Identifying BRCA wild-type ovarian cancers with a homologous recombination deficiency phenotype (101)

Abstract

<b>Objectives:</b> It is unknown how best to evaluate <i>BRCA</i> wild-type ovarian carcinomas for homologous recombination deficiency (HRD), and previous studies did not show a survival advantage for <i>BRCA1</i> methylation. We sought to further characterize <i>BRCA1/RAD51C</i> promoter methylation in ovarian cancer (OC). We hypothesized that a combined assay with quantitative methylation and mutation assessment would identify <i>BRCA</i> wild-type ovarian carcinomas with an HRD phenotype. <b>Methods:</b> Digital droplet PCR primers and probes were designed to target CpG sites in the promoter regions of <i>BRCA1</i> and <i>RAD51C</i> known to be associated with transcriptional repression. Multiple aspects of the assay were tested and optimized in a CLIA environment. The fraction of methylated alleles was corrected for neoplastic cellularity and ≥60% was considered highly methylated. The methylated ddPCR (MddPCR) assay was then applied to DNA from banked epithelial ovarian cancer specimens previously assessed by methylation-sensitive PCR (MSP) and BROCA next-generation sequencing. RNA expression of a subset of <i>BRCA1</i> methylated samples was determined using RNA-seq, and expression was compared with methylation using LOESS regression. Clinical characteristics were then compared between highly methylated and unmethylated samples using a t-test for continuous variables, Fisher's exact test for categorical variables, and the log-rank test for overall survival (OS). Additional cases with germline/somatic mutations in homologous recombination repair (HRR) genes identified by BROCA sequencing were included in an overall survival analysis using the log-rank test. <b>Results:</b> Total 370 epithelial OC specimens were tested by MddPCR. MddPCR had 100% concordance with MSP. Among the samples, 9.5% were positive for <i>BRCA1</i> promoter methylation, and 3% were positive for <i>RAD51C</i> promoter methylation ≥5%. RNA expression analysis of a sub-set of <i>BRCA1</i> methylated samples demonstrated that the degree of <i>BRCA1</i> methylation was negatively correlated with <i>BRCA1</i> expression levels (r=-0.82, p<0.001). There were 26 highly methylated neoplastic <i>BRCA1</i> or <i>RAD51C</i> samples, 14 samples with low methylation (5-59% methylated alleles), and 294 samples with no methylation or HRR mutation. The median overall survival was 76 months for highly methylated cases compared to 50 months for unmethylated cases (HR: 0.66, p=0.20). Highly methylated samples were significantly less likely to have had prior chemotherapy exposure (12% vs 31%, <i>p</i>=0.04). Paired primary/recurrent samples were evaluated, two of which contained a highly methylated initial sample and showed a decrease in methylation at the time of recurrence. In the expanded cohort including HRR mutated OC, median OS was 80 months in cases with either high methylation or a non-<i>BRCA</i> HRR mutation compared to 50 months in unmethylated wild-type samples (HR: 0.63, p=0.02), with a survival curve similar to <i>BRCA</i> mutated samples (median OS: 66 months, HR: 0.76, p=0.01). <b>Conclusions:</b> Ovarian cancer that was either highly methylated or had a non-<i>BRCA</i> HRR mutation was associated with increased OS compared to unmethylated/wild-type cases, suggesting an HRD phenotype. Methylated samples had similar OS to HRR mutated cancers; however, survival analysis was limited by the low total number of highly methylated samples. Exposure to chemotherapy may decrease promoter methylation, and analysis of additional paired samples is ongoing. Quantitative methylation analysis could augment tumor sequencing in identifying OC with a favorable response to PARP inhibitors.Fig. 1