The Search for Genomic Biomarkers of Response to Immunotherapy in Ovarian Cancer

In this issue of Clinical Cancer Research, Landen and colleagues report on an exploratory biomarker substudy of the placebo-controlled randomized phase III IMagyn050 trial evaluating the PD-L1 inhibitor atezolizumab combined with chemotherapy and bevacizumab for ovarian cancer (1). In this substudy, the investigators demonstrate that there was no significant relationship between mutations in BRCA1/2 or homologous recombination deficiency (HRD) and increased benefit to atezolizumab, despite an increase in tumor mutation burden (TMB) and an association with PD-L1 status. Moreover, alterations in BRCA2, RB1, NF1, and CCNE1 were found to be prognostic regardless of the treatment administered.

High-grade serous ovarian cancer (HGSOC) is the deadliest gynecologic malignancy and is characterized by chromosomal structural variations and copy number aberrations (2). Patients with HGSOC undergo surgical debulking and treatment with carboplatin and paclitaxel, along with bevacizumab in appropriately selected patients. Patients may also be offered maintenance therapy with Poly-ADP Ribose Polymerase inhibitors (PARPi), although the benefit is primarily observed in homologous recombination–deficient tumors (3, 4). Despite aggressive multimodality treatment, most patients will experience recurrent and incurable disease.

Based on preclinical data, there was initial excitement that immune checkpoint inhibitors (ICI) might be an effective treatment option in HGSOC. In a metanalysis of 10 studies, investigators found that lower rates of tumor associated lymphocytes (TIL) were associated with worse overall survival (OS; pooled HR, 2.24; 95% CI; 1.71–2.91). Intriguingly, the protective effect of TILs was reduced in patients with optimal debulking (5). Unfortunately, clinical studies of single agent and combined ICI have thus far been disappointing, with objective response rates (ORR) of approximately 10% across clinical trials (6, 7). PD-L1 expression seems to enrich for response; for example, in KEYNOTE-100, patients with a combined positive score (CPS) ≥10 had ORRs of 11.6% (3.9, 25.1) in cohort A and 18.2% (5.2, 40.3) in cohort B, respectively. The combination of nivolumab with or without ipilimumab was also studied in a randomized phase II study; the addition of ipilimumab resulted in a superior response rate and slightly longer PFS, but at the cost of additional toxicity (8). Other novel combinations that have been evaluated include the PARPi niraparib plus pembrolizumab, which demonstrated an ORR of 18% (9), as well as the triplet combination of bevacizumab, oral metronomic cyclophosphamide, and pembrolizumab, which had an ORR of 45% and a median progression-free survival (PFS) of 10 months (10).

Given that PD-L1 expression has only a modest performance as a biomarker in HGOSC, there has been ongoing work to identify other potential genomic biomarkers. Early work in BRCA1/2 mutated HGSOC found that high TMB in pretreatment samples was associated with improved median PFS and OS (HR = 0.82, P = 0.002 and HR = 0.83, P = 0.011, respectively); this was not observed in BRCA wild-type (WT) ovarian cancer (11). Thus, it was hypothesized that BRCA1/2–deficient tumors might be more sensitive to ICI. However, in a single institution retrospective study comparing the response rate to ICI in tumors with pathogenic BRCA1/2 variants versus WT, response to ICI did not vary according to BRCA1/2 status (20% in BRCA1/2 WT vs. 16% in BRCA1/2 mutated; P = 0.796; ref. 12). One caveat of these data is that patients with BRCA1/2 alterations had more prior lines of treatment (median, 5 vs. 4 lines, respectively; P = 0.018) and longer duration of disease (53.6 vs. 37.8 months) compared with WT patients. Therefore, it is challenging to determine whether the lack of response in BRCA1/2 tumors is due to the inherent nature of heavily pretreated cancers.

The author's work builds on prior work to identify genomic biomarkers of response to ICI in HGSOC. In the IMagyn050 study, patients with newly diagnosed ovarian cancer were randomized to receive either atezolizumab or placebo with standard chemotherapy and bevacizumab. PD-L1 positivity was assessed along with HRD status. There was no improvement in PFS in patients with BRCA2-mutated or HRD tumors, but the authors did note a trend toward improved PFS in patients who received atezolizumab and harbored BRCA1-mutated tumors (1, 13). However, there were no clear statistically significant signals to indicate that there is a role for immunotherapy in HGSOC in treatment-naïve patients irrespective of HRD status.

Why has ICI thus far disappointed in HGSOC? One possibility is that only a subgroup of patients may benefit from ICI alone or in combination, and that this group remains to be identified. Both KEYNOTE 100 and the IMagyn050 found that patients with higher rates of PD-L1–positive cells exhibited a better response to immunotherapy. In addition, there has been a positive signal in endometriosis-associated ovarian cancers including endometroid and ovarian clear cell carcinoma (14). TMB and microsatellite instability (MSI) have also been assessed in ovarian cancer histologic subtypes; endometroid ovarian cancer was found to have the highest prevalence of MSI-H and high TMB (15).

Another possibility is that assessing HRD and BRCA1/2 mutations may be too blunt of an instrument, as not all forms of genome instability are the same and do not produce the same tumor microenvironment. A recent multimodal investigation of HGSOC demonstrated that the tumor microenvironment (TME) is heavily influenced by mutational processes. Single-cell whole-genome sequencing distinguishes between BRCA1 (HRD duplication subtype) and BRCA2 deficiency (HRD deletion subtype) within HRD tumors and between the homologous recombination–proficient CCNE1 amplification (fold back inversions) and CDK12-associated tumors (tandem duplications) based on the pattern of copy number variation. Indeed, the genomic pattern appeared to be associated with immune TME. Intriguingly, loss of the MHC class I genes was also frequently noted in HGSOC, which is a common mechanism of immunotherapy resistance (Fig 1). Thus, chromosomally unstable tumors may lose the HLA locus early, facilitating immune evasion and resistance to immunotherapy.

Another possibility is dysregulation of cGAS/STING, a cytoplasmic DNA sensing pathway. Cytoplasmic DNA can stem from ruptured micronuclei formed from chromosomally unstable tumors and activate an innate immune response through the cGAS effector, STING. This provides a link between chromosomal instability and inflammatory signaling. However, in an orthotopic mouse model inactivation of cGAS resulted in reduced metastasis (16), and activation of STING through the action of STING agonists have thus far largely been unsuccessful in clinical trials (17). Another possibility is that chronic activation of the cGAS/STING pathway in HGSOC results in downregulation of the innate immune response, thereby facilitating immune evasion (Fig 1). The emerging preclinical evidence suggests that chromosomally unstable tumors represent difficult to treat malignancies due to the generation of an immunosuppressive microenvironment.

The question remains: can we outsmart HGSOC with immunotherapy? The preponderance of clinical evidence suggests that there is little benefit for single agent ICI or ICI in combination with chemotherapy in this disease. Moreover, biomarkers that are predictive in other disease types, including PD-L1 expression, are lacking in HGSOC. Therefore, the future of immunotherapy in HGSOC cancer has two prospective paths. The first is to continue the translational work to understand the genomic subclassifications of HGSOC and how the molecular drivers impact response to therapy. The second is to develop novel combination therapies, including bispecific antibodies and combinations with tyrosine kinase inhibitors. For now, we have a definitive answer that ICI is ineffective in combination with chemotherapy and bevacizumab in HGSOC, irrespective of HRD status.

C.F. Friedman reports other support from Genentech/Roche, Puma, Daiichi, AstraZeneca, and Merck and personal fees and other support from Seagen and Bristol-Myers Squibb outside the submitted work. No disclosures were reported by the other authors.

D.H. Al-Rawi is supported by the Mary Jane Milton Endowed Fellowship in Gynecologic Oncology and National Institutes of Health, under Award Number T32-CA009207. This manuscript was supported in part by the MSK Cancer Center Support Grant P30 CA008748. C.F. Friedman is a member of the Parker Institute for Cancer Immunotherapy at MSK.