,40. GENIE data were restricted to panels that profiled SNVs, CNAs, and SVs. Accordingly, BRCA2i samples in GENIE and TCGA were known to not harbor any SNV, CNA, or SV that affected BRCA2. In order to preserve an adequate sample size in ICGC, no analogous restrictions regarding profiling were required. BRCA2i samples in ICGC did not have any SNV, CNA, or SV profiling of BRCA2. A summary of patients with BRCA2 PSV and PSVs is provided in Supplementary Table 1.
Single nucleotide variant (SNV) analyses
Sixty-seven candidate genes were identified for inclusion in somatic mutation analyses from the literature10,11,43,44. These included 18 hereditary prostate cancer genes, 21 genes that are known interactors with BRCA2, and 28 genes that are recurrently mutated in prostate cancer (Supplementary Table 2). The a priori identified candidate genes were supplemented with genes that had among the top 20 highest pathogenic SNV frequencies in the ICGC, GENIE, and TCGA data for primary or metastatic samples. SNV frequencies were defined as the proportion of individuals with profiling of a particular gene who harbored a variant affecting the gene of interest. The sample size for frequency calculations varied across genes, as not all candidate genes were profiled in the GENIE panels. To ensure sufficient power, only genes profiled in at least 40% of tumors were carried forward for analysis. 76 total candidate genes were studied in primary tumors and 51 total candidate genes in metastatic tumors. SNV frequencies were calculated including all mutations and for likely pathogenic mutations only based on Sequence Ontology and the resulting predicted impact by Ensembl (http://uswest.ensembl.org/info/genome/variation/prediction/predicted_data.html). SNV frequencies were visualized using modified Oncoplots, accounting for the variable gene profiling coverage across individuals42. Differences in candidate gene mutation frequency by BRCA2 alteration status and between primary and metastatic BRCA2d tumors in the GENIE data were evaluated.
Enrichment of oncogenic pathways was evaluated using maftools in primary tumors from TCGA and ICGC42,45. In addition to the eight pathways pre-selected in maftools, a chromatin remodeling pathway was included and consisted of: SWI1, SWI2, SNF2, SWI3, SWI5, SWI6, HDAC1, HDAC2, RbAp46, RbAp48, MTA1, MTA2/MTA3, MBD3, MBD2, CHD3, CHD4, INO80, and SWR1. The proportion of samples harboring a pathogenic SNV within the pathway of interest was compared by tumor BRCA2 status.
Tumor mutational burden (TMB) was calculated using SNV data for primary tumors for the ICGC data with whole-genome sequencing (WGS) profiling. Data from TCGA and GENIE, which used whole-exome sequencing (WES) and targeted panels, respectively, were not included to ensure comparability of TMB measures across sites. TMB was defined as the total number of somatic mutations present in a tumor sample per megabase (Mb)46. Total TMB captured all SNVs regardless of gene or variant type. Pathogenic TMB was limited to variants in known genes meeting pathogenicity criteria (http://uswest.ensembl.org/info/genome/variation/prediction/predicted_data.html). The capture size for the WGS datasets was set to 3000 Mbp. Differences in pathogenic TMB and total TMB by tumor BRCA2 status were assessed.
The number of total and pathogenic transitions and transversions were calculated for patients with primary tumors in ICGC with WGS SNV profiling. Differences in the number of total and pathogenic transitions and transversions by tumor BRCA2 status were assessed.
Enrichment for Catalog of Somatic Mutations in Cancer (COSMIC) single base substitution (SBS), doublet base substitution (DBS), and small insertion and deletion (ID) mutational signatures was performed for primary BRCA2d and BRCA2i tumors using SigMiner in the ICGC and TCGA data using the default parameters17,47. SNV data were converted to maf format using maftools and a matrix was constructed for non-negative matrix factorization decomposition using the sig_tally function. The optimal number of signatures was automatically extracted using the sig_auto_extract function. Signatures with cosine similarity >0.9 and known etiologies were reported for BRCA2d and BRCA2i tumors.
Copy number alteration analysis
CNA analyses were conducted using data from GENIE (primary and metastatic) and TCGA (primary). Candidate genes for CNA analyses were identified in a similar manner as for the SNV analyses. 67 genes identified in the literature were supplemented with genes among the top 20 most commonly copy number-altered in primary or metastatic tumors (Supplementary Table 2). Only genes profiled in at least 40% of individuals were carried forward for analysis, resulting in 75 candidate genes for primary tumors and 57 candidate genes for metastatic tumors. CNA frequencies were defined as the proportion of individuals with profiling of a particular gene who harbored a CNA affecting the gene of interest. Differences in candidate gene CNA frequency by BRCA2 alteration status were evaluated using Fisher’s exact test separately for primary and metastatic samples. The frequency of pathogenic CNAs was also compared between primary and metastatic BRCA2d tumors in the GENIE data.
TMPRSS2 and ETS structural variant (SV) analyses
Structural variant (SV) analyses were conducted in GENIE and TCGA samples (Fig. 1) to investigate the association between BRCA2 deficiency and the occurrence of TMPRSS2- and ETS-related SVs. Alteration frequencies were calculated for harboring any SV as well as for SVs affecting TMPRSS2, ERG, or ETS family genes, accounting for gene coverage across the GENIE panels. Samples with SV profiling for at least one ETS family gene were included in estimating frequencies of ETS family SVs. ETS genes considered included: ELF1, ELF2, ELF3, ELF4, ELF5, ELK1, ELK3, ELK4, ERF, ERG, ESE3, ETS1, ETS2, ETV1, ETV2, ETV3, ETV4, ETV5, ETV6, ETV7, GABPA, FLI1, FEV, SPDEF, SPI1, SPIB, and SPIC. To account for potential misannotation of SVs, SVs involving neighboring genes within 2mbp of TMPRSS2 or an ETS family gene were classified as affecting TMPRSS2 or the ETS gene, as appropriate. SVs of interest were classified as TMPRSS2-ETS or ETS-other. Differences in alteration frequencies for these two SV categories by BRCA2 alteration status as well as differences in the frequency of TMPRSS2-ERG, TMPRSS2-ETV1, and TMPRSS2-intragenic SVs were tested.
Hypothesis testing
Distributions of clinical covariates, SNVs, CNAs, and SVs by BRCA2 alteration status were compared using t-tests or Wilcoxon rank-sums tests for continuous variables and Fisher’s exact test for categorical variables. The Benjamini–Hochberg procedure was used to control the false discovery rate (FDR ≤ 5%)48.
Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this article.
Data availability
The data that support the findings of this study are available from the AACR Project GENIE (https://www.synapse.org/#!Synapse:syn7222066/wiki/405659), the International Cancer Genome Consortium (https://dcc.icgc.org/releases/release_28), and The Cancer Genome Atlas (https://www.cbioportal.org/study/summary?id=prad_tcga_pan_can_atlas_2018).
Code availability
Data analysis was conducted using R (version 4.0.5) and Python (version 3.8.5). Maftools (version 2.8.05), sigminer (version 2.1.1), and custom scripts were used to complete the analyses. Code is available upon request.
References
Kote-Jarai, Z. et al. BRCA2 is a moderate penetrance gene contributing to young-onset prostate cancer: implications for genetic testing in prostate cancer patients. Br. J. Cancer 105, 1230–1234 (2011).
Consortium, T. B. C. L. Cancer risks in BRCA2 mutation carriers. JNCI: J. Natl Cancer Inst. 91, 1310–1316 (1999).
Oh, M. et al. The association of BRCA1 and BRCA2 mutations with prostate cancer risk, frequency, and mortality: a meta-analysis. Prostate 79, 880–895 (2019).
Castro, E. et al. Germline BRCA mutations are associated with higher risk of nodal involvement, distant metastasis, and poor survival outcomes in prostate cancer. J. Clin. Oncol. 31, 1748–1757 (2013).
Page, E. C. et al. Interim results from the IMPACT study: evidence for prostate-specific antigen screening in BRCA2 mutation carriers. Eur. Urol. 76, 831–842 (2019).
Mitra, A. et al. Prostate cancer in male BRCA1 and BRCA2 mutation carriers has a more aggressive phenotype. Br. J. Cancer 98, 502–507 (2008).
Taylor, R. A. et al. Germline BRCA2 mutations drive prostate cancers with distinct evolutionary trajectories. Nat. Commun. 8, 13671–13671 (2017).
Thorne, H. et al. Decreased prostate cancer-specific survival of men with BRCA2 mutations from multiple breast cancer families. Cancer Prev. Res (Philos.) 4, 1002–1010 (2011).
Castro, E. et al. PROREPAIR-B: a prospective cohort study of the impact of germline DNA repair mutations on the outcomes of patients with metastatic castration-resistant prostate cancer. J. Clin. Oncol. 37, 490–503 (2019).
Robinson, D. et al. Integrative clinical genomics of advanced prostate cancer. Cell 161, 1215–1228 (2015).
Cancer Genome Atlas Research Network. The molecular taxonomy of primary prostate cancer. Cell 163, 1011–1025 (2015).
Sandhu, S. K. et al. Poly (ADP-ribose) polymerase (PARP) inhibitors for the treatment of advanced germline BRCA2 mutant prostate cancer. Ann. Oncol. 24, 1416–1418 (2013).
Castro, E. et al. High burden of copy number alterations and c-MYC amplification in prostate cancer from BRCA2 germline mutation carriers. Ann. Oncol. 26, 2293–2300 (2015).
Campbell, P. J. et al. Pan-cancer analysis of whole genomes. Nature 578, 82–93 (2020).
AACR Project GENIE Consortium. AACR Project GENIE: powering precision medicine through an international consortium. Cancer Discov. 7, 818 (2017).
Hoadley, K. A. et al. Cell-of-origin patterns dominate the molecular classification of 10,000 tumors from 33 types of cancer. Cell 173, 291–304.e6 (2018).
Alexandrov, L. B. et al. The repertoire of mutational signatures in human cancer. Nature 578, 94–101 (2020).
Armenia, J. et al. The long tail of oncogenic drivers in prostate cancer. Nat. Genet. 50, 645–651 (2018).
Rubin, M. A. & Demichelis, F. The genomics of prostate cancer: emerging understanding with technologic advances. Mod. Pathol. 31, S1–S11 (2018).
Decker, B. et al. Biallelic BRCA2 mutations shape the somatic mutational landscape of aggressive prostate tumors. Am. J. Hum. Genet 98, 818–829 (2016).
Safran, M. et al. GeneCards Version 3: the human gene integrator. Database (Oxf.) 2010, baq020 (2010).
Lv, S. et al. Histone methyltransferase KMT2D sustains prostate carcinogenesis and metastasis via epigenetically activating LIFR and KLF4. Oncogene 37, 1354–1368 (2018).
Chen, J. J., Silver, D., Cantor, S., Livingston, D. M. & Scully, R. BRCA1, BRCA2, and Rad51 operate in a common DNA damage response pathway. Cancer Res. 59, 1752s–1756s (1999).
Sztupinszki, Z. et al. Detection of molecular signatures of homologous recombination deficiency in prostate cancer with or without BRCA1/2 mutations. Clin. Cancer Res. 26, 2673–2680 (2020).
Kari, V. et al. Loss of CHD1 causes DNA repair defects and enhances prostate cancer therapeutic responsiveness. EMBO Rep. 19, 1609–23 (2018).
Hjorth-Jensen, K. et al. SPOP promotes transcriptional expression of DNA repair and replication factors to prevent replication stress and genomic instability. Nucleic Acids Res. 46, 9891 (2018).
Boysen, G. et al. SPOP mutation leads to genomic instability in prostate cancer. Elife https://doi.org/10.7554/eLife.09207 (2015).
Melamed, J., Einhorn, J. M. & Ittmann, M. M. Allelic loss on chromosome 13q in human prostate carcinoma. Clin. Cancer Res. 3, 1867–1872 (1997).
Ittmann, M. M. & Wieczorek, R. Alterations of the retinoblastoma gene in clinically localized, stage B prostate adenocarcinomas. Hum. Pathol. 27, 28–34 (1996).
Chakraborty, G. et al. Significance of BRCA2 and RB1 co-loss in aggressive prostate cancer progression. Clin. Cancer Res. 26, 2047–2064 (2020).
Hernández, M. & Almeida, T. A. Is there any association between nek3 and cancers with frequent 13q14 deletion? Cancer Invest 24, 682–688 (2006).
Kluth, M. et al. 13q deletion is linked to an adverse phenotype and poor prognosis in prostate cancer. Genes Chromosom. Cancer 57, 504–512 (2018).
Khemlina, G., Ikeda, S. & Kurzrock, R. Molecular landscape of prostate cancer: implications for current clinical trials. Cancer Treat. Rev. 41, 761–766 (2015).
Henrique, R. et al. High promoter methylation levels of APC predict poor prognosis in sextant biopsies from prostate cancer patients. Clin. Cancer Res. 13, 6122–6129 (2007).
Luedeke, M. et al. Predisposition for <em>TMPRSS2-ERG</em> fusion in prostate cancer by variants in DNA repair genes. Cancer Epidemiol. Biomark. Prev. 18, 3030 (2009).
Lozano, R. et al. Association between BRCA2 alterations and intraductal and cribriform histologies in prostate cancer. Eur. J. Cancer 147, 74–83 (2021).
Cai, C. et al. ETV1 is a novel androgen receptor-regulated gene that mediates prostate cancer cell invasion. Mol. Endocrinol. 21, 1835–1846 (2007).
Chakravarty, D. et al. OncoKB: a precision oncology knowledge base. JCO Precis. Oncol. https://doi.org/10.1200/PO.17.00011 (2017).
Landrum, M. J. et al. ClinVar: public archive of relationships among sequence variation and human phenotype. Nucleic Acids Res. 42, D980–D985 (2014).
Cingolani, P. et al. A program for annotating and predicting the effects of single nucleotide polymorphisms, SnpEff: SNPs in the genome of drosophila melanogaster strain w1118; iso-2; iso-3. Fly. (Austin) 6, 80–92 (2012).
Chen, H. et al. Comprehensive assessment of computational algorithms in predicting cancer driver mutations. Genome Biol. 21, 43 (2020).
Mayakonda, A., Lin, D.-C., Assenov, Y., Plass, C. & Koeffler, H. P. Maftools: efficient and comprehensive analysis of somatic variants in cancer. Genome Res. 28, 1747–1756 (2018).
Huang, F. W. et al. Exome sequencing of African–American prostate cancer reveals loss-of-function ERF mutations. Cancer Discov. 7, 973–983 (2017).
Yuan, J. et al. Integrative comparison of the genomic and transcriptomic landscape between prostate cancer patients of predominantly African or European genetic ancestry. PLoS Genet 16, e1008641 (2020).
Sanchez-Vega, F. et al. Oncogenic signaling pathways in the cancer genome Atlas. Cell 173, 321–337.e10 (2018).
Ang, C. et al. Prevalence of established and emerging biomarkers of immune checkpoint inhibitor response in advanced hepatocellular carcinoma. Oncotarget. 10, 40 (2019).
Wang, S. et al. Copy number signature analysis tool and its application in prostate cancer reveals distinct mutational processes and clinical outcomes. PLoS Genet. 17, e1009557 (2021).
Benjamini, Y. & Hochberg, Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J. R. Stat. Soc.: Ser. B (Methodol.) 57, 289–300 (1995).
Acknowledgements
This work was supported by the National Cancer Institute of the National Institutes of Health under award numbers R01-CA207365 (to TRR) and R00-CA245900 (to KHK). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. The results presented here are in whole or part based upon data generated by the TCGA Research Network: https://www.cancer.gov/tcga. The authors would like to acknowledge the American Association for Cancer Research and its financial and material support in the development of the AACR Project GENIE registry, as well as members of the consortium for their commitment to data sharing. Finally, the authors would like to acknowledge the members of the International Cancer Genome Consortium. Interpretations are the responsibility of study authors.
Author information
Authors and Affiliations
Contributions
KHK, SB, and TRR conceived and designed the study. SB and KHK performed the analyses. KHK, SB, and TRR interpreted the data and wrote the initial draft of the manuscript. KHK, SB, SP, and TRR provided critical feedback on the revised manuscript, and approved it for submission. KHK and SB contributed equally to this work and are consider co-first authors.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary information
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Kensler, K.H., Baichoo, S., Pathania, S. et al. The tumor mutational landscape of BRCA2-deficient primary and metastatic prostate cancer. npj Precis. Onc. 6, 39 (2022). https://doi.org/10.1038/s41698-022-00284-6
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41698-022-00284-6
- Springer Nature Limited