Co-occurring rare germline DNA repair gene variants in BRCA1/BRCA2 implicated hereditary breast cancer families Authors: Wejdan M. Alenezi, Neil Recio, Caitlin T. Fierheller, Corinne Serruya, Timothée Revil, Anne-Marie Mes-Masson, Diane Provencher, Celia M. T. Greenwood, Jiannis Ragoussis, Patricia N. Tonin Published: 2026 ABSTRACT As gene-panels expand to include candidate breast cancer predisposing genes (CPGs) involved in diverse DNA repair pathways, we investigated an index breast cancer (BC) case from 56 BRCA1/BRCA2 implicated high-risk hereditary BC families using gene-based approach for co-occurring, rare germline variants predicted to be damaging and clinically relevant in 276 DNA repair genes (DRGs). Using whole exome sequencing analyses, we identified a total of 287 variants in 55% of DRGs, of which 24 loss-of-function and 36 other predicted damaging variants were identified in 72% and 76% of BRCA1 and BRCA2 implicated cases, respectively. We also identified 60 variants of clinical interest in various known CPGs, including those involved in risk to other cancers, in 72% of BRCA1 and 60% of BRCA2 cases. The number of variants predicted to be deleterious in diverse DRGs raises questions about the interpretability of findings as panel testing expands beyond clinically established breast CPGs. Introduction Heterozygous carriers of germline pathogenic variants (PVs) in BRCA1 or BRCA2 have a high life-time risk for female breast cancer (BC) estimated at 55-72%, depending on the gene involved1. PVs in these major BC predisposing genes (BCPGs) have been implicated in a significant proportion of Hereditary Breast and/or Ovarian Cancer (HBC/HBOC) syndrome families which exhibit an autosomal dominant mode of inheritance1. Germline genetic testing identifies carriers of PVs in BRCA1 or BRCA2 who may benefit from cancer screening, prevention and therapeutic management strategies1,2. Other BCPGs implicated in HBC syndrome families include ATM, BARD1, CHEK2, PALB2, RAD51C and RAD51D3,4. Though CDH1, NF1, PTEN, STK11 and TP53 are involved in high-risk BC, these CPGs have been associated with cancer syndromes distinct from HBC and HBOC1,3,4. Though all these BCPGs have been included in gene-panel tests for BC risk assessment and management, BRCA1 and BRCA2 remain the major risk genes accounting for about one-third of all HBC families, with other BCPGs collectively accounting for less than 5% of such families3,5. With the increased use of gene-panel testing and sequencing-based screening, there have been reports of rare carriers of PVs in at least two BCPGs6–8. Although the implications are unknown for BC risk, phenotype or disease progression, these observations are intriguing given that the established BCPGs play critical roles in safeguarding the genome through DNA damage response repair9. Prompted by our earlier report of the co-occurrence of PVs in both PALB2 and BRCA1 in a familial BC case of French Canadian (FC) ancestry10, we used targeted-variant analysis in index carriers of PVs in BRCA1 or BRCA2 that were selected from high-risk BC families11,12. The FC of Quebec Canada exhibit a significant frequency of specific PVs in BRCA1, BRCA2, PALB2 and RAD51D due to genetic drift, a genetic architecture that has facilitated the interpretation of rare variants identified in candidate hereditary cancer risk genes applicable to other populations13–16. Here, we report the analyses of whole exome sequencing (WES) to investigate co-occurring rare germline variants in 276 DNA repair genes (DRGs)17, which include the established BCPGs and CPGs implicated in other hereditary cancers. Using a gene-based approach, we present predicted damaging variants based on selected, high-performing in silico tools and those clinically classified as PVs/LPVs based on the American College of Medical Genetics and Genomics (ACMG) guidelines18,19. Results Rare predicted damaging variants Rare variants in 276 DRGs17 were identified in WES data from 18 BRCA1 and 38 BRCA2 implicated BC (BRCA1/BRCA2) cases from HBC families using a previously established variant selection and annotation pipeline20 (Methods; Supplementary Tables 1-2). A total of 571 variants were identified with minor allele frequencies (MAFs) ≤ 0.01 or were not reported in The Genome Aggregation Database v.2.1.1. (gnomAD), where 485 had read depths ≥10 and variant allele frequencies (VAFs) ≥ 0.2. All PVs in the 56 BRCA1/BRCA2 cases were re-identified in our WES data and excluded from further analyses. From the remaining total of 377 variants that met our filtering criteria, 287 were unique predicted damaging variants by at least 1 out of 12 best performing in silico tools21 (Supplementary Table 3) and analyzed further. The gene-based approach generated 287 unique co-occurring rare germline variants, including newly detected variants in BRCA1/BRCA2, were identified in 55% (153/276) of DRGs involved in diverse pathways17: Base Excision Repair (BER), Fanconi Anemia (FA), Homologous Recombination (HR), Non-Homologous End Joining (NHEJ), and Mismatch Repair (MMR) (Supplementary Table 3). All 56 BRCA1/BRCA2 cases carried 2 to 13 (median 6) of these unique variants. All but 6 (2%; 6/287) variants were identified in a heterozygous state. About 30% (87/287) of variants had MAFs of ≤10-4 which are as rare as MAF of PVs in BRCA1/BRCA2 reported in the general population according to gnomAD, and about 10% (30/287) have not been reported in this genetic database. These 117 rare variants were identified in 34% (94/276) of DRGs, and were found in 94% (17/18) of BRCA1 and 92% (35/38) of BRCA2 cases. About 8% (24/287) of variants were predicted to incur loss-of-function (LoF): 11 frameshift, 9 stop-gain, 3 canonical splice-site (delta score ≥0.5 by SpliceAI22) and 1 start-lost variant. About 39% (7/18) of BRCA1 and 50% (19/38) of BRCA2 cases carried 1 to 2 of these 24 LoF variants. The remaining 92% (263/287) of variants were predicted to have a moderate effect: 226 missense, 25 non-canonical splice-site, 7 in-frame deletions and 5 synonymous variants. Among these 263 variants, 32 missense variants were predicted to be damaging by REVEL (score ≥0.7)23, and 3 non-canonical splice site and 1 synonymous variant were predicted to effect splicing by SpliceAI. One to 4 of these 36 predicted damaging variants were identified in 50% (9/18) of BRCA1 and 45% (17/38) of BRCA2 cases. Therefore, of the total of LoF and predicted damaging variants, 21% (60/287) were identified in 72% (13/18) of BRCA1 and 76% (29/38) of BRCA2 cases, where each carried 1 to 5 (median 2) of any these variants. About 7% (21/287) of unique co-occurring rare germline variants were identified in 9 of the total19 established BCPGs1 (Supplementary Table 1 and 3). These included variants in ATM, BARD1, BRCA1, BRCA2, CHEK2, PALB2, RAD51C, RAD51D and TP53, 9 of which had MAFs ≤10-4. They were identified in 33% (6/18) of BRCA1 and 39% (15/38) of BRCA2 cases. There were 3 LoF (ATM c.3802del; p.Val1268Ter, BRCA2 c.9976 A > T; p.Lys3326Ter and CHEK2 c.247del; p.Gln83LysfsTer54), 1 non-canonical splice-site and 17 missense variants. Three of these variants were predicted to be damaging: BRCA2 c.4585 G > A; p.Gly1529Arg, BARD1 c.1339 C > G; p.Leu447Val and CHEK2 c.667 C > T; p.Arg223Cys. Clinical classification of rare variants To infer the clinical consequences of carrying damaging variants in DRGs, the 287 unique variants were classified according to the ACMG guidelines18,19 (Methods; Supplementary Table 3). A total of 60 co-occurring rare germline variants of clinical interest 5 PVs, 10 likely pathogenic (LPVs) and 45 variants of uncertain significance (VUSs) were identified in 18% (51/276) of DRGs (Fig. 1). They were identified in 72% (13/18) of BRCA1 and 60% (23/38) of BRCA2 cases, where each case carried 1 to 4 (median of 2) of such variants. Many of these variants (93%; 56/60) had MAFs ≤10-4 or have not been reported in gnomAD.Fig. 1Oncoprint of predicted damaging variants in DNA repair pathway genes of clinical interest identified in BRCA1 and BRCA2 implicated index BC cases.Genes in bold are established breast cancer (BC) predisposing genes according to the National Comprehensive Cancer Network Clinical Practice in Oncology Guidelines (Version 3.2025)—Genetic/Familial High-Risk Assessment: Breast, Ovarian and Pancreatic1. Variants classified by The American College of Medical Genetics and Genomics (ACMG) guidelines18 (Methods; Supplementary Table 3) indicated with colour coded square (based on variant type) where black dot indicates pathogenic (PV) or likely pathogenic variant (LPV), and all other colour coded squares as variants of uncertain clinical significance (VUS). Variants in genes indicated with an asterisk (*) were identified in the heterozygous state, while ACMG guidelines classify variants in these genes as pathogenic when present in the homozygous state. Variants in genes clinical interest as a known breast cancer predisposing gene (BCPG) yet not involved in DNA pathway repair gene (DRG) were also included. To protect anonymity of participants approximate ages of diagnoses of cancers a provided to the nearest 10th-year age range at diagnoses. BER Base Excision Repair, BC Breast Cancer, DDR Direct DNA Repair, DRG DNA repair genes, FA Fanconi Anemia, HR Homologous Recombination, NER Nucleotide Excision Repair, NHEJ Non-homologous End Joining, MMR Mismatch Repair, Translesion Synthesis. Oncoprint of predicted damaging variants in DNA repair pathway genes of clinical interest identified in BRCA1 and BRCA2 implicated index BC cases. Genes in bold are established breast cancer (BC) predisposing genes according to the National Comprehensive Cancer Network Clinical Practice in Oncology Guidelines (Version 3.2025)—Genetic/Familial High-Risk Assessment: Breast, Ovarian and Pancreatic1. Variants classified by The American College of Medical Genetics and Genomics (ACMG) guidelines18 (Methods; Supplementary Table 3) indicated with colour coded square (based on variant type) where black dot indicates pathogenic (PV) or likely pathogenic variant (LPV), and all other colour coded squares as variants of uncertain clinical significance (VUS). Variants in genes indicated with an asterisk (*) were identified in the heterozygous state, while ACMG guidelines classify variants in these genes as pathogenic when present in the homozygous state. Variants in genes clinical interest as a known breast cancer predisposing gene (BCPG) yet not involved in DNA pathway repair gene (DRG) were also included. To protect anonymity of participants approximate ages of diagnoses of cancers a provided to the nearest 10th-year age range at diagnoses. BER Base Excision Repair, BC Breast Cancer, DDR Direct DNA Repair, DRG DNA repair genes, FA Fanconi Anemia, HR Homologous Recombination, NER Nucleotide Excision Repair, NHEJ Non-homologous End Joining, MMR Mismatch Repair, Translesion Synthesis. About 8% (5/60) of variants of clinical interest were identified in 5 different BCPGs. These include ATM c.3802del; p.Val1268Ter and CHEK2 c.247del; p.Gln83LysfsTer54 that conferred LoF and were classified as PV/LPVs and 3 missense variants in ATM, BARD1 or PALB2, which were classified as VUSs. These variants were identified in 6% (1/18) of BRCA1 and 11% (4/38) of BRCA2 cases. There were 16 other variants in 9 BCPGs that were classified as benign or likely benign involving BRCA1, BRCA2, RAD51C, RAD51D, CHEK2, PALB2, RAD51C, RAD51D and TP53. Among these variants, two were LoF: BRCA2 c.9976 A > T; p.Lys3326Ter and CHEK2 c.320-5 T > C, and two were missense BRCA2 c.4585 G > A; p.Gly1529Arg and CHEK2 c.667 C > T; p.Arg223Cys, although they were predicted to be damaging by REVEL. About 11% (7/60) of variants of clinical interest were identified in 5 established CPGs involved in other types of hereditary cancer. Three VUSs were identified in a heterozygous state in 4 BC cases: MLH1 c.415 C > G; p.Pro139Ala or MSH2 c.2118 C > G; p.Asp706Glu and POLE c.320 C > T; p.Ala107Val. Two PV and LPV: MUTYH c.1178 G > A; p. Gly393Asp, c.691 G > A; p.Val231Met and c.1178 G > A; p.Gly393Asp. Three BC cases were identified carrying a LPV c.2858 A > G; p.Lys953Arg or a VUS c.919 C > T; p.Pro307Ser in SMARCA4. None of carriers of these variants had a personal history of cancer or reported a family history of cancer at other sites predictive of the involvement of any of these CPGs (Supplementary Table 2). The remaining 80% (48/60) of variants of clinical interest were identified in 42 different DRGs that have not been associated with cancer risk. They were identified in 72% (13/18) of BRCA1 and 50% (19/38) of BRCA2 cases. They include 9 PVs/LPVs in ASCC3, ERCC6, EXO1, NEIL1, NSMCE2, POLG, POLI, RTEL1 and TDG that were identified in 11% (2/18) of BRCA1 and 21% (8/38) of BRCA2 cases, where there were 3 carriers of TDG c.1090+1 G > T. There was one homozygous carrier of XPA p.Met1Ter, which was classified as PV in the homozygous state. This result is interesting given that there is no indication that this carrier had Xeroderma Pigmentosum-Group A (XPA), an autosomal recessive condition with variable expressivity associated with homozygous carriers of PVs in XPA. Carriers of VUSs in ASCC3, EXO1 and RTEL1 were also identified as were 2 carriers each of VUSs in BLM, ERCC4, POLN, RTEL1 and TELO2. Our analysis of WES data also included a separate investigation of CDH1, NF1 and STK11 as PVs in these known BCPGs are routinely included in gene panel testing for BC or other cancers though they are not DRGs (Supplementary Table 1). Using our analysis pipeline, we identified one BRCA1 carrier of LPV STK11 p.Gly215Val and VUS STK11 p.Asp207Glu, and a BRCA2 carrier of the same VUS (Fig. 1, Supplementary Table 3). Both carriers reported a family history of colorectal cancers, which may be indicative cancer phenotypes in STK11 implicated Peutz–Jeghers Syndrome cancer families (Supplementary Table 2). Age of diagnoses of breast cancer in carriers of co-occurring variants in DRGs Although are our study is limited by sample size (Supplementary Table 2, Fig. 1), we have reviewed the ages of diagnoses of primary breast cancer and contralateral or among carriers of clinically relevant variants in BRCA1or BRCA2 and another DRG. Among the 16 BRCA1 carriers diagnosed with BC under age 50 years, 12 had co-occurring PV/LPV/VUS allele in another DRG; whereas among 2 BRCA1 carriers diagnosed with BC over/equal to age 50 years, 1 had at least one co-occurring PV/LPV/VUS allele in another DRG. (Fisher two-sided p = 0.49). Among the 30 BRCA2 carriers diagnosed with BC under age 50 years, 18 had co-occurring PV/LPV/VUS allele in another gene; whereas among 8 BRCA1 carriers diagnosed with BC over/equal to age 50 years, 5 had at least one co-occurring PV/LPV/VUS allele in another DRG. (Fisher two-sided p = 0.69). Although our data does not distinguish contralateral versus secondary BC disease (Supplementary Table 2; Fig. 1), we observed the following. Among the 18 BRCA1 carriers: 5/6 with contralateral/secondary disease had co-occurring PV/LPV/VUS allele in another DRG; and 8/12 with unilateral disease had co-occurring PV/LPV/VUS allele in another DRG (Fisher two-sided p = 0.44). Among the 38 BRCA2 carriers: 4/8 with contralateral/secondary disease had co-occurring PV/LPV/VUS allele in another gene; and 19/30 with unilateral disease had co-occurring PV/LPV/VUS allele in another DRG (Fisher two-sided p = 0.38). Discussion Our analyses of WES data from 56 BRCA1/BRCA2 BC cases from HBC families identified 287 unique variants predicted to be damaging or clinically relevant in 153 of 276 (55%) diverse DRGs using gene-based approach. These variants were identified in the other established BCPGs1: ATM, BARD1, CHEK2, PALB2, RAD51C, RAD51D and TP53, but not in PTEN. Variants of clinical interest were also identified in POLE, MLH1 and MSH2, CPGs involved in hereditary colorectal, endometrial and ovarian cancers24, and in SMARCA4, a CPG involved in hereditary small cell carcinoma of the ovary, hypercalcemic type and rhabdoid tumour25, though both personal and family history of cancer from index carriers would not necessarily predict involvement of these CPGs. Interestingly, heterozygous carriers of PV/LPVs in MUTYH were identified, a CPG involved in the autosomal recessive form of hereditary colorectal cancer24. Though monoallelic MUTYH variants have been reported in BC cases4,24,26 where one study estimated a 1.4-fold increased risk for BC among carriers27 the findings are not equivocal8,24,28,29. A separate analyses of CDH1, NF1 and STK11, also known BCPGs that are not known to be involved in DNA repair pathways, also identified BRCA1/BRCA2 carriers of clinically relevant variants in STK11. Both carriers reported a modest family history of colorectal cancers though not suggestive of Peutz–Jeghers Syndrome1. Our findings suggest that BC cases from high-risk BC families harbouring PVs in BRCA1 or BRCA2 may also carry clinically relevant variants in other established CPGs. With increasing use of gene panel testing are reports of rare cancer cases harbouring PVs in more than one clinically relevant CPG. Multi-locus inherited neoplasia syndrome (MINAS) refers to the co-occurrence of PVs in different CPGs6. Recently, the mean age at BC diagnosis in co-occurring BRCA1/BRCA2 PV carriers was found similar or moderately younger than carriers with a BRCA1 PV alone though findings were limited due to sample size7. As with our findings, co-occurring PVs in BRCA1/BRCA2 and other established BCPG6,7 such as ATM or CHEK2 were observed though our results due not indicate a younger age of BC diagnoses in co-occurring carriers of clinically relevant variants in BRCA1 or BRCA2 and DRGs. A recent analysis of the BRIDGES dataset for variants of interest in 11 BCPGs among 55,815 BC cases, reported that the co-occurrence of PVs in BRCA1, BRCA2 and PALB2 was less frequent than expected compared to the other BCPGs analyzed30. The authors suggested that the co-occurrence of VUSs in BRCA1, BRCA2 or PALB2 with PVs in other BCPGs may provide supportive evidence against pathogenicity of those VUSs30. Our study also identified co-occurring rare germline variants of interest in DRGs that are not established CPGs, such as PV/LPVs in ASCC3, ERCC6, EXO1, NEIL1, NSMCE2, POLG, POLI, RTEL1, TDG and XPA based on ACMG guidelines19,31. However, genetic studies have implicated ERCC6, EXO1, POLG, RTEL1 and XPA in hereditary BC cases negative for BRCA1/BRCA2 PVs. Predicted damaging rare missense variants in ERCC6 have been reported in BC cases of non-Northwestern European populations using WES32–35. Rare LoF PVs in EXO1 has been identified in WES analyses of Finnish early-onset BC case36, and Non-Finnish European HBC families37,38. Interestingly, genome wide association studies (GWAS) analyses of common single nucleotide polymorphic (SNPs) variants in ERCC639 and EXO140,41 were associated with BC cases versus controls. A predicted damaging rare missense variant in POLG was reported as two times more frequent in BC cases as compared to controls in a study of DNA damage response genes in Finnish BC cases42. Likely deleterious missense variants in RTEL1 were found significantly associated with BC when compared to controls43. In contrast, a meta-analysis revealed a SNP in XPA was least likely associated with BC in both non-Asians and population-based controls44. In our study, although the identification of a homozygous carrier of PV XPA could be indicative of XP, we are not aware of this phenotype in our index carrier. While XP is considered an autosomal recessive disorder with 100% penetrance due to the underlying causal genotypes in XPA, the specific clinical presentation varies among patients, whereby some variants allow for some residual DNA repair activity (hypomorphic variant) often with less severe clinical features compared to those with a total loss. Without any molecular assays of this start-loss variant from this index case, we cannot be certain if the c.3 G > T; p.Met1Ter variant shared by both parental chromosome alleles is completely silenced by this variant and/or alternative isoforms are expressed to compensate for effects as suggested by a recent study of XPA germline variants45. Our findings suggest that the frequency of BC carriers with clinically relevant co-occurring variants in BRCA1/BRCA2 and other BCPGs may be higher than expected. One large population-based study reported that 5.03% of BC cases versus 1.63% of controls carry PV in any of 12 established BCPGs though this carrier frequency may be higher in ancestrally defined populations3. Though reports of recurrent, founder PVs in BRCA1, BRCA2, CHEK2, PALB2 and TP53 in FC BC cases not selected for family history of BC revealed carrier frequencies ranging from 0.3% to 6.4% versus <1% of controls depending on the BCPGs investigated15,46,47, we are only aware of our report of a single familial BC case harbouring PVs in both BRCA1 and PALB2 from a FC family10. With respect to other DRGs investigated in our study, it is interesting that our group also reported different PVs in EXO1 and NEIL1 in WES data from familial ovarian cancer cases of FC ancestry not accounted for by BRCA1/BRCA220. Establishing co-occurring allele frequencies in the FC population would be important to determine clinical relevance, especially given the founder effects exhibited in this population48,49. An important consideration in the interpretation of the variant burden in our study cohort is the potential influence of Berkson’s paradox, so-called collider bias50. As our BC cases were selected based on having a strong family history of BC diagnosed in close relatives in the same familial branch, there exists an inherent selection bias favouring individuals who are more likely to carry additional rare germline variants in other genes. This non-random sampling may artificially inflate the observed co-occurrence of BRCA1/BRCA2 PVs with other clinically relevant variants in DRGs that were identified in our study, potentially leading to a spurious inverse or exaggerated association between these variants and other genetic or family history features. Recognizing this phenomenon is critical for contextualizing variant enrichment and highlights the need for cautious interpretation of genetic interactions in highly selected hereditary cancer study groups. Furthermore, though not available, determining the segregation of clinically relevant DRGs in families of our BRCA1 or BRCA2 carriers might have provided insight into association with other cancer phenotypes reported in their families. It is important to highlight that this study was not designed to estimate penetrance or BC risk attributable to the identified co-occurring rare germline variants, nor to establish gene– or variant–disease associations for BC predisposition. This study, rather, illustrates the challenges in interpreting such variants in the context of BC index cases carrying BRCA1/BRCA2 PV/LPVs (Fig. 1; Supplementary Table 3). We, therefore, applied an integrated gene-based approach, combining clinical classification based on ACMG guidelines18,19 with computational prediction using selected, high-performing in silico tools (including predicted effects on splicing or protein function) to systematically identify and categorize co-occurring rare variants across genes with heterogeneous levels of evidence for BC predisposition. This challenge has clear implications for medical genetics, where multigene testing increasingly returns multiple rare findings per individual8,30, and where clinical practice requires careful separation of variant-level classification, gene-level validity for BC predisposition, and the evidence-basis for disease-specific risk attribution. The same observations are relevant to research, where co-occurring rare variants across DNA repair pathways may be informative for characterizing the phenotypic consequences of MINAS6,51, including potential effects on cancer initiation and progression, yet require systematic evidence integration supported by segregation, functional, and independent genetic association analyses before attributing BC predisposition relevance. In this study, the complexity of such variant interpretation is illustrated by recurrent scenarios observed in our study group. First, variants with well-established pathogenicity in one inheritance context may be encountered in another, such as monoallelic MUTYH variants, for which pathogenicity is classically defined in the biallelic setting (MUTYH-associated polyposis), whereas BC predisposition for monoallelic carriers remains incompletely resolved24. Second, we identified variants of potential clinical interest in DNA repair and genome maintenance genes that are not established BC predisposition genes, including variants classified as pathogenic/likely pathogenic or predicted damaging in ASCC3, ERCC6, EXO1, NEIL1, NSMCE2, POLG, POLI, RTEL1, and TDG; these variants are presented with reference to prior reports in BC cohorts without implying causality, BC risk association, or penetrance for these loci. Third, even within canonical CPGs, certain variants illustrate that mutational consequence and clinical classification do not map cleanly onto high-penetrance risk. BRCA2 p.Lys3326Ter, for example, has been associated with low increase in BC risk in epidemiological studies52. Other functional studies provide supporting biological evidence that mouse embryonic stem cell–based assay demonstrated that cells expressing p.Lys3326Ter exhibit hypersensitivity to cisplatin while still complementing the lethal phenotype caused by loss of endogenous Brca2, suggesting that the C-terminal truncation may partially impair DNA repair function without completely abolishing BRCA2 activity53. Collectively, these findings emphasize that co-occurring rare germline variants spanning established BCPGs, other cancer risk genes, and DNA repair genes without established BC predisposition create a spectrum of interpretive challenges in both clinical and research settings, highlighting the need for structured evidence-based integration and appropriately designed follow-up studies. Questions remain concerning the biological significance of carrying more than one PV in established BCPGs. Studies of heterozygous carriers with co-occurring PVs in BRCA1/BRCA2 suggested that BC tumours most likely develop from the loss of the functioning allele but not both implying that the biological effects at the somatic level are additive rather than synergistic6. BC is a heterogeneous disease with a high degree of somatic molecular genetic diversity both between and within tumours which together with the heritable germline landscape likely affects disease progression and sensitivity to chemo- and radiotherapy beyond the impact of carrying a CPG risk allele54. With respect to our findings, several reports have suggested that variants in or near ERCC6, EXO1, NSMCE2, and POLG may be clinically relevant to BC. ERCC6 SNPs may modify occupational radiation exposure and BC risk55. Increased EXO1 expression has been associated with breast carcinogenesis and poor prognosis56. A newly developed Capture Hi-C assay identified protein-coding genes including NSMCE2 as potential targets of GWAS identified loci associated with BC risk57. Interestingly, NSMCE2, a super-enhancer regulated gene, when expressed at high levels was linked to poor prognosis and therapy resistance in BC58. Similarly, SNPs in POLG have been associated with tumour phenotype and disease outcome in BC patients59. Notably, the participants in our study group could have benefited from poly(ADP-ribose) polymerase (PARP) inhibitor therapy since they carry PVs in BRCA1/BRCA2 that affect the HR pathway. The clinical implications for participants who also carried variants in DRGs involved in other pathways is unknown. An investigation of an ovarian cancer cell line model system suggested that responses to PARP inhibitors may be due to deficiencies in DRGs in defined pathways such as HR and either NER or MMR60. Further research is needed on the vulnerabilities of BC tumours arising in carriers of co-occurring PVs affecting functionally diverse DRGs. In conclusion, our study was not designed to estimate penetrance or risk to BC. We investigated 56 index BC cases carrying BRCA1/BRCA2 PVs from high-risk HBC families with shared ancestry for co-occurring, rare germline variants across 276 DRGs using gene-based approach. We identified 21 unique variants predicted to be damaging by a curated set of high performing in silico tools in 9 DRGs that are established BCPGs, found in 33% and 39% of BRCA1 and BRCA2 cases, respectively. In addition, we identified 60 unique clinically classified variants (15 PVs/LPVs and 45 VUSs) based on the ACMG guidelines in 51/276 DRGs, found in 72% and 60% of BRCA1 and BRCA2 cases, respectively. Overall, the frequency of these carriers may be higher than expected and warrants further investigation of other populations at high risk for hereditary BC, including segregation with cancer phenotypes within high-risk cancer families and tumour characteristics. The number of predicted damaging and clinically relevant variants identified in different DRGs raises questions concerning the interpretability of findings as panel testing expands beyond established BCPGs. Methods Study subjects The 56 index BC cases for WES analyses were selected from 56 high-risk BC families that have been extensively characterized by our group and previously reported to carry PVs in BRCA1 (n = 18) or BRCA2 (n = 38)11,12 (Supplementary Table 2). Briefly, index cases were diagnosed with BC < 66 years of age and had a family history of at least two confirmed cases of invasive BC diagnosed <66 years of age in first-, second-, or third-degree relatives in the same familial branch by review of pathology reports or death certificates; and reported FC of Quebec ancestry for all four grandparents as described previously11,12. The ages of diagnosis of BC and personal history of contralateral or secondary BC or other cancer types, and reported family history of other cancer types in 10-year intervals to protect anonymity of participants are described in Supplementary Table 2. Peripheral blood lymphocyte (PBL) DNA samples, ages of diagnoses of personal and family history of cancer from index cases were available from an established biobank resource following ethical board approval from the McGill University Health Centre Research Ethics Board (MP-37-2019-4783): the Banque de tissus et données of the Réseau de recherche sur le cancer of the Fond de recherche du Québec – Santé (RRCancer biobank) (rrcancer.ca/) where participants had been previously recruited with written informed consent to participate in this biobank resource. The project was conducted at the Research Institute of the McGill University Health Centre and approved by the McGill University Health Centre Research Ethics Board (MP-37-2019-4783), in accordance with its guidelines and the Declaration of Helsinki. WES analysis of DNA repair genes WES was performed using PBL DNA from the 56 index BC cases at the McGill Genome Centre and rare germline variants were identified using a previously described variant selection and annotation pipeline20. In brief, for primary and secondary analysis, NimbleGen SeqCap® EZ Exome v3.0 library kit (Roche, US), followed by paired-end sequencing on different Illumina HiSeq platforms was performed. Reads were aligned to the human reference genome assembly GRCh37/hg19 using Burrows-Wheeler aligner v0.7.17, followed by PCR deduplication using Picard v2.9.0. Realignment around small insertions and deletions was performed, and germline variants were called using HaplotypeCaller using Genome Analysis Toolkit (GATK) v3.5. Variants were then filtered for base sequencing quality score ≥30 and annotated using Ensembl Variant Effect Predictor (VEP) and GEMINI v0.19.1. Sequencing data for all 276 genes met our quality control criteria and therefore were included in the study. For tertiary analysis, the list of annotated variants were filtered for those: (1) occurring in protein coding and splice-site regions of 276 DRGs17 including BRCA1/BRCA2 (Supplementary Table 1); (2) having a MAF ≤ 0.01 in the general population based or not reported in gnomAD (gnomad.broadinstitute.org/); and (3) having a read depth ≥10 and variant allele frequency (VAF) ≥ 0.2. The resulting list of 470 unique variants, excluding the 15 known PVs in the 56 BRCA1 or BRCA2 cases, were annotated using 9 selected in silico tools according to their predictions for being damaging: (Combined Annotation Dependent Depletion v1.4 (CADD [Phred score ≥20]), Eigen (score ≥0.0), Meta-analytic Logistic Regression v4.2 (MetaLR [score ≥0.5]), Meta-Analytic Support Vector Machine v4.2 (MetaSVM [score ≥0.0]), MetaRNN 4.2 (score ≥0.5), Rare Exome Variant Ensemble Learner v4.2 (REVEL [score ≥0.7)), Variant Effect Scoring Test v4.2 (VEST [score ≥0.5]), Protein Variation Effect Analyzer v4.0 (PROVEAN v4.0 [score ≤−2.5]) and AlphaMissense (score ≥0.6)61 or four selected tools for affecting splicing (Maximum Entropy Estimates of Splice Junction v2.0 (MaxEntScan [score≥0.5]), Database Splicing Consensus Single Nucleotide Variant (dbscSNV): AdaBoost v4.0 (ADA [score ≥0.5]) and Random Forest v4.0 (RF [score≥0.5]) and SpliceAI (score ≥0.5) using in silico tools that were selected for their best performance as previously described21. The final list of 287 unique variants, which were predicted to be damaging by at least one of these in silico tools, were classified as PV, LPV, VUS, benign or likely benign based on the ACMG18 guidelines as previously described20. This list of 287 variants was verified manually using the Integrative Genomics Viewer (IGV) v2.8. (igv.org). To capture clinically relevant variants in all known BCPGs, a separate analysis of WES data for CDH1, NF1 and STK11, known non-DRGs, was performed using the same pipelines of analyses described for the analyses of DRGs. Ethics approval This study was approved by the Research Ethics Board of the Research Centre of the McGill University Health Centre, Montreal, Canada (MP-37-2019-4783). Supplementary information Supplementary_njp_V10_06032026 Supplementary_njp_V10_06032026