BRCA loss of function including BRCA1 DNA-methylation, but not BRCA-unrelated homologous recombination deficiency, is associated with platinum hypersensitivity in high-grade ovarian cancer

Abstract

Background

In high-grade ovarian cancer (HGOC), determination of homologous recombination deficiency (HRD) status is commonly used in routine practice to predict response to platinum-based therapy or poly (ADP-ribose) polymerase inhibitors (PARPi). Here we tested the hypothesis that BRCA loss of function (LOF) due to epigenetic or genetic aberrations is a better predictor for the clinical outcome than HRD. One hundred thirty-one HGOC tissues were tested for BRCA DNA-methylation, BRCA mutations, HRD and BRCA1 mRNA expression, followed by a comprehensive survival analysis.

Results

BRCA1-methylation was detected in 11% of the tumors, exclusively in BRCA1-wild-type (wt) HGOCs. BRCA1-methylated tumors (BRCA1-meth) had HRD-scores similar to those of BRCA-mutated (mut) tumors, and higher compared to unmethylated-BRCA-wt tumors (BRCA-wt-unmeth; P < 0.001). Platinum-refractory or -resistant HGOCs at first recurrence were all BRCA-unmeth cancers. Only one of the BRCA-mut cancers had a platinum-resistant recurrence. Thus, 99% of relapses in cancers with epigenetic or genetic BRCA-alterations were platinum-sensitive. Multivariate analysis confirmed BRCA-LOF as an independent predictor of progression-free survival (PFS) and overall survival (OS), whereas HRD-status had no predictive value for PFS and OS. Patients with BRCA-wt-unmeth cancers had the worst outcome compared to patients with cancers harboring epigenetic or genetic BRCA-alterations (PFS: P = 0.007; OS: P = 0.022). Most importantly, the BRCA-wt-unmeth subfraction of HRD-positive HGOCs exhibited the same poor survival as the entire HRD-negative cohort.

Conclusion

In HGOC BRCA mutational status together with BRCA1-methylation exhibit the best predictive power for favorable clinical outcome and thus high sensitivity to platinum-based therapy, whereas BRCA-unrelated HRD positivity was not associated with improved platinum sensitivity.

Background

Epithelial ovarian cancer (EOC) is the eighth most common cancer in women and is the leading cause of death from gynecologic cancers [1, 2]. Most EOCs are diagnosed at an advanced stage (FIGO-stage III–IV), of which approximately 85% are HGOC.

Treatment options for EOC have evolved significantly in recent years with the use of targeted therapies, such as antiangiogenic agents and PARPi [3,4,5,6]. Homologous recombination repair (HRR) deficiency (HRD) is considered to be a predictive biomarker for platinum and PARPi therapy response in HGOC [7, 8]. A well-recognized intrinsic biomarker of PARPi response in EOC is platinum sensitivity, thus leading to the approval of olaparib, niraparib and rucaparib monotherapy as maintenance therapy in platinum-sensitive recurrent EOC [9,10,11]. HRD occurs in about half of HGOC as shown by pathway analyses or clinical trials [12, 13]. The main causes of HRD are germline or somatic mutations in the BRCA1 and BRCA2 genes (13–20%), but also other genes of the HRR-pathway (e.g. PALB2, RAD51C/D) and a number of currently unknown factors [12, 14,15,16,17,18]. Identification of additional mechanisms leading to HRD may increase the number of patients who might benefit from PARPi-therapy. BRCA1 DNA-methylation, another mechanism of BRCA1-inactivation, has been reported in 5–12% in OCs [12, 19,20,21,22,23]. To date, BRCA-methylation determination is not separately performed in routine analysis. Kalachand et al. recently described in a meta-analysis including data from 2636 OC patients across 15 studies that BRCA1-methylated (BRCA1-meth) OC showed similar clinicopathological features as BRCA1-mut OC in terms of high grade and young age at diagnosis, but unlike BRCA-mutated (BRCA-mut) cancers, BRCA1-meth OC was not associated with improved survival [24]. Recently, we described that BRCA1/2 mRNA expression has been shown to be of clinical significance in OC [25]. Here, we investigated BRCA loss of function (LOF) via epigenetic or genetic aberrations in relation to HRD and BRCA1 mRNA expression to gain a better understanding of the clinical significance of BRCA-LOF emphasizing also BRCA1 promotor methylation for platinum sensitivity and its impact on the prognosis of OC patients.

Methods

Patients and samples

EOC tissue samples from 131 patients obtained during primary debulking (n = 123) or biopsy prior to neoadjuvant chemotherapy (n = 8) were collected and processed at the Department of Obstetrics and Gynecology, Medical University of Innsbruck (MUI) between 1989 and 2021. Staging was performed according to the International Federation of Gynecology and Obstetrics (FIGO) classification system. Only patients who received platinum-based first-line chemotherapy were eligible for this study; a total of 54 patients received bevacizumab maintenance therapy; PARPi were not yet approved for routine use at that time. All patients were monitored within the outpatient follow-up program of our department. Clinical, pathological and follow-up data were stored in a database according to the privacy policy of our hospital. The study was approved by the Ethics Committee of the MUI (reference numbers: AN2015-0038 346/4.17; 1054/2019) and conducted following the Declaration of Helsinki. An overview of the molecular analyses performed on the patient samples and the corresponding case numbers is shown in Supplementary Fig. 1.

DNA-isolation

Genomic DNA was isolated from pulverized, quick-frozen specimens (n = 131) using the DNeasy Tissue Kit (Qiagen, Hilden, Germany).

BRCA-methylation analysis

Bisulfite-modification of all 131 tumor DNA samples was performed using the EZ DNA-Methylation-Gold-Kit (Zymo Research, CA, USA) according to the manufacturer's instructions. Primers and probes for BRCA1- and BRCA2 promoter-methylation analysis were used as previously described [26]. MethyLight-analysis was performed and the percentage of methylated reference (PMR) values was calculated, as previously reported [25], using a BRCA1 PMR > 10 as the cut-off for promoter-hypermethylation, as described by Weisenberger et al. [26]. PMR-values were adjusted for the proportion of tumor cells (this proportion was calculated within each sample based on the variant allele frequencies (VAF) of SNPs in regions of LOH in neoplastic cells).

BRCA-mutation analysis

Targeted NGS was conducted for all 131 tumor samples using the TruSight Cancer Sequencing Panel (Illumina, San Diego, USA) on the Illumina MiSeq®- and the NextSeq-system (Illumina, CA, USA). Mutation analysis was performed using SeqNext® Software (JSI medical systems GmbH, Ettenheim, Germany). Variants with a VAF of at least 10% were classified according to the consensus recommendations of the American College of Medical Genetics [27].

Array analysis

The DNA from 120 tumor samples was analyzed using the Global Screening Array (Illumina) according to the manufacturer’s protocol. To determine the tumor cell proportion, data were analyzed with Illumina GenomeStudio 2.0 and NxClinical (Bionano, San Diego, CA, USA, SNP-FASST2-Segmentation Algorithms). For the determination of tumor cell proportion, regions of copy number neutral loss of heterozygosity (LOH) were selected and the tumor cell proportion (aberrant cell fraction) was calculated based on the variant allele frequency of the SNPs. For the determination of the HRD-score, regions of LOH, allelic imbalance and copy number variations were quantified.

HRD-assessment

HRD-data were available for 120 patients analyzed in this study. For HRD-quantification, a LOH-score and an Aneuploidy-Normalized-Telomeric-Imbalance (ANTI)-Score were used, based on the LOH-score published by Frampton et al. and using a modification of the Telomeric-Allelic-Imbalance (TAI)-score by Birkbak et al. [28, 29]. Both variables were equally weighted and integrated into a common HRD-score with a diagnostic cut-off value of ≥ 1.77 to discriminate between HRD-positive (HR-deficient) and HRD-negative (HR-proficient) tumors [30]. To determine genome-wide copy number variation and allelic variation, the DNA was analyzed by the Global Screening Array (Illumina) according to the manufacturer’s protocol, and data analysis was performed using Illumina GenomeStudio 2.0 and NxClinical (Bionano, SNP-FASST2-Segmentation Algorithms) software. This HRD-test has been validated by paired analyses using Myriad MyChoice DX (Myriad Genetics). A correlation-coefficient of 0.86 was revealed between the respective scores, with 100% agreement in the classification of samples as HRD-positive and HRD-negative [30].

BRCA1 mRNA expression analysis

RNA-isolation, reverse-transcription and quantitative real-time PCR were performed as previously described [25]. Primers and probe for BRCA1 were purchased from Applied Biosystems (Foster City, CA, USA, Applied Biosystems Assay ID: Hs01556193_m1). BRCA1-mRNA expression was adjusted to tumor cell proportion (n = 63).

Statistics

Comparisons between two continuous variables were made using the Mann–Whitney test and between more continuous variables using the Kruskal–Wallis test. The Chi-square test was used to test for differences in categorical variables. Survival analysis was performed using Kaplan–Meier curves and the log-rank test. Cox-regression analysis was used for multivariate survival analysis (all variables from in the univariate analysis which revealed P values < 0.2 were included in the multivariate analysis). For analyses of BRCA1 DNA methylation in relation to platinum sensitivity, progression-free patients who were lost to follow-up within 6 months after the last platinum treatment (n = 9) were excluded. All statistical analyses were performed using SPSS (version 29.0; SPSS Inc., Chicago, IL, USA).

Results

BRCA1 methylation status in relation to BRCA mutations and clinical pathological characteristics

Analysis of BRCA1- and BRCA2-methylation in 131 HGOC tissue samples revealed BRCA1-methylation in 11% (14/131) of the tumors. No BRCA2-methylation was observed in any of the samples. Furthermore, in none of the BRCA1-mut cancers a BRCA1 promotor methylation was identified, suggesting that epigenetic silencing of BRCA1 and BRCA1-mutations may be mutually exclusive. BRCA1-methylation was almost exclusively found in 19% of BRCA-wild-type (wt) tumors (13/70), whereas only one BRCA2-mut tumor was BRCA1-meth (Table 1). Furthermore, 83% (10/12) of BRCA1-meth tumors were positively associated with HRD-positivity, whereas BRCA-wt-unmethylated (BRCA-wt-unmeth) tumors were predominantly HRD-negative (73%; 41/56) (P < 0.001). For completeness, in general, HRD-positive scores were detected in 63% of the samples analyzed (76/120) and in 98% of BRCA-mut tumors (51/52). The respective HRD-scores of the different molecular subgroups are depicted in Fig. 1A. Notably, the HRD-scores did not differ between BRCA1-meth (median HRD-score = 2.65) and BRCA1-mut tumors (median HRD-score = 2.73). Furthermore, lower BRCA1 mRNA expression was observed in BRCA1-meth or BRCA1-mut tumors in comparison to BRCA1–wt tumors. (Fig. 1B; P = 0.005, P = 0.031).

Table 1 Associations of BRCA DNA-methylation and clinicopathological and genetic features

Fig. 1

BRCA1 DNA-methylation in relation to BRCA-mutations, HRD and BRCA1 mRNA expression. (A) HRD-scores in different OC subgroups according to BRCA epigenetic or genetic aberrations (n = 120). HRD-score threshold was defined as ≥ 1.77 (red line). The HRD-scores above this threshold were considered HRD-positive (HR-deficient) and values below the threshold were considered HRD-negative (HR-proficient). Note: The BRCA2-mut-meth tumor is included in the BRCA-mut group. (B) BRCA1 mRNA expression in relation to BRCA1 epigenetic or genetic aberrations (n = 63) Note: BRCA2-mut tumors are included in the BRCA1-wt group, the BRCA2-mut-meth tumor is included in the BRCA1-meth group

BRCA-LOF by epigenetic or genetic aberrations was associated with younger age (P = 0.007), but with no other classic clinicopathological features as shown in Table 1.

BRCA-LOF by epigenetic or genetic aberrations and platinum sensitivity

Platinum sensitivity was defined by the interval between the last platinum administration and the first recurrence: disease progression within six months was considered platinum-resistant and after six months platinum-sensitive. Referring selectively to the BRCA-methylation and -mutational status, 71 of 72 cancers with epigenetic (n = 12) or genetic (n = 59) BRCA-aberrations were found to be platinum-sensitive (positive predictive value (PPV) = 99%; P < 0.001), and 92% (11/12) of HGOC patients classified as platinum-resistant had BRCA-wt-unmeth tumors, whereby it should be noted that none of the platinum-resistant cancers showed BRCA1-methylation. We also analyzed this separately in HGOC-patients with residual disease after primary surgery (n = 35) in order to obtain better comparability and a more precise estimation of platinum-responsiveness. Again, 24 out of 25 tumors with epigenetic (n = 5) or genetic (n = 19) BRCA-aberrations were classified as platinum-sensitive (PPV = 96%; P = 0.001) and 83% (5/6) of HGOC patients with assumed platinum-resistant cancers had BRCA-wt-unmeth tumors.

However, based on the common diagnostic approach combining the BRCA-mutational and HRD-status, 80 of 84 BRCA-mut and/or HRD-positive tumors were platinum-sensitive (PPV = 95%; P = 0.011) and only 64% (7/11) of patients considered as platinum-resistant had BRCA-wt, HRD-negative tumors. In patients with residual disease, 25 of 27 BRCA-mut and/or HRD-positive tumors were platinum-sensitive (PPV = 93%; P = 0.005) and 67% (4/6) of cancers defined as platinum-resistant were BRCA-wt, HRD-negative. When considering HRD-status alone, the PPV for platinum-sensitive recurrent disease was 91% (20/22) in cases with residual disease.

BRCA-LOF by epigenetic or genetic aberrations and survival

Univariate survival analysis revealed that patients with tumors harboring BRCA-LOF had a favorable progression-free survival (PFS) and overall survival (OS) (median-PFS: 2.7 years, 95% CI 1.5–3.9; median-OS: 7.8 years, 95% CI 6.0–9.6) compared to patients with BRCA-wt-unmeth cancer (median-PFS: 1.8 years, 95% CI 1.1–2.4; median-OS: 5.4 years, 95% CI 3.5–7.3) as shown in Table 2 and Figs. 2A and 2B (PFS: P = 0.011; OS: P = 0.047). This was also confirmed in the multivariate analysis (Table 3A; PFS: HR 0.5, 95% CI 0.3–0.8, P = 0.007 and OS: HR 0.6, 95% CI 0.3–0.9, P = 0.022).

Table 2 Univariate survival analysis in 131 HGOC-patients

Fig. 2

BRCA-LOF by epigenetic or genetic aberrations and survival. (A) Analysis of progression-free survival (PFS) and (B) overall survival (OS). Subgroup analyses of HRD-positive BRCA-wt-unmeth HGOC for (C) PFS and (D) OS compared with HRD-negative BRCA-wt-unmeth and with BRCA-LOF HGOC and for (E) PFS and (F) OS compared with the entire HRD-negative cohort

Table 3 Multivariate analysis in HGOC patients

Interestingly, the subfraction of patients with HRD-positive tumors which are unrelated to BRCA-LOF (HRD-pos. BRCA-wt-unmeth) representing 20% (15/76) of all HRD-positive tumors; with a median-PFS of 1.8 years, (95% CI 1.4–2.2) and a median-OS of 6.6 years, (95% CI 0.2–13.1) did not differ in survival from patients with HRD-negative BRCA-wt-unmeth tumors (median-PFS: 1.8 years, 95% CI 0.5–3.0; median-OS: 5.4 years, 95% CI 3.2–7.6) (Table 2; Fig. 2C, D), nor did they differ from patients of the entire HRD-negative cohort (median-PFS: 2.1 years, 95% CI 1.0–3.3, P = 0.533; median-OS: 5.4 years, 95% CI 3.3–7.5, P = 0.743; Fig. 2E, F). Survival analysis of these patients with HRD-positive tumors unrelated to BRCA-LOF in relation to BRCA-LOF tumors, revealed a significantly worse PFS (Fig. 2C; P = 0.048), but no difference in OS (Fig. 2D; P = 0.131).

Neither HRD- nor BRCA mutational status as usually determined in routine practice had any predictive value in our cohort (Table 2; Table 3B).

A subgroup analysis revealed that superior survival in patients with tumors harboring BRCA-LOF was only observed in patients without the addition of bevacizumab to chemotherapy as a maintenance-therapy (median-PFS: 2.5 years, 95% CI 0.5–4.5, median-OS: 7.8 years, 95% CI 5.8–9.7) compared to patients with BRCA-wt-unmeth HGOC (median-PFS: 1.3 years, 95% CI 0.8–1.9; median-OS: 3.9 years, 95% CI 0.0–8.5) (Table 2; PFS and OS each P = 0.030). This survival advantage disappeared in patients treated with bevacizumab (Table 2). These findings were confirmed in the multivariate analyses shown in Table 3C for PFS (HR 0.4, 95% CI 0.2–0.8, P = 0.007) and OS (HR 0.5, 95% CI 0.3–0.9, P = 0.012).

Discussion

Based on data from The Cancer Genome Atlas (TCGA), Yang et al. showed loss of BRCA1 mRNA expression by BRCA1-methylation in 10% (33/316) of HGSOC and no promoter-hypermethylation of BRCA2 [23]. This was confirmed by Abkevich et al. [22] and is also consistent with our results. According to our data, BRCA1-methylation appears to affect the DNA HRR-mechanism to the same extent as pathogenic mutations of both BRCA-genes, substantiated by an almost identical extent of genome-wide scarring (HRD-score) in both cancer subgroups. This is in agreement with the findings of Kalachand et al. [24].

In recent years, the prognostic and predictive significance of BRCA1-methylation in OC has been the subject of intense debate. A decade ago, TCGA-data on OC-patients treated with platinum-based chemotherapy showed a significant OS superiority in BRCA-mut compared to BRCA-wt cancers, but this was not the case for BRCA1-wt-meth cancers, which behaved like BRCA-wt cancers [12].

A recent meta-analysis with data from 2,636 OC patients showed that BRCA1-methylation was not associated with increased survival [24]. However, in our work we focus on BRCA-LOF due to epigenetic or genetic aberrations and compared this combined assessment with clinicopathological features and the determination of HRD-status to predict clinical outcome. BRCA-LOF was detected predominantly in tumors from younger patients, but there was no association between BRCA1 methylation and age. It would appear that the status of the BRCA mutation is the primary driver of this effect, with germline BRCA mutations potentially associated with this finding. In terms of survival, it is noteworthy that the subfraction of patients with tumors exhibiting HRD-positivity unrelated to BRCA-LOF had the same poor survival as patients with HRD-negative cancers. Overall, our data show that from the generally expected 50% HRD-positive HGOCs, only the group of BRCA-LOF cancers, consisting of BRCA-mut or BRCA1-meth cancers represent those with exceptionally high sensitivity to platinum-based chemotherapy.

Interestingly the proportion of patients with platinum-resistant recurrence among BRCA1/2 mutant cases in our cohort is very low compared to the literature -Alsop et al. report a percentage of 14.9% of platinum-resistant patients with BRCA1/2 mutations [31]. Multifactorial causes such as BRCA mutation type and prevalence, surgical radicality, the interval between surgery and the start of chemotherapy, chemotherapy dose density, the use of combination chemotherapy or chemo-monotherapy could explain this difference.

Platinum-sensitivity of cancers is closely related to their sensitivity to PARPi and is therefore considered to be an intrinsic biomarker for PARPi-efficacy. Overlapping gene dependencies for carboplatin- and PARPi-response have been previously described by Coehlo et al. [32]. Therefore, we are tempted to speculate that the combined assessment of epigenetic or genetic BRCA-inactivation may also be an important surrogate to reliably predict PARPi response. However, this hypothesis must be evaluated in separate clinical trials of PARPi therapy.

In our survival analyses, the BRCA-LOF due to epigenetic or genetic BRCA-inactivation was a more powerful predictor of HGOC-outcome than HRD-testing. Patients with BRCA-wt-unmeth tumors had the most unfavorable clinical outcome, regardless of HRD-status, presumably due to impaired platinum-sensitivity. The fact that none of platinum-resistant cancers in our investigations showed BRCA1-methylation supports this hypothesis.

Interestingly, we also observed that in the group of patients receiving bevacizumab maintenance-therapy, the survival disadvantage in patients with BRCA-wt-unmeth tumors disappeared. It may be assumed that this group of patients with less platinum sensitive tumors will derive the greatest benefit from antiangiogenic therapy.

In a recent post-hoc exploratory biomarker analysis of pre- and post-platinum samples from ARIEL2-study (a single-arm, open-label, phase-2 trial testing efficacy of the PARPi rucaparib in recurrent HGOC), Swisher et al. described that high-level methylation of the BRCA1-promoter measured by quantitative methylation-sensitive digital droplet polymerase chain reaction (MS-ddPCR) was a strong biomarker of rucaparib-sensitivity, with similar power as BRCA-mutations [33]. Also Kondrashova et al. identified homozygous or hemizygous BRCA1 methylation as a predictor for rucaparib clinical response using HGSOC patient-derived xenografts (PDX) and archival tumor and pre-treatment biopsy samples from 23 patients from ARIEL2-study [34]. Blanc-Durand et al. recently showed that OCs with high BRCA1-hypermethylation were very likely to exhibit high genomic instability scores as an indicator of HRD which is in line with our observations [35]. They identified BRCA1-methylation in 19% of BRCA-wt tumors and concluded that these patients would therefore be good candidates for PARPi-treatment. They proposed to differentiate between biallelic BRCA1-promoter methylation (high-methylation; ≥ 70%) leading to a homozygous gene silencing and a monoallelic methylation with incomplete silencing of BRCA1 (low-methylation; 30–69%). Despite this distinction, these authors did not find a statistically significant difference in PFS.

In our analysis, we did not use a cutoff to dichotomize our small cohort of 14 methylated cancers into high and low BRCA1 DNA methylation categories. But the median adjusted PMR-value among these BRCA1 methylated specimens was quite high at 70.6, with a minimum PMR-value of 32.7 (data not shown). However, we could show in our cohort that BRCA1-methylation lead to a significant loss of BRCA1 mRNA-expression and consequently to a BRCA-LOF, the primary end-point of our study.

It should be noted that the number of cases included in our study represents a limitation together with the fact that none of the included patients was treated with PARPi. Therefore, further validation in a larger cohort, preferably in patients with HGOC treated with first-line PARPi maintenance, would be beneficial.

Conclusion

There is a high risk that parts of the clinically relevant subgroup of BRCA1-meth OC will remain unrecognized if only HRD-testing is performed in addition to BRCA mutation analysis. Distinguishing this subgroup is particularly important because completely BRCA-unrelated HRD-positive cancers have been shown to lack high sensitivity to platinum-based chemotherapy and probably also to PARPi treatment. Therefore, we strongly advocate the additional determination of BRCA1 methylation status to BRCA mutation status and HRD status in HGOC for a more accurate prediction of sensitivity to platinum and, presumably, to PARPi.

Further retrospective validation of a joint determination of BRCA loss by epigenetic or genetic aberrations in HGOC patients treated with first-line PARPi maintenance therapy is urgently warranted. This could validate our further hypothesis that the beneficial effect of PARPi maintenance therapy in HRD-positive BRCA-wt HGOCs shown in large randomized phase III clinical trials is predominantly due to the BRCA1-meth subfraction of the tumors.