Background

Genetic testing for germline pathogenic variants is fast becoming routine practice for men presenting with high-risk prostate cancer (PCa) in Western countries.¹ Rare pathogenic variants in medium- to high-penetrance genes not only have therapeutic implications but independently or together with common variants predict disease susceptibility and adverse outcomes for the patient and family.² Studies of PCa germline are generally focused on European ancestral populations and have linked PCa risk and advanced disease to rare or low-frequency pathogenic variants within DNA repair and cancer predisposition genes.^3,4 Consequently, commercially available panels for PCa germline genetic testing include a combination of up to 20 genes, namely BRCA1, BRCA2, ATM, CHEK2, PALB2, TP53, MLH1, MSH2, MSH6, PMS2, MUTYH, RAD50, RAD51C, RAD51D, APC, EPCAM, HOXB13, NBN, BRIP, and FANCA.

It is well established that African ancestry is a significant risk factor for advanced PCa, with the lifetime risk of dying from PCa reported to increase by 2.3- to 5-fold for African Americans compared with all other ethnic groups within the United States.⁵ For sub-Saharan Africa, PCa mortality rates are almost 2.7-fold greater than global estimates.⁶ Along with a lack of germline genetic screening, studies focused on rare/low-frequency pathogenic variants within Africa have been lacking. According to the Philadelphia PCa Consensus Conference, no agreement could be reached about germline testing of people with African ancestry due to a lack of data.⁷ A single study merging data from African American males with data from males from East Africa, specifically Uganda, associated rare BRCA2, ATM, PALB2, and NBN pathogenic variants with aggressive PCa, identifying novel African-specific predicted deleterious variants (PDVs).⁸ The latter study highlighted the need for further evaluation of these European-biased panels across different regions and populations represented within sub-Saharan Africa.

Home to the most genetically diverse population,⁹ southern Africa has PCa mortality rates that are 1.4-fold greater than those in eastern Africa and 2.7-fold greater than those in the United States, at 22 per 100,000,⁶ whereas Black South African males are at a 2.1-fold increased risk for advanced PCa at presentation compared with African American males (adjusted for age).¹⁰ Our first-of-its-kind study aimed to determine whether current PCa germline screening panels have clinical benefit for males from southern Africa.

Patients and Methods

The study cohort included 113 South African males diagnosed with predominantly advanced PCa with a Gleason score or an International Society of Urological Pathology Grade Group (ISUP GG) biased toward high-risk ISUP GG ≥4 PCa (72/113, 63.7%), with an almost even distribution of intermediate-risk (ISUP GG 2/3, 18.6%) and low-risk (ISUP GG 1, 17.7%) disease; a mean age of 67 years (range, 45–99 years); a prostate-specific antigen (PSA) level of 370 ng/mL (range, 8–4,841 ng/mL) at diagnosis; and a family history of prostate or any cancer (Table 1). The elevated PSA levels observed within our study cohort have previously been reported for Black South African males.¹⁰ Patients provided informed consent to participate in the study and were recruited as part of the Southern African Prostate Cancer Study (SAPCS), with approval granted by the University of Pretoria Faculty of Human Research Ethics Committee (HREC #43/2010, including US Federalwide Assurance FWA00002567 and IRB00002235 IORG0001762) in South Africa. Molecular genetic research for patients from the SAPCS bioresource was approved by the St. Vincent’s Hospital Human Research Ethics Committee in Sydney, Australia (#SVH15/227).

Table 1.

Patient Characteristics

View Table

Through genome-wide interrogation of 7,472,833 biallelic single-nucleotide variants (SNVs), all patients were classified genetically as being of African ancestry. As previously described¹¹ and in brief, DNA was extracted from whole blood and 2 × 150 cycle paired-end whole genomes were sequenced (Illumina HiSeq X Ten or NovaSeq) to an average of 46× coverage (range, 30–97×) and aligned to a GCRh38 reference, SNVs and small insertions and deletions (indels; <50 base pairs) were called using the Genome Analysis Toolkit (Broad Institute),¹² and larger high-confidence structural variants (SVs) were called using Manta (version 1.6.0, Illumina).¹³ The 20 genes included in this study were selected based on the latest NCCN Clinical Practice Guidelines in Oncology (NCCN Guidelines) for Prostate Cancer (Version 4.2022),¹⁴ multigene prostate cancer-specific panels,⁷ and the recent suggestion for APC, MUTYH, and RAD50 as PCa germline gene panel candidates.¹⁵ Variant data for the 20-gene panel (including 2,000 bases upstream and downstream) were made available for this study through the SAPCS Data Access Committee and sequence data were deposited in the European Genome-Phenome Archive (https://ega-archive.org) under study accession number EGAS00001006425 and dataset accession number EGAD00001009067. The variant data are available via the European Variation Archive under project number PRJEB54721.

After the removal of common variants (minor allele frequency [MAF] ≥0.05), rare and low-frequency variations were further defined as a PDV if they were classified using ≥1 of 2 prediction tools as deleterious (Sorting Intolerant From Tolerant [SIFT]) or probably/possibly damaging (PolyPhen-2 [Polymorphism Phenotyping v2]), and/or they resulted in a stop-gain, missense, or splice-site donor variant, as previously described.¹⁶Variants identified as benign or likely benign in ClinVar were then excluded from the study. Further refinement for pathogenicity, based on the guidelines of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology (ACMG/AMP)¹⁷ using InterVar (Keck School of Medicine, USC), led to further exclusion and PDV refinement. Finally, the Cancer Genome Interpreter¹⁸ tool was used to establish potential oncogenic status (high-risk gene status), specifically for PDVs with either undetermined pathogenicity or a lack of pathogenic evaluation, with further oncogenic potential (likely pathogenic) determined as a result of family history and/or early-onset disease presentation. Furthermore, SVs leading to major gene disruption were further defined as a predicted deleterious SV. All variations were manually observed and inspected through the Integrative Genomics Viewer¹⁹ and patients’ clinical records were assessed for each PDV; patients presenting with germline potentially oncogenic MUTYH variants were interrogated for loss of heterozygosity (LoH) within the matched tumor using a TitanCNA²⁰ inferred copy number data generated previously.¹¹

Results

Interrogating a total of 1,503,374 bases across the 20 genes from 113 patients of African ancestry with PCa, we identified 21,899 SNVs, 4,626 indels, and 73 SVs. SNVs and indels were assessed as PDVs and for pathogenic relevance using a 4-phase identifier system (Figure 1). Excluding for common variants (MAF ≥0.05) based on National Center for Biotechnology Information–derived African and European population identifiers, 78 SNVs were identified using SIFT and/or PolyPhen-2 as potentially deleterious (supplemental eTable 1, available with this article at JNCCN.org), of which BRCA2 p.Ile2944Phe (c.8830A>T; MAF, 0.0619) and APC p.Ser26Arg (c.78C>A; MAF, 0.0663) were common within our study population. Although their potential contribution as low-penetrance susceptibility alleles could not be assessed, notably, APC p.Ser26Arg occurred in 13 patients, of whom 12 (92.3%) presented with ISUP GG ≥4 and 1 presented with ISUP GG 3, which was greater than expected for the study distribution (Table 1). HOXB13 was the only gene with no SNVs identified during phase I analysis, which included absence for the most recently identified HOXB13 p.Ter285Lys (c.853delA; rs77179853) associated with advanced PCa in West African men²¹ and the well-established HOXB13 p.Gly84Gluz.²²

View Full Size

Phased workflow for identifying PDVs, including SNVs, small indels, and SVs/PDSVs, and filtered for MAF <0.05 and variant annotation including known pathogenic, predicted oncogenic, and high-risk POV/SVs in a cohort of 113 African patients with PCa for establishing gene-inclusion criteria for African-specific 20-gene panel (blue) germline testing, resulting in 12 candidate genes: 6 with high risk (pathogenic and/or oncogenic: ATM, BRCA2, CHEK2, TP53, FANCA, and RAD51C) and 6 with possible high risk (potentially oncogenic: APC, MSH6, MUTYH, PALB2, PMS2, and RAD50).

Abbreviations: indel, insertion and deletion; MAF, minor allele frequency; PCa, prostate cancer; PDSV, predicted deleterious structural variant; PDV, predicted deleterious variant; POSV, potentially oncogenic structural variant; POV, potentially oncogenic variant; SNV, single-nucleotide variant; SV, structural variant.

Citation: Journal of the National Comprehensive Cancer Network 21, 3; 10.6004/jnccn.2022.7097

During phases II and III we excluded for variants determined not to be pathogenic. Specifically, 26 variants reported to be “benign” or “likely benign” in ClinVar (phase II) and 12 variants categorized as likely benign using the ACMG/AMP criteria (phase III) were removed. The 38 remaining PDVs found in 52 of the 113 patients were distributed across 16 of the selected genes; 3 were confirmed as pathogenic—CHEK2 p.Arg95Ter (c.283G>A; ISUP GG 3), ATM p.Arg3047Ter (c.9139C>T; 2 patients, ISUP GG 4/5), and TP53 p.Arg282Trp (c.844C>T; ISUP GG 4)—and 1 was likely pathogenic: BRCA2 p.Trp31Arg (c.91T>C; ISUP GG 5 and family history of breast cancer), leaving 34 with undetermined pathogenicity.

Considering that a lack of African-relevant data would drive a higher proportion of PDVs of unknown or unconfirmed pathogenicity, we used the power of the Cancer Genome Interpreter (phase IV) to further interrogate the 34 undefined PDVs. We identified 12 predicted oncogenic variants spanning APC, ATM, BRCA2, CHEK2, MUTYH, FANCA, MSH6, PALB2, PMS2, RAD50, and RAD51C, of which 2—BRCA2 p.Leu3038Ile (c.9112C>A; ISUP GG 3) and RAD50 p.Leu34Phe (c.102G>C; ISUP GG 4)—were novel (Table 2). Although age-related clonal hematopoiesis of intermediate potential (CHIP) was recently reported not to be a risk factor for PCa in males of European ancestry and not significantly associated with potentially pathogenic/deleterious DNA repair gene variants,²³ to ensure that pathogenic/oncogenic variants were inherited rather than CHIP-derived, read counts were used to determine variant allele frequencies, with all well above the 10% CHIP upper threshold (average, 48.6%; range, 20.9%–66.7%). Through further interrogation of SV break points, we observed an additional novel variant that disrupted the 3′ end of MUTYH via translocation with chromosome 2 (chr2:203987190), supported by split-read evidence (16/74 reads; variant allele frequency = 21.6%; ISUP GG 4). Predicting that RNA transcript breakage would result in defective base excision repair, we assumed oncogenic potential. Although studies suggest that MUTYH requires biallelic germline presentation for pathogenicity,²⁴ a large study of 10,389 patients with monoallelic MUTYH pathogenic cancer (33 tumor types) compared with >100,000 healthy individuals showed monoallelic variants that not only were increased in patients with cancer but also, together with a second somatic hit through LoH, promoted tumorigenesis.²⁵ Although somatic LoH was not detected in the patient with the MUTYH translocation, of the 2 patients presenting with the splice donor MUTYH variant (rs140288388 G>A), 1 (ISUP GG 2) presented with and 1 (ISUP GG 5) presented without matched tumor MUTYH LoH.

Table 2.

Pathogenic and Potentially Oncogenic Germline Variants Identified

View Table

Overall, the 4 pathogenic and 13 potentially oncogenic variants were identified in 12 of the 20 genes (60%), which included 2 each in ATM, BRCA2, APC, CHEK2, and MUTYH, and 1 each in FANCA, MSH6, PALB2, PMS2, RAD50, RAD51C, and TP53. Further clinical observations for the pathogenic and potentially oncogenic variants in 20 (17.7%) patients showed that 75% presented with advanced PCa (ISUP GG ≥4), which increased to 90% when we included patients with ISUP GG 3. All but a single oncogenic variant were rare; specifically, ATM p.Asp44Gly (c.131A>G) was found in 4 patients (MAF, 0.017) all diagnosed with ISUP GG 4 PCa and lay within the telomere-length maintenance and DNA damage repair domain, which is critical for telomere maintenance function and cell viability²⁶ (Figure 2A). Notably, the patient with the novel BRCA2 p.Leu3038Ile variant that lies within the central oligonucleotide/oligosaccharide single-stranded DNA binding fold of the highly conserved DNA binding domain, critical for mediating homologous recombination and maintaining genome stability²⁷ (Figure 2B), presented at an age 13 years younger at diagnosis than the study mean (54 vs 67 years of age). In addition, patients presenting with FANCA Arg504Cys (c.1510G>A; ISUP GG 4; age 65 years) and RAD51C Arg260Gln (c.779G>A; ISUP GG 3; age 64 years) reported a positive family history of PCa.

View Full Size

Amino acid position, MAF, and variant type, including pathogenic/likely pathogenic (ClinVar/InterVar), potentially oncogenic (CGI), predicted deleterious (SIFT/PolyPhen), splice variant, or novel, identified in 113 patients of African ancestry with PCa for the genes presenting with the highest number of potentially impactful germline variants in our study: (A) ATM and (B) BRCA2.

Abbreviations: CGI, Cancer Genome Interpreter; MAF, minor allele frequency; PCa, prostate cancer; PolyPhen, Polymorphism Phenotyping v2; SIFT, Sorting Intolerant From Tolerant.

Citation: Journal of the National Comprehensive Cancer Network 21, 3; 10.6004/jnccn.2022.7097

Discussion

Although PCa germline testing gene panels have almost exclusively been identified through studies focused on non-African populations, African ancestry is a significant risk factor for PCa and adverse outcomes.^5,6,8 Aside from a single East African study,⁸ further inclusion across sub-Saharan Africa has been lacking. In addition, studies are recurrently reporting germline variations in genes associated with risk for aggressive PCa.^2,8 However, identifying high-penetrance genes that mediate the genetic pathways and influence the risk and course of the disease is challenging given the high numbers of sporadic cases of disease and the rarity of pathogenic variations.² In this first-of-its-kind study for southern Africa, we focused on Black South African males presenting with largely advanced disease, identifying 38 PDVs and one predicted deleterious SV in 16 of the 20 most common genes included in PCa screening panels. Noting that 10.2% of the PDVs/SVs were novel, we find that the bias toward variants of uncertain pathogenic significance, together with the identification of a single known PCa pathogenic variant, CHEK2 p.Arg95Ter,²⁸ further emphasizes the need for further African-specific investigation to establish tailored PCa screening panels.

Overall, 5 patients presented with a known rare pathogenic variant. Except for CHEK2 p.Arg95Ter, pathogenic variants were novel to PCa and included ATM p.Arg3047Ter, which is located in the FATC domain, was previously reported to block lymphocyte development (Figure 2A), and is considered pathogenic for hereditary cancer-predisposing and ataxia-telangiectasia syndromes²⁹; TP53 p.Arg282Trp, most predominantly associated with Li-Fraumeni and hereditary cancer-predisposing syndrome; and BRCA2 p.Trp31Arg, shown to be likely pathogenic in hereditary breast and ovarian cancer. Notably, the African patient presenting with the pathogenic BRCA2 variant reported a family history of breast cancer. The presence of the ATM pathogenic variant in 2 patients with advanced PCa calls for further consideration for population-specific relevance. Overall, we observed a lower prevalence of rare pathogenic variants among our cohort (5.6%; 4/72 patients with ISUP GG 4/5) compared with the 11.8% reported for non-African patients with confirmed metastatic PCa,³⁰ whereas our data are comparable to those reported for East African patients (5.7%, unknown metastatic status) and African American patients (3.4%), including ATM and BRCA2 presenting as African-relevant contributing pathogenic candidates.⁸

To address the lack of African-relevant data and bias toward PDVs of unconfirmed pathogenicity, we further annotated for oncogenic potential, identifying together with the pathogenic variants 13 additional potentially oncogenic variants, which taken together impacted 12 genes and 17.7% of the patients. Compared with the overall study, patients presenting with pathogenic and potentially oncogenic variants were biased toward advanced PCa, defined as ISUP GG ≥4 (63.7% vs 75%) or ISUP GG ≥3 (77.0% vs 90%), respectively. Although the true pathogenicity of the oncogenic variants is yet to be determined, the observed early-onset novel BRCA2 and PCa familial RAD51C and FANCA variants provide additional merit for their pathogenic potential. Irrespective of pathogenicity, a recent US study suggested that the RAD family of genes (although excluding for RAD51C) is significantly more likely to harbor a germline variant in African (n=259) versus European (n=272) ancestral patients presenting with a bias toward low-risk disease (ISUP GG <3, 79.6% vs 83.5%, respectively).³¹ As the most recent addition to the PCa gene panels, our study raises additional considerations for the inclusion of FANCA, at least when considering men of African ancestry. When we considered the early-onset and familial PCa associating rare oncogenic variants with known pathogenic variants, the prevalence increased to 6.9% (5/72) of patients presenting with high-risk or very-high-risk PCa, defined using the current NCCN Guidelines (ISUP GG ≥4)¹⁴ or 9.2% (8/87) using the expanded criteria (ISUP GG ≥3).

Appreciating the limitation of the study size (113 patients) yet the bias toward advanced disease, notable exclusions included the lack of pathogenic/oncogenic BRCA1, MSH2, and HOXB13 variants. Unlike BRCA2, an association between PCa risk and BRCA1 mutation has been less consistent^32,33; however, the NCCN Guidelines state that eligible patients should be evaluated for both BRCA2 and BRCA1 status.¹⁴ More recently, the relative risk for BRCA1 PCa-associated pathogenic variants has been associated with younger age of diagnosis.³² Notably, our cohort was biased toward older age at presentation (mean age, 67 years; range, 45–99 years). In contrast to our study, pathogenic BRCA1 variants in African Americans have been associated with early-onset³⁴ and metastatic PCa,³⁵ whereas both pathogenic and BRCA1 variants of unknown significance are reportedly more frequent in African versus European ancestral Americans with PCa.^31,36,37 Compared with our study, the latter were overall biased toward patients presenting with lower-risk disease. MSH2 is another gene that is known to frequently harbor pathogenic variants associated with advanced PCa³⁸; however, only one PDV was found in the southern African population in our study and was identified as likely benign according to the ACMG/AMP classification. Although the 2017 Philadelphia Prostate Cancer Consensus Conference recommended testing for HOXB13 variants, especially for suspected hereditary PCa,³⁹ we observed a high conservation of HOXB13 in our limited study.

Comparing our data with the larger East African study (n=651),⁸ although notable differences included the lack of pathogenic (or oncogenic) NBN variants in southern African patients and conversely TP53 for East African patients, similarities included a notable exclusion for MLH1 and MSH2 across 764 sub-Saharan African patients. One should further appreciate that in addition to geographic and ethnic differences, there were notable differences in patient age (mean age, 67 vs 70 years) and tumor pathology (ISUP GG ≥4, 63.7% vs 47.2%) at presentation between the southern and East African studies (the latter of 441 patients with known pathology), respectively. Although the East African study was supported by African American data (n=1,447; ISUP GG ≥4, 24.7%),⁸ the recent African American study³⁶ with a larger representation of patients with advanced PCa (n=237; ISUP GG ≥4, 32.9%) and interrogating a 14-gene panel concurred with the pathogenic relevance for BRCA2 and ATM while further confirming the East African–identified PALB2, denoted as a high-risk gene candidate in our study. Additional overlaps with African American data most recently reported for 276 DNA damage repair genes³¹ include the potential relevance for PMS2 and another RAD family member, RAD50, as high-risk gene candidates. We concur that men of African ancestry present with a narrower spectrum of pathogenic variants compared with non-African men, highlighting the limitations for African-relevant translation. Conversely, after observing 12 of the 20 most-common tested genes presented with potentially oncogenic variants in southern African patients with PCa, we call for further clarification for pathogenicity. Notably, more studies need to be conducted within southern Africa and across the African diaspora to validate current findings and develop germline testing panels that are more aligned to the genetic profile of this population.

Conclusions

The high PCa mortality rates reported for southern African males, and in turn the high number, novelty, and expanse of oncogenic variants identified in our study, including pathogenic variants spanning ATM, BRCA2, CHEK2, and TP53, with a further familial link to FANCA and RAD51C and an additional early-onset link to BRCA2 (although the pathogenicity of variants in APC, MSH6, MUTYH, PALB2, PMS2, and RAD50 cannot be excluded), highlight the immediate benefits for including African patients in largely routine Westernized germline carrier screening programs. Conversely, we highlight the limitations of the current 20-gene panel approach for men of African ancestry. Our data support the notion that alternative unknown gene targets could be playing a significant role for males of African ancestry, and as such, we call for additional African inclusion and genome-wide interrogation. Ultimately, guidelines focused on African inclusion need to be established to ensure that the clinical benefit for PCa screening through prevention or targeted therapy is available to all globally.