Comparative analysis of distinct genomic landscapes in young-onset gBRCA1/2 breast cancer

Approximately 12% of young-onset breast cancer patients (first diagnosis at age ≤40 years) and 30% of very-young-onset breast cancer patients (first diagnosis at age ≤30 years) carry germline BRCA1 or BRCA2 pathogenic variants (gBRCA1/2 PV) (1–4). gBRCA1/2 PV carriers have an elevated lifetime breast cancer risk of 60%–80% (5–9). The Prospective Study of Outcomes in Sporadic versus Hereditary Breast Cancer (POSH) is a prospective cohort study investigating differences in outcomes between hereditary (gBRCA1/2-associated) and sporadic young-onset breast cancer (10). Enrollment of more than 2,700 women diagnosed with breast cancer at age ≤40 (<50 if a known gBRCA1/2 PV carrier) was performed in the United Kingdom between 2000 and 2008. The study demonstrated a 5-year survival advantage but no survival difference by year 8 after diagnosis for women with estrogen receptor–positive (ER-positive) breast cancer compared with those with ER-negative disease and similar overall survival (OS) between young-onset female breast cancer patients with and without gBRCA1/2 PVs (3, 11). Genomic features such as allele-specific loss of heterozygosity (asLOH) were not considered in the primary analysis.

BRCA1 and BRCA2 are critical for repairing double-stranded DNA breaks through homologous recombination, and complete loss of BRCA1/2 function through asLOH is associated with homologous recombination deficiency (HRD) (12–14). Although both proteins function in the same pathway, some broad differences in somatic features between gBRCA1 and gBRCA2 tumors have been reported. gBRCA1 tumors are mostly ER negative, whereas gBRCA2 tumors are mostly ER positive (15, 16). Compared with gBRCA2 tumors, gBRCA1 tumors harbor high frequencies of TP53 PVs and greater amounts and distinct patterns of copy number variants (CNVs) (17–22). Losses of chromosome 2q (chr2q), chr4p, chr4q, chr5q, and chr12q are frequent in gBRCA1-associated tumors, whereas losses of chr6q and chr13q occur in gBRCA2-associated tumors (21–23). Furthermore, gBRCA1-associated tumors exhibit greater activation of genes involved in DNA repair than gBRCA2-associated tumors (24). However, the full landscape of differences between gBRCA1 and gBRCA2 tumors remains incomplete, as previous studies have been limited by low numbers of carriers, with the largest having only 64 gBRCA1/2 carriers, and by the use of older comparative genome hybridization technologies (17, 21). It is still unknown whether there are any molecular differences between gBRCA1 and gBRCA2, other than ER status, that have potential therapeutic utility.

gBRCA1/2-associated tumors are sensitive to poly(ADP-ribose) polymerase inhibitors (PARPi) in both metastatic and high-risk early breast cancers (25–29). Adjuvant PARPi significantly improves OS in early-stage, high-risk disease (30) and improves progression-free survival in metastatic disease (27, 28). In ER-positive, HER2-negative tumors, adjuvant cyclin-dependent kinase 4/6 inhibitors (CDK4/6i) combined with endocrine therapy improve invasive disease–free survival in early-stage disease, and significantly increase OS in metastatic disease, compared with endocrine therapy alone (31–34). Thus, both PARPi and CDK4/6i confer clinically meaningful benefit in early and metastatic breast cancer.

Despite the efficacy of both PARPi and CDK4/6i, primary and secondary resistance to each therapy has been observed (28, 30, 35–40). PARPi resistance most commonly arises through reversion mutations of gBRCA1/2 (41). Absence of asLOH (37) also may confer primary resistance to PARPi in gBRCA1/2 PV carriers (14, 42). We previously reported differences in asLOH frequency between gBRCA1 and gBRCA2 breast cancers from PV carriers unselected for age (90% vs. 54%); in the current study, all patients were under age 50 (92.6% diagnosed ≤40) (14). CDK4/6i impede cell proliferation by inhibiting the phosphorylation of the retinoblastoma protein (Rb) during the cell cycle (43, 44). Genomic alterations, such as RB1 loss and MYC amplification, have been implicated in resistance to CDK4/6i (40, 45).

In young-onset pretreatment gBRCA1/2-associated breast cancer, the frequency of asLOH and its impact on survival remain unclear, as does the landscape of genomic alterations associated with response to CDK4/6i. An improved understanding of the distinct genomic landscapes of treatment-naive gBRCA1 and gBRCA2 PV–associated breast cancer may inform current treatment selection, particularly in ER-positive, HER2-negative tumors for which both PARPi and CDK4/6i are treatment options, and may inform strategies for possible future therapy development. Thus, we performed comprehensive genomic profiling of treatment-naive, primary non-metastatic breast tumors using whole-exome sequencing (WES) of matched tumor-germline DNA in 136 breast cancer patients with gBRCA1/2 PVs from the POSH cohort.

Cohort description. We evaluated WES results of 136 treatment-naive matched tumor-germline samples from gBRCA1/2-positive women diagnosed with young-onset breast cancer participating in the POSH study (Figure 1, Supplemental Figure 1, and Supplemental Tables 1 and 2; supplemental material available online with this article; https://doi.org/10.1172/jci.insight.203005DS1). Eighty-six (63.2%) and 50 (36.8%) women had gBRCA1 and gBRCA2 germline PVs, respectively. The median age at diagnosis was 36 years, and 92.6% of women were age ≤40 (Table 1). Overall, 65 women (47.8%) had ER-positive tumors. Demographic, tumor pathology, and treatment data are shown in Table 1.

Figure 1

CONSORT diagram of POSH study samples.

Table 1

Characteristics of study participants

Molecular features in breast cancers. We evaluated the genomic landscapes of the breast cancers, including asLOH, HRD, tumor mutational burden (TMB), and single-base substitution (SBS). Tumors had high rates of asLOH when stratified by gene (93% for gBRCA1 and 96% for gBRCA2) and by ER status (94.4% for ER negative and 93.8% for ER positive) (Supplemental Table 3). HRD scores were significantly higher in tumors with asLOH compared with tumors without asLOH (nonLOH) in both gBRCA1 (57.4 ± 1.3 vs. 22.6 ± 6.1, P < 0.0001) and gBRCA2 (43.7 ± 1.5 vs. 23.5 ± 6.5, P = 0.005) PV carriers and when grouped by ER status (ER-negative tumors: 57.6 ± 1.3 vs. 22.8 ± 9.2, P < 0.0001; ER-positive tumors: 46.4 ± 1.6 vs. 22.8 ± 4.1, P < 0.001) (Figure 2A and Supplemental Table 3). gBRCA1 tumors with asLOH had significantly higher HRD scores than gBRCA2 tumors with asLOH (57.4 ± 1.3 vs. 43.7 ± 1.5, P < 0.0001) (Figure 2A).

Figure 2

Molecular features of breast cancer in germline BRCA1 and BRCA2 pathogenic variant carriers with overall survival. (A) Homologous recombination deficiency (HRD) scores in tumors with allele-specific loss of heterozygosity (asLOH) compared with those without, stratified by germline variant gene and ER status using 2-sided t tests. Dotted black line is HRD score = 42. (B) Median of the proportion of single-base substitution signature 3 (SBS3), SBS1, SBS26, and SBS18 in tumors with asLOH by gene and in tumors with asLOH and those without, compared using Mann-Whitney non-parametric tests. Error bars are interquartile range (IQR). HR, homologous recombination; MMR, DNA mismatch repair. (C) Median tumor mutational burden (TMB) by asLOH status, germline gene variant, and ER status, compared using Kruskal-Wallis and Mann-Whitney non-parametric tests. Error bars are IQR. (D) Overall survival by asLOH status, evaluated using the Kaplan-Meier estimator and log-rank test. *P < 0.05, **P < 0.01, ***P < 0.001, ****P <.0001.

SBS3, the signature associated with gBRCA1/2 PVs (37), was the most abundant signature in all tumors with an average contribution of 32.8% (Supplemental Figure 2). SBS1 (aging-associated), SBS26 (mismatch repair deficiency–associated), and SBS18 (reactive oxygen species–associated) were the next most abundant signatures (37). gBRCA1 tumors had significantly higher median percentage proportional contributions of SBS1 (12.9 vs. 7.3, P = 0.013) and SBS18 (1.4 vs. 0, P = 0.007), and significantly lower proportions of SBS3 (27.3 vs. 42.6, P = 0.002) and SBS26 (5.9 vs. 9.4, P = 0.049), than gBRCA2 tumors. The median percentage proportional contributions of SBS3 and SBS18 were significantly higher in tumors with asLOH than in nonLOH tumors (SBS3: 38.5 vs. 15.9, P = 0.023; SBS18: 0 vs 0, P = 0.023) (Figure 2B and Supplemental Table 4).

Tumors with gBRCA1 asLOH had higher median TMB than those with gBRCA2 asLOH (3.4, IQR 1.6–11.2, vs. 1.5, IQR 0.9–3.4, P = 0.001). ER-positive tumors with asLOH had significantly lower median TMB than ER-negative tumors with asLOH (1.7, IQR 0.9–3.9, vs. 3.4, IQR 1.6–11.0, P = 0.008) (Figure 2C). However, there was no difference in median TMB between tumors with and without asLOH. Tumors with TMB ≥ 10 have been found to be responsive to treatment with the anti–PD-1 monoclonal antibody pembrolizumab (46, 47). In our cohort, slightly more tumors with asLOH (24 of 128 [18.8%]) had TMB ≥ 10 than those without asLOH (1 of 8 [12%]), but it was not notable.

We evaluated breast cancer OS by tumor molecular features. Survival in women with nonLOH tumors was 100% throughout the 8.2-year median follow-up period but not statistically significantly different from that of women with tumors with asLOH (Figure 2D). Favorable survival for women with nonLOH tumors remained when the analysis was stratified by germline variant gene or by ER status (Supplemental Figure 3). The clinical characteristics of the nonLOH tumors are shown in Supplemental Table 5. Although not significantly so, women with HRD-high (≥42) tumors and those with SBS26 tended to do worse (Supplemental Figure 4, A and B). Women with tumors with SBS3 had significantly lower OS than those without (HR 4.46, P = 0.033). Similarly, when stratified by gene and ER status, women with SBS3 tumors had a decreased OS, albeit not to a significant degree (Supplemental Figure 4, C–G).

Copy number variation in tumors. WES enabled copy number analysis, which revealed chromosome arm chr1q and chr8q gains, both prevalent in breast cancers (48), in all tumors (Figure 3A). Eighty-eight and 108 tumors had single-copy loss of RB1 and TP53, respectively. Copy loss of RB1 and BRCA2, both located on chromosome 13, occurred on the same segment significantly more frequently than expected by chance (56/88, P = 0.010), whereas copy loss of TP53 and BRCA1, both on chromosome 17, occurred on the same segment significantly less frequently than expected by chance (22/108, P < 0.001) (Figure 3A). Thus, it appears that RB1 and BRCA2 loss are non-independent, whereas copy number loss of TP53 and BRCA1 are independent of each other. The loss of RB1 (RB1 asLOH) occurred significantly more frequently in gBRCA2 asLOH tumors than in gBRCA1 asLOH tumors; however, there was no difference in the frequency of RB1 PVs or RB1 biallelic loss. Conversely, TP53 asLOH was not different between gBRCA1 asLOH and gBRCA2 asLOH tumors, although gBRCA1 asLOH tumors had significantly more TP53 PVs and TP53 biallelic loss than gBRCA2 asLOH tumors (Figure 3A).

Figure 3

Copy number variation in tumor samples. (A) Landscape of copy number variants in tumors. Inset shows zoomed-in images of chromosomes 13 and 17 and the quantified values of times that RB1 shares a copy number loss segment with BRCA2 and TP53 shares one with BRCA1, evaluated using a χ² test. Vertical bar graph shows the frequency of RB1 and TP53 pathogenic variants (PVs), asLOH, and biallelic loss in BRCA1 and BRCA2 tumors with asLOH, evaluated using Firth logistic regression. *P < 0.05, ***P < 0.001. (B) Enrichment of chromosome arm gains and losses in tumors with BRCA1 asLOH compared with those with BRCA2 asLOH, adjusted for ER status. Enrichment was performed using Firth logistic regression. Dashed line is –log₁₀(P value) = 1.3.

We observed differences in chromosome arm copy number between tumors with gBRCA1 asLOH and gBRCA2 asLOH (Supplemental Figure 5). After adjustment for ER status (49), chr6q, chr19p, chr2q, chr7q, and chr10p gains and chr5q copy number were enriched significantly in tumors with gBRCA1 asLOH compared with those with gBRCA2 asLOH (Figure 3B).

Enrichment of gene-level alterations. To further define the complete genomic landscape of the tumors, we assessed gene-level somatic alterations (pathogenic single-nucleotide variants/indels and CNVs) in 598 cancer-associated genes (Supplemental Table 2). We found 4,146 pathogenic single-nucleotide variants (SNVs)/indels and 4,205 pathogenic CNVs in 572/598 genes (Supplemental Table 6). Pathogenic SNVs/indels in the 50 most commonly altered genes are shown in Supplemental Figure 6.

When we compared pathogenic SNVs/indels and CNVs in tumors from gBRCA1 PV carriers with those from gBRCA2 PV carriers, with adjustment for ER status, we observed a larger number of somatic alterations significantly enriched in gBRCA1 tumors (Figure 4A and Supplemental Table 7). Pathway analysis revealed enrichment of pathogenic SNVs/indels/CNVs in genes in multiple hallmark pathways (50), including the reactive oxygen species (ROS) pathway, DNA repair, and epithelial-mesenchymal transition, in gBRCA1 tumors compared with gBRCA2 tumors (Figure 4A).

Figure 4

Differential enrichment of somatic variants, inclusive of pathogenic single-nucleotide variants, indels, and copy number alterations, and hallmark pathway enrichment in breast cancers. Enrichment of somatic variants and enrichment of Hallmark Pathway alterations by germline variant gene adjusted for ER status (A); by ER status adjusted for germline variant gene (B); by HRD (high ≥42) adjusted for asLOH (C); and by asLOH stratified by germline gene variant (D). Enrichment analysis was performed using Firth logistic regression. Dashed line is –log₁₀(P value) = 1.3.

With adjustment for gBRCA1/2 PV carrier status, ER-positive tumors had greater enrichment of pathogenic somatic SNVs/indels/CNVs compared with ER-negative tumors (Figure 4B and Supplemental Table 8) and were enriched for alterations in genes in the coagulation and mTORC1 signaling hallmark pathways. ER-negative tumors were enriched for alterations in the interferon-γ response pathway (Figure 4B). The total number of genes with SNVs/indels/CNVs and with SNVs/indels was higher in ER-negative tumors compared with ER-positive tumors when the analysis was performed without adjustment for gBRCA1/2 PV status, as seen in previous studies (Supplemental Figure 7) (51).

More genes were enriched in SNVs/indels and CNVs in HRD-high tumors compared with HRD-low tumors, adjusted for LOH status (Figure 4C and Supplemental Table 9). Genes in the E2F targets and WNT/β-catenin signaling pathways were the most highly enriched in HRD-high tumors (Figure 4C). Tumors without asLOH were enriched for alterations in the oxidative phosphorylation pathway compared with those with asLOH (Figure 4D and Supplemental Table 10).

Presence of CDK4/6 resistance alterations. Genetic alterations implicated in acquired and intrinsic resistance include alterations in RB1, TP53, AURKA, MYC, CCNE1, SPEN, FAT1, ARID1A, PTEN, FGFR1, and EGFR (40, 45, 52–58). We evaluated the presence of somatic alterations (SNVs/indels; and CNVs, including single-copy loss) implicated in resistance to CDK4/6i in ER-positive, HER2-negative tumors within the POSH cohort, and we found that all tumors contained at least one alteration implicated in resistance (Figure 5) (54). Genes implicated in CDK4/6i resistance exhibited various patterns of alteration in gBRCA1 and gBRCA2 PV carriers in ER-positive, HER2-negative tumors from the POSH cohort (Figure 5). Alterations in RB1, primarily consisting of single-copy loss (asLOH), were more frequent in gBRCA2 tumors (33/38, 86.8%) than in gBRCA1 tumors (8/12, 66.7%). Additionally, 4 tumors in gBRCA2 PV carriers had frameshift PVs in RB1. PVs in TP53 were more frequent in gBRCA1 tumors (6/12, 50%) than in gBRCA2 tumors (5/38, 13%), although the overall prevalence of TP53 alterations was similarly high in both groups (83.3% in gBRCA1 tumors and 86.8% in gBRCA2 tumors) when single-copy-number loss was considered. Other CDK4/6i resistance–associated alterations observed in gBRCA1 and gBRCA2 tumors were in SPEN (gBRCA1: 50%; gBRCA2: 52.6%), AURKA (gBRCA1: 25%; gBRCA2: 26.3%), ARID1A (gBRCA1: 33.3%; gBRCA2: 50.0%), FAT1 (gBRCA1: 83.3%; gBRCA2: 55.3%), PTEN (gBRCA1: 50.0%; gBRCA2: 34.2%), MYC (gBRCA1: 58.3%; gBRCA2: 47.4%), EGFR (gBRCA1: 8.3%; gBRCA2: 2.6%), and CCNE1 (gBRCA1: 8.3%; gBRCA2: 0%).

Figure 5

Enrichment of somatic variants associated with resistance to CDK4/6 inhibitors in ER-positive, HER2-negative tumors from germline BRCA1/2 carriers in POSH compared with noncarriers from TCGA. Oncoprint of somatic variants in tumors, frequency of somatic variants, and enrichment of variants in POSH versus TCGA samples. Enrichment analysis was performed using Firth logistic regression. Dashed line is –log₁₀(P value) = 1.3; q is the false discovery rate–adjusted P value.

To determine whether gBRCA1/2 carriers were more likely than noncarriers to have somatic alterations in CDK4/6i resistance–associated genes, we compared rates of alteration between carriers from POSH and noncarriers from The Cancer Genome Atlas (TCGA) (Supplemental Figure 8), all with ER-positive, HER2-negative tumors. To mitigate any potential bias in sequencing coverage across the 2 cohorts, we limited our analysis to genes with at least 1 alteration detected in the TCGA cohort (Supplemental Table 8). We found statistically significant enrichment of alterations in several genes in tumors from the POSH cohort compared with those from TCGA, including RB1 (OR 6.3, 95% CI 2.8–15.4, adjusted P value [P_adj] = 0.001), TP53 (OR 4.6, 95% CI 1.9–12.1, P_adj = 0.017), FAT1 (OR 3.9, 95% CI 1.84–8.7, P_adj = 0.013), and MYC (OR 4.0, 95% CI 1.8–9.1, P_adj = 0.017), all of which are associated with resistance to CDK4/6i (Figure 5).

We performed WES on treatment-naive matched tumor germlines from 136 young gBRCA1/2 PV carriers from the POSH cohort and found high levels of asLOH, over 90%, in all tumors regardless of germline variant gene and tumor ER status. No deaths were seen in women with a nonLOH breast cancer over the 8.2-year follow-up period; however, these numbers were small, and survival differences with asLOH tumors were not statistically significant. Presence of SBS3 was associated with a significantly lower OS, though numbers were small. We observed numerous significant differences between the genomic landscape of gBRCA1 and gBRCA2 PV–associated breast cancers, with implications for future therapeutic selection. When evaluating ER-positive, HER2-negative treatment-naive breast cancers, we found significant enrichment of genetic changes associated with CDK4/6i resistance mechanisms.

Primary analysis of POSH found no difference in breast cancer OS in gBRCA1/2 PV carriers compared with noncarriers. Other studies also have found similar breast cancer OS comparing gBRCA1/2 PV carriers and noncarriers following treatment with anthracyclines and taxane chemotherapy, although improved pathological complete response rates to chemotherapy in gBRCA1 PV carriers are seen (59, 60). When BRCA1/2 asLOH status was considered, breast cancer OS for nonLOH tumors was better than but not significantly different from OS for asLOH tumors; OS for asLOH and nonLOH tumors was significantly better than OS for nonBRCA1/2 tumors (14). In ovarian cancer, asLOH tumors had significantly better OS than nonLOH tumors and non-BRCA1/2 tumors after platinum treatment (14). Our findings are suggestive of better outcomes for women with nonLOH tumors but are limited by their low sample count and are not statistically significant. We also observed significantly improved OS in tumors without SBS3. It is possible that tumors with asLOH and SBS3 have genomic instability that promotes progression despite initial responses to chemotherapy. Larger studies are needed to clarify whether nonLOH tumors indeed have more favorable outcomes.

Comparing gBRCA1 and gBRCA2 breast cancers with asLOH, gBRCA1 tumors had significantly higher TMB scores (although the median TMB was low at 3.4), HRD scores, and proportions of aging-associated SBS1 and ROS-associated SBS18. gBRCA2 asLOH tumors had higher proportions of SBS3 and SBS26, which are associated with defective DNA repair through homologous recombination and mismatch repair, respectively (37, 61). gBRCA1 tumors, compared with gBRCA2 tumors, had enrichment of multiple chromosome arm gains and amplifications; chr6q was the most significant even after correction for ER status. Other studies also have found greater numbers of CNVs in gBRCA1 than in gBRCA2 tumors, focal gains and amplifications of chr6q in gBRCA1 tumors, and deletions of chr6q in gBRCA2 tumors (22, 23, 62, 63). We show that the deletions of chr6q that are significantly enriched in gBRCA2 tumors are dependent on ER status, as the association is lost with adjustment for ER status, and that chr6q gains are gBRCA1 specific. Chr6q contains the cancer-associated genes ROS1 (64); FOXO3, which, in addition to promoting breast cancer growth and metastasis, regulates ROS (65, 66); and RSPO3, which drives the formation of hormone receptor–negative breast cancer and is associated with epithelial-mesenchymal transition (EMT) (67). We find significantly more copy number gains of ROS1, FOXO3, and RSPO3 and significantly more FOXO3 amplifications in gBRCA1 compared with gBRCA2 tumors. Compared with gBRCA2 tumors, gBRCA1 tumors were enriched for alterations in ROS, EMT, and DNA repair pathways. Loss of BRCA1 has been shown to promote EMT, increase ROS, and dysregulate DNA repair pathways in other studies (24, 68–70). The observed molecular differences suggest that gBRCA1 and gBRCA2 tumors may respond differently to molecularly targeted therapies such as ROS1 inhibitors (64, 71) and anti-RSPO3 antibodies (72, 73), which are under investigation for cancer treatment in the preclinical and clinical settings.

It has been suggested that RB1 asLOH occurs simultaneously with BRCA2 asLOH in tumors with gBRCA2 PVs owing to the proximity of the genes on chr13 (17, 54, 74). Inagaki-Kawata et al. suggested that TP53 loss on chr17 co-occurred with BRCA1 loss but did not determine whether the loss occurred on the same or different segments (17). We found that RB1 asLOH was significantly more frequent in gBRCA2 tumors than in gBRCA1 tumors and co-occurred on CNV segments with BRCA2 asLOH. The rate of RB1 asLOH was greater than 70% in both gBRCA1 and gBRCA2 treatment-naive tumors. This finding suggests that RB1 loss in the context of HRD may confer an intrinsic fitness benefit that is not necessarily BRCA2 specific; however, loss of RB1 likely happens through differing mechanisms in gBRCA1 versus gBRCA2 tumors, the latter being due to simultaneous loss of BRCA2 and Rb1 as they are in close physical proximity. There was no difference in the frequency of TP53 asLOH between gBRCA1 and gBRCA2 tumors; the frequency of TP53 asLOH was greater than 90% in both tumor types, and TP53 asLOH was rarely on the same segments as BRCA1 asLOH. A previous study suggested that during gBRCA1-associated breast cancer development, TP53 asLOH occurred before BRCA1 loss (75). Thus, RB1 asLOH is dependent on BRCA2 asLOH but TP53 asLOH is independent of BRCA1 asLOH in breast cancer.

The availability of FDA-approved targeted therapies for the treatment of gBRCA1/2-driven tumors and for ER-positive, HER2-negative tumors presents a treatment decision in the management of gBRCA1/2 carriers with ER-positive, HER2-negative tumors (27, 28, 31–34, 76). Previous studies of breast tumors unselected for age at diagnosis, with ages ranging from 26 to 80 years, showed levels of asLOH as low as 54% in gBRCA2 PV carriers (14). In this study of young women, we found high levels of asLOH, greater than 90%, in both gBRCA1 and gBRCA2 PV carriers, regardless of tumor ER status. Additionally, women with nonLOH breast cancers had improved survival, although to a non-significant degree, likely owing to small numbers. Similarly, women whose breast cancers did not demonstrate SBS3 consistently had improved survival but non-significantly when stratified by gene or ER status. This finding suggests that most breast cancers in young women with gBRCA1/2 PVs may be responsive to PARP inhibition. It is likely that in populations with older gBRCA1/2 carriers, sporadic tumors are mixed in with gBRCA1/2-associated breast cancers, but young patients have mostly gBRCA1/2-associated tumors. Two retrospective studies in patients with advanced breast cancer have suggested that gBRCA2 PV carriers with advanced breast cancer have attenuated responses to CDK4/6i plus endocrine therapy (77, 78). Bruno et al. reported lower progression-free and overall survival in carriers of gBRCA1/2, gATM, and gCHEK2 PVs compared with noncarriers, and Kim et al. identified gBRCA2 PVs as being associated with lower progression-free survival following treatment with CDK4/6i, albeit with only 5 gBRCA2 PV carriers (77, 78). A larger study in patients with metastatic breast cancer similarly reported lower progression-free survival rates for gBRCA2 PV carriers treated with CDK4/6i compared with noncarriers (54). Our findings in young women with early breast cancer suggest that genomic mechanisms that underlie resistance to CDK4/6 inhibitors, including SNVs/indels and CNVs in RB1, TP53, FAT1, and MYC, are already present in treatment-naive breast cancers. These findings support the need for prospective clinical trials to test the optimal choice between PARPi and CDK4/6i, such as the ongoing clinical trial (NCT06380751) testing PARPi versus CDK4/6i in gBRCA1/2 PV carriers with metastatic breast cancer. Additionally, this study provides information that may be useful for clinicians making treatment decisions before prospective trials are completed (76).

The main limitation was the low number of participants with nonLOH tumors, which may have limited our ability to detect significant survival differences. We evaluated only pretreatment samples and thus can only make predictions about potential treatment responses that will need to be tested in subsequent studies.

In conclusion, we show that tumors from young gBRCA1/2 PV carriers with breast cancer almost always have asLOH. Women with tumors without LOH or SBS are likely to have improved survival at baseline. We also show differences between the molecular features of gBRCA1 and gBRCA2 tumors that may impact the suitability of molecularly targeted therapies. Importantly, we show enrichment of gene alterations associated with resistance to CDK4/6 inhibition in ER-positive, HER2-negative tumors from gBRCA1/2 PV carriers compared with noncarriers in treatment-naive breast cancers. Given the high levels of asLOH and presence of CDK4/6i resistance–associated alterations, our data suggest that PARPi may be preferable over CDK4/6i in young gBRCA1/2 carriers with ER-positive, HER2-negative breast cancer when both therapies are considered in the adjuvant and metastatic setting, supporting the need for prospective trials testing these drugs.

Sex as a biological variable. All participants in the study were female, reflecting the higher breast cancer risk that female gBRCA1/2 PV carriers have compared with males.

Study description. The POSH prospective cohort study (United Kingdom, 2000–2008) recruited over 3,000 women with invasive breast cancer, age ≤40, to investigate the impact of gBRCA1/2 PVs on breast cancer outcomes (10). The primary outcome analysis was published in 2018 with a median follow-up of 8.2 years (3).

Participants with gBRCA1/2 PVs were identified using sequencing and multiplex ligation probe analysis in patients meeting UK thresholds for clinical germline genetic testing at the time (3, 79). Three hundred ninety-one patients with a gBRCA1/2 PV were identified; 55 patients were excluded, as they received neoadjuvant chemotherapy. An additional 135 patients were excluded because tumor blocks were unavailable.

Sequencing and bioinformatics analysis. All H&E-stained sections were reviewed to ensure breast cancer diagnosis by a board-certified pathologist. After exclusion of patient samples with insufficient tumor as judged by the pathologist, failed DNA extraction, or failed library preparation, and those without tumor-germline matches, we performed WES of 160 formalin-fixed, paraffin-embedded (FFPE) tumor blocks and 194 germline DNAs (Figure 1). After tumor-germline matching, removal of duplicate tumor samples, and exclusion of 3 tumors that were metastatic at diagnosis, 2 tumors from patients without gBRCA1/2 PVs, and 1 tumor sample with poor sequencing coverage, we had 136 matched tumor-germline sequenced samples (Figure 1). The only significant differences between the 136 and the full set of 391 gBRCA1/2 PV carriers from POSH were receipt of neoadjuvant chemotherapy, pathological M stage, and unknown pathological T stage, all of which were due to predetermined study eligibility criteria (Supplemental Table 1). The eligibility for the overall study included women diagnosed with breast cancer between ages 41 and 50 years if known to be gBRCA1/2 positive. We called and filtered somatic single-nucleotide variants (SNVs), indels, and copy number variants (CNVs) in 598 cancer genes (Supplemental Table 2) using a custom computational pipeline (Supplemental Figure 1).

We selected 81 women with treatment-naive, non-metastatic, ER-positive, HER2-negative invasive breast cancer diagnosed before age 50 years without gBRCA1/2 PVs (14) in TCGA for use as noncarrier comparators. We retrieved BAM files from the National Cancer Institute Genomic Data Commons (80), excluding 6 patients with germline PVs in CHEK2, LZTR1, MSH6, PALB2, and POT1. Somatic SNVs, indels, and CNVs were called using the same custom computational pipeline used for POSH samples in 66 unique tumor-germline samples (Supplemental Figures 1 and 8).

Details of asLOH determination, HRD score calculation, single-base substitution (SBS) signature fitting, and TMB calculation are given in Supplemental Methods.

Statistics. We performed Firth logistic regression to determine the enrichment of chromosomal arm CNVs and alterations in genes by germline variant gene, ER status, and HRD status (threshold HRD-high ≥42) in R (81). Analyses were adjusted for ER status, germline variant gene, asLOH status, and multiple testing as appropriate. We compared mean HRD scores using 2-sided t tests and compared median SBS proportions and TMB scores using Mann-Whitney and Kruskal-Wallis non-parametric tests. Survival analysis was performed using the Kaplan-Meier survival estimator; we estimated hazard ratios (HRs), 95% confidence intervals (CIs), and P values using Cox regression with Firth’s penalized method.

We assigned genes to pathways based on hallmark pathway (50) classification and performed enrichment analysis using Firth logistic regression, as above, with the hallmark pathways as dependent variables. Analyses were adjusted for ER status, germline variant gene, and asLOH status as appropriate.

P < 0.05 was considered statistically significant.

Study approval. Written informed consent was obtained from all participants at recruitment for further analysis of their tissue and clinical data; ethical approval was granted in 2000 by the South West Multicentre Research Ethics Committee (MREC00/6/69).

Data availability. All data necessary for the analyses performed in this study are provided in the Supplemental Tables and the Supporting Data Values file. Tumor sequencing data are available in the NCBI’s Sequence Read Archive database (PRJNA1374212). Analytic code is available at https://github.com/nathanson-lab/POSH_WES_JCI_Insight_2026, commit ID d83a320. Anonymous germline sequencing data from the POSH study will be available upon request starting 3 months after publication of the article, for the purpose of an approved proposal, to researchers who have provided a completed Data Sharing request form that describes a methodologically sound proposal and, if appropriate, signed a Data Sharing Agreement. Proposals will be reviewed by the POSH study committee. Data will be shared once all parties have signed relevant data sharing documentation, covering the study steering committee conditions for sharing, and, if required, an additional Data Sharing Agreement from the sponsor. Proposals for POSH study data should be directed to the chief investigator of the POSH study at D.M.Eccles@soton.ac.uk.