Abstract 277: High agreement between immune profiling assays on the detection of infiltrating immune cells in ovarian tumor tissue

Abstract

Abstract Detecting immune cells and estimating their heterogeneity in the tumor microenvironment (TME) is critical to inform assessments of cancer prognosis and response to immunotherapies. Many immune activity assays to detect immune cells are available, including multiplex immunofluorescence (mIF) of tissue slices to assess co-localization of immune cell markers and traditional immunohistochemistry (IHC) staining. More recently, statistical approaches to define cell types from RNAseq data (e.g., deconvolution, gene set enrichment) have been widely adopted due to the wide availability of transcriptomics profiles. Here, we interrogated the level of agreement between immune cells in the ovarian TME assessed by mIF (Vectra images processed via InForm/Halo), IHC, and deconvoluted whole exome RNAseq. We leveraged data from epithelial ovarian tumor cores embedded in tissue microarrays from participants in the Nurses’ Health Study I and II and the New England Case-Control Study (mIF, n=668 women; IHC, n=467; RNAseq, n=241). To explore agreement between technologies, we calculated the overall and compartment-specific percentages of cells positive for mIF markers CD3, CD4, CD8, CD19, CD20, CD138 and FOXP3, including cell specific phenotypes based on these markers. For IHC, a pathologist generated scores based on proportion of positive cells and/or aggregation into four levels (0, 1, 2, or 3) for markers CD8, CD68, and CD163. Deconvolution of transcriptomic profiles was performed with CIBERSORT (22 cell types) and xCell (64 cell types). Spearman correlation between technologies for cell specific markers (i.e., cytotoxic T cells, regulatory T cells (Tregs), B cells, macrophages, monocytes, among others) was used to assess agreement. We observed that detection of CD8+ and CD3+CD8+ cells from mIF (cytotoxic T cells) was consistent with the other assays (r >0.25), and the highest correlation was observed with IHC CD8+ cells (r=0.52). For deconvoluted transcriptomic predictions, CD8 T cells (xCell, CIBERSORT), and memory CD8 T cells (xCell) showed high agreement with mIF CD8+ and CD3+CD8+ cells (r=0.25 to 0.48). CD4+ and CD3+CD4+ cells detected with mIF also showed correlation with xCell CD4 memory T cells (r=0.25 to 0.34). As expected, only general class B cells from xCell showed agreement with mIF CD20+ cells (r=0.35). In addition, xCell Tregs correlated with mIF FOXP3+ cells (r=0.34). These results highlight the importance of considering the predictive power of immunological assays for the study of the TME. Compared to mIF, the robustness of deconvoluted RNAseq data varies with the cell type being predicted. Finally, RNAseq should be considered a complementary technology to staining of archival tissues given that the former does conserve spatial distribution of cell types. Citation Format: Oscar E. Ospina, Mary K. Townsend, Tianyi Wang, Naoko Sasamoto, Paul Stewart, Jose Conejo-Garcia, Kathryn Terry, Shelley S. Tworoger, B L. Fridley. High agreement between immune profiling assays on the detection of infiltrating immune cells in ovarian tumor tissue [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 277.