Ovarian Cancer in the Era of Immune Checkpoint Inhibitors: State of the Art and Future Perspectives

Simple Summary Ovarian cancer (OC) represents the fifth leading cause of cancer-related deaths among women. In the advanced disease setting, OC recurrence after chemotherapy is over 70% in the first 2 years, with few therapeutic options. Immunotherapy with the immune checkpoint inhibitors (ICIs) showed high efficacy and changed the therapeutic scenario of many tumors in the last 10 years. With our review, we aimed to summarize the clinical trials of ICIs in OC. In OC, ICIs clinical trials have reported poor outcomes in terms of patient response and survival, with some studies failing to reach their objectives. Combining immunotherapy with drugs targeting different pathways might enhance efficacy and overcome cancer resistance. The search for biomarkers predicting ICIs response is essential for the identification of patients most likely to benefit from ICI therapy. Abstract Background: Ovarian cancer (OC) represents the eighth most common cancer and the fifth leading cause of cancer-related deaths among the female population. In an advanced setting, chemotherapy represents the first-choice treatment, despite a high recurrence rate. In the last ten years, immunotherapy based on immune checkpoint inhibitors (ICIs) has profoundly modified the therapeutic scenario of many solid tumors. We sought to summarize the main findings regarding the clinical use of ICIs in OC. Methods: We searched PubMed, Embase, and Cochrane Databases, and conference abstracts from international congresses (such as ASCO, ESMO, SGO) for clinical trials, focusing on ICIs both as monotherapy and as combinations in the advanced OC. Results: 20 studies were identified, of which 16 were phase I or II and 4 phase III trials. These trials used ICIs targeting PD1 (nivolumab, pembrolizumab), PD-L1 (avelumab, aterolizumab, durvalumab), and CTLA4 (ipilimumab, tremelimumab). There was no reported improvement in survival, and some trials were terminated early due to toxicity or lack of response. Combining ICIs with chemotherapy, anti-VEGF therapy, or PARP inhibitors improved response rates and survival in spite of a worse safety profile. Conclusions: The identification of biomarkers with a predictive role for ICIs’ efficacy is mandatory. Moreover, genomic and immune profiling of OC might lead to better treatment options and facilitate the design of tailored trials.


Introduction
Ovarian cancer (OC) accounts for about 2% of tumors, representing the eighth most common cancer among the female population. The incidence is around 11 cases/100,000 inhabitants/year, and it is higher among white women [1,2]. The frequency of OC rises with

Pembrolizumab
In the KEYNOTE-100 (NCT02674061) phase II study, pembrolizumab 200 mg q3w was administered to two cohorts of patients with recurrent ovarian cancer (ROC): cohort A enrolled 285 patients after one to three prior therapies with a treatment-free interval (TFI) of 3-12 months; cohort B included 91 progressive patients with up to six previous lines of therapy with a TFI of at least 3 mos. The primary endpoint was overall response rate (ORR) by Response Evaluation Criteria in Solid Tumors (RECIST) criteria and according to PD-L1 expression. Secondary endpoints included: duration of response (DoR), disease-control rate (DCR), progression-free survival (PFS), overall survival (OS), and safety. The combined ORR of the two cohorts was 8.0%, the overall DCR 37%, and around 1/3 of responses lasted more than 6 months. The mDoR was 8.2 mos in cohort A and not reached in cohort B. The mPFS was 2.1 mos [12]. PD-L1 positive patients (defined as a combined positive score-[CPS] ≥ 10) reached better results than PD-L1 negative, in terms of both ORR (17.1%) and mOS (21.9 mos-cohort A, and 24.0 mos-cohort B) [12,13]. The most common adverse events (AEs) were fatigue (33.8%), nausea (15.4%), and decreased appetite (10.6%), with 19.7% of women experiencing >G3 AEs. The most common immune-related AEs (irAEs) were thyroid disorders (17.5%). Two treatment-related deaths were recorded, and 5.1% of patients discontinued the treatment due to toxicity [12].
In the Keynote-028 (NCT02054806) multi-cohort phase Ib trial, only PD-L1 positive patients were included. Twenty-six women were treated with pembrolizumab in the OC cohort. The ORR represented the primary endpoint. After a median follow-up of 15.4 mos, ORR was 11.5%, mPFS 1.9 mos, and mOS was 13.8 mos. A total of 73.1% of patients experienced at least one treatment-related adverse event (TRAE): arthralgia (19.2%), nausea (15.4%), and pruritus (15.4%) were the most common. One G3 hypertransaminasemia was recorded, while no deaths or treatment discontinuation for toxicity occurred [14].

Nivolumab
As a single agent, nivolumab was administered to 20 patients with PR-ROC in the UMIN000005714 phase II trial, evaluating the best overall response (BOR) as a primary endpoint: two complete responses (CR) were recorded, the DCR was 45%, the mPFS 3.5 mos, and the mOS 20.0 mos. Of note, ≥G3 TRAEs occurred in 40% of patients. Two patients (10%) experienced serious TRAEs, and 11% of patients discontinued Nivolumab treatment mainly due to treatment-related thyroid disorders [18].
The combination of nivolumab plus bevacizumab was tested in 38 platinum-progressing epithelial ovarian cancer (EOC) patients in the NCT02873962 phase II study. ORR was the primary endpoint, while secondary endpoints were ORR according to platinum sensitivity and PD-L1 expression, PFS, and safety. The combination of nivolumab plus bevacizumab resulted in an ORR of 28.9%, ranging from 16.7% in the platinum-resistant (n = 18) to 40.0% in the platinum-sensitive patients (n = 20). Median PFS was 9.4 mos and 12.1 mos in the overall and platinum-sensitive population, respectively. Of note, better response rates were observed in patients with PD-L1 negative than PD-L1 positive disease. A total of 89.5% of patients developed AEs, of whom the most common were fatigue (47.4%), headache (28.9%), myalgia (28.9%), serum amylases increase (28.9%), aspartate aminotransferase level increase (26.3%), hypertension (26.3%). Four pneumonitis (10.5%) and two colitis (5.3%) cases were reported [20].
The combination of avelumab plus chemotherapy was tested in two randomized phase III trials. The JAVELIN Ovarian 100 (NCT02718417) evaluated carboplatin-paclitaxel chemotherapy alone versus chemotherapy plus avelumab followed by avelumab maintenance versus chemotherapy plus avelumab in the front line OC setting. Nine hundred and ninety-eight stage III-IV patients were enrolled. The primary endpoint was PFS. However, after a median follow-up of 11 mos, PFS was not improved in both avelumab arms, and the trial was stopped after meeting futility criteria [23]. Based on the absence of benefit from Avelumab in unselected patients, the JAVELIN Ovarian PARP 100 (NCT03642132) phase III study, with three arms consisting of chemotherapy plus avelumab followed by maintenance with avelumab plus talazoparib, chemotherapy followed by talazoparib maintenance, and chemotherapy plus bevacizumab followed by bevacizumab, was also terminated [24].

Durvalumab
Durvalumab was tested in combination with chemotherapy, PARPis, or anti-VEGF agents. In the NCT02431559 phase I/II trial, 40 PR-ROC women received durvalumab plus PLD, reaching an ORR of 15% and a 6-mos PFS rate of 47.7%. The most frequent TRAEs were palmar-plantar erythrodysesthesia syndrome (PPES), stomatitis, fatigue, abdominal pain, nausea, fever. G3 TRAEs occurred in at least two patients and included lymphopenia, anemia, increased lipase, rash, and stomatitis [28]. The combination of durvalumab and olaparib was tested in three phase II studies. The MEDIOLA study aimed to evaluate 12 w DCR and safety as primary endpoints, plus 28 w DCR, ORR, DOR, PFS, and OS as secondary endpoints. Initially, 32 women with BRCAmutant PS-ROC were included. The 12 w DCR was 81% [29]. After a median follow-up of 20.4 mos, 28 w DCR was 65.6% with mPFS 11.1 mos, ORR 71.9%, and mOS was not reached [30]. Subsequently, the study included 63 BRCA-wild type patients. Thirty-two patients received durvalumab plus olaparib, 31 patients were treated with olaparib plus durvalumab plus bevacizumab. The doublet cohort reached an ORR of 31.3% (95% CI 16.1-50.0%) vs. 77.4% in the triplet cohort (95% CI 58.9-90.4%). The mPFS was 5.5 mos for the doublet cohort and 14.7 mos for the triplet cohort, respectively. The 24 w DCR was 28.1% in the doublet cohort and 77.4% in the triplet cohort. The most common ≥G3 AEs were anemia, neutropenia, and lipase increased in both cohorts, while in the triplet cohort, hypertension and fatigue were also registered. Six percent and sixteen percent of patients discontinued the treatment in the double and triplet cohorts, respectively [31]. In the doseescalation phase I/II NCT02484404 trial, among 35 patients with ROC, a DCR of 53% was observed with durvalumab plus olaparib or cediranib (5 PR, 13 stable disease [SD]). ≥G3 AEs included anemia (26%) and lymphopenia (14%) [32]. In a third single-center study (NCT02484404), 35 patients with PR-ROC were included. The (primary endpoint) ORR was 14%. Exploratory analyses showed that an increased gamma-interferon (γ-IFN) production was associated with longer PFS (p = 0.023), whereas increased vascular endothelial growth factor receptor (VEGFR)-3 levels determined shorter PFS (p = 0.017). Haematologic toxicity caused the highest ≥G3 AEs (most frequently anemia, 31%) [33].

Anti CTLA-4
Few trials have explored the activity of single agents anti-CTLA-4 Ipilimumab or Tremelimumab in the advanced/recurrent OC with unsatisfactory results. In the NCT01611558 phase II trial with ipilimumab at the dose of 10 mg/kg, 38 out of 40 PR-ROC patients did not complete treatment due to PD, severe toxicity, or death [34]. The combination with PARPis is still at an early stage but seems to be tolerated and induces anti-tumor responses. More specifically, 24 PR-ROC patients received tremelimumab alone or combined with olaparib in the NCT02485990 phase II trial, with 1 PR and 9 SD. No G4 AEs were reported, while the most common G3 toxicities were rash (13%), hepatitis, and colitis (both 8%) [35]. The same combination was administered to three BRCA-mutant OC patients in the NCT02571725 phase I trial, with a good safety profile (only G1/2 AEs were reported) and decreased tumor size after three cycles [36].

Discussion
Given the impact on morbidity and mortality among the female population, the search for new therapeutic options represents an unmet need for OC. Immunotherapy has revolutionized the treatment landscape of many solid tumors in the last ten years, and it now represents the first therapeutic approach with impressive survival benefits in diseases such as lung cancer, melanoma, renal cell carcinoma [10]. However, limited benefits have emerged in OC, even leading to premature termination due to the futility of some studies. Different components of the OC tumor microenvironment (TME) contribute to this failure, such as myeloid-derived suppressor cells (MDSCs), tumor-associated macrophages (TAMs), T-cells, cytokines, and soluble factors [37][38][39][40]. MDSCs exert immunosuppressive functions, such as the inhibition of T-effector and natural killer (NK)-cells, and are induced under proinflammatory cytokines, IFNγ, tumor necrosis factor-alpha (TNFα), interleukin (IL)-6 [41]. In OC, IL-6 plays a negative prognostic role and is associated with high MDSCs, and tumor progression [42,43]. The inflammatory cytokines cooperate to induce cyclooxygenase-2 (COX-2) and lead to prostaglandin E2 (PGE2) synthesis, which limits T-cell recruiting at tumor sites, together with VEGF [44,45]. TAMs are recruited at ovarian tumor sites, and IL-6, IL-10, transforming growth factor (TGF)-β promote their differentiation in M2 macrophages, associated with tumor invasiveness, spread, and angiogenesis [46][47][48]. M2 macrophages increase with the OC stage when contemporary M1 macrophages decrease, playing a negative prognostic role [49][50][51]. Moreover, they promote immunosuppression by producing cytokines (IL-1R, IL-10, C-C Motif Chemokine Ligand [CCL]17, CCL20, CCL22) that inhibit T-effectors proliferation and enhance Tregs function [52][53][54]. Treg cells are associated with advanced stages of OC and have a negative prognostic and immunosuppressive role [54]. They produce IL-10 and TGFβ, contributing to the inhibition of effector T-cells [55]. High levels of immunosuppressive elements within OC TME can also weaken dendritic cells and antigen-presenting cells (APCs) activity [56]. More accurate knowledge of the TME of the primary tumors and the metastatic sites will facilitate the design of more effective treatment combinations (Figure 1). cells, and are induced under pro-inflammatory cytokines, IFNγ, tumor necrosis factoralpha (TNFα), interleukin (IL)-6 [41]. In OC, IL-6 plays a negative prognostic role and is associated with high MDSCs, and tumor progression [42,43]. The inflammatory cytokines cooperate to induce cyclooxygenase-2 (COX-2) and lead to prostaglandin E2 (PGE2) synthesis, which limits T-cell recruiting at tumor sites, together with VEGF [44,45]. TAMs are recruited at ovarian tumor sites, and IL-6, IL-10, transforming growth factor (TGF)-β promote their differentiation in M2 macrophages, associated with tumor invasiveness, spread, and angiogenesis [46][47][48]. M2 macrophages increase with the OC stage when contemporary M1 macrophages decrease, playing a negative prognostic role [49][50][51]. Moreover, they promote immunosuppression by producing cytokines (IL-1R, IL-10, C-C Motif Chemokine Ligand [CCL]17, CCL20, CCL22) that inhibit T-effectors proliferation and enhance Tregs function [52][53][54]. Treg cells are associated with advanced stages of OC and have a negative prognostic and immunosuppressive role [54]. They produce IL-10 and TGFβ, contributing to the inhibition of effector T-cells [55]. High levels of immunosuppressive elements within OC TME can also weaken dendritic cells and antigen-presenting cells (APCs) activity [56]. More accurate knowledge of the TME of the primary tumors and the metastatic sites will facilitate the design of more effective treatment combinations (Figure 1). OC encompasses a heterogeneous group of malignancies that in over 95% of cases have an epithelial origin and are more frequently represented by high grade serous OC encompasses a heterogeneous group of malignancies that in over 95% of cases have an epithelial origin and are more frequently represented by high grade serous ovarian carcinoma (HGSOC) (70% of cases), followed by endometrioid ovarian cancer (EOC) (10%), clear cell OC (ccOC) (10%), low-grade serous OC (LGSOC, less than 5%), and mucinous OC (MOC, around 3%) [3]. Among them, the ccOC seems to be the most immunogenic: it more frequently carries the DNA microsatellite instability (MSI), has higher CD8+ tumor-infiltrating lymphocites (TILs), CD8+/CD4+ ratio, and higher PD-L1 levels [57,58]. Effectively, it is five times more responsive to ICIs than other OC subtypes [19]. Even among HGSOC, at least four different genomic classes were identified in The Cancer Genomic Atlas registry, differing for immunoreactivity. A unique subtype expresses genes related to immune sensitivity such as Toll-like receptor (TLR), TNF and is characterized by higher TILs infiltration [59,60]. Moreover, proteomics studies showed that the four subclasses of HSGOC are characterized by different expressions of proteins involved in DNA replication, ECM and cellular interaction, and cytokine signaling that contributes to immune responsiveness [61]. In our opinion, the different ICIs response observed among OC patients is rooted in the inter-tumor heterogeneity. Therefore, a deeper insight into the genomics characteristics of OC and their relationship with the immunological profile could allow us to better clarify the predictive factors for ICIs response. Ideally, specific immunogenomic scores could be developed for more accurate patients selection.
OC has been indicated as potentially more immune responsive when carrying BRCA mutations or homologous recombination deficiency (HRD). In fact, the impaired DNA repair leads to neo-antigens production, resulting in a higher tumor mutational burden (TMB) (even if <10 mutations per megabase are usually detected) and recruiting TILs at tumor sites. However, HRD or BRCA mutations were not linked to a higher sensitivity to ICIs in the IMagyn050 nor in the Javelin Ovarian 100 trials [23,27]. BRCA-mutant/HRD OC is associated with higher CD3+ and CD8+ TILs, PD1/PD-L1 levels, and genes related to cytotoxicity, such as T-Cell Receptor (TCR), γ-IFN, and TNF-Receptor pathway [62][63][64][65]. As proof of this, in the NCT02484404 trial, durvalumab plus olaparib determined a longer PFS in case of increased γ-IFN production [33]. Another mechanism of immune responsiveness is represented by the mismatch repair (MMR) deficiency, harboring the DNA MSI. MSI tumors produce neo-antigens, with a 10-100-fold higher TMB than MS stable (MSS)-tumors, resulting in high immunogenicity. Some genes triggering MSI were also identified in a percentage ranging from 17% to 59% of OC (more commonly in non-serous subtypes): the oncosuppressor TP53; Dihydropyrimidinase-related protein (DPYSL)-2, involved in microtubules function; Alpha Kinase (ALPK)-2, with a role in apoptosis and DNA repair [66]. In Lynch syndrome, a germline mutation of the MMR genes MutL homolog (MLH)-1, MutS homolog (MSH)-2 and -6, PMS1 homolog (PMS)-2 leads to an increased risk to develop some cancer subtypes, including OC [67]. Therefore, these tumors may be good candidates for ICIs treatment. Other genes could be involved in ICI response, justifying the different results observed among OC patients. The SWItch/Sucrose Non-Fermentable (SWI/SNF) complex consists of around 15 subunits, acting as a chromatin remodeler. In other tumor subtypes, the loss of function of the SWI/SNF complex predicts ICI response, increasing MMR deficiency, TMB, and neo-antigens production [68]. SWI/SNF complex mutations were frequently detected in OC [69]. We can assume that genetic diversity contributes to different ICI responses among OC patients. A more extensive genetic characterization could allow more accurate identification of responders and non-responders.
The possible relationship between platinum-and immunotherapy-sensitivity/resistance is also a field that merits further investigation [70]. A series of genetic and epigenetic elements were identified to drive platinum response: alterations of p53, specific microR-NAs, elements driving the epithelial-to-mesenchymal transition (EMT), HRD, and BRCA mutations [71]. Since BRCA mutation and HRD were proposed to correlate with platinum sensitivity in contemporary deficient nucleotide excision repair, the co-administration of PARPis and ICIs in PS-ROC could result in higher ORR and survival rates [29][30][31]. PARPis enhance ICIs activity because they induce the release of neoantigens, increasing the TMB, promote PD-L1 expression, and directly activate the IFN genes; however, this was determined in OC [17,[29][30][31]72,73]. Many ongoing trials are addressing this combination strategy in the advanced setting (Table 2).  As ICIs monotherapies showed only minimal results in terms of response rate and survival in OC, the combination with agents with different mechanisms of action appears a promising strategy to increase efficacy. Although chemotherapy represents a cornerstone in the treatment of advanced OC, it was historically perceived to play an immunosuppressive role. On the contrary, more recently, it has emerged that platinum derivatives promote APCs and their function, activating the immune response [74][75][76][77]. Doxorubicin plays an immunomodulatory effect, reducing the immunosuppressive state and improving tumor sensitivity to NK and CD8+ T-cells [78]. Low-dose cyclophosphamide also holds immunomodulatory properties, such as Tregs reduction and CD8+ cells induction [79,80]. However, the studies conducted so far did not lead to survival improvements. Besides the immunological potential, timing and schedule should be more deeply investigated and optimized for improving efficacy. The combination of ICIs and anti-VEGF agents seems attractive because the anti-angiogenic drugs directly influence OC TME [20,31,32,[81][82][83]. Other combinations with multikinase inhibitors targeting VEGF/VEGFRpathway, such as cabozantinib or lenvatinib, are now under evaluation. The association with other agents with immunotherapeutics role, such as the anti-Lymphocyte-activation gene 3 (LAG-3) Relatlimab, as well as monoclonal antibodies such as the anti-Cluster of differentiation (CD)27 Varlilumab, the anti-CD47 Magrolimab, is under investigation (Table 2). Actually, overcoming the immunosuppressive pathways in the TME could represent a complementary way to potentiate ICIs effect on the immune system. Therapeutic vaccines were administered in OC, inducing cellular and humoral responses but rarely survival improvement as monotherapies [84]. Hence, several tumor-associated antigens were found in OC, such as p53, folate receptor (FR), New York Esophageal Squamous Cell Carcinoma-1 (NY-ESO-1), and Ca125 [85][86][87][88]. Therefore, combinations of ICIs and vaccines need to be explored. New approaches such as autologous TILs, cancer cell therapy, and adoptive cell therapy (ACT) also represent future possibilities for improving ICIs efficacy (Table 2).
Currently, a uniformly accepted predictive role of PD-L1 for ICIs response was not yet identified in solid tumors, including OC. PD-L1 expression varies between primary tumors and metastases, implying heterogeneity [89]. However, even if PD-L1 positivity was retrieved in around 1/3 OCs, the clinical impact was not elucidated, with conflicting results regarding the association with higher tumor stage/grade or shorter survival [90][91][92][93][94]. Indeed, some of the published trials reported better results for PD-L1 positive than PD-L1 negative patients [12,13,27]. In other studies, PD-L1 positivity was not predictive of ICIs response [19,20]. Recent research has focused on the post-transcriptional modifications of PD1, and even more PD-L1, which N-glycosylation of specific sites functionally modulates. PD-L1 and PD1 N-glycosylation ensure stability, prevents clearance, and influences mutual interactions [95,96]. The N-glycosylation of the PD1/PD-L1 receptors and its aberrations should be better investigated as possible immune resistance mechanisms in OC since specific glycoproteomic signatures were found in HGSOC: the immunoreactive subtype was richer in mannose than the mesenchymal, which was mainly fucosylated [97]. Moreover, it was evidenced that the antibodies used in the immunohistochemical analysis for PD-L1 accessed the highly glycosylated PD-L1 with difficulty, resulting in a certain percentage of PD-L1 false-negative results partially explaining ICIs efficacy also in PD-L1 negative patients [98]. More profound knowledge of the post-transcriptional status of PD1/PD-L1 and the search for biomarkers with a predictive role for ICIs' efficacy is warranted to ensure the best patient selection.

Conclusions
Thus far, OC remains one of the few malignancies in which ICIs have not changed the standard of care, and neither monotherapies nor combinations have been approved. Effectively, significant heterogeneity was identified across OC patients at the genomic, proteomic, glycoproteomic, and immunologic levels, that in our opinion, should be further investigated to improve ICIs efficacy. We also believe that the combinations of ICIs with agents with different mechanisms of action will strengthen ICIs efficacy in OC. The combinations of ICIs with anti-VEGF agents or PARP-inhibitors represent potentially very effective associations, and several studies examine this strategy. However, schedule and timing should be optimized in order to preserve tolerability. Combinations with other agents, such as multikinase inhibitors, immunotherapies targeting the immunosuppressive network in the TME, or vaccines, should be further explored to maximize the efficacy with minimal toxicity.
Besides PD-L1, biomarkers with a predictive role to ICIs should be investigated. Integrating such biomarkers with genomic and immunologic profiling will provide a comprehensive understanding of OC, guiding clinical trials towards rational therapy combinations and sequencing.