Stratification of ovarian tumor pathology by expression of programmed cell death-1 (PD-1) and PD-ligand- 1 (PD-L1) in ovarian cancer

The early signs of ovarian cancer are asymptomatic and thus approximately 75 % of cases are detected in the advanced metastatic stages. Conventional management strategies for advanced disease include cytoreductive surgery and chemotherapy. Most current treatments are not curative for patients with advanced disease and hence survival for this category of patients is low [1]. It is estimated that in 2017 there will be 22,440 new cases of ovarian cancer in the Unites States, and that 14,080 patients will die due to this disease [2]. Approximately 80 % of patients diagnosed with late stage ovarian cancer die within five years.

To provide more effective treatment options for patients, several clinical trials are ongoing using novel single and combination regimens to improve survival. For cancer therapy, there have been several distinct landmarks in the development of new therapies and FDA approved treatments over the last decade [3]. However, even with the current treatment options a considerable number of patients are not yet receiving adequate therapy for the management of advanced stage ovarian cancer and other malignancies.

The development and optimization of the use of novel therapies such as immunotherapy, requires an in-depth understanding of specific target molecules and cellular interactions in tumors. Early efforts in immunotherapy can be traced to 1891, in which administration of intra-tumoral injections of bacteria led to a shrinkage of patients’ tumor [4, 5]. Since then, significant progress has been made in the field [6]. One of the recent highlights in novel treatment options for cancer has been the targeting of immune checkpoint inhibitory molecules [7‐9]. Immune checkpoints are critically important in health and disease. They represent co-signaling pathways which are either costimulatory or coinhibitory. In the body, linkage of coinhibitory receptor and ligand suppresses T cell receptor signaling, and limits immune responses. Whereas this function of checkpoint inhibitory molecules is beneficial during resolution of infection, or in the development of self tolerance to prevent autoimmune conditions [10‐12], ligation of checkpoint inhibitory molecules can be a powerful and unwanted mechanism of immunosuppression in cancer [13‐15]. Since the successful introduction and FDA approved use of an antibody targeting checkpoint inhibitory molecule cytotoxic T lymphocyte associated-4 (CTLA-4) (Ipilimumab; Yervoy®) in patients with unresectable or metastatic melanoma in 2011 [16], this agent is now in use in over 40 countries. Attention has more recently focused on another checkpoint inhibitory molecule programmed cell death-1 (PD-1) and its ligand programmed cell death-1 ligand (PD-L1) [17‐19].

Antibodies inhibiting PD-1 and PD-L1 have recently been FDA approved for the treatment of cancer. For example, the agent nivolumab (Opdivo®) is approved for unresectable or metastatic melanoma, non-small cell lung cancer (NSCLC), Hodgkin’s lymphoma and renal cell carcinoma. Pembrolizumab (Keytruda®) is FDA approved for melanoma and NSCLC, and a blocking anti-PD-L1 antibody Atezolizumab (Tecentriq®) is also FDA approved for unresectable bladder cancer and for NSCLC. Blockade of this pathway is particularly useful in patients as it is applicable to a wide range of cancers, and because it induces anti-tumor immune responses capable of targeting mutated proteins [20]. Importantly, treatment targeting PD-1 signaling has fewer high grade toxicities than other immunotherapies [13, 21].

Medical centers are currently utilizing these agents in ongoing clinical trials for various cancers, including ovarian cancer [7, 22, 23]. Initial reports of some trials show promising objective response rates (ORR) for the treatment of ovarian cancer with anti-PD-1 antibody nivolumab (ORR of 15%, n = 20 patients), and pembrolizumab (ORR 11.5%, n = 49), or an anti-PD-L1 antibody avelumab (ORR 10%, n = 124) [3, 24]. Those responding often had durable responses, suggesting that if we could identify the subgroup that might typically respond, we could advance the therapeutic options in this subgroup of ovarian cancer patients.

PD-1 is primarily expressed on CD4+ and CD8+ T cells and is associated with T cell exhaustion [11, 12, 14]. PD-L1 is expressed on many cell types including tumor cells and macrophages, including those with an immunosuppressive phenotype [12, 25, 26]. Ligation of PD-L1 on tumor cells with PD-1 on T cells, for example, abrogates T cell proliferation, diminishes T cell activation and leads to a predominance of a T helper 2 (Th2) cytokine tumor microenvironment, with a pro-tumor propensity. Antibody blocking of PD-1 or PD-L1 restores T cell proliferative and cytotoxic functions, and induces a T-helper 1 (Th1) phenotype, thereby re-invigorating T cells, with resulting potent anti-tumor capacity [14, 27, 28].

The immune mechanisms of disease improvement with administration of checkpoint inhibitory molecules are not well understood. Clinically, there is also ongoing debate over which patients will benefit from this therapy, whether patients who respond initially will continue to show complete responses (CR) or partial responses (PR), and whether patients’ tumors need to express PD-1 and/ or PD-L1 in abundance, to predict beneficial responses to checkpoint inhibitory molecule blocking therapy targeting these molecules. At the present time, there are more questions than answers.

As a study of the immune microenvironment of ovarian cancer patient tumors offers insight into the baseline immune landscape associated with patient survival and tumor pathology, and implicates broader scope for targeting these molecules in combination studies with conventional therapy and with other novel therapies, we undertook these present investigations. We primarily selected advanced disease ovarian cancer patients for study, since this group typically have poor outcome with standard therapy, and our future goal in translational medicine is to address the need for novel alternative treatment options in this patient sector. We evaluated the expression and localization of PD-1 and PD-L1 in a cohort of ovarian cancer formalin fixed paraffin embedded (FFPE) tumor sections, and investigated whether the relative expression levels of these molecules can be relevant patient prognostic indicators. We also studied the impact of tumor infiltrating lymphocytes (TILS) along with these checkpoint molecules, on patient status including tumor grade, disease stage and survival post diagnosis.

Ovarian cancer is usually diagnosed in the advanced metastatic stages. Treatment of advanced stage disease with conventional therapies is only sufficiently effective in a limited number of patients, thus in about 80 % of these patients there is disease progression or disease recurrence and death, within five years of diagnosis. In many cancers, investigators are focusing on the development of novel therapies as alternative and more robust options to existing therapies. Whereas conventional therapies primarily focus on the destruction of tumor cells, many novel therapies are designed to stimulate immune cells to elaborate augmented anti-tumor immune responses. In this respect, checkpoint immune inhibitory molecules have come full circle over the last decade for cancer immunotherapy.

In a normal functioning immune system, T cell activating and inhibitory receptors balance immune tolerance, and the amplification of immune responses. In the body, immune checkpoints are designed to reduce autoimmune responses, or to attenuate immune responses which were elaborated after infections [10‐12, 14, 27, 28]. In cancer, blocking of immune checkpoint molecules with antibodies is a novel and promising therapy, as it potentiates anti-tumor immune responses in patients [9, 13, 15, 19].

The first of these checkpoint inhibitory molecules to be targeted for blocking in therapy, and is now FDA approved for cancer therapy, is CTLA-4 [16, 17]. Therapy design is based on the following principle. Binding of costimulatory molecules CD80 or CD86 on antigen presenting cells to CD28 on T cells delivers a positive costimulatory signal contributing to T cell activation. On the contrary, linkage of CD80 or CD86 to CTLA-4 (a molecule closely related to CD28) results in inhibition of immune responses, and exhausted T cells, which are less able to proliferate or to secrete T helper 1 (Th1) cytokines [29] . Blocking of this inhibitory pathway with anti-CTLA-4 antibodies, results in re-invigorated T cells with greater proliferative function, and durable anti-tumor potency [30, 31]. By similar rationale, there are several other immune checkpoints in the immune system, which are now being targeted in cancer clinical trials. In the case of blocking PD-1 signaling, there are currently about nine different antibodies in cancer clinical trials targeting the PD-1/ PD-L1 pathway [3, 17].

This study was designed to better understand the ovarian cancer tumor microenvironment (TME) with relation to the localization and frequency of PD-1, PD-L1 and TILS in the tumors of ovarian or related cancer patients, diagnosed with advanced disease. Firstly, results showed that patients’ age was an independent prognostic factor in survival, with patients over 60 years of age more likely to die than those diagnosed when younger than 60 years. This may be due to the fact that younger patients can tolerate more aggressive surgery and chemotherapy than older patients. Additionally, as expected, disease stage was also an independent prognostic factor in outcome, such that patients diagnosed with advanced disease had a lower median survival than patients diagnosed with early stage disease. These findings are in agreement with those of other investigators [32].

To shed light on the relevance of PD-1 and PD-L1 in ovarian cancer outcome, we studied the abundance of these molecules in the TME. PD-1 was compartmentalized in the stroma and in the tumor epithelium, and this molecule was expressed in 87% of tumors. PD-L1 was only present in the tumors of 33% of patients. Patients who expressed PD-L1 had a trend towards survival, as did those expressing PD-1 or CD3, even though these trends were not significant. In our cohort we did not find a significant association with FoxP3 and survival. The presence of PD-L1 and FoxP3 together in high grade tumors showed the same level of association as the presence of PD-L1 alone. Some studies have reported that FoxP3 positive cells in ovarian tumors is negatively associated with outcome, however a meta-analysis of 7 ovarian cancer studies with a total of 869 patients, did not find FoxP3 TILS in ovarian cancer to be a significant prognostic indicator [33].

In cancer there are conflicting reports concerning the expression patterns of PD-1 in patients’ tumors and the association with survival, with either positive or a negative association [34‐36]. One recent report found that PD-1 positive TILS and /or PD-L1 positive tumor cells had a positive association with survival of ovarian cancer patients [37].

The expression of PD-L1 in tumors was shown to be positively associated with survival in NSCLC [38] and in ovarian cancer [37]. On the contrary, others report a negative prognostic impact of PD-L1 expression in ovarian cancer [39, 40] and breast cancer [41]. In a review and meta-analysis of 17 studies using data of 2869 head and neck cancer (HNC) patients, authors found that there was no significant association between the expression of PD-L1 on survival in HNC patients [42]. Additionally, similar analysis of reports with NSCLC patients also did not show a significant association with PD-L1 expression and survival [43]. Taken together, this indicates that the prognostic impact of PD-1/ PD-L1 expression in tumors is not yet established. Here, our studies did not find a significant association between survival and PD-1 or PD-L1 expression in ovarian cancer.

Differences in reports of the expression of these molecules and associations with survival may be attributed to several reasons. Firstly, it is possible that there may be different survival outcomes due to the site of cancer. For example, Paulsen and colleagues [38] found that whereas in a cohort of patients a high density of PD-1 and PD-L1 had a favorable impact on NSCLC, this association was not present when these molecules were studied at metastatic sites such as lymph nodes of the same patients. This is highly likely because each cancer site has a different immune landscape, and levels of soluble molecules such as IFN-γ which is a strong regulator of PD-L1 expression [44, 45].

Secondly, differences in reports may be due to variations in staining protocols across individual laboratories. Many investigators report difficulty in IHC staining for PD-L1. In these present investigations, we initially used tumor arrays to study the expression of PD-L1 and PD-1 in tumors. When optimizing our staining protocol, we found that it was difficult to select cores which had a good representation of tumor and stromal areas for accurate visualization of the density of these molecules in tissue arrays. Therefore, in this study we used whole tumor sections for the identification of these molecules, as is done for patient diagnosis. For molecules such as PD-L1 especially, which is not widely expressed in tumor tissues, reports in which IHC staining was performed using tumor cores may give different findings than reports from other laboratories in which staining was performed using whole tissue sections.

Thirdly, the use of different primary antibody clones to identify PD-1 or PD-L1 in tumor tissue in each laboratory, may also lead to variability in staining of sections and in interpretation. Finally, manual staining protocols in comparison with automated staining may further contribute to variations in staining interpretation.

Finally, additional parameters which may alter the expression of these molecules in tumors, is the administration of treatments such as chemotherapy to patients prior to surgery. However, at our center, for patients diagnosed with ovarian and related gynecologic cancer, the primary treatment is most often surgery for removal of tumors. Patients then undergo courses of standard therapy such as chemotherapy. Consistent with this practice, we found that a study of our patient cohort treatment plans revealed that only 3 of 55 (5.5%) patients had chemotherapy in the interval before surgery.

A low frequency of PD-1 in tumors was associated with advanced disease. This association between low PD-1 density and advanced disease was only significant when measuring S-PD-1 or combined PD-1, whereas low T-PD-1 density alone was not associated with advanced disease. Although beyond the scope of this study, this finding raises the possibility that T-PD-1 and S-PD-1 positive cells may perform unique immunosuppressive roles in the ovarian TME.

PD-L1 expression was almost exclusively restricted to high grade tumors, such that there was a positive and significant association between PD-L1 and high grade tumors. This finding may be of translational significance in selecting patients for therapy blocking PD-1/ PD-L1 signaling, and we suggest that patients with high grade tumors, with pre-existing PD-L1 expression might be excellent candidates for therapy blocking this pathway. In support of this idea, a recent report shows that in an ongoing study of urothelial bladder cancer patients, treatment with durvalumab (MED14736; an anti-PD-L1 antibody) resulted in improved outcome in PD-L1 tumor positive patients. In pre-treatment tumor biopsies patients, 40 patients were PD-L1 positive and 21 patients negative for PD-L1. In 42 evaluable patients, the ORR was 31.0% (95% CI, 17.6 to 47.1), the ORR was 46.4% (95%CI, 27.5 to 66.1) in the PD-L1 positive patient subgroup, and 0% (95% CI, 0.0 to 23.2) in patients negative for PD-L1 [46].

We suggest that due to the conflicting reports concerning the impact of PD-1 and PD-L1 on survival in cancer patients, a future larger study is needed investigating these molecules in ovarian tissue, with standardized protocols and defined cut off points for positive staining and scoring criteria across centers, to minimize study variations. Even so, the potency of patient responses to PD-1/ PD-L1 blocking antibody therapy may be influenced by the density of other pre-existing or emerging checkpoint molecules in tumors, including T-cell immunoglobulin and mucin-domain containing-3 (TIM-3), lymphocyte-activation gene 3 (LAG-3) and V-domain Ig suppressor of T cell activation (VISTA). Other parameters such as the presence of myeloid derived suppressor cells, levels of Th2 cytokines (for example IL-10) and of indoleamine 2,3-dioxygenase (IDO) can also limit anti-cancer immune responses to therapy blocking PD-1/ PD-L1 [19, 47, 48]. Furthermore, genetic alterations within the tumor (including DNA rearrangements, mutations, deletions and insertions) alter tumor mutational loads, and it is reported that tumors with high mutational loads have the greatest response to checkpoint inhibitory blockade therapy [17, 49, 50].

Finally, due to the multiplicity of factors regulating ORR, we believe that antibody therapy targeting the PD-1/ PD-L1 pathway in ovarian cancer will be of maximum efficacy when used in combination with other treatment regimens. Such treatments include standard therapy, immunotherapy blocking other checkpoint inhibitory molecules, dendritic cell vaccines, chimeric antigen receptor (CAR) T cell therapy, or targeted therapy, all of which can downregulate other immune suppressive mechanisms in patients, concomitantly.

Treatments inhibiting PD-1 and PD-L1 are beneficial only in some ovarian cancer patients. Our studies reveal that a low density of PD-1 and of PD-L1 expressing cells in tumor tissue are significantly associated with advanced disease, and that PD-L1 is expressed significantly more in high grade tumors than in low grade tumors. We conclude that a subgroup of advanced disease ovarian cancer patients with high grade tumors, bearing PD-L1, may be the best candidates for immunotherapy targeting PD-1 and/ or PD-L1.