The correlation of PD-L1 expression in cytological and histological material of serous high-grade ovarian cancer

Introduction: Neoplastic cells in peritoneal lavage express various proteins with significant prognostic and therapeutic potential. Such expression could differ from the expression in a primary tumor or in metastases. In this research, we compared PD-L1 (programmed cell death ligand-1) expression on ovarian cancer cells in cytological material with its expression on peritoneal metastases and a primary tumor.

Materials and methods: The study included 30 patients who had been operated on for high-grade serous ovarian cancer (HGSC) in FIGO IIIC, over the period of one year. Cytoblocks, cytological and tissue microarrays were assembled and immunostained with PD-L1 antibody. For each tumor compartment we determined four PD-L1 expression categories: negative, low, moderate, and strong expression, according to the percentage of membrane positive tumor cells. Moderate and strong positivity together were considered as high PD-L1 expression.

Results: Moderate PD-L1 expression was the most frequent pattern in primary HGSC (50%) and in peritoneal metastases (omentum) (60%). Cytological samples mostly showed low PD-L1 expression (57%). Statistical analysis did not show a significant difference in PD-L1 expression between the study groups. We found a positive correlation of PD-L1 expression between different, matched tumor samples in every patient, with statistical significance (p < 0.05) between all the analyzed samples.

Conclusion: PD-L1 expression was similar in all three tumor compartments. This could point to similar peritumor regulatory mechanisms of HGSC in primary tumor tissue and cytology tumor samples. Immunohistochemical analysis of the assembled cytoblocks is sufficiently reliable in the assessment of PD-L1 expression on cancer ovarian cells from cytological material.

INTRODUCTION

Ovarian cancer is the eighth most common gynecologic cancer [1]. Malignant ascites in women is a common consequence of ovarian cancer [2]. Cancer staging is usually an indication for peritoneal lavage procedure. According to the International Federation of Gynecology and Obstetrics (FIGO), in stages I and II, the frequency of malignant tumor cells in peritoneal lavage is about 7%, which increases to 89% in stages III and IV [3],[4]. High-grade serous ovarian cancer (HGSC) is the most common gynecologic malignant tumor with the presence of malignant tumor cells in the peritoneal lavage [3].

Programmed cell death ligand-1 (PD-L1) is a molecule with heterogenic expression which is not specific only for tumor cells [5],[6]. In this research we analyzed PD-L1 expression on ovarian cancer cells in HGSC, in different sample types. One of the main immunosupressive interactions is that of PD-L1 protein on tumor cells with programmed cell death-1 (PD-1) on T-lymphocytes. This relation leads to suppression of T-cells and inactivation of their effector functions [5],[6],[7].

Peritoneal effusion in ovarian cancer has an immunosuppressive effect. Previous studies reported various immunosuppressive factors in samples of malignant ascites [2],[8]. PD-1 and PD-L1 molecules are certainly present in malignant peritoneal effusion. They could be significant as potential prognostic and therapeutic markers [8].

Tumor tissue in the omentum point to advanced stages of ovarian cancer and more aggressive biological behavior. The omentum probably has worse antitumor regulatory immune system than the primary site. Such tumor cells are often less differentiated compared to the primary tumor [9].

In this study we analyzed the differences between PD-L1 cancer cell expression in cytological material of peritoneal lavage, in peritoneal metastases, and in primary HGSC.

MATERIALS AND METHODS

Patient cohort

The study was designed as a cross-sectional study, including patients who had surgery due to high-grade ovarian cancer over the period of one year, at the Clinic for Gynecology and Obstetrics, University Clinical Centre of Serbia. We selected 30 patients with HGSC in FIGO IIIC, with peritoneal metastases and peritoneal lavage for immunohistochemical study.

Material collection and processing

All cytological samples (5 ml) were centrifugated for 10 minutes at 1800 rpm. Cytological paraffin blocks were made using the Plasma-Thrombin method [10]. To centrifugated samples we first added 0.5 ml of plasma. In the next step, we put 0.3 ml of thrombin and the tubes were left on 37°C. A clot was formed immediately. The clot was transferred to a tissue bag, fixed in formalin, and subjected to routine histological procedure, like the samples taken from a primary tumor and metastases. In each case, a representative tissue block was selected for further immunohistochemical analysis.

Tissue and cytological microarray

According to tissue and cytological microarray procedures, cylinders were sampled from each paraffin block using a 3 mm puncture needle. A cytological cylinder is the most cellular part of a sample. We counted hot spot areas with more than 50 tumor cells for appropriate evaluation. Tumor tissue cylinders taken from primary sites and metastases contained most homogeneous vital parts of the tumor, with the least necrosis. Cylinders were then moved to recipient paraffin blocks. A series of micro samples were formed. Placental tissue was used as an internal control of immunohistochemical analysis and block orientation [11],[12].

Immunohistochemical analysis

Immunohistochemical staining was done on the Autostainer Link 48, Agilent, Denmark at the Institute of Pathology, Faculty of Medicine, Belgrade. PD-L1 antibody (clone 22C3) and EnVision FLEX visualization system K8023, Agilent were used for immunohistochemical analysis. EnVision FLEX solution pH 6.1, K8005, Agilent was used for epitope unmasking. When it comes to PD-L1 antibody, it was a primary, monoclonal, anti-human, M3653, Agilent, dilution ratio 1:30. Considering PD-L1 immunoscore, we defined four categories for evaluation of a primary tumor site and metastatic tissue PD-L1 expression levels: negative (with no positive cells/with a single positive cell (30 cells). We compared PD-L1 expressions in different samples (the primary site, a metastasis, peritoneal lavage) of each tumor. PDL1 expression in different samples was correlated with available histopathology and clinical parameters [14].

STATISTICAL ANALYSIS

Statistical analysis was done using descriptive methods and adequate statistical tests (Student t-test, Fisher`s exact test, Mann–Whitney, Kruskal–Wallis, ANOVA) for the level of significance of 0.05, using SPSS21 for Windows. The comparison of PD-L1 expression between different tumor samples was performed using Spearman correlation analysis (r_s).

RESULTS

PD-L1 expression in primary high-grade serous ovarian cancer

The mean age of patients in this study group was 57.5±10.14 years. Moderate expression was the most frequent pattern of PD-L1 expression in primary HGSC (50%). One third (33%) of primary tumors had low expression. Strong PD-L1 expression and the absence of PD-L1 expression were less common, with similar frequency (Figure 1).

Figure 1. Categories of PD-L1 expressions in primary high-grade serous ovarian cancer

High PD-L1 expression (moderate and strong expression) in primary tumors was found in 18 (60%) cases. Clinical data and tumor size in high and low PD-L1 expression groups are shown in Table 1.

Table 1. Categories od PD-L1 expression according to clinical parameters and tumor size

There was no statistically significant difference in PD-L1 expression according to menopausal status (p = 0.741 Fisher’s test). The mean tumor size was not significantly different according to PD-L1 expression in tumor cells of primary HGSC (p = 0.864 Kruskal-Wallis).

PD-L1 expression in tumor metastases

PD-L1 expression in peritoneal metastases (omentum) most frequently (60%) showed moderate expression (18 patients). Eight samples (27%) had strong positivity, while other PD-L1 categories showed lower frequencies.

The mean age of patients with high PD-L1 expression in peritoneal metastases was 64.9±10.14 years. There were 23 patients in menopause (88%). In the group of patients with high PD-L1 expression in peritoneal metastases, the mean size of primary HGSC was 40 mm (18-190 mm).

There was not a statistically significant difference in the mean age of patients according to PD-L1 expression categories in metastases of HGSC (p = 0.586 ANOVA). There was no statistically significant difference of PD-L1 expression considering the menopausal status of patients (p = 0.699 Fisher’s test). The size of the primary tumor was not significantly different in relation to PDL1 expression in peritoneal metastases (p = 0.073 Kruskal-Wallis). Statistically significant difference was found in the size of the primary tumor between low and moderate PD-L1 expression (p = 0.006 Mann Whitney test).

PD-L1 expression in cytological material from peritoneal lavage

Adequate cellularity for analysis was present in 28 patients. Low expression was found in 16 samples (57%), while 9 samples (32%) had moderate PD-L1 expression. There was no strong PD-L1 expression in cytological material (Figure 2).

Figure 2. Categories of PD-L1 expression in peritoneal lavage

Patients with high PD-L1 expression in cytological material were 64.9±10.31 years. Most of them were menopausal (89%). The mean size of primary HGSC considering high PD-L1 expression in cytological material was 72.5 mm (18-190 mm).

There were no statistically significant differences in high-grade PD-L1 expressions considering age (p = 0.938 ANOVA), the menopausal status of patients (p = 0.699 Fischer test), and mean tumor size (p = 0.132 Kruskal-Wallis).

Correlation of PD-L1 expression in different samples of HGSC

Categories of high PD-L1 expression in the omentum (87%) and high PD-L1 expression in primary HGSC (60%) were compared. There were no statistically significant differences. Ovarian cancer cells in peritoneal lavage had lower PD-L1 expression compared to cancer cells in omental metastases. Comparing the frequency of moderate expression levels between these two samples, no significant difference was found (12 vs.18; p = 0.273).

Immunohistochemical analysis showed various frequencies of PD-L1 positivity in different materials. The most frequent PD-L1 expression in primary HGSC and metastases was moderate. In cytological material low PD-L1 expression was dominant (Figure 3).

Figure 3. The most common categories of PD-L1 expression in primary HGSC (A) x100, (B) x400, in metastases (omentum) (C) x100, (D) x400, and in cytological material (E) x100, (F) x400 are presented

In this research we analyzed the correlation of PD-L1 expression between different tumor samples, matched according to patients, and found statistically significant results between all study groups. We detected a significant positive correlation in PD-L1 expression between primary HGSC and metastases (p < 0.001; r_s = 0.620). PD-L1 expression between primary HGSC and cytology samples also showed a positive correlation (r_s = 0.599), that was statistically significant as well (p < 0.001). Metastases showed PD-L1 expression in positive correlation according to cytology samples (r_s = 0.817), also statistically significant (p < 0.001).

DISCUSSION

Malignant ascites has a characteristic tumor microenvironment with different antitumor regulatory mechanisms which could stimulate cancer cell proliferation, and the ability of these cells to metastasize. Various cytokines, proinflammatory and immune factors in the peritoneal effusion alter the host’s immune system. Numerous molecules which are expressed in peritoneal lavage have prognostic and therapeutic significance. Their expression could be different from the one in a primary tumor site or in metastases [15].

There are not many studies on PD-L1 expression on cytological specimens and their correlation with PD-L1 expression in standard histopathological samples. Current reference associations (Papanicolaou Society of Cytopathology and Pulmonary Pathology Society) are very reticent about the idea of using cytological preparations to examine PD-L1 expression. Not enough papers and research on this topic, and undefined interpretation of results, make clinicians and researchers doubt reliability of this analysis [16]. Current studies compare the expression levels of PDL1 markers in histological and cytological material on lung cancer tissue [17].

Our research is based on the analysis of differences in PD-L1 expression in ovarian cancer cells located in different tumor microenvironments. Since most primary tumor samples show high PD-L1 expression, the PD-L1 regulatory mechanism is thought to be significantly active in HGSC. We confirmed that high PD-L1 expression is associated with poorer differentiation of ovarian cancer (grade 3) and with an advanced cancer stage (FIGO IIIC).

In our study, PD-L1 expression on ovarian cancer cells from peritoneal lavage showed some frequency differences in expression compared to primary cancer tissue. There was a lower number of patients (32%) with high PD-L1 expression on tumor cells in peritoneal lavage than in a primary tumor (60%). Lower expression of PD-L1 markers on tumor cells in peritoneal lavage could be explained by additional immunoregulatory mechanisms which certainly affect the tumor microenvironment. The assumption is that these mechanisms are differently regulated in peritoneal lavage according to primary cancer tissue. Signal pathways such as LAG-3 (lymphocyte activation gene 3) and TIM-3 (T cell immunoglobulin and mucin domain-containing 3) affect each other as well as the expression of PD-L1 markers. The activity of proinflammatory cytokines (TNF-α, IL-6), galectin-9, and similar factors is also important in the regulation of these processes [18]. A detailed analysis of the interaction of these and similar immune control pathways could lead to more comprehensive explanations of the antitumor immunity process.

The immunohistochemical method using cytoblocks is very representative for adequate detection of ovarian cancer cells from peritoneal lavage [19]. This could be very significant for inoperable patients where diagnostic and therapeutic options are limited. Isolating cancer cells that way could be less invasive and reliable enough to define further procedures. Cancer cells in cytoblocks show the identical type of membrane PD-L1 expression as in histological samples, which further confirms the validity of this method. Using cytoblocks, the material shows a significant correlation of immunohistochemical expression with standard histological preparations. Considering the applicability of cytological material for interpreting PD-L1 expression, there was acceptable reliability of results by cytoblock formation, in contrast to conventional cytological smears and the Liquid-Based method [20].

We found that frequencies of high PD-L1 expression are higher in omentum cancer cells (87%) than in primary cancer (60%) and cytological material (32%). In matched samples, we found a positive correlation between PD-L1 expression in different tumor samples. We got statistically significant results which showed PD-L1 expression on ovarian tumor cells from HGSC with enough reliability, regardless of whether a primary tumor, a metastasis, or a cytological sample were analyzed. These results demonstrate the possibility of applying PD-L1 immunohistochemical analysis to different types of materials with equal reliability.

A limitation of this research is certainly a small sample size. Considering the prospective study design and availability of adequate cytology samples we found this study significant enough to point to further similar and more comprehensive research projects in future. Tumor microenvironment is a great and unexplored field. Comparing its regulatory mechanisms and different molecular behavior could lead to a better understanding of antitumor interactions such as immunosuppression.

CONCLUSION

PD-L1 expression was similar in all three tumor compartments. This could point to similar peritumor regulatory mechanisms between primary tumor tissue and cytology tumor samples. Considering these results, a less invasive procedure such as peritoneal washing, could be used in PD-L1 immunoanalysis and reliably represent characteristics of a primary tumor site. For confirming these findings, we need more comprehensive studies in future.