Clear cell ovarian tumors display a unique tumor immune microenvironment

Abstract

<h3>Objectives:</h3> Based on differences in chemosensitivity and clinical outcomes, we hypothesize that clear cell ovarian carcinoma (CCCO) may display a unique tumor immune microenvironment compared to high grade serous ovarian carcinoma (HGSOC). In addition, micro-satellite instability (MSI) and ARID1DA mutations are more common in clear cell ovarian carcinoma suggesting improved outcomes with checkpoint blockade immunotherapy use. Thus, our objective is to compare the immune landscape of CCCO to HGSOC using imaging mass cytometry (IMC). <h3>Methods:</h3> Primary untreated ovarian tumors were collected under an IRB-approved protocol. Formalin fixed paraffin embedded (FFPE) tissue sections were stained with 34 metal-tagged antibodies to detect various cell specific and immune related markers using the Fluidigm protocol. IMC data was obtained by the Fluidigm Helios CyTOF instrument utilizing the Hyperion Imaging System laser ablation module. Images from a 1mm² area of each tissue section were processed and converted to tiff files using MCD Viewer (Fluidigm). The multiplexed images were analyzed using a deep learning based data analytic pipeline beginning with cell segmentation then multiple rounds of phenotype clustering for cell type annotation. Various cell phenotypes were then measured for cell density analysis, or the frequency of a cell phenotype, within a tumor enriched area. Spearman correlation was performed between cell densities of CCCO and HGSOC. <h3>Results:</h3> Samples from 51 patients were used for the study: 10 with clear cell histology and 41 with high grade serous histology. In the CCCO group, stage of disease ranged from stage I (70%) to III (10%), while all the HGSOC had stage III (92.7%) or IV (7.3%) disease (p<0.001). Images were analyzed using 18 markers, based on high signal-to-noise ratios. Following cell segmentation and phenotype clustering, 39 cell subtype clusters were identified including 6 macrophage cell subtypes, 11 T cell subtypes, and 12 tumor cell subtypes. CCCO had a lower density of intratumoral activated CD8+ T cells (CD44+CD8a+CD45RO+) than the HGSOC population (p=0.048). Tumor cell subtypes in the tumor-enriched areas differed between the two groups. CCCO had a higher density of B7-H4midKi67hi tumor cells (p=0.004), but had a lower density B7-H4hiKi67midKeratin8_18+ (p=0.027), B7-H4hiKeratin8_18+ (p=0.004), and B7-H4hiKi67midKeratin8_18+ tumor cells (p=0.007). CCCO had a lower density of CD31+CD73mid cells (p=0.034) but higher density of CD73hi (p=0.024) compared to HGSOC. <h3>Conclusions:</h3> Clear cell ovarian tumors harbor a distinct tumor immune microenvironment compared to high grade serous ovarian tumors. In our cohort, despite most having stage I disease, CCCO demonstrated features associated with poor prognosis including decreased intratumoral CD8+ T cells, higher density of a subtype of actively dividing B7-H4mid ki67hi cancer cells, and higher density of CD73hi cells. Since B7-H4 and CD73 both act as immune checkpoint inhibitors that can confer chemoresistance, therapeutic strategies targeting these molecules may enhance CD8+ T cell mediated antitumor immune response and chemosensitivity to ultimately improve survival rates in patients with CCCO.