Clinical Relevance of BRCA1 Promoter Methylation Testing in Patients with Ovarian Cancer

Ovarian cancer is a dreadful disease representing the fifth cause of women mortality in developed countries (1). Significant progress in ovarian cancer management has been made in the last decade with the discovery of a new therapeutic class called PARP inhibitors (PARPi) that exploits the molecular weakness of ovarian cancer (2). Indeed, PARPis act by trapping inactive PARP onto the single-strand breaks (SSB) preventing their repair (3, 4). Therefore, the accumulation of SSBs during DNA replication results in DNA double-strand breaks that would normally be repaired by the homologous recombination (HR) DNA repair machinery. Thereby, cells deficient in HR (HRD), are exquisitely sensitive to PARPis, supporting the concept of synthetic lethality. This targeted killing was first demonstrated in cells lacking functional BRCA1 or BRCA2 (5, 6), and led to the clinical development of PARPis in patients with BRCA1/2-mutated ovarian cancer. As a result, PARPis were first approved as maintenance therapy in first-line or relapsed BRCA-mutated (BRCAmut) ovarian cancer bringing unprecedented survival benefit (7, 8).

Numerous clinical and in vitro studies support that PARPi might have a broader effect beyond BRCA-mutated ovarian cancer (9). Genomic instability scores (GIS) designed to detect genetic “scars” in HRD tumors have demonstrated their capacity to predict PARPi benefit. Several large randomized phase III trials in newly diagnosed advanced high-grade ovarian cancer have shown that PARPi maintenance resulted in a remarkable improvement in progression-free survival (PFS) compared with placebo in patients with BRCA wild-type (WT), high-GIS ovarian cancer. Whereas PARPi benefit in the remaining BRCA WT, low-GIS population is more limited (10–12). On the basis of these large clinical trials, PARPi (olaparib, niraparib, or rucaparib) are now recommended as first-line maintenance, in case of BRCA mutation (Category 1 recommendation) or high-GIS (Category 2A), or more broadly after an objective response to platinum-based chemotherapy (Category 2A) in patients with advanced ovarian cancer (13).

Another mechanism that can lead to BRCA inactivation is promoter hypermethylation (14, 15). This silencing event only occurs on the promoter of BRCA1 and is mutually exclusive of BRCA1 mutation. Early studies suggested that BRCA1 hypermethylation (BRCAmeth) was observed in around 15% of BRCA WT ovarian tumors but did not seem to impact platinum sensitivity or prognosis contrarily to BRCA mutation (16–18). However, in a retrospective comprehensive genomic analysis of the ARIEL2 trial, Swisher and colleagues demonstrated that a high BRCA1 methylation score (suggestive of homozygous BRCA1 silencing) was associated with high LOH (surrogate marker of GIS) and sensitivity to rucaparib and could also serve as a predictive biomarker of HRD (19). Furthermore, the decrease or loss of methylation under platinum chemotherapy may lead to PARPi resistance (19).

Our study aimed to evaluate the clinical utility of BRCA1 promoter methylation testing on a large cohort of high-grade ovarian cancer treated in our institute to predict HRD status. We then correlated the methylation status (high vs. low vs. none) to clinical behavior and platinum benefit in particular.

All patients were provided written informed consent authorizing the use of residual tumor tissue obtained during their routine diagnosis and treatment for translational research. The study protocol was approved by local ethics committee (no. 2022-159). Study was conducted in accordance of the Declaration of Helsinki.

This is a nonrandomized retrospective analysis. The last 100 tumors routinely tested for GIS by MyChoice CDx (Myriad Genetics) and with sufficient DNA were included. Samples were eligible for inclusion in the study if the patient was an adult woman of at least 18 years of age with high-grade ovarian cancer (any histology included). GIS was performed on diagnostic samples, pretreatment biopsies preferably. If tumor surface was <7 mm² or if GIS test result was nonconclusive, a second attempt was performed on interval surgery samples.

Clinical and pathologic data were retrospectively collected from electronic medical record and secured in an online RedCap database hosted in Institut Gustave Roussy. The achievement of complete cytoreduction (CC0) at interval debulking surgery was assessed.

PFS was calculated from the date of initial diagnostic to disease progression or death. Overall survival (OS) was calculated from the date of diagnosis to death or loss of follow-up.

Tumor genomic DNA was extracted using QIAsymphony DSP DNA Kits (Qiagen). Two hundred nanograms of DNA were treated by bisulfite (EpiTect Bisulfite Kits, QIAGEN).

The quantification of the BRCA1 promoter methylation was performed by droplet digital PCR (ddPCR, Stilla Technologies). Nine CpG site of BRCA1 promoter (c.-220 to c.-96) were coved by the primers and probes. A positive control was included in each experience (EpiTect PCR Control DNA Set, QIAGEN). BRCA1 primers designed are reported in Supplementary Table S1.

BRCA1 and TP53 status ware analyzed in a large pan-cancer panel of 85 genes. All the exonic regions were covered and large rearrangement were detected. The homologous recombination repair (HRR) genes were included: ATM (NM_000051.3; exons 2 to 63), BAP1 (NM_004656.3; exons 1 to 17), BARD1 (NM_000465.3; exons 1 to 11), BRCA1 (NM_007294.3; exons 2 to 3 and 5 to 24), BRCA2 (NM_000059.3; exons 2 to 27), BRIP1 (NM_032043.2; exons 2 to 20), CHEK2 (NM_007194.3; exons 2 to 15), FANCA (NM_000135.2; exons 1 to 43), NBN (NM_002485.4; exon 1 to 16), PALB2 (NM_024675.3; exons 1 to 13), RAD51 (NM_001164269.1; exon 2 to 10), RAD51B (NM_133509.3; exons 2 to 11), RAD51C (NM_058216.2; exon 1 to 9), RAD51D (NM_002878.3; exon 1 to 10), RAD54L (NM_001142548.1; exons 1 to 18). Other genes included were RB1 (NM_000321.2; exon 1 to 27), TP53 (NM_000546.5; exon 1 to 11), CCNE1 (NM_001238.3; exons 2 to 12), CDK12 (NM_016507.3; exons 1 to 14), PTEN (NM_000314.8; exon 1 to 9), NF1 (NM_001042492.3; exon 1 to 58). The library was prepared with SureSelect XT HS Target Enrichment System (Agilent Technologies) and underwent by next-generation sequencing (NGS) on a NextSeq 500 (Illumina).

BRCA1 methylation zygosity was calculated as the ratio of BRCA methylated allelic frequency to TP53 mutation allelic frequency, a robust surrogate of tumor cellularity in TP53-mutated high-grade ovarian cancer. BRCA1 methylation homozygosity was defined using a cutoff of 70% for high-methylation status and deemed to be associated with complete gene silencing (9, 19).

The HRD status was determined on slides sent to Myriad Genetics and tested with the MyChoice CDx panel. The cutoff for the GIS score was 42 as recommended by test manufacturer (20).

Statistical analysis was performed on the whole cohort; no formal calculation of power or sample size was needed. Descriptive statistics were used to summarize patient demographic and clinical characteristics.

Comparisons between categorical variables were performed using a nonparametric Fisher exact test. A P value of <0.05 was considered significant. OS and PFS were estimated using the method of Kaplan and Meier and presented with Rothman 95% confidence intervals (CI) at 5 and 10 years. HRs with 95% CIs were calculated using Cox models with no adjustment.

All statistical analyses were carried out using RStudio (version 1.4.1103) and BlueSky Statistics package (version 7.30).

Anonymized data are provided as Supplementary Materials and Methods (Supplementary Table S3), and sequencing data have been deposited on Gene Expression Omnibus public database (GSE229315).

A total of 100 patients were prospectively tested for GIS between December 2020 and August 2022. Median age was 62.4 years. Most patients presented advanced (FIGO III/IV) disease and only 5 patients had FIGO I/II stages (6.2%). A total of 92% were high-grade serous ovarian cancers and 11 patients (11%) harbored a deleterious BRCA mutation (seven BRCA1 mutations and four BRCA2 mutations). GIS was considered positive (score ≥ 42) for 52 patients (52%) and negative for 48 patients (48%). Complete cytoreductive surgery (CC0) rate was 84% (64/76); however, the information was not available for 24 patients. Patients’ characteristics are summarized in Table 1 and Supplementary Table S2. With a median follow-up of 22.6 months, median PFS was 27.8 months (95% CI, 13.6–34.8) and median OS was not reached.

BRCA1 promoter methylation was assessable on 97 samples (96%). Three samples were noncontributive as no TP53 mutation was detected (whereas a deleterious TP53 mutation was reported in another tumor block) suggesting lack of tumor cells. In addition, there were 7 patients with TP53wt tumors (one clear cell and two undifferentiated tumors in particular), and none of them presented BRCA1 methylation. For the further analyses, only TP53-mutated patients (N = 90) were considered to properly assess BRCA1 methylation levels.

As expected, we did not detect any promoter hypermethylation in the 11 BRCAmut patients confirming that these are mutually exclusive events. Overall, 26 of 79 BRCAwt patients presented BRCA1 promoter hypermethylation (BRCAmeth = 33%). BRCAmut and BRCAmeth were significantly younger in comparison with BRCAwt patients (59.0 and 59.6, respectively vs. 66.3 years old, P = 0.006); however, they did not demonstrate any other obvious clinical or pathologic differences (Supplementary Table S3).

Using the previously published 70% cutoff of zygosity (9, 19), 15 BRCAwt patients had high-methylation levels (19%) and 11 low-methylation levels (14%). All of the highly methylated tumors (15/15 = 100%), were classified HRD by MyChoice CDx assay (score ≥ 42; Fig. 1). Biallelic BRCA1 methylation (high-BRCAmeth) correlated perfectly with HRD status with a positive predictive value of 100% (15/15). Mean GIS was 61.5 for BRCAmut, 66.4 for high-BRCAmeth, 58.9 for low-BRCAmeth, and 33.3 for BRCAwt unmethylated (P < 0.001; Fig. 2).

While follow-up was relatively short, median PFS was 59 months, 41.4 months, non-reached and 25.0 months in the BRCAmut, high-BRCAmeth, low-BRCAmeth and BRCAwt populations, respectively. The difference did not reach statistical significance, P = 0.45.

Intriguingly, all the low-methylated cohort (N = 11) were also classified as HRD by MyChoice CDx performed at the same timepoint. Half of these samples were collected after neoadjuvant platinum chemotherapy (5/11) and three of those developed early relapse [with a platinum-free interval (PFI) ≤ 6 months], despite a high GIS suggestive of HRD. Among the six samples with low-BRCAmeth in pretreatment samples, none have relapsed so far (follow-up range: 2 to >60 months).

Overall, out of the 51 HRD tumors with contributive methylation assay, 11 were BRCAmut (22%), 15 were highly methylated (29%), and 11 were low methylated (21%). Among the remaining 14 patients with BRCAwt/HRD tumors, we dectected one deleterious ATM mutation and one deleterious BRIP1 mutation that could possibly explain their phenotype (Fig. 3).

There has been substantial debate over the prognostic and predictive relevance of BRCA1 promoter methylation with many early studies showing no association with survival or platinum sensitivity (21). However, most of these studies were conducted years ago and used very heterogeneous assays for methylation. Those studies were also focused on treatment of ovarian cancer after relapse with a methylation status determined on primary tumor. Moreover, recent studies have highlighted the importance of demonstrating high-level methylation to confirm homozygous silencing (19). Here, we report, for the first time, the clinical relevance of BRCA1 promoter hypermethylation status on a large cohort of advanced ovarian cancer tumors (with known BRCA mutation and GIS status) using a methylation-specific PCR assay and validated criteria to define highly methylated samples.

Using quantitative ddPCR, BRCA1 methylation was detected in 33% of BRCAwt patients (29% of the overall population) and it was perfectly correlated with HRD status by MyChoice CDx assay. Complete silencing of both alleles is necessary for HRD (homozygous event). Using TP53 allelic frequency and a predetermined zygocity cutoff of 70%, 18% of patients were considered highly-methylated (homozygous). High-BRCA1 methylation was associated with a positive predictive value of 100%. These results suggest that the systematic implementation of BRCA1 promoter methylation testing in patients with BRCAwt tumors can effectively detect around 20% of patients harboring high-BRCAmeth tumors, and undoubtedly deficient HR. These results may be highly relevant to other tumor types with suggestion of HRD or PARPi benefit beyond BRCA1/2 mutations such as triple-negative breast cancer (22).

In addition, we detected 14%, of low-BRCAmeth tumors (1%–69% methylated tumor cells) which can be in favor of a monoallelic event (9). All low-BRCAmeth tumors were also classified as HRD according to GIS suggesting that they harbor genomic scars of unknown origin in which heterozygous BRCA1 methylation may have been implicated. Some samples were exposed to platinum-chemotherapy and therefore, may have harbored homozygous BRCA1 methylation in the past accounting for the high-GIS. Indeed, as it has been demonstrated in vitro and retrospectively, under therapeutic pressure, loss of BRCA1 methylation can occur with platinum chemotherapy, leading to a HR-proficient state even in the context of extensive genomic scarring. To our knowledge, we report for the first time, 3 patients with low BRCA1 methylation status after neoadjuvant chemotherapy and confirm their aggressive behavior with early relapse post-platinum despite high genomic instability. Whether these tumors were previously highly methylated on pretreatment samples is currently being investigated. These results reinforce the relevance of measuring the evolution of BRCA1 methylation zygocity as a real-time assessment of HR proficiency and an effective biomarker for platinum and PARPi sensitivity (9, 14).

While it is admitted that loss of homozygous methylation is associated with PARPi resistance, the clinical relevance of low-methylation status in untreated samples remains unknown. We detected 6 patients with low BRCA1 methylated tumors, all with high-GIS. While follow-up is short, none have relapsed to date. The molecular mechanisms behind HRD in this situation remain unknown but we can hypothesize several explanations. First, ovarian cancer is known to harbor high levels of intratumor heterogeneity (23, 24). Herein, we can easily imagine that these low-BRCAmeth tumors may be composed of a mix of homozygous BRCA1 methylated and nonmethylated clones accounting for these results. In addition, ovarian cancer cells can demonstrate extreme intrinsic plasticity that may result in partial loss of BRCA1 methylation as a spontaneous tumor fitness reaction (25, 26). Finally, other posttranscriptional events on the nonmethylated allele could account for complete BRCA inactivation and HRD.

This study suffers from several limitations. The relatively low number of patients, as well as the relatively low number of methylation events, hampered our statistical power. Moreover, we were not able to analyze both the untreated and the postchemotherapy samples to assess methylation evolution under therapeutic pressure. Also, we lack sufficient follow-up to robustly compare outcomes of the different subgroups to detect any survival differences. Finally, the analysis of methylation was not extended to RAD51C promoter methylation which could represent the fourth cause of HRD after BRCA1/2 mutations, BRCA1 promoter methylation, and RAD51C/D mutations (27).

To conclude, this is a large cohort of patients with first-line ovarian cancer, representative of newly diagnosed high-grade ovarian cancer, clinically and genomically annotated, evaluating the clinical relevance of BRCA1 promoter methylation. We demonstrated that approximately 20% of BRCAwt patients demonstrate high-BRCA1meth, are likely to be HRD and therefore good candidates to PARPi maintenance. Their prognosis is also similar to BRCAmut ovarian cancer. Similarly to others, we detected some low-BRCAmeth samples after chemotherapy exposure, suggesting partial BRCA1 methylation loss, and these patients also relapsed prematurely following first-line platinum and would likely be PARPi resistant. Finally and intriguingly, the low-methylated, untreated samples showed high genomic instability levels. Extensive molecular characterization and long follow-up of these patients might reveal new insights that can, with time, impact the therapeutic strategy. Our results support the clinical usefulness of BRCA1 methylation testing to determine the etiology of HRD in all patients with advanced ovarian cancer.

P. Pautier reports other support from AstraZeneca and GSK during the conduct of the study as well as other support from MSD and ESAI and personal fees from PharmaMar outside the submitted work. J. Michels reports personal fees from GSK board and Brenus Pharma and grants from Merck outside the submitted work. A. Elies reports personal fees from GSK outside the submitted work. E. Rouleau reports grants and personal fees from AstraZeneca and grants from Clovis, Roche, GSK, BMS, and MSD outside the submitted work. A. Leary reports grants from Sanofi and Inivata; personal fees and other support from AstraZeneca; personal fees from GSK Tesaro and Clovis; and non-financial support from Biocad, Seattle Genetics, and Ability outside the submitted work. No disclosures were reported by the other authors.

F. Blanc-Durand: Conceptualization, formal analysis, validation, investigation, writing–original draft, project administration, writing–review and editing. R. Tang: Investigation, methodology, writing–review and editing. M. Pommier: Formal analysis, methodology. M. Nashvi: Formal analysis, methodology. S. Cotteret: Formal analysis, supervision, validation. C. Genestie: Resources, supervision. A. Le Formal: Formal analysis, methodology. P. Pautier: Conceptualization, supervision. J. Michels: Conceptualization, supervision. M. Kfoury: Conceptualization, supervision. R. Hervé: Supervision. S. Mengue: Supervision. E. Wafo: Supervision. A. Elies: Supervision. G. Miailhe: Supervision. J. Uzan: Supervision. E. Rouleau: Conceptualization, formal analysis, supervision, validation, methodology, writing–review and editing. A. Leary: Conceptualization, formal analysis, supervision, validation, methodology, writing–review and editing.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.