Activity of durvalumab plus olaparib in metastatic castration-resistant prostate cancer in men with and without DNA damage repair mutations

Despite success in other solid tumors, checkpoint inhibitors have not shown overall improvements in survival as monotherapy for metastatic castration-resistant prostate cancer (mCRPC). Two phase III trials of ipilimumab failed to meet their endpoint of improved overall survival [1, 2]. Despite this, there were signs of activity, suggesting that a subset of patients may benefit from immunotherapy. Results from patients previously treated with enzalutamide suggest that enzalutamide may increase expression of programmed death ligand 1 (PD-L1) [3]. Indeed, in a separate ongoing trial, 3/12 patients who previously progressed on enzalutamide had objective radiographic responses to programmed cell death protein 1 (PD-1) blockade with pembrolizumab, and 5/28 patients had a prostate-specific antigen (PSA) decline of ≥50% [4]. Mutations in mismatch repair (MMR) genes, one of many pathways involved with DNA damage repair (DDR), are associated with microsatellite instability in advanced prostate cancer and may serve as a possible biomarker of response to immune-blocking antibodies, as seen in other solid tumors. MMR mutations, however, are thought to occur in < 5% of prostate cancer patients [5].

Mutations in other DDR genes appear to occur much more frequently. These include alterations in the homologous recombination repair pathway genes, including BRCA2 and ATM, among others. Alterations in these additional DDR-related genes occur in approximately 20–25% of mCRPC patients, with approximately 12% harboring germline alterations in DDR genes [6, 7]. Alterations in these genes appear to predict response to PARP inhibitors. In a phase II trial of single-agent olaparib, 16/49 patients had either an objective response according to RECIST v.1.1, a reduction of at least 50% in the PSA level or a confirmed reduction in circulating tumor cells (CTCs) (response rate, 33%; 95% CI, 20–48) [8]. The majority of patients who had a response had homologous recombination pathway gene alterations. The phase II KEYNOTE-199 study of single-agent pembrolizumab in docetaxel-refractory mCRPC (NCT02787005) showed patients with somatic mutations in BRCA1/2 or ATM had higher responses rates (9). Whole-exome sequencing was conducted on 6/9 responders with available data, and 4/6 patients had mutations in DDR genes [9].

Mounting evidence from trials in other solid tumors suggests alterations in DDR genes beyond the MMR pathway may also predict response to immunotherapy [10, 11]. There is also increasing rationale for combining PARP inhibition with immunotherapy, even though the mechanism of synergy is not fully understood. Foremost among candidate intracellular pathways is STING (stimulator of interferon genes), an innate immune response activated by cytosolic DNA (perhaps a consequence of DNA damage) that can lead to enhanced interferon (IFN) production [12]. It is evident that, in some patients, mutational burden is associated with response to PD-1/PD-L1 inhibition. It has been suggested that PARP inhibition can potentiate DNA damage and inefficient repair in tumors, and could lead to immunologically relevant mutations [13, 14].

In this study, the PARP inhibitor olaparib is combined with durvalumab to treat mCRPC, regardless of patients’ mutational status. Durvalumab is a human IgG1-K monoclonal antibody that selectively binds human PD-L1 and is approved by the U.S. Food and Drug Administration for the treatment of urothelial cancer and non-small cell lung cancer. In phase I of this study, durvalumab was safely given every 4 weeks at a fixed dose of 1500 mg i.v. in combination with 300 mg of olaparib tablets p.o. every 12 h. This regimen was selected for further study in the phase II cohorts [15]. Here we present findings for the phase 2 mCRPC cohort.

The preliminary PFS data and response rate for this cohort of mCRPC patients treated with durvalumab plus olaparib indicate deep and sustained responses with this therapy combination, with 53% of patients having radiographic and/or PSA responses, median PSA decline of 85% among the 9 responders, and median duration of response of 16.1 months in those patients with mutations in DDR genes. One third of responders harbored germline mutations in DDR genes, one third had detectable biallelic somatic alterations in DDR genes, and one third had neither.

Mutational burden, DDR status, prior therapies, and other variables can critically affect the cancer-immune set point, the peripheral immune phenotype, and response to therapy [19]. The data presented here suggest that durvalumab plus olaparib can affect both innate and adaptive immunity in patients with mCRPC, and that engagement of these 2 types of immunity may be associated with prolonged PFS. Our data also suggest that durvalumab plus olaparib for mCRPC may promote dendritic cell (DC) maturation, enhance CD4⁺ T-cell activity, and reactivate CD8⁺ T-cell antitumor immunity.

Early immune responses were associated with benefit. It has been suggested that DCs must be optimally stimulated to overcome the immunosuppressive tumor microenvironment and achieve robust and durable responses to immunotherapy [16]. Durvalumab plus olaparib may enhance maturation of DCs into CD83⁺CD141⁺ mDCs that are competent to present antigen and direct adaptive immune responses. Therefore, our finding of increased mDCs at C1D15 is consistent with increased activation of CD4⁺ and CD8⁺ T cells in response to durvalumab plus olaparib. Dendritic cells (DCs) that express CD141 produce IFN-γ and support CD4⁺ T-cell polarization to a Th1 phenotype [16]. Expression of the functional marker CD83 on CD141⁺ DCs is associated with fully mDCs that present antigen and induce T cell-mediated immune responses [17]. In our cohort, patients with increased expression of CD83 on CD141⁺ mDCs from C1D1 to C1D15 had prolonged PFS (Fig. 3d).

Changes in CD8⁺ and CD4⁺ cell populations also predicted response. Patients with > median percentage of Ki67⁺PD-1⁺ cells among total CD8⁺ T cells in response to therapy had prolonged PFS (Fig. 3e), as did patients with > median percentage of Ki67⁺PD-1⁺ cells among total CD4⁺ T cells (Fig. 3f). Analysis of expression of HLA-DR, another T-cell activation marker, showed that patients with > median percentage of Ki67⁺HLA-DR CD8⁺ and CD4⁺ T cells at C3D1 had prolonged PFS.

In two recent studies, patients with melanoma [20] or non-small cell lung cancer [21] had increased levels of Ki67⁺PD-1⁺CD8⁺ T cells in peripheral blood after treatment with a PD-1-targeting agent, which was considered a potential indicator of activated tumor-specific CD8⁺ T cells. Kamphorst et al. [21] propose that early PD-1⁺CD8⁺ T-cell responses may be associated with clinical outcome. They note that it may be difficult to detect cycling PD-1⁺CD8⁺ T cells in peripheral blood due to their transient appearance and the timing of peripheral blood analyses. In our study, our time points appear to facilitate detection of this transient immune population. Furthermore, at C1D15 we see apparent associations of Ki67⁺PD-1⁺CD8⁺ T cells with improved PFS. At C3D1 the association of Ki67⁺CD8⁺ T cells with PFS is no longer apparent, but 2 new associations appear to correspond with PFS: Ki67⁺HLA-DR⁺CD8⁺ T cells and Ki67⁺PD-1⁺CD4⁺ T cells.

The efficacy seen in this trial may be due to synergy between the 2 agents. Recent data show that DNA damage plays a role in priming the type I IFN system, where DNA damage results in enhanced production of type I IFNs via the cytosolic DNA sensor STING, which can prime the innate immune system for an amplified response [22]. STING is involved in controlling the transcription of host defense genes, including type I IFNs, following recognition of DNA in the cell cytosol [12]. Leaked DNA in the cytosol binds to cyclic GMP-AMP synthase (cGAS), leading to an upregulation of the STING pathway. This in turn activates type I IFNs and cytosolic DNA sensors such as cGAS [23], which potently prime antitumor T cells [12].

PARP inhibition may also render tumor cells more amenable to immune attack through immunogenic modulation, as has been demonstrated with other targeted therapies [24]. Preclinical in vivo and in vitro breast cancer models have shown that PARP inhibition can upregulate PD-L1 by inactivating GSK3β, and that subsequent blockade of PD-L1 resensitizes PARP inhibitor-treated cells to T-cell killing [25]. Additionally, recent data suggest that expression of PD-L1 is upregulated in response to DNA double-strand breaks, and that in BRCA2-depleted cells, PARP inhibition upregulates PD-L1 expression [26]. Combination therapy may be effective, regardless of DDR mutational status, if DNA damage signaling in tumors is down-regulated, thus increasing PD-L1 expression in tumors.

In this study, 2/3 of responders had DDR gene alterations that would predict response to olaparib. Therefore, the depth and duration of response in this trial may simply be an enhancement and prolongation of olaparib’s efficacy when given in combination with durvalumab. However, DDR alteration biomarkers were identified in only 2/3 of responders compared to 88% of patients in the TOPARP-A trial [8]. Emerging data in metastatic urothelial cancer suggest that tumor mutational burden is associated with DNA replication and DDR pathways [11]. Tumors with mutations in DDR genes had higher tumor mutational burden and response rates to PD-L1 blockade with atezolizumab [11]. The deep and sustained responses seen with the combination of durvalumab plus olaparib may reflect the possible association between DDR mutational status and increased potential for immunogenicity [11].

Some of the DDR biomarker-negative responders had shallow deletions of BRCA2. Given the responses, one hypothesis is that the shallow deletion in these responders reflects deep deletion in a subset of tumor cells rather than shallow deletion in all cells, as has been previously proposed [27]. The checkpoint inhibitor may extend this response to additional, unaltered tumor cells. In the TOPARP-A trial, the only responder with a shallow BRCA2 deletion also had a deletion in PALB2.

The range of immune-related toxicities seen in this study raises the possibility of additive or synergistic toxicity. In addition to myelosuppression and nausea attributable to olaparib, 4/17 patients (24%) experienced side effects consistent with irAEs, including sudden-onset hearing loss, optic neuritis, and a case of RS3PE of the hands and upper extremities, all successfully treated with transient steroids. While these toxicities have been reported with PD-L1 inhibition, the high number of apparent irAEs in such a small cohort is noteworthy given the broader clinical experience with these agents [28–30]. It is unclear if these toxicities are the byproduct of antitumor immunologic synergy or if PARP inhibitors immunologically modify off-target healthy tissue, perhaps rendering them more vulnerable to immunologic attack.

Though not detected in this cohort, DNA MMR mutations resulting in microsatellite instability are also associated with higher mutational burden and neoantigens [31]. The only MMR alteration we noted was an indel in PMS2 in a patient harboring BRCA2 homozygous deletion. We did not see any evidence of a hypermutation phenotype in this or any other patient. Using genomic analysis of mCRPC samples, recent data has identified a subtype of mutant tumors in advanced prostate cancer with biallelic loss of CDK12 [32]. CDK12 biallelic loss is associated with increased immunogenicity due to higher neoantigen burden compared to other molecular subtypes of prostate cancer and is mutually exclusive with mutations in DDR [32]. Patients with CDK12 mutations may benefit from immune therapies. CDK12 mutations were not detect in this cohort of patients.

The limitations of our study, such as a small cohort of patients, require cautious interpretation of findings. This is a single-arm study in which all patients receive the same combination therapy. It is not clear to what extent there is synergy of the combination vs. additive effects of 2 individual drugs. In this small patient cohort, pharmacodynamic markers are exploratory and hypothesis-generating. Thus, we found that baseline immune characteristics, in addition to alterations of DDR genes, may predict outcomes in patients treated with durvalumab plus olaparib. In addition, early changes in immune characteristics may be early indicators of clinical outcome. Preliminary data are being reported due to the nature of results seen thus far, particularly the sustained PSA responses. To further investigate combination therapy with durvalumab plus olaparib, we plan to expand the mCRPC cohort up to an additional 65 patients to obtain more detailed response information and conduct additional correlative studies. Patients will continue to undergo mandatory on-study biopsies of a metastatic site of disease, with additional biopsies planned. The expansion cohort will enable in-depth correlative analysis of this combination therapy to better define a potential therapeutic synergy and facilitate biomarker development.

The clinical development of inhibitors of PARP and PD-1/PD-L1 has had a substantial impact on the treatment of advanced malignancies. This is the first study to demonstrate activity for the combination of these agents in prostate cancer patients without biallelic inactivation in DDR pathways, and deep responses in patients with known mutations. While the study is limited by a small patient cohort, the 12-month PFS is 51.5% in patients with advanced metastatic disease, > 50% of whom are taxane-refractory. The future of treatment for mCRPC may take us beyond androgen suppression to combination therapies such as PARP inhibition plus immunotherapy.