Background

Triple-negative breast cancer accounts for approximately 15% of all breast cancer cases and is characterized by a lack of expression of the estrogen and progesterone receptors and the absence of human epidermal growth factor receptor 2 overexpression/amplification [1, 2]. Although good response rates are observed when triple-negative tumors are treated with systemic chemotherapy, these tumors are often associated with early relapse, visceral metastasis, reduced survival, and poor prognosis compared with those of other subtypes of breast cancer [3, 4].

Germline pathogenic variants in the BRCA1 and BRCA2 tumor suppressor genes are associated with increased risk of breast and ovarian cancers [5]. The products of these genes, the BRCA1 and BRCA2 proteins, work together in response to DNA damage to maintain genomic integrity, particularly during the process of homologous recombination. Homologous recombination is a high-fidelity DNA repair mechanism responsible for DNA double-strand break repair using the sister chromatid as a template. Although BRCA2 function is related to homologous recombination exclusively through its ability to recruit RAD51 to double-strand break sites, BRCA1 is a multifunctional protein that links the DNA damage-sensing process to the repair itself through multiple domains that interact with a wide range of proteins. BRCA1 plays a major role in homologous recombination through the formation of protein complexes that bind to double-strand breaks and recruit protein effectors responsible for its resection and repair [6,7,8,9,10]. Hence, defects in BRCA1/2 protein functions lead to genomic instability, thereby promoting tumorigenesis.

Patients harboring BRCA1 germline pathogenic variants display an increased risk of triple-negative breast cancer (60–80%) [11,12,13]. Moreover, sporadic triple-negative tumors often display molecular and histopathological features that resemble tumors developed by BRCA1 mutation carriers [14], suggesting that defects in other mechanisms could similarly impact the tumorigenic process. Defining groups of patients with BRCA-impaired triple-negative breast cancer is important for the clinical management of patients because several new treatment strategies are being evaluated for BRCA1/2 mutation-related tumors. For instance, tumors with defects in BRCA1/2 proteins may be particularly sensitive to DNA damage-inducing agents, such as platinum salts [15], because the impairment of these proteins results in defective DNA repair by homologous recombination. Moreover, poly(ADP-ribose)-polymerase (PARP) inhibitors were found to be effective for the treatment of BRCA1/2-mutated tumors [16]. Recently, olaparib, a PARP inhibitor, was approved by the U.S. Food and Drug Administration for the treatment of advanced ovarian cancer patients whose tumors are deficient in BRCA1/2 [17, 18]. Few studies have evaluated PARP inhibitor treatment in triple-negative breast cancer [19, 20]; therefore, its efficacy in the treatment of these tumors remains unclear.

Recent studies have reported wide variation (range 8.5–30%) in the prevalence of germline BRCA1 pathogenic variants in patients with triple-negative tumor in a manner that depends on the screening method and population characteristics [21, 22]. A high rate of germline BRCA1 pathogenic variants has been observed in early-onset triple-negative tumors in multiple cohorts [13, 23, 24]. However, in Brazil, most studies have focused on evaluating the germline variants in breast cancer patients with a positive family history for cancer regardless of subtype [25,26,27,28,29,30], and few studies have evaluated tumor histology and molecular characteristics [13, 31].

In this study, we evaluated BRCA impairment resulting from genetic and epigenetic events to provide a broader analysis of the role of BRCA1/2 in hereditary and sporadic triple-negative breast cancer. We screened somatic and germline BRCA1/2 loss-of-function pathogenic variants (by assessing point genetic variation and chromosomal rearrangements) and BRCA1 promoter methylation status in tumor tissues from an unselected cohort of triple-negative tumor samples from a single institution. We classified the samples as hereditary, i.e., tumors from patients who were carriers of germline pathogenic variants, and as sporadic, i.e., tumors without germline pathogenic mutation in both BRCA1/2 genes. Our results showed that BRCA1 impairment was a recurrent event in both sporadic and hereditary triple-negative tumors diagnosed in young women. Additionally, we demonstrated similar survival rates in both hereditary and sporadic BRCA1-impaired cases, and these rates were increased compared to those in patients with BRCA1-proficient tumors. These data could be used for the improvement of the clinical management of triple-negative breast cancer patients.

Methods

Samples

Triple-negative breast cancer samples from patients diagnosed between 2000 and 2014 were retrieved from the Tumor Biobank [32, 33] of the A. C. Camargo Cancer Center. The samples were selected irrespective of age at diagnosis or family history. Tumors were classified by immunohistochemistry results showing that less than 1% of the tumor cells stained positive for estrogen receptor and progesterone receptor and that human epidermal growth factor receptor 2 staining was scored negative or equivocal if no amplification was detected by fluorescent in situ hybridization (gene to centromere ratio < 2.0) [34, 35]. Additional prognostic or basal cell (p53, p63, EGFR, CK5/6, CK14, and Ki67) immunohistochemistry staining was performed as part of a diagnostic routine in our Surgical Pathology Department. Clinical and pathological data were collected from the patients’ electronic medical records.

For chemotherapy treatment, a subset of 33 (25%) patients received preoperative chemotherapy, whereas 86 (66%) underwent post-operative chemotherapy. This information was not available for 12 (9%) patients. Most patients (82%) received anthracycline, cyclophosphamide, and taxane-based chemotherapy. This study was approved by the Research Ethics Committee of A. C. Camargo Cancer Center under protocol number 1746/13 and performed in accordance with the ethical standards as laid down in the Declaration of Helsinki.

Frozen tissues were cut in 5-micrometer-thick sections, fixed, and stained with hematoxylin and eosin for histological analyses. An experienced breast pathologist evaluated the hematoxylin and eosin slides to ensure representative tumor selection. Manual tissue macrodissection, using the slide as a guide, was performed for samples with at least 70% tumor tissue. DNA extraction was performed by the A. C. Camargo Cancer Center Biobank. The QIAsymphony DNA Mini Kit (QIAGEN, Hilden, NRW, Germany) was used for DNA extraction according to the manufacturer’s instructions.

Screening for loss-of-function mutations in BRCA1 and BRCA2 genes

A total of 131 triple-negative breast cancer samples were submitted for sequencing of the entire coding sequence and the exon–intron boundaries of the BRCA1 and BRCA2 genes by next-generation sequencing (Ampliseq™ BRCA1 and BRCA2 panel-Ion PGM Torrent; Thermo Fisher Scientific, Waltham, MA, EUA). Reference sequences indexed as LRG_292 (BRCA1) and LRG_293 (BRCA2) by the Locus Reference Genomic database (http://www.lrg-sequence.org/) were used for alignment and variant calling using CLC Genomics Workbench 6 software (QIAGEN, Hilden, NRW, Germany). Criteria for calling genetic variants included at least 50 × coverage and variant frequency greater than 5%. Genetic variants were classified according to the ClinVar (https://www.ncbi.nlm.nih.gov/clinvar/), Breast Cancer Information Core (http://research.nhgri.nih.gov/bic), and Leiden Open Variation Database - International Agency for Research on Cancer (http://hci-exlovd.hci.utah.edu/home.php) databases. We considered loss-of-function variants as those classified as pathogenic in the aforementioned databases or novel variants that lead to a predicted premature stop codon in the resulting protein (frameshift, nonsense, or splice site). Pathogenic genetic variants were confirmed by capillary sequencing of DNA samples (BigDye Terminator V3.1-ABI 3130xl; Thermo Fisher Scientific, Waltham, MA, EUA). For determining the germline or somatic nature of the pathogenic mutations, the variants were evaluated in DNA from leukocytes or adjacent normal tissue samples when this biological material was available in the A. C. Camargo Cancer Center Biobank. For two samples, this determination was performed by independent genetic testing results available in the patients’ records.

BRCA1 chromosomal rearrangement that results in copy number variation within the gene was evaluated by Multiplex ligation-dependent Probe Amplification (MLPA – P087, MRC-Holland, Amsterdam, NL) in 99 patients out of the 131 from the whole cohort, and experimental procedures were conducted according to the manufacturer’s recommendations using leukocyte and/or tumor DNA. Coffalyzer software (MRC-Holland, Amsterdam, NL) was used at default settings for data analyses.

Assessment of BRCA1 promoter hypermethylation in triple-negative breast cancer

For BRCA1 promoter methylation analysis, bisulfite conversion of tumor DNA was performed using the EZ DNA Methylation-Gold (Zymo Research, Irvine, CA, USA) kit according to the manufacturer’s protocol; converted DNA was eluted in 20 μL. Primers were designed to encompass 19 CpG dinucleotides in the region −191 to +199 bp from the transcription start site (GRCh37/hg19 chr17:41,277,302-41,277,691). PCR products were generated using Platinum Taq DNA Polymerase High-Fidelity (Thermo Fisher Scientific, Waltham, MA, EUA) in a final volume of 50 uL; the product of this reaction was used for subsequent library construction. Libraries were prepared using a custom strategy as previously described by our group [36] in conjunction with the 454 GS Junior platform. Briefly, after PCR amplification, end repair, and A-tailing, unique Y-shaped barcoded DNA adaptors were added to each sample with T4 DNA ligase. The ligated fragments were then amplified by low-cycle PCR to generate the library for sequencing. An inhouse pipeline was used to calculate the CpG methylation levels and bisulfite conversion rate. Maximally selected rank statistics were applied, determining 16.1% as a threshold for stratifying samples according to the methylation status. A full description of the method is available in the Electronic Supplementary Material (ESM 1).

Association between BRCA impairment and clinicopathological features

Chi-square or Fisher’s (categorical) and Mann–Whitney (numerical) tests were used to analyze the association between the presence of BRCA impairment and patient characteristics, such as age at diagnosis, family history of breast cancer (defined as the presence of breast cancer in any relative up to third degree), and tumor features (tumor size, Scarff–Bloom–Richardson histologic grade, lymphovascular and perineural invasion, lymph node status, and expression levels of CK5/6, CK14, EGFR, p53, and p63). Patients diagnosed at 40 years of age (yo) or younger were considered to have an early age at diagnosis. Survival curves were calculated using the Kaplan–Meier method. The log-rank (Mantel–Cox) test was used for survival comparisons. All analyses were performed using GraphPad Prism 5 software, and the significance level was set at 5%.

Results

BRCA1/2 loss-of-function mutation and its association with demographic characteristics

The age at diagnosis for the 131 patients selected for this study ranged from 18 to 87 yo (median 49 and mean 51 yo). Pathogenic point variations were detected in 18 (13.7%) tumor samples, including 16 (88.8%) in BRCA1 and 2 (11.2%) in BRCA2. Copy number variations within the BRCA1 gene were investigated in 99 cases showing no chromosomal rearrangements either in tumor or leukocyte DNA. Sixteen patients showed 14 different point variations in the BRCA1 gene (four missense, five frameshift insertion/deletion, four stop codons, and one splice site) (Table 1). The BRCA1 founder pathogenic variant c.181T > G;p.(Cys61Gly) was observed in three unrelated cases. The two pathogenic variants detected in the BRCA2 gene were both frameshifts and were found for the first time. Five out of the 16 (31.3%) pathogenic mutations identified were detected for the first time in this study (3 in BRCA1 and the 2 in BRCA2—Table 1). A total of 113 samples (82.3%) were negative for BRCA1/2 pathogenic variants, including samples from seven patients from whom variants of uncertain significance were identified (four in BRCA1 and three in BRCA2). One sample harboring a BRCA1 pathogenic variant also displayed a concurrent variant of uncertain significance (Table esm2).

Table 1 Pathogenic mutations identified in 131 unselected triple-negative breast cancer (TNBC) samples
Full size table

The germline or somatic nature of pathogenic variants was determined for all pathogenic variants identified in our cohort. Interestingly, germline variants accounted for 94.4% (17/18) of the pathogenic variants. The only somatic pathogenic variant detected was a nonsense mutation in the BRCA1 gene, c.5503C > T;p.(Arg1835*), which was reported as a somatic mutation in the Catalogue of Somatic Mutations in Cancer database (COSM78883-http://cancer.sanger.ac.uk/cosmic) and as a germline variant in the ClinVar database (55601-http://www.ncbi.nlm.nih.gov/clinvar/). We detected 13.0% of samples with germline variants (17/131 – 15 in BRCA1 and 2 in BRCA2) and 0.8% (1/131 in BRCA1) with somatic pathogenic variants. This analysis allowed the classification of triple-negative as hereditary tumors, i.e., those tumors from patients who were carriers of germline pathogenic variants (N = 17), or as sporadic tumors, i.e., tumors with BRCA1/2-negative or somatic pathogenic variants (N = 114).

Hereditary triple-negative was, as expected, mostly related to BRCA1 (88.2% – 15/17). A significant association of hereditary tumors was observed with a young age at diagnosis (p = 0.006), and a marginally significant association was observed with a family history of breast and ovarian cancer (p = 0.0509). A total of 28% of the family history-positive patients presented with BRCA1 mutation compared with only 12% in the family history-negative group. The mean age at diagnosis in the hereditary group was 43 yo versus 53 yo for sporadic (median 41 vs. 50). When stratified by age groups (up to 40, 41–50, and 50 + yo), the BRCA1/2 germline pathogenic variant rate was significantly increased in patients diagnosed in the up to 40 yo group (26%) compared with those in the other groups (13 and 7%; p = 0.017; Fig. 1a).

Fig. 1
figure 1

a BRCA1 germline pathogenic variant rate in triple-negative breast cancer (TNBC) patients diagnosed at up to 40, 41–50, and more than 50 yo (χ 2 test, p = 0.018). b Mean BRCA1 promoter methylation level in triple-negative breast cancer (TNBC) samples with respect to BRCA mutation status. BRCA-mutated triple-negative breast cancers (TNBCs) are indicated by red bars. BRCA-negative triple-negative breast cancer (TNBC) are illustrated by light gray bars. Methylation level was calculated as the mean methylation of the 19 CpG dinucleotides analyzed in the region −191 to +199 bp relative to the BRCA1 transcription start site. Samples with a mean methylation greater than the calculated threshold (from maximally selected rank statistics analysis) of 16.09% were classified as BRCA1-hypermethylated. c BRCA1 impairment rate in triple-negative breast cancer (TNBC) patients diagnosed at up to 40, 41–50, and more than 50 yo (χ 2 test, p = 0.0011)

Full size image

BRCA1 promoter methylation screening in sporadic triple-negative and its association with demographic characteristics

Because BRCA1 germline pathogenic variants play a major role in hereditary triple-negative, accounting for 88.2% of cases in our study, we investigated BRCA1 gene promoter CpG methylation in tumor samples as a potential somatic mechanism of gene silencing. BRCA1 promoter methylation was assessed in a subset of 123 tumors of the 131 samples from the whole cohort (17 hereditary – 15 BRCA1 and 2 BRCA2 germline pathogenic variants; and 106 sporadic – 01 BRCA1 somatic pathogenic variant and 105 BRCA1/2-negative). Based on the results of the maximally selected rank statistics analysis, the value of 16.1% of methylation in CpG dinucleotides was established as the threshold for categorizing the samples as hypermethylated and nonhypermethylated. The BRCA1 promoters of 27 (20.6% – 27/131) tumors scored above this limit and were considered hypermethylated. Interestingly, all BRCA1-hypermethylated tumors were from the sporadic group, and none overlapped with the hereditary group (Fig. 1b) (see Table esm1 and Table esm2 in the Electronic Supplementary Material for complete results).

Additionally, we assessed the association of BRCA1 promoter hypermethylation in the group of 105 sporadic triple-negative tumors (BRCA1/2 wild-type pathogenic variant) with demographic variables (age at diagnosis and familial history). For this analysis, we compared BRCA1-hypermethylated (N = 27) and BRCA1-nonhypermethylated (N = 78) tumor samples. As expected, BRCA1 promoter hypermethylation was not associated with family history (22% of BRCA1-hypermethylated vs. 23% BRCA-proficient, p = 1.000). However, despite the fact that BRCA1 promoter hypermethylation was not significantly associated with age at diagnosis (mean 49 vs. 54 yo, p = 0.962), the hypermethylation rate was increased, but at nonsignificant levels (p = 0.131), in women diagnosed at up to 40 yo (39%) when compared to that in women diagnosed from 41 to 50 yo or older than 50 yo (24 and 18%, respectively).

Association between BRCA1 impairment and clinicopathological features

Next, we investigated the association between BRCA1 gene deficiency and clinicopathological features. For this analysis, we excluded from the hereditary triple-negative group two samples with BRCA2 germline pathogenic variants. Tumors were classified as BRCA1-impaired if either a pathogenic variant in the BRCA1 gene or BRCA1 promoter hypermethylation was detected or as BRCA1-proficient for the group of BRCA1/2-negative pathogenic variants and BRCA1 nonhypermethylated tumors. A total of 32.8% (43/131) of tumors displayed BRCA1 impairment.

BRCA1-impaired tumors were significantly associated with a young age at diagnosis (p = 0.002; mean age 47 vs. 54 yo). A significantly increased proportion of patients diagnosed at up to 40 yo also presented with BRCA-impaired tumors (55% in women up to 40 yo, 33% in women 41-50 yo, and 22% in women older than 50 yo; p = 0.007—Fig. 1c). Moreover, although high histological grade (grade 3) was frequently observed in triple-negative in general, our data showed that BRCA1-impaired tumors were statistically associated with grade 3 tumors (98% vs. 80%, p = 0.011) (Table 2). BRCA1-impaired tumors were also associated with p63 immunohistochemistry positivity (50% vs. 20%, p = 0014), a marker of myoepithelial (basal) cells. Nevertheless, no association was observed with basal-like features, as most tumors with available data were positive (93%) for at least one basal cell marker. No other pathological characteristics exhibited statistically significant associations with BRCA1 impairment (Table 2).

Table 2 Clinicopathological features of BRCA-impaired (positive for BRCA1/2 mutation or BRCA1 promoter methylation) triple-negative breast cancer (TNBC) and BRCA-proficient (negative for mutation or hypermethylation) triple-negative breast cancer (TNBC) unselected and selected for early-onset triple-negative breast cancer (TNBC) (≤ 40 yo)
Full size table

Trends towards better overall (p = 0.081) and disease-free survival (p = 0.074) were observed in BRCA1-impaired tumors compared with those in BRCA-proficient tumors (Supplemental Fig. esm2 b and d), even when restricted to patients who underwent adjuvant chemotherapy (p = 0.103; p = 0.075; Supplemental Fig. esm2 e–f). Moreover, patients with BRCA1-mutated and BRCA1-hypermethylated tumors exhibited similar outcomes (Supplemental Fig. esm2 a and c), supporting the hypothesis, to some extent, that BRCA1-impaired triple-negative breast cancer, both sporadic and hereditary, tends to be more sensitive to DNA-damaging chemotherapy.

BRCA1 impairment in early-onset triple-negative breast cancer

Based on the strong association observed between BRCA1 impairment and young age at diagnosis (BRCA1 inactivation was enriched among women diagnosed up to 40 yo), we compared clinical variables in this subset of patients. Intriguingly, BRCA1-impaired tumors were associated with negative lymph node status (p = 0.003). Tumor infiltration of the axillary lymph nodes was observed in 6% (1/17) of patients diagnosed with BRCA1-impaired tumors in contrast to 62% (8/13) of patients in the BRCA-proficient group (Table 2). Consistent with this observation, significantly better overall and close to significantly better disease-free survival rates were observed in patients with BRCA1-impaired tumors compared to those in patients with BRCA-proficient tumors (p = 0.046; p = 0.052—Fig. 2b and d). Additionally, we investigated whether a positive family history of cancer could eventually favor early triple-negative diagnosis or the survival rate by evaluating both variables as a function of family history, although no significant association was found with the survival rate or age at diagnosis (Supplemental Fig. esm3 and Fig. esm4).

Fig. 2
figure 2

a Overall survival curves for patients diagnosed with early-onset (≤ 40 yo) BRCA1-pathogenic variant (red line), BRCA1-hypermethylated (green line), and BRCA-proficient (blue line) triple-negative breast cancer (TNBC) (log-rank test, p = 0.1272). b Overall survival of patients diagnosed with early-onset (≤ 40 yo) BRCA1-impaired (orange line) or BRCA-proficient (blue line) triple-negative breast cancer (TNBC) up to 40 yo (log-rank test, p = 0.0461. c Disease-free survival curves for patients diagnosed with BRCA1-pathogenic variant (red line), BRCA1-hypermethylated (green line), or BRCA-proficient (blue line) triple-negative breast cancer (TNBC) up to 40 yo (log-rank test, p = 0.1103). d Disease-free survival for patients diagnosed with BRCA1-impaired (orange line) or BRCA-proficient (blue line) triple-negative breast cancer (TNBC) up to 40 yo (log-rank test, p = 0.0520)

Full size image

Discussion

To our knowledge, this is one of the few studies to comprehensively investigate point mutations, large rearrangements, and promoter methylation in the BRCA1 gene in triple-negative breast cancer [37, 38] and represents the largest cohort of this subtype of tumor screened for loss-of-function alterations in the BRCA1/2 genes in an under-represented population, the Brazilians. The mean age at diagnosis in this study (51 yo) was similar to those in other studies, including those of the METABRIC study (52.7 yo) (Molecular Taxonomy of Breast Cancer International Consortium) [39], The Cancer Genome Atlas (54.2 yo) [40], and others (51 yo) [21]. As described in the literature, we observed that tumors often displayed a high grade (86.5% Scarff–Bloom–Richardson Grade 3) and a high proliferation index (93.5% of tumors with Ki67 > 15% by immunohistochemistry), which is consistent with the histological aggressiveness commonly observed in this breast cancer subtype [41, 42]. Furthermore, the majority of tumors were positive for at least one basal marker (93.3% – 84/90), reinforcing the strong association observed between triple-negative and the basal-like phenotype [43], irrespective of BRCA1 status.

Investigating a cohort of early-onset breast cancer patients unselected for family history in Brazil, Carraro et al. [13] reported that 50% of triple-negative breast cancer patients up to 35 yo harbored BRCA1 germline pathogenic variants; however, the sample group was small. Studying a larger cohort, we identified pathogenic variants in 13.7% of tumors (12.2% in BRCA1 and 1.5% in BRCA2), which were predominantly (93.8%) germline pathogenic variants, mostly in the BRCA1 gene (88.9%). Exclusively considering early-onset tumors, the pathogenic variant frequency increased to 28.1% in our cohort. This information reinforces the recommendation for genetic testing in women diagnosed with triple-negative breast cancer at a young age because they have an increased risk of carrying germline BRCA1 mutations [23].

A significant proportion of samples displayed somatic inactivation of the BRCA1 gene (20.6% by BRCA1 promoter hypermethylation) in our study. Other studies analyzing BRCA1 promoter methylation in triple-negative tumors of patients unselected for age or family history reported approximately one-third of the tumors as showing hypermethylation at this gene [37, 44]. Curiously, in our study and others [37], no concomitant inactivation of BRCA1 by germline pathogenic variants and somatic promoter methylation was detected, suggesting that these are mutually exclusive events. By combining data from mutation and methylation analysis, we found that more than half (55%) of triple-negative of patients diagnosed before 40 yo showed BRCA1 impaired tumors. When considered in combination with recent evidence of potential benefits in treating BRCA-deficient tumors with platinum-based salts and PARP-1 inhibitors [15, 16], this information is particularly important and could help guide the selection of patients who may benefit from these drugs.

None of the hereditary triple-negative cases showed concomitant BRCA1-promoter hypermethylation, prompting the question of what mechanism is responsible for the inactivation of the wild-type allele in the BRCA1 gene. In fact, others have investigated this subject and shown that loss of heterozygosity is frequently observed in BRCA1-mutated tumors [45, 46]. We observed that in 62.5% (10/16) of the BRCA1-mutated tumors, the pathogenic variant was present in > 60% of the reads (Table 1), in concordance with higher peaks of allelic variants detected in Sanger sequencing, suggesting partial loss of the wild-type allele in tumors. These data could be underestimated given the presence of normal infiltrating tissue, which in our case was estimated to be at most 30% (criterion of A. C. Camargo Cancer Center Biobank is a minimum of 70% tumor tissue). However, this observation is speculative, and more detailed investigation is needed to obtain a definitive conclusion about loss of heterozygosity. Moreover, DNA methylation seems not to be responsible for this phenomenon, as we and others [37, 47] have shown. In addition, BRCA1 somatic point mutations are rare in breast cancer [48]. Hence, the second molecular event responsible for the inactivation of BRCA1 remains unclear.

Clinical trials evaluating the activity of PARP inhibitors rely on the fact that patients with tumors lacking functional homologous recombination DNA repair respond to the treatment, whereas patients with homologous recombination DNA repair-proficient tumors do not. Therefore, it is important to properly identify BRCA-impaired tumors for inclusion in clinical trials and to avoid misallocation of patients in study arms to obtain more accurate results. Moreover, prior to treatment with PARP inhibitors, it is essential that tumors be screened for loss-of-function alterations related to this pathway, which could help optimize treatment regimens. In this context, given that PARP inhibitor treatment is a promising strategy for targeting tumors with defective DNA damage repair capacity and that BRCA1 impairment is a recurrent event in young patients with triple-negative tumors, the data provided here offer prospects to increase the efficacy of PARP inhibitor therapy prescription if screening of BRCA1 loss-of-function includes both germline mutation and somatic promoter hypermethylation. Unfortunately, several studies have focused only on BRCA germline mutation screening as an indicator for PARP inhibitor therapy, which could result in the grouping together of different breast tumors (www.clinicaltrials.gov).

Previous studies have demonstrated no significant differences in survival for BRCA-mutated triple-negative patients compared with BRCA-wild-type triple-negative patients [49, 50]. Conversely, in our study, significantly better overall rate was observed in patients diagnosed with early-onset (≤ 40 yo) BRCA1-impaired tumors compared to those in patients with early-onset BRCA-proficient tumors. Interestingly, all women diagnosed with early-onset triple-negative breast cancer were treated with anthracycline/cyclophosphamide/taxane-based chemotherapy, which makes this group homogenous in terms of therapy. Therefore, BRCA1-impaired tumors in younger patients could be more sensitive to treatment given that anthracycline and cyclophosphamide ultimately induce DNA damage and that these tumors lack functional homologous recombination DNA repair machinery [51, 52]. Perhaps, the association with survival was not observed for older patients because this population is not enriched for BRCA-inactivating events, and other molecular mechanisms are likely responsible for triple-negative tumorigenesis. Nonetheless, these findings should be interpreted cautiously given that a positive association between lymph node metastasis, a classic breast cancer marker of poor survival, and BRCA proficiency was identified. Moreover, we cannot discard the influence of clinical surveillance measures on the survival of patients with BRCA1 pathogenic variants.

In this study, almost all tumor samples were obtained from surgical specimens (pre-surgical biopsies were not available from all patients). Given that BRCA mutations are associated with higher rates of a complete pathological response to neoadjuvant chemotherapy [3], it is possible that BRCA-impaired tumors were under-represented in our study cohort. However, by using surgical specimens, we avoided issues concerning tumor representation.

Conclusions

Here, we demonstrate that BRCA1 pathogenic variants (especially of germline origin) and promoter methylation (somatic) are critical mechanisms of BRCA1 inactivation in triple-negative breast cancer in young patients. Investigations into these two inactivation mechanisms provide better characterization of these tumors and offer prospects for new drugs targeting tumors with deficiencies in DNA repair.