Mutations in Homologous Recombination Genes and Outcomes in Ovarian Carcinoma Patients in GOG 218: An NRG Oncology/Gynecologic Oncology Group Study

BRCA1 and BRCA2 are key components of the BRCA–Fanconi anemia DNA repair pathway that controls DNA repair via homologous recombination (1–3). Germline and somatic mutations affecting homologous recombination repair (HRR) genes are relatively common, found in approximately one third of ovarian carcinoma patients, with the majority of mutations occurring in BRCA1 and BRCA2 (4, 5). Ovarian carcinoma patients with inherited mutations in BRCA1 and BRCA2 have longer 5-year survival, likely due to improved sensitivity to platinum chemotherapy (6–8). This may also be true for somatic and germline mutations in other HRR genes, although the number of cases with non-BRCA HRR mutations studied to date is small (5). In addition, BRCA1- and BRCA2-mutated ovarian carcinomas may have a lower angiogenic profile (9) and more favorable immunophenotype than those without mutations, which might also contribute to better outcomes (10, 11).

GOG-0218 was a randomized, phase III trial for patients with primary advanced ovarian, peritoneal, and fallopian tube carcinoma (ovarian carcinoma), examining the addition of the antiangiogenesis drug bevacizumab to standard chemotherapy with carboplatin and paclitaxel. Patients in this study who received extended bevacizumab had a significant prolongation in median progression-free survival (PFS) of 3.8 months (12). In the parallel European ICON-7 trial, a non-prespecified subgroup analysis suggested that extended bevacizumab was of greatest benefit for patients with poor prognostic features, such as stage IV disease, or unresectable or suboptimally resected tumors, demonstrating a benefit in overall survival (OS) in these “high-risk” patients (13).

We hypothesized that ovarian carcinoma patients with inherited mutations in HRR genes, including a select set of genes beyond BRCA1 and BRCA2, have longer survival and improved platinum sensitivity and may derive less benefit from extended bevacizumab. We tested this hypothesis using tissues from patients in GOG-218 who had provided consent for translational research.

GOG-0218 clinical trial details have been described previously (12). DNA from tumor and blood from patients on GOG-0218 were sequenced using BROCA, a targeted capture, massively parallel sequencing test developed at the University of Washington (Seattle, WA; ref. 14). All available germline DNA samples extracted from blood were sequenced (n = 788). These patients have been described previously (8). To improve detection of HRR deficiency in this group by detection of somatic mutations in HRR genes, we also sequenced DNA from neoplastic tissue from a subset of these patients with negative germline testing (n = 324). Finally, we sequenced DNA from neoplastic tissue from 407 additional cases that did not have available DNA from blood, detecting both germline and somatic mutations (bringing the total number of sequenced patients to 1,195). Distinguishing between germline and somatic mutations could not be done with certainty in the group with only tumor sequencing. As the effect on outcome was predicted to be similar, germline and somatic mutations were combined for analyses. All patients provided written informed consent as approved by an Institutional Review Board, in accordance with ethical guidelines as described in the U.S. Common Rule. As patients were not specifically consented for open access genomic data, complete sequencing data, such as BAM files, cannot be shared publicly.

A subset of genes predicted to impact homologous recombination repair (HRR) when mutated was selected prior to analysis based on review of the available literature and expert opinion, including ATM, ATR, BARD1, BLM, BRCA1, BRCA2, BRIP1, CHEK2, MRE11A, NBN, PALB2, RAD51C, RAD51D, RBBP8, SLX4, and XRCC2. Sequencing reads were aligned to the human reference genome (hg19). Variants were identified using GATK37 and Pindel after indel realignment and base quality calibration. Copy number variations were detected in germline samples as described previously (14–16). Missense mutations were only included if proven to be damaging in functional assays or classified as likely pathogenic on ClinVar (https://www.ncbi.nlm.nih.gov/clinvar/; ref. 17).

Clinical information was collected as per the GOG-0218 protocol (12). All pathology was centrally reviewed. PFS and OS were defined as the time between enrollment and progression or death, respectively. Proportional hazards models were used to provide estimates of the relative hazards for progression and death by genotype, adjusted by clinical characteristics. The effect of mutation status on the impact of bevacizumab on progression was assessed with a test of interaction. All P values are 2-sided.

Of 1,873 women in GOG 218, 1,195 (63.8%) had DNA available for sequencing. Baseline characteristics of sequenced and not sequenced patients are described in Table 1. The distribution of race was significantly different between the groups, with more non-Hispanic Whites and fewer Asian/Pacific Islanders in the sequenced group (P < 0.001, χ²). There were more optimally debulked stage III patients and less stage IV patients in the sequenced group (P < 0.001, χ²). The distribution of histologic subtypes also differed, with fewer patients with clear cell carcinoma in the sequenced group (P = 0.012, χ²). The median PFS duration for those who were not sequenced was 10.5 months compared with 13.5 months for those sequenced (P < 0.001) and the median overall survival for those not sequenced was 30.9 months compared with 46.2 months for those who were sequenced.

In total, 307 of 1,195 (25.7%) sequenced patients had damaging mutations in one of the selected genes affecting HRR. There were 148 (12.4%) with mutations in BRCA1, 78 (6.5%) with mutations in BRCA2, and 81 (6.8%) with mutations in other HRR genes. Ten patients (0.8%) had more than one mutation. Of these 10, three were in the BRCA1 mutation group, with additional mutations in RBBP8 (one) and NBN (two), and four were in the BRCA2 group (one with CHEK2, one with RBBP8, one with both ATM and CHEK2, and one with BRIP1 and CHEK2). Of the 81 patients in the other HRR group, there were 84 mutations in the following genes: 12 in ATM, five in ATR, three in BARD1, five in BLM, 19 in BRIP1, two in CHEK2, two in MRE11A, three in NBN, six in PALB2, seven in RAD51C, seven in RAD51D, eight in RBBP8, two in SLX4, and three in XRCC2. Details of the patients with non-BRCA HRR mutations are provided in Supplementary Table S1.

In the subset of 324 patients with negative germline testing who had subsequent sequencing of neoplastic DNA, there were 32 (9.9%) somatic mutations in HRR genes [(17 (5.2%) in BRCA1, seven (2.2%) in BRCA2, and eight (2.5%) in other HRR genes]. In a small exploratory analysis, these 32 “definitely somatic” mutation carriers had significantly lower hazards of both progression and death when compared with 174 patients with “definitely germline” mutations (Supplementary Fig. S1). Mutation rates by the source of sequenced DNA are described in more detail in Supplementary Table S2.

The proportion of patients in each mutation group by clinical characteristics is shown in Table 2. Mutation status did not differ by disease stage or primary site. Mutations in HRR genes were found in all histologic subtypes. The overall mutation rate in the high-grade serous carcinomas was 27.0% (262/971), which was not significantly different from the rate in unspecified carcinoma (22/101, 21.8%, P = 0.29), endometrioid (10/42, 23.8%, P = 0.73), clear cell (6/28, 21.4%, P = 0.67), or mucinous carcinoma (2/7, 28.6%, P = 1.0). The low-grade serous carcinomas had a lower HRR mutation rate of 10.9% [5/46, 95% confidence interval (CI), 4.8–23.1; P = 0.02, Fisher exact] when compared with high-grade serous carcinomas.

Damaging mutations in BRCA1, BRCA2, or other non-BRCA HRR genes were all associated with longer PFS and OS relative to cases without mutations (Fig. 1). After adjusting for study treatment, stage of disease, size of residual disease, and initial performance status, adjusted HRs for both progression and death were significantly lower in cases with HRR mutations when compared with those with no mutations (Fig. 1). This effect was strongest for BRCA2 mutations and was similar for BRCA1 and non-BRCA HRR mutations.

For this analysis, only patients from arm 1 (carboplatin/paclitaxel with placebo) and arm 3 (carboplatin/paclitaxel with bevacizumab and bevacizumab maintenance) were included (n = 809). In patients with no mutations (n = 581), extended bevacizumab significantly prolonged PFS (15.7 months vs. 10.6 months; HR = 0.71; 95% CI, 0.60–0.85; P = 0.0001; Fig. 2A). In those with mutations (n = 228), extended bevacizumab conferred a median PFS of 19.6 months versus 15.4 months (HR = 0.95; 95% CI, 0.71–1.26; Fig. 2C). Using a test of interaction, mutation status did not significantly modify the effect of extended bevacizumab on progression [(0.95/0.71) = 1.33; 95% CI, 0.95–1.85; P = 0.10].

GOG-0218 was a large randomized, placebo-controlled phase III clinical trial with standardized treatment, central pathology review, and follow-up, providing a unique opportunity to assess the impact of HRR deficiency on clinical outcomes. In this trial, patients with ovarian carcinoma with defective HRR, defined by damaging somatic or germline mutations in one of 16 genes, had significantly prolonged PFS and OS when compared with those without mutations (Fig. 1). This is the first study in ovarian carcinoma large enough to separately assess the impact on outcomes of non-BRCA HRR mutations, which are less common than mutations in BRCA1 and BRCA2, although our smaller single institution study showed a trend toward improved OS with mutations in a subset of non-BRCA genes (5). We carefully selected a list of 14 non-BRCA genes a priori that we thought would be most likely to impact outcomes based on in vitro data. Ovarian carcinomas with mutations in other HRR pathway genes such as FANCM, FANCA, and FANCI were not classified as HRR deficient. Interestingly, the non-BRCA HRR mutations were associated with PFS and OS curves that were almost identical to BRCA1 mutations, supporting our a priori selection of HRR genes as meaningful.

As expected, BRCA1 and BRCA2 mutations were associated with longer survival, with the most profound effect seen for BRCA2 mutations with a median survival advantage of 33 months compared with those without mutations. These patients were treated years before PARP inhibitors became available, and these outcome differences would likely be even greater with current treatments. Previous studies have also demonstrated a better outcome for BRCA2 compared with BRCA1 mutations (5–8), but ours is the first large study to include somatic mutations and to assess outcomes relative to mutation status for patients treated on a phase III clinical trial with standardized treatment and follow-up. Given the meaningful impact of HRR deficiency on both PFS and OS, the presence of HRR mutations, including BRCA1 and BRCA2, should be carefully considered in the design and analyses of ovarian carcinoma clinical trials.

Mutations in HRR genes were found in all histologic subtypes of ovarian carcinoma, and rates were only significantly lower in low-grade serous carcinomas, 5/46 (10.9%) versus 262/971 (27%) in high-grade serous carcinomas, P = 0.02, (Table 2). All tumors in GOG-0218 underwent centralized review by gynecologic pathologists, minimizing pathologic misclassification. These results are similar to our recent observation that germline HRR mutations are distributed across many histologic subtypes without a clear predilection for high-grade serous carcinomas (8). Therefore, HRR status cannot be assumed by histologic type. Some clinical trials targeting HRR deficiency, such as those using PARP inhibitors, are restricted to patients with high-grade serous carcinomas (18), or to those with either high-grade serous or high-grade endometrioid histology (19). Our data do not support restricting access to these clinical trials or to germline genetic testing or tumor sequencing for HRR defects based on histologic subtype alone.

The role of antiangiogenesis-targeted therapy in ovarian carcinomas with and without HRR deficiency has not been previously defined. In the subset of patients with HRR mutations in GOG-0218, extended bevacizumab did not confer a statistically significant prolongation in PFS (Fig. 2C). However, there was insufficient evidence that bevacizumab had a different effect on PFS or OS in HRR-mutated versus nonmutated cases, using a test of interaction (Fig. 2). The power of this analysis was limited by not having translational samples from all participants in GOG-0218; therefore, the number of mutation carriers was relatively small. Sequenced patients were less likely to be stage IV and had better PFS and OS, which also may have reduced our ability to detect a difference. Within the similar ICON7 trial, a subset of 359 cases subjected to expression profiling, proliferative, and mesenchymal subtypes may have derived greater benefit from bevacizumab therapy; however, tests of interaction were not done to formally test that hypothesis. The mutation status of the ICON7 cancers was not determined (20). Further studies are needed to identify ovarian carcinomas best treated with bevacizumab, but our data do not currently support BRCA mutation status as a determinant.

This study has several limitations. Although germline sequencing data are available on all patients (by sequencing DNA from blood or tumor), around 40% of patients did not have tumor sequenced. This likely underestimated the true mutation frequency by missing some somatic mutations. In addition, as germline and somatic HRR mutations are rarely present in the same patient, utilizing a “germline negative” group (n = 324/1,195, 27.1%) for additional tumor sequencing likely overestimated the somatic mutation rate by excluding those with germline mutations who were unlikely to have somatic mutations (Supplementary Table S2). Exome sequencing data from The Cancer Genome Atlas (TCGA) demonstrated somatic or germline mutations in DNA repair genes in around 40% of patients with high-grade serous histology ovarian carcinomas (4), compared with 27% in patients with high-grade serous carcinomas in this trial. This difference is likely due to a combination of (i) our preselected shorter HRR gene list chosen to maximize likelihood of therapeutic impact; (ii) our incomplete tumor sequencing data in 40% of cases; and (iii) our more stringent definition of damaging mutations compared with TCGA. Finally, mutations in non-BRCA “other” HRR genes were too few to allow a survival analysis by individual gene, necessitating combining these patients into one group for analysis.

In conclusion, HRR deficiency is a strong predictor of both PFS and OS in ovarian carcinoma, including mutations in 14 non-BRCA HRR genes (ATM, ATR, BARD1, BLM, BRIP1, CHEK2, MRE11A, NBN, PALB2, RAD51C, RAD51D, RBBP8, SLX4, and XRCC2). Histologic subtype does not provide sufficient information to predict either inherited or somatic HRR mutations. There is insufficient evidence that HRR mutations should be a criterion, either positive or negative, in the decision to use bevacizumab maintenance therapy for advanced ovarian carcinoma.

T. Walsh is a consultant/advisory board member for Color Genomic. R.A. Burger is a consultant/advisory board member for Amgen, Astra Zeneca, Clovis, Genentech, Gradalis, Invitae, Janssen, Merck, Morphotek, NuCana, Tesaro, and VBL Therapeutics. K.S. Tewari reports receiving speakers bureau honoraria from Roche and is a consultant/advisory board member for Genentech. F. Backes is a consultant/advisory board member for Tesaro. M.C. King is a consultant/advisory board member for Color Genomics. No potential conflicts of interest were disclosed by the other authors.

The funders/sponsors had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Conception and design: B.M. Norquist, M.F. Brady, R.S. Mannel, M.C. King, M.J. Birrer, E.M. Swisher

Development of methodology: B.M. Norquist, M.F. Brady, T. Walsh, M.C. King, E.M. Swisher

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): B.M. Norquist, M.F. Brady, M.I. Harrell, T. Walsh, R.A. Burger, K.S. Tewari, R.S. Mannel, G. Glaser, C. Bailey, S. Rubin, J. Soper, H.A. Lankes, N.C. Ramirez, M.J. Birrer, E.M. Swisher

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): B.M. Norquist, M.F. Brady, M.I. Harrell, T. Walsh, M.K. Lee, S. Gulsuner, S. Casadei, R.A. Burger, K.S. Tewari, F. Backes, M.C. King, M.J. Birrer, E.M. Swisher

Writing, review, and/or revision of the manuscript: B.M. Norquist, M.F. Brady, M.I. Harrell, R.A. Burger, K.S. Tewari, F. Backes, R.S. Mannel, G. Glaser, S. Rubin, J. Soper, H.A. Lankes, N.C. Ramirez, M.C. King, M.J. Birrer, E.M. Swisher

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M.F. Brady, S.S. Bernards, H.A. Lankes, M.C. King, E.M. Swisher

Study supervision: B.M. Norquist, M.C. King, M.J. Birrer, E.M. Swisher

This work was supported by NCI grants to the Gynecologic Oncology Group Administrative Office (CA 27469), the Gynecologic Oncology Group Statistical and Data Center (CA 37517), the Gynecologic Oncology Group Tissue Bank (U10 CA27469, U24 CA114793, and U10 CA180868), NRG Oncology Grant (1 U10 CA180822), R01CA157744 (M. King), RO1CA175716 (T. Walsh and M. King), and by the Ovarian Cancer Research Foundation Alliance (E.M. Swisher and B.M. Norquist), The Breast Cancer Research Foundation (M.C. King), the Department of Defense Ovarian Cancer Research ProgramOC120312 (E.M. Swisher), the V Foundation Translational Research Award (E.M. Swisher), The Women's Reproductive Health Research Career Development Award5K12HD001264-13 (B.M. Norquist), The Nancy Kelley Jensen Faculty Fellowship for Gynecologic Oncology Research (B.M. Norquist), and the Wendy Feuer Research Fund for the Prevention and Treatment of Ovarian Cancer (E.M. Swisher).

The following Gynecologic Oncology Group member institutions participated in the primary treatment studies: Cancer Trials Support Unit, University of Oklahoma Health Sciences Center, University of California Medical Center at Irvine-Orange Campus, Fred Hutchinson Cancer Research Center, Ohio State University Comprehensive Cancer Center, Gynecologic Oncology Network/Brody School of Medicine, Mayo Clinic, Metro-Minnesota CCOP, Abramson Cancer Center of the University of Pennsylvania, University of North Carolina at Chapel Hill, Abington Memorial Hospital, University of Alabama at Birmingham, Rush University Medical Center, Walter Reed National Military Medical Center, Washington University School of Medicine, University of Kentucky, University of Iowa Hospitals and Clinics, Roswell Park Cancer Institute, Duke University Medical Center, University of Colorado Cancer Center – Anschutz Cancer Pavilion, University of Chicago, Cleveland Clinic Foundation, Women's Cancer Center of Nevada, University of Hawaii, Mount Sinai School of Medicine, Northwestern University, Fox Chase Cancer Center, Yale University, Women and Infants Hospital, University of California at Los Angeles Health System, Memorial Sloan Kettering Cancer Center, Indiana University Hospital/Melvin and Bren Simon Cancer Center, University of Minnesota Medical Center-Fairview, University of Mississippi Medical Center, University of Pittsburgh Cancer Institute, University of New Mexico, Case Western Reserve University, The Hospital of Central Connecticut, University of Texas-Galveston, Cancer Research for the Ozarks NCORP, Cooper Hospital University Medical Center, University of Virginia, Moffitt Cancer Center and Research Institute, State University of New York Downstate Medical Center, University of Wisconsin Hospital and Clinics, Carle Cancer Center, Northern Indiana Cancer Research Consortium, Penn State Milton S. Hershey Medical Center, Wake Forest University Health Sciences, Stony Brook University Medical Center, University of Massachusetts Memorial Health Care, Fletcher Allen Health Care, Michigan Cancer Research Consortium Community Clinical Oncology Program, MD Anderson Cancer Center, Gynecologic Oncology of West Michigan PLLC, Central Illinois CCP, Delaware Christiana Care CCOP, Virginia Commonwealth University, University of Cincinnati, Tufts-New England Medical Center, Scot and White Memorial Hospital, Tacoma General Hospital, Northern New Jersey CCOP, Virginia Mason CCOP, New York University Medical Center, Aurora Women's Pavilion of Aurora West Allis Medical Center, Wisconsin NCI Community Oncology Research Program, Evanston CCOP-North Shore University Health System, Cancer Research Consortium of West Michigan NCORP, Missouri Valley Cancer Consortium CCOP, William Beaumont Hospital, Colorado Cancer Research Program NCORP, Saint Louis-Cape Girardeau CCOP, University of Texas Southwestern Medical Center, University of Illinois, Kalamazoo CCOP, Upstate Carolina CCOP, Heartland Cancer Research CCOP and Wichita CCOP.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.