We conducted a retrospective study of 172 PCa patients with BRCA1/2 germline alterations ( Table S1 ) and 312 PCa patients without BRCA germline mutations ( Table S2 ) who underwent genomic profiling with a hybridization capture-based next-generation sequencing (NGS) assay between February 2018 and June 2020 collected from the database of GloriousMed Technology Co., Ltd. (Shanghai, China). These patients were mostly from four hospitals (Renji Hospital of Shanghai Jiaotong University School of Medicine, The First Affiliated Hospital of Wenzhou Medical University, General Hospital of Eastern Theater Command, The First Affiliated Hospital of Nanjing Medical University). The study was approved by the Committee for Ethics of the First Affiliated Hospital of Wenzhou Medical University and informed consent was obtained from each patient. We collected 123 circulating tumor DNA (ctDNA) samples from and 59 biopsied tumor tissue samples from the 172 patients ( Table S1 ).

The samples underwent NGS testing at GloriousMed Clinical Laboratory Co., Ltd. Cell-free DNA (cfDNA, from plasma), tumor formalin-fixed, paraffin-embedded (FFPE) DNA, and genomic DNA (gDNA, from white blood cells) were extracted according to standard procedures using a QIAamp Circulating Nucleic Acid Kit (Qiagen), QIAamp DNA FFPE Tissue Kit (Qiagen), and Blood Genomic DNA Mini Kit (cwbiotech), respectively. From each sample, 200–500 ng FFPE DNA, 20–100 ng cfDNA, or 500 ng gDNA were used for library preparation and quantification according to KAPA HyperPrep protocols (KAPA). The genes’ coding regions were captured using custom-designed DNA enrichment panels (50/66/620/642 panels). For analysis, we focused on the common 50 genes ( Supplementary Table 2 ). Library pools (5–6) were hybridized to the capture panel according to standard procedures. Then, the libraries were purified and quantified using AMPure XP (Beckman Coulter) and a Qubit™ dsDNA HS Assay Kit (Thermo Fisher Scientific). The final libraries were sequenced on Illumina NextSeq 500 (75-bp paired-end reads [PE75]) or NovaSeq 6000 (PE150) instruments.

The raw data were trimmed using Trimmomatic (9). Then, the reads were aligned with the human reference genome (hg19) using Burrows-Wheeler Aligner (10). Duplicated reads were removed using Picard (http://broadinstitute.github.io/picard/). Mapped reads were realigned to the genome using Genome Analysis Tool Kit (GATK) (11). Germline mutations were called using GATK’s HaplotypeCaller (11) with a paired workflow. Variants were then annotated using ANNOVAR (12) and an in-house-developed code. The human identity concordance of the paired samples was verified using an in-house script. Germline mutations considered deleterious (frameshift insertions, nonsense/stop-gains, splice site variants, deletions, or reported as pathogenic or likely pathogenic in the ClinVar database) were included for analysis. Here, “pathogenic alterations” includes pathogenic or likely pathogenic alterations; “non-pathogenic” represents variants of uncertain significance (VUS).

The assessment of clinical characteristics between different cohorts, including age at diagnosis, Gleason score, et al., were based on the Wilcoxon rank sum test. Graphpad Prism V8 (GraphPad Software, Inc.) and R v3.6.1 (www.R-project.org) were used for data analysis. A two-sided P value <0.05 was considered significant.

The 172 patients with BRCA germline mutations comprised patients diagnosed with PCa (NA), localized prostate cancer (LAPC), metastatic hormone-sensitive prostate cancer (mHSPC), or mCRPC ( Table 1 ). Castration resistance was defined according to the European Association of Urology (EAU) Guidelines on Prostate Cancer (2021 edition). Significant difference was found in median age between those with and without BRCA1 mutation (69 years; range, 53–89 years vs. 66 years; range, 44–98 years, p < 0.05), but not in BRCA2 mutation(65.5 years; range, 34–85 years vs. 66 years; range, 44–98 years, p > 0.05). Overall, there was no significant difference in PSA value (13.6; range, 0–1000 vs. 0; range, 0–5000, p > 0.05) and Gleason score (p > 0.05) between the patients with and without BRCA mutation. The baseline comparison between the patients with pathogenic and non-pathogenic BRCA1/2 mutations was performed, significant difference was found in median age between them (65 years; range, 34-82 years vs. 67.5 years; range, 53-89 years, p < 0.05). Similarly, there was no significant difference in PSA value (6.95; range, 0–1000 vs. 10.85; range, 0–905, p > 0.05) and Gleason score (p > 0.05) between the patients with pathogenic and non-pathogenic BRCA1/2 mutations.

The frequency distribution of BRCA1/2 germline pathogenic and non-pathogenic (VUS) mutations differed significantly ( Figure 1 ). BRCA1 germline mutations were mainly VUS (27.91%, 48/172) while BRCA2 germline mutations were mainly pathogenic (41.86%, 72/172). The frequency of BRCA2 germline mutations in our cohort (69.19%, 121/172) was much higher than that of BRCA1 germline mutations (32.56%, 57/172) ( Table 1 ). Based on pathological stage, BRCA1 and BRCA2 mutations were mainly found in mCRPC, especially BRCA2, with a frequency of 16.86% (29/172) and 37.79% (65/172), respectively. While it was similar between LAPC and mHSPC, BRCA1 was 5.81% (10/172) and 6.4% (11/172), BRCA2 was 12.21% (21/172) and 9.88% (17/172).

The overall analysis showed that the BRCA1/2 variants were distributed in most exon regions of the BRCA1 and BRCA2 genes, and no new hot spot variants were found ( Figure 2 and Figure S1 ). A total of 208 BRCA germline variants were identified: 63 and 145 in BRCA1 and BRCA2, respectively. c.2726A>T (p.N9091) was the most frequently mutated variant (3.37%, 7/208). Most of the variants occurred only once (58.65%, 122/208); 41.35% (86/208) of the variants with >1 occurrence were mainly distributed in mCRPC (55/86). Among the 63 BRCA1 variants, c.2726.4>T (p.N9091) was the most common (7/63), while c.5722-5723DELCT (p.L1908FS) was the most common (5/145) among the 145 BRCA2 variants.

In the variant type analysis, frameshift, missense, and splicing were the common BRCA1 and BRCA2 variants. Frameshift and missense were the most advantageous variants, occurring in 90.7% (132/145) and 80% (50/63) of BRCA1 and BRCA2 variants, respectively. The difference was that in-frame and nonsense variants only appeared in BRCA2 and BRCA1, respectively.

Here, we report the distribution of the pathogenic and non-pathogenic variants in the major exons of BRCA1 and BRCA2 ( Figures 3A, B ). There were 54.55% (6/11) and 69.51% (57/82) pathogenic variants in BRCA1 exon 10 and BRCA2 exon 10/11, respectively. The number of variants per exon was normalized according to the exon length ( Figures 3C, D ). Exon 4 and exon 13 had the most variants in BRCA1 and BRCA2, respectively. Exon 4 and exon 5 had the most pathogenic variants in BRCA1 and BRCA2, respectively. Among all variants, the overall pathogenic rates for BRCA1 and BRCA2 were 17.46% (11/63) and 56.55% (82/145), respectively ( Figures 3E, F ). The frameshift variants were pathogenic both in BRCA1 and BRCA2. Similarly, missense variants also showed the same trend in BRCA1 and BRCA2. The difference was that splicing variants were non-pathogenic and pathogenic in BRCA1 and BRCA2, respectively.

We identified somatic alterations in the AR pathway genes, DDR pathway genes, and tumor suppressor genes (TP53/RB1) in the patients ( Figure 4 ). AR (26%, 32/123), TP53 (20%, 25/123), FOXA1 (15%, 18/123), NCOR2 (12%, 15/123), and PTEN (10%, 12/123) were the top five somatic mutation genes associated with ctDNA sequencing. In the tissue samples, the top seven somatic mutation genes were FOXA1 (34%, 20/59), TP53 (15%, 9/59), AR (15%, 9/59), NCOR2 (14%, 8/59), FANCA (12%, 7/59), RB1 (12%, 7/59), and SPOP (10%, 6/59). Among the somatic mutations associated with BRCA2 germline mutations, the most frequent were FOXA1 (ctDNA sequencing, 7.4% [4/54]; tissue samples, 52% [11/21) and NCOR2 (ctDNA sequencing, 7.4% [4/54]; tissue samples, 24% [5/21]) mutations; TP53 was the dominant somatic mutation associated with BRCA1 germline mutations (ctDNA sequencing, 25% [1/4]; tissue samples, 17% [1/6]).

Further, there were some differences in the blood and tissue profiles, so we conducted consistency analysis on mutation data from 10 patients using both tissue and matched blood samples ( Figure S2 and Table S3 ). Most of the 10 patients had a relatively high degree of consistency between the mutations in the tissue and matched blood samples (e.g. Patients #2, 6, 9). However, the sample size (10 patients) was small, which limited consistent comparison of the occurrence frequency of the related genes in specific tissue and blood pathways.

Research targeting BRCA1/2 mutations has received increasing attention in recent years, in part because of the success of PARPi in clinical studies. While these studies were mostly focused on foreign populations, there have been few studies on Chinese populations. Studying BRCA1/2 mutations in the Chinese PCa population will enable more comprehensive understanding of BRCA1/2 mutations in this population, and further insightful analysis of the characteristics of these mutations will ultimately provide a more optimal treatment plan for patients.

Although our cohort was smaller than that in another prospective study of BRCA1 and BRCA2 germline pathogenic mutations in a Chinese population (172 vs. 316) (13), our study has a larger Chinese cohort with BRCA1 and BRCA2 germline pathogenic mutations (9 vs. 2, 72 vs. 20). The BRCA2 germline mutation carriers in the present study were at an earlier age, i.e., 67 years (range, 34–89 years), which was similar to previous reports of patients with PCa with BRCA2 mutations having an earlier age of diagnosis (14). Furthermore, most of the clinical stages were concentrated in the mCRPC stage and had high Gleason scores, and the frequency of BRCA1 and BRCA2 germline mutations during the metastatic PCa (MPC) stage (mHSPC and mCRPC) was higher than that in the localized stage. These results confirm that BRCA1/2 mutation carriers are more likely to have lymph node involvement and distant metastases (15). BRCA-positive PCa populations often have higher Gleason scores (≥8) and higher tumor-node-metastasis (TNM) stage (15). These findings could provide more comprehensive evidence for novel endocrine therapy treatments for PCa. New endocrine therapy has a better effect on BRCA mutation carriers compared with non-carriers in the mCRPC population, and PCa populations with BRCA1 or BRCA2 mutations could benefit from abiraterone or enzalutamide treatment (16).

Understanding the distribution of pathogenic variants in BRCA1/2 in key domains (exons or introns) and the role of each specific variant is of great significance for PCa treatment. Here, BRCA1/2 variant analysis revealed no distinct hotspot mutation. In patients with PCa, the BRCA2 gene has a higher risk of mutations in the c.756-c.1000 and c.7914+ regions (17, 18). However, our results yield no similar conclusions: the frequency of BRCA2 mutation was 16.55% (24/145) in the c.7914+ region, and was 8.97% (13/145) in the in c.756-c.1000 region.

Notably, each variant type showed different characteristics. Here, the top three BRCA variants were frameshift, missense, and splicing variants. Meta-analyses investigating the presence of BRCA genes in patients with cancer found that the missense variant was the most frequent in patients with BRCA1 and BRCA2 variants (19). However, our results show that the frameshift variant was the most frequent in BRCA2 variant carriers, and that all of the variants were pathogenic. Therefore, it is necessary to develop new specific tests for exon or intron–exon boundaries for more accurate PCa clinical diagnosis and treatment.

We collected BRCA1/2 gene germline mutation data from the Chinese PCa population, and found that the BRCA2 mutation frequency (69.19%) was much higher than that of BRCA1 (32.56%), and that most mutations occurred at the MPC (mHSPC and mCRPC) stage. This suggests that there might be different tumor gene expression patterns in BRCA2. This is similar to previous studies reporting that the characteristics of BRCA2 mutated tumors were more similar to those of mCRPC than of LAPC (20–22). In 6,902 men with BRCA1 or BRCA2 mutations who developed cancer, especially breast, prostate, and pancreatic cancer, and multiple primary tumors, there was an association with a higher rate of BRCA2 mutations (23). Moreover, clinical trial data (TRITON2 and PROfound) have shown that patients with BRCA2 mutations could benefit more from PARPi than those with BRCA1 mutations (24). These results suggest that patients with PCa with BRCA2 mutations might receive higher prognostic benefit than BRCA1 carriers.

There is an association between patients with PCa with BRCA1/2 mutations carrying other mutations (e.g., TP53) with poorer prognosis and PARPi sensitivity (25). The TRITON2 study observed 62% and 42% BRCA1 and BRCA2 mutation carriers with TP53 mutations, respectively (26). The cBioPortal database, which contains publicly available genomic information, shows that harmful TP53 mutations are more common in patients with PCa carrying BRCA1 mutations than in patients carrying BRCA2 mutations (39% vs. 23%) (26). In the present study, ctDNA sequencing showed that the TP53 mutation frequency in the somatic mutant along with BRCA1 germline mutations was much higher than that of BRCA2 germline mutation (25% [1/4] and 5.6% [3/54]). In tissue sequencing, the frequency was 17% (1/6) and 5% (1/21), respectively.

We also found that, except TP53, FOXA1 and NCOR2, along with BRCA1 germline mutations, were more frequent than BRCA2 germline mutations. Tumors with FOXA1 mutations accompanied by higher Gleason scores, shorter biochemical relapse time, and faster metastatic disease progression (27). NCOR2 could interact with AR, thereby inhibiting the transcriptional activity of AR (28, 29). Recent studies have also found that patients with FOXA1 and NCOR2 mutations had poor prognosis (22). This evidence suggests that FOXA1 and NCOR2 somatic mutations may affect disease progression in patients with BRCA germline mutation. However, our results lack follow-up information for the patients, and the relationship between FOXA1 and NCOR2 mutations and the prognosis of patients carrying BRCA germline mutations should be explored in the future.

Our study has some limitations. Our data contain many variants of VUS (53.85%, 112/208), which prevents elucidation of the pathogenicity of some mutations, consequently delaying the selection of appropriate therapies. Therefore, new database updates or more information mining of mutations are needed. In addition, the panel we used can only capture exon regions and may have missed some meaningful intron mutations.