Long-Term Responders on Olaparib Maintenance in High-Grade Serous Ovarian Cancer: Clinical and Molecular Characterization

Epithelial ovarian cancer remains the leading cause of death from gynecologic malignancies, with high-grade serous ovarian cancer (HGSOC) the most common histologic subtype (1). Following cytoreductive surgery and platinum-based chemotherapy, approximately 70% of patients with HGSOC relapse despite an initial response to therapy (2). Given the practical certainty of recurrence in relapsed HGSOC, a maintenance approach was proposed to delay subsequent progression via continuation of treatment beyond cytotoxic chemotherapy. The PARP inhibitor (PARPi) olaparib (Lynparza) is the first validated molecularly targeted agent for HGSOC. Ledermann and colleagues demonstrated the benefit of maintenance olaparib after response to platinum-based chemotherapy in women with HGSOC recurrence (NCT00753545; ref. 3). Patients were randomized to placebo or olaparib, introduced within 8 weeks of completion of the last dose of platinum-based chemotherapy and maintained until progression. Olaparib maintenance was associated with significantly longer progression-free survival [PFS; median, 8.4 vs. 4.8 months; HR, 0.35; 95% confidence interval (CI), 0.25–0.49; P < 0.001; data maturity 58%]. Patients with a BRCA1 mutation (BRCA1m) or a BRCA2 mutation (BRCA2m) had significantly greater PFS benefit with olaparib maintenance than those receiving placebo (median, 11.2 vs. 4.3 months; HR, 0.18; 95% CI, 0.10–0.31; P < 0.0001; data maturity 52%; ref. 4). Overall survival improved by 3 months, which was not significant, possibly because of posttrial cross-over, whereby 22.6% of placebo patients received PARPi in subsequent clinical trials following progression on placebo. Olaparib is now approved in Europe, Australia, and Canada for the maintenance treatment of women with relapsed, BRCA1/2m-positive (germline or somatic) HGSOC who have had complete or partial response to platinum-based chemotherapy (5). Olaparib is also approved in the United States for monotherapy in patients with germline BRCA1/2m advanced HGSOC who have been treated with ≥3 prior lines of chemotherapy (6). Proteins encoded by the BRCA1/2 genes are crucial effectors of double-strand break DNA repair (7); as a result, BRCA1/2m carriers are known to be highly responsive to DNA-damaging agents, including platinum-based chemotherapies (8, 9) and PARPi (8), although mechanisms of action and resistance to PARPi are not fully understood (10). Other than deleterious BRCA1m or BRCA2m, there are no predictive biomarkers for sensitivity to olaparib. Data from The Cancer Genome Atlas (TCGA) and functional studies indicate that approximately half of HGSOC are homologous recombination repair (HR) defective (11–13). PARP is involved in the repair of single-strand DNA breaks, and its inhibition can result in replication-associated double-strand breaks. Such HR-deficient cells as those found in BRCA1/2-mutated tumors, whether repaired via error-prone pathways or persistent without repair, cause further genetic instability and can lead to cell death (14).

We hypothesize that studying patients with HGSOC with prolonged response to olaparib may identify additional biomarkers of response beyond BRCA1/2m. The objective of this study was to identify and characterize long-term (LT) responders to olaparib maintenance in comparison with short-term (ST) responders in terms of clinical and molecular profile to pinpoint additional markers of response and explore potential resistance factors.

We investigated the molecular and clinical characteristics of LT and ST responders randomized to maintenance olaparib or placebo in the phase II, randomized, double-blind study of olaparib in patients with platinum-sensitive relapsed HGSOC following treatment with ≥2 lines of platinum-based chemotherapy (NCT00753545; Study 19; ref. 3). This trial enrolled 265 patients, with 136 patients assigned to olaparib and 129 to placebo. Given that the greatest PFS benefit was at 11.2 months, we defined LT responders, whatever their BRCA1/2 status, as having a double PFS benefit, that is, of >2 years. ST responders were defined as having PFS of <3 months, given the PFS observed in the placebo group of 4.8 months.

A second comparison/confirmation cohort from the open-label, randomized, phase II study of olaparib/carboplatin/paclitaxel with olaparib maintenance versus carboplatin/paclitaxel/observation in patients with platinum-sensitive recurrent HGSOC (NCT01081951; Study 41) was evaluated (15). In the combination phase, 156 patients were treated (81 in the olaparib-plus-chemotherapy group and 75 in the chemotherapy-alone group), and 121 continued to maintenance or no further treatment (66 in the olaparib-plus-chemotherapy group and 55 in the chemotherapy-alone group). Given that patients in this study received olaparib with chemotherapy, the ST responders were defined as patients with PFS of <6 months, taking into account the time over which olaparib was administered with chemotherapy to assess the maintenance period.

Clinical trial data were prospectively obtained for all treated patients (Table 1) and archival tumor specimens, predominantly taken at the time of initial diagnosis, from patients who consented to further use of their specimens. Previously performed germline BRCA1/2 testing (Myriad Genetics) results were obtained.

Assessment of somatic BRCA1/2 mutations (Myriad tumor testing and Foundation Medicine), determination of the three-biomarker HR deficiency (HRD) score (myChoice HRD test, Myriad Genetics; ref. 16), and BRCA1 methylation (Myriad tumor testing) were conducted (17). HRD scores were determined using an assay that calculates scores for whole-genome tumor LOH, telomeric allelic imbalance and large-scale state transition. The HRD score is the sum of the three scores, where ≥42 is a predefined threshold validated as a high HRD score (16). The Methyl-Profiler DNA Methylation PCR Assay System (SABiosciences) was used to quantify methylation levels, with a used quantitative cut-off of promoter methylation of >10%. HRD status was determined on the basis of the combination of the dichotomized HRD score using the predefined HRD threshold and tumor BRCA1/2 status (scored as mutated if deleterious or suspected deleterious mutations in germline or somatic BRCA1/2 were present). Therefore, HRD was defined as a high HRD score (≥42) and/or BRCA1/2-mutated tumor (16).

Mutational profile determined by the Foundation Medicine T5 panel (entire coding sequence of 287 cancer-related genes plus select introns from 27 genes) and other genetic alterations (deletions and functional rearrangements) were assessed on archival tumor tissue (18). Classification of TP53 mutations (TP53m) was determined using the WHO International Agency for Research on Cancer Database.

Pathology review for tumor cell content was performed for all patients with molecular analyses, whereas pathology review of histologic subtype by an expert pathologist in gynecologic cancer, blinded to cohort or outcome, was only performed for patients with remaining tissue.

Descriptive analyses from all LT and ST responders to olaparib/placebo were assessed for statistical significance. Fisher exact or χ² tests were used to test for associations between individual explanatory variables and response duration (LT vs. ST) as appropriate. SPSS and Excel were used for all analyses. Given the exploratory data analysis, no type 1 error correction was performed.

Data were collected from 265 total patients on study for the patients with LT and ST responses to olaparib therapy or placebo as part of Study 19. Thirty-seven patients were identified as LT responders, of whom 32 (86.5%) had received olaparib. Of the 61 ST patients identified, 21 (34.4%) had received olaparib. The main characteristics are summarized in Tables 1 and 2. LT responders on olaparib (32 patients) had a median of three prior lines of therapy, with one-third having relapsed within 6 to 12 months of their penultimate platinum-based chemotherapy. Of the LT responders on olaparib, 44% (14/32) had partial responses to their most recent platinum-based chemotherapy, with residual disease evident at the time of olaparib maintenance (Table 1). The other 18 of 32 (56%) LT responders on olaparib had complete response to the most recent platinum-based chemotherapy, in comparison with only five of 21 (24%) olaparib ST responders (Table 1). Complete response to platinum-based chemotherapy at the time of olaparib maintenance was associated with LT response to olaparib (P = 0.026, univariate analysis). The treatment-free interval following the penultimate platinum-based chemotherapy did not correlate with LT response (Supplementary Table S1). Sixty-five of 98 archival tumor samples (37 LT and 61 ST) were from primary tumors, and nine of 98 were from metastases (this information was unavailable for the other 24 patients).

Receipt of olaparib over placebo was significantly associated with LT response (32/53, 60.4% vs. 5/45, 11.1%; P < 0.0001). More patients treated with olaparib were LT than ST responders (P = 0.052).

From the LT responders, 27 of 37 (73%) had loss-of-function mutations in BRCA1/2. Among the 32 LT olaparib responders, BRCA1m and BRCA2m were found in 10 and 13 patients, respectively, with one patient showing deleterious mutations in both BRCA1 and BRCA2 (Tables 2 and 3). All five LT responders receiving placebo were BRCA1/2m positive (Table 2). A greater number of BRCA1/2m was found in LT responders compared with ST responders (Table 3). In contrast, among the 21 olaparib ST responders, 10 were found to carry a deleterious BRCA1/2m (seven of 10 were in BRCA1; Tables 2 and 3).

We further analyzed the type and location of BRCA1/2m in the olaparib cohort between LT and ST responders (Fig. 1 and Supplementary Fig. S1). Of the 17 patients on olaparib who had BRCA1m (LT and ST), nine had founder mutations (E23fs* or Q1756fs*), whereas only one patient had a founder BRCA2m (S1982fs*; ST responder group). Interestingly, among patients on olaparib, mutations in the DNA-binding domains of BRCA1 (n = 1) or BRCA2 (n = 4) were only seen in the LT responder group.

BRCA1 methylation status was available for 27 of 37 (73%) LT responders, all of whom were negative. In contrast, BRCA1 methylation status was available for 42 of 61 (69%) ST responders, eight of whom (19%) had BRCA1 methylation. Methylation of BRCA1 was not associated with LT olaparib response.

No significant difference was seen between LT and ST responders according to the HRD score (Table 3). Most of the patients enrolled in the study were characterized by HR-deficient status (81% of the LT and ST on olaparib). Among data available for 26 of the 32 (81%) LT responders on olaparib, 25 patients (96%) had HRD, in contrast to 76% of the ST olaparib responders (Table 3). Of the 21 patients with high HRD score in the LT responder group, three were BRCA1/2 wild-type. Of the nine ST patients with a high HRD score, two (22%) were BRCA1/2 negative in the ST responder group. However, a significant number of HRD scores are missing for the BRCA1/2 wild-type group (Table 2).

Taking together high HRD score (≥42) and/or BRCA1/2-mutated tumor, there was a trend toward HRD in LT versus ST responders, with 96% of LT responders and only 76% of ST responders showing HRD (P = 0.07).

Next-generation sequencing was performed on archival tumor DNA from 44 patients [28 (87%) LT and 16 (76%) ST responders on olaparib]. From 44 patients, over 600 identified alterations were classified by type and likely functional significance by Foundation Medicine (Supplementary Table S2). A total of 242 likely functional generic alterations in 99 genes were found, with TP53, BRCA1, and BRCA2 mutations being the most common (89%, 34%, and 36% of patients, respectively; Fig. 2 and Supplementary Table S2).

TP53 signaling was considered aberrant in most cases (90%), as expected for the HGSOC histology subtype (27 of 28 patients with available data analyzed in the LT responder group, 12 of 16 in the ST responder group; Supplementary Fig. S1). The patient with no TP53m in the LT responder group had pathology reviewed and HGSOC confirmed. Of the four patients with no TP53m in the ST responder group, two had confirmed HGSOC. Other types of mutations and gene amplifications were observed, particularly in genes encoding proteins involved in DNA repair and damage response, regulation of cell cycle, apoptosis, and MAPK/PI3K signaling (Fig. 2 and Supplementary Table S2). Furthermore, four patients had alterations in PTEN (two homozygous deletions, one somatic variant, and one functional intergenic truncation), all of whom were LT responders on olaparib. These PTEN alterations co-occurred with BRCA1/2 mutation for three of the patients, only 1 LT patient had a PTEN mutation in the absence of a BRCA1/2 mutation. In contrast, no PTEN alterations were seen in the ST responder group (Supplementary Table S3). Interestingly, three patients randomized to the placebo arm had PTEN alterations and were ST responders (Supplementary Table S3).

LT and ST responders from Study 41 were identified and analyzed (Fig. 3 and Supplementary Table S4). In total, 19 LT responders were identified, and all were in the olaparib arm. Olaparib maintenance was also significantly associated with LT responders (P < 0.0001). BRCA1/2m was statistically correlated with LT response to olaparib (Supplementary Table S4). BRCA1m and BRCA2m were observed in six and five LT responders, respectively. The 11 ST patients were all BRCA1/2 wild-type.

There are limited data describing patients with ovarian cancer who experience prolonged benefit from PARP inhibition, other than evidence for the role played by deleterious mutations in BRCA1/2. Defective DNA repair via HR repair deficiency is a fundamental vulnerability in HGSOC and can be exploited with PARPi, such as olaparib, by induction of cancer-specific synthetic lethality (19). Examination of broader clinical and molecular data of extreme responders may uncover potential biomarkers of response (20). We identified LT and ST responders to PARPi from women with HGSOC entered into two olaparib maintenance trials, Studies 19 and 41, and have molecularly characterized these exceptional responders. We found that 32 and 19 patients with recurrent platinum-sensitive HGSOC achieved LT (defined as >2 years) response to olaparib maintenance as part of the studies, respectively. Germline and somatic BRCA1/2m were observed to be associated with LT response to olaparib, with an enrichment of BRCA2m in the LT olaparib responders, compared with the frequency of BRCA1m and BRCA2m in HGSOC detected at the start of the trial and the BRCA1/2 ratio observed in the general population (21). Our finding is in agreement with data suggesting differences in outcome and response to therapy between BRCA1 and BRCA2 genotypes (22). Previous studies have shown that BRCA2m is associated with prolonged survival in invasive epithelial ovarian cancer (23). In 2012, Liu and colleagues showed that the presence of a BRCA2m was associated with longer survival and better therapy response than a BRCA1m in HGSOC (24). Many LT olaparib responders had a BRCA2m in the RAD51-binding domain, described as a frequent site of BRCA2m by TCGA (11), but also in DNA-binding sites (Fig. 3). As such, mutations in the RAD51 region are expected to attenuate or abolish interactions with RAD51, resulting in failure to load RAD51 to DNA-damage sites (24). Our data also suggest that silencing of BRCA1 through promoter methylation does not result in improvement in response to platinum-based chemotherapy or to sequential chemotherapy and maintenance olaparib therapy, as previously suggested by TCGA and other publications showing a lack of survival benefit and correlation with platinum sensitivity (11, 25).

However, our study did not identify a potential mechanism involved in the small group of BRCA1/2 wild-type patients who had a LT benefit to olaparib maintenance, currently not eligible for olaparib in clinical practice. Beyond BRCA1/2m, there have been a number of mechanisms of HRD described that may correlate with platinum and PARP response (11), and newly developed homologous recombination repair panels have assessed several additional novel genes, including NBN, MRE11, RAD50, RAD51C, PALB2, BARD1, and BRIP1 (19, 26, 27). Our results show that the majority of patients enrolled in the study were HR deficient, a potential enrichment due to the selection of patients with HGSOC based on platinum-sensitivity recurrence and objective response to platinum. The phase II study ARIEL2 investigating rucaparib (another PARPi) monotherapy in patients with recurrent platinum-sensitive HGSOC has confirmed BRCA1/2m as a biomarker of response, as well as genomic LOH, a potential predictive surrogate marker for HRD (28). It was hypothesized that the inability of the cell to perform HR repair leads to genomic scarring and LOH, thus enabling the use of high LOH as a signature of HRD. However, no data are currently available for LT responders and PARPi progression. Recently, the maintenance phase III study of niraparib, another PARPi, showed increased PFS in all patients—germline BRCA1/2 patients (the group with greatest benefit), and non-BRCA1/2 carriers (comprising of both HRD-positive and HRD-negative tumors; ref. 29). Eligible patients had to have achieved response following four to six cycles of platinum-based chemotherapy with a CA-125 in the normal range or reduced by 90% for at least 7 days, an absence of measurable disease greater than 2 cm at study commencement. Consistent with our findings, the PFS benefit to PARPi is driven by minimal residual disease (complete response prior maintenance), BRCA1/2 mutation, and HRD-positive tumors, though not exclusively. The current HRD assays available did not completely identify biomarkers involved in response or resistance to PARPi.

Interestingly, four genetic alterations in the PTEN gene were observed, of which three were associated with BRCA1/2m in the LT responders on olaparib but none in the ST responder group. Moreover, in the placebo arm, three patients with PTEN alterations had disease progression within three months (Supplementary Table S3). The significance of this finding is not clear but warrants further investigation. Although many LT responders to olaparib harbor a BRCA1/2m, our data show the occurrence of ST responders to olaparib in patients with BRCA1/2m. This finding highlights the limitation of the analysis on archival rather than tumor tissue at the time of disease progression. There is evidence that in cells deficient in DNA-damage repair, such as those with BRCA1/2m, additional mutations can restore function and allow effective DNA repair (30). Reversion mutations in BRCA1/2 have been described and associated with olaparib and platinum resistance (31), although this effect was supposed to be minimized by the selection of patients with response to platinum-based chemotherapy. Understanding therapeutic resistance requires comprehensive disease assessment at the specific time of therapeutic intervention; timing and treatment strategy are imperative to efficacy. Several mechanisms of resistance have been described related to the HR pathway (30). Although tumors harboring a BRCA1/2m lack the HR repair pathway required for error-free repair of DNA double-strand breaks (32), other DNA-repair pathways exist that can become engaged, ultimately leading to olaparib resistance but platinum sensitivity, as with non-homologous end-rejoining alterations (33, 34). As such, determining the disease- and therapy-specific HRD signature is important. Reports on this signature show varying gene lists, and these differences are likely attributed to variances in methodologies. Peng and colleagues identified a HR defect gene signature that can be used to functionally assess HR status and predict clinical outcomes (27). Pennington and colleagues reported that germline and somatic mutations in 13 HR genes predict platinum response and survival in ovarian, fallopian tube, and peritoneal carcinomas (26). These signatures need prospective validation.

This is the first study contrasting LT with ST responders to PARPi in terms of clinical and molecular data. Our results show that LT response to olaparib has been observed in platinum-sensitive recurrent HGSOC. This durable response may be multifactorial and driven by germline and somatic BRCA1/2m. This pilot study warrants a larger cohort to characterize LT responders. A study is ongoing to identify LT and ST responders to olaparib and allows for additional tumor tissue collection for analysis (NCT02489058).

D. Hodgson, J.D. Robertson, and J.C. Barrett hold ownership interest (including patents) in AstraZeneca. K.M. Timms and J.S. Lanchbury hold ownership interest (including patents) in Myriad Genetics Inc. S.B. Kaye and J. Ledermann are consultant/advisory board members for AstraZeneca. C. Gourley reports receiving commercial research grants from AstraZeneca and Novartis; speakers bureau honoraria from AstraZeneca and Roche; and is a consultant/advisory board member for AstraZeneca and Clovis Oncology. C.L. Scott reports receiving speakers bureau honoraria from Prime Oncology, and is a consultant/advisory board member for AstraZeneca and Clovis Oncology. No potential conflicts of interest were disclosed by the other authors.

Conception and design: S. Lheureux, B.A. Dougherty, S. Runswick, D.R. Hodgson, S. Kaye, C. Gourley, D. Bowtell, E.C. Kohn, C. Scott, M.J. O'Connor, J.D. Robertson, J. Ledermann, J.C. Barrett, T.W. Ho, A.M. Oza

Development of methodology: S. Lheureux, B.A. Dougherty, S. Runswick, D.R. Hodgson, J.S. Lanchbury, T. Panzarella, C.B. Gilks, J. Ledermann, J.C. Barrett, T.W. Ho, A.M. Oza

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Lheureux, B.A. Dougherty, S. Runswick, D.R. Hodgson, K.M. Timms, S. Kaye, C. Gourley, C. Scott, U. Matulonis, C.B. Gilks, J.D. Robertson, J. Ledermann, J.C. Barrett, A.M. Oza

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Lheureux, Z. Lai, B.A. Dougherty, S. Runswick, D.R. Hodgson, S. Kaye, D. Bowtell, E.C. Kohn, C. Scott, T. Panzarella, J.V. Burnier, C.B. Gilks, J.D. Robertson, J.C. Barrett, T.W. Ho, A.M. Oza

Writing, review, and/or revision of the manuscript: S. Lheureux, Z. Lai, B.A. Dougherty, S. Runswick, D.R. Hodgson, K.M. Timms, J.S. Lanchbury, S. Kaye, C. Gourley, D. Bowtell, E.C. Kohn, C. Scott, U. Matulonis, T. Panzarella, K. Karakasis, J.V. Burnier, C.B. Gilks, M.J. O'Connor, J.D. Robertson, J. Ledermann, J.C. Barrett, T.W. Ho, A.M. Oza

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Lheureux, B.A. Dougherty, S. Runswick, D.R. Hodgson, K.M. Timms, T.W. Ho, A.M. Oza

Study supervision: S. Lheureux, K.M. Timms, A.M. Oza

This research was conducted in collaboration with AstraZeneca.

This research was conducted in collaboration with AstraZeneca. Professor Oza did not receive any funding for this study, and the molecular analysis was performed by AstraZeneca.

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.