2.1 NCCN Guidelines for Germline Testing

Recommendations for germline testing were based on three NCCN guidelines: Prostate Cancer (Version 1.2025–December 4, 2024) [16], Prostate Cancer Early Detection (Version 2.2024–March 6, 2024) [4], and Genetic/Familial High-Risk Assessment: Breast, Ovarian, Pancreatic, and Prostate (Version 3.2025–March 6, 2025) [5]. Genes recommended for germline testing include 11 genes associated with PCa susceptibility and 19 genes related to response to PARPi. Six of these genes overlap between the two categories, including ATM, BRCA1, BRCA2, CHEK2, MLH1, and PALB2.

We performed association tests of P/LP mutations in 11 NCCN-recommended PCa susceptibility genes with both PCa diagnosis and aggressiveness in two large cohorts, a population-based cohort and a hospital-based cohort. The first cohort was the UK Biobank (UKB), a population-based cohort from across the United Kingdom, aged between 40 and 69 years at recruitment [31]. A PCa diagnosis was obtained from national cancer registries (ICD10: C61) and self-report. PCa-specific death was identified from death registries. Based on the last date of access on 2023-06-12, there were 15,928 PCa cases (including 1458 who died of PCa) and 191,055 men without a diagnosis of PCa. Access to the UKB data was approved under Application Number 50295.

The second cohort was of PCa patients (N = 6124) from the Brady Urological Institute and Comprehensive Cancer Center, Johns Hopkins School of Medicine, Baltimore, MD (Hopkins). The majority of these PCa patients were recruited from those who underwent radical prostatectomy for treatment of clinically localized PCa since 1987 [32]. Pathological evaluation and tumor grading were consistently performed by Hopkins pathologists. Radical prostatectomy cases with preoperative PSA ≥ 20, pathological Gleason score ≥ 8, metastatic disease, or who suffered death from PCa were classified as aggressive PCa (N = 1808), otherwise, as nonaggressive PCa (N = 4316) (Table 1). The study cohort was approved by the Institutional Review Boards at Johns Hopkins School of Medicine.

P/LP variants in each gene were annotated from whole-exome sequencing data using the ClinGen sequence variant interpretation (CVI) criteria for all subjects in both cohorts [33]. The ClinGen CVI refines and standardizes the application of the American College of Medical Genetics (ACMG)/American Molecular Pathology (AMP) guidelines for variant interpretation and is used in clinical genetic testing. Association tests of each gene with risk for PCa diagnosis were performed in the UKB by comparing the aggregated carrier rate of P/LP variants in a gene between cases and controls. Association testing of each gene with aggressiveness of PCa was performed by comparing aggregated carrier rate of P/LP variants between lethal (PCa-specific death) and nonlethal cases in the UKB and between aggressive and nonaggressive cases in the Hopkins cohort separately. All association tests were performed using logistic regression, adjusting for age and genetic background (the top 10 principal components estimated from genome-wide SNPs). Results were then combined using meta-analysis.

4.1 Genes Associated With Risk of PCa Diagnosis

The NCCN guidelines of PCa and early detection of PCa recommend germline testing for 10 genes, including those involved in HRR (ATM, BRCA1, BRCA2, CHEK2, PALB2), MMR (MLH1, MSH2, MSH6, PMS2), and a prostate-specific gene (HOXB13) [4, 16]. In addition, the NCCN guidelines of genetic/familial high-risk assessment recommend testing TP53 for PCa susceptibility along with five other genes that overlap with the guidelines of PCa and early detection of PCa mentioned above (ATM, BRCA1, BRCA2, CHEK2, and HOXB13) [5].

Deleterious mutations in most of these genes affect DNA repair capacity and can increase PCa risk. These mutations have been reported in PCa patients and potentially contributed to their development of PCa [17]. However, some mutations in DNA repair genes may be identified incidentally and do not contribute to PCa risk, as patients are often tested due to a family history of hereditary cancer syndromes. For genes where nominal statistical associations with PCa risk were reported, not all findings have been consistently replicated in independent studies [18]. The lack of consistent associations for some genes is likely due to several factors, including false positives from multiple tests, publication bias, and limited statistical power to detect low-frequency mutations with moderate effects [32]. When testing multiple genes, an even larger sample size is required to meet more stringent significance levels, such as those adjusted using the Bonferroni correction to reduce the likelihood of false-positive findings. Therefore, only well-designed studies with larger sample sizes and appropriately adjusted P-values for multiple testing can yield reliable results. Among the 11 guideline-recommended genes, potential deleterious variants in BRCA2, ATM, CHEK2, MSH2, and HOXB13 have been associated with PCa risk and replicated in several large studies [19-23]. No consistent association with PCa risk has been found for other genes, including BRCA1.

Results from our association tests were consistent with findings from large, published studies. Furthermore, the use of P/LP annotations in our study, as recommended by the ACMG/AMP, enhances the accuracy and clinical relevance of our findings. The OR estimates for each significant gene derived from this population-based cohort offer more reliable information for genetic counseling.

Overall, results from our analysis and large published studies provide statistical evidence that P/LP mutations in HOXB13, BRCA2, ATM, CHEK2, and MSH2 are associated with PCa risk, with estimated ORs for PCa between 1.81 and 4.05. In contrast, no statistical evidence currently supports an association between PCa risk and other guideline-recommended genes, including BRCA1. These findings have important clinical implications for the interpretation of germline testing results and subsequent clinical management. For instance, if a PCa patient carries a P/LP mutation in one of these significantly associated genes, it may provide urologists with insights into the genetic basis of the cancer and allow for informed counseling of the patient's male family members regarding their own PCa risk and the importance of early detection. Conversely, if a PCa patient carries a P/LP mutation in one of the nonsignificant genes, no such inference regarding PCa risk should be made for the patient or their family members. However, the presence of a P/LP mutation in these genes may indicate an increased risk for other cancers, warranting a referral to a genetic counselor for further assessment and risk management.

4.2 Genes Associated With Aggressive PCa and Outcomes

Germline mutations in genes implicated in PCa progression have potential clinical utility by supplementing clinical information and somatic DNA alterations to guide treatment decisions. This includes determining the suitability of active surveillance versus definitive treatments (such as surgery or radiation) and predicting disease progression on local therapy and decreased overall survival. However, specific genes associated with PCa progression remain largely unknown. Although the association of PCa progression with several PCa-related genes has been reported in various studies [23-29, 34-40], most have not been consistently replicated, including BRCA1, CHEK2, PALB2, MLH1, MSH2, MSH6, PMS2, and TP53. The limited number of genes associated with PCa aggressiveness is likely due to the smaller sample sizes available for association testing in case-only analyses compared to studies in larger case-control cohorts, as well as challenges to accurately and consistently distinguish aggressive from nonaggressive PCa [32]. Citing the inconsistent and uncertain results between germline mutations and prognosis of PCa from retrospective studies with moderate-to-high risk of bias, the NCCN guidelines do not recommend them for risk stratification [1-3, 16].

However, consistent evidence associating BRCA2 and ATM with PCa progression has been reported in several large studies [23-29], including the largest study to date involving 4207 aggressive and 16,170 nonaggressive PCa cases. This study reported association results for 9 of the 11 genes [23]. BRCA2 and ATM were found to be significantly associated with aggressive PCa, p < 0.0055 for performing 9 tests (0.05/9 genes). Additionally, a nonsignificant higher risk for aggressive PCa was observed for six other genes, while PMS2 was associated with a lower risk. Furthermore, results from our combined analysis of UKB and Hopkins cohorts provided further support for the association of BRCA2 and ATM with PCa aggressiveness.

In addition, emerging evidence suggests that the African-ancestry HOXB13 variant (X285K) is associated with aggressive PCa [41-43]. In a larger study comprising 22,361 African-ancestry men (including 11,688 PCa cases), the X285K variant was significantly associated with PCa risk (OR = 2.4, 95% CI: 1.5–3.9, p = 2E-04) [42]. Importantly, patients with more aggressive and advanced disease (Gleason score ≥ 8, stage T3/T4, and/or metastatic disease) had a higher risk compared to unaffected controls. Furthermore, a recent functional study revealed a unique gain-of-function oncogenic mechanism of X285K protein in regulating E2F/MYC signatures [43].

Overall, findings from our analysis and well-designed large studies consistently demonstrate that P/LP mutations in BRCA2 and ATM are associated with aggressive PCa. Understanding which tested genes are associated with aggressive PCa has critical clinical implications for treatment decisions. For patients carrying a P/LP mutation in a gene significantly associated with PCa aggressiveness, such as BRCA2 and ATM, urologists may consider more intensive treatment strategies based on disease stage. For localized disease, definitive treatments such as surgery or radiation therapy may be preferred over active surveillance [28], Conversely, if a PCa patient carries a P/LP mutation in a gene that has not been significantly associated with aggressive PCa, no treatment decisions can be inferred specifically from an inherited genetic perspective alone.

4.3 Genes Associated With PARPi Sensitivity

The NCCN guidelines of PCa recommend genetic testing of 19 genes involved in PARPi sensitivity for considering treatment of metastatic CRPC (mCRPC) using olaparib, rucaparib, talazoparib, and niraparib [1-3, 16]. These genes are primarily in the HRR pathway, including ATM, ATR, BARD1, BRCA1, BRCA2, BRIP1, CDK12, CHEK1, CHEK2, FANCA, FANCL, MLH1, MRE11A, NBN, PALB2, RAD51B, RAD51C, RAD51D, and RAD54L. PCa patients with deleterious mutations in these genes—whether somatic or germline—may exhibit enhanced sensitivity to PARPi due to synthetic lethality [6, 7, 9-15].

These 19 genes were recommended for testing primarily based on results from four clinical trials where HRR alterations were evaluated, including PROfound [9], TALAPRO-2 [12], TRITON3 [14], and MAGNITUDE [15]. They were included in these trials as sequencing assays for these genes in tumor DNA were readily available. Key findings from these clinical trials and other studies can be summarized as the following [44-49], (1) PARPi are effective for treatment of mCRPC, particularly for patients with HRR alterations, (2) among HRR gene alterations, evidence is consistent and strongest for BRCA2, weak for BRCA1, and lacking for other genes, (3) gene alterations were analyzed by combining somatic and germline origin and most were likely somatic [50], and (4) germline mutations are rare and statistical evidence is lacking. The low frequency of germline P/LP mutations in most of these genes from our study, except for BRCA2, further highlights the challenge.

Overall, despite the primary objective of the NCCN guidelines for germline testing being to identify patients with mutations that may enhance their response to PARPi, evidence associating germline mutations in HRR genes with PARPi response remains limited. Understanding these findings has critical clinical implications for treatment decision-making. Patients with germline BRCA2 mutations are expected to respond well to PARPi, but the same assumption has not been supported for germline BRCA1 or other HRR gene mutations.

4.4 Beyond NCCN Guidelines

The NCCN guidelines recommend specific criteria to identify a subset of PCa patients for germline testing, based on the premise that these individuals are more likely to carry mutations that could impact clinical management. However, findings from the PROCLAIM trial revealed no significant difference in the prevalence of P/LP variants between those who met the germline testing criteria (8.8%) and those who did not (6.6%) among 958 PCa patients [51]. These results suggest that the NCCN criteria may miss up to 42% of patients carrying P/LP variants, promoting calls for universal germline genetic testing in all patients diagnosed with PCa [52, 53].

Beyond the NCCN-recommended genes, germline variants in additional genes have also been reported to be associated with PCa risk, aggressiveness, and response to treatment. Many of these associations have not been consistently replicated in independent studies, likely due to false discoveries in initial reports and insufficient statistical power to confirm rare mutations.

However, recent large-scale genome and exome studies have implicated several novel genes for PCa risk (BIK, SAMHD1, and FAM111A) [22, 23], poor prognosis (KLK3 and AOX1) [23, 54], and novel genes involved in sensitivity to PARPi, such as MMS22L [55, 56]. Validation of these novel genes in independent populations is needed.

In addition to these monogenic genes, various polygenic risk scores (PRS)—which aggregate many established common PCa risk-associated SNPs—have been consistently associated with the risk to develop PCa [57-60]. In particular, their performance of identifying high-risk men for PCa screening was recently demonstrated in the BARCODE1, a large prospective PCa screening trial [61]. However, while PRSs identified high-risk men for PCa, including aggressive and lethal PCa, they do not distinguish between aggressive and nonaggressive disease among PCa patients.