Abstract PR-006: Spatially resolved, single cell assessment of pancreatic ductal adenocarcinoma expression subtypes reveals mixed and hybrid basal-classical marker expression with prognostic significance and discrete spatial localization

Abstract

Abstract Background: Pancreatic adenocarcinoma (PDAC) is a rapidly progressive disease, with few molecular markers that stratify patients for survival time or treatment response. Bulk transcriptional analyses of PDAC have identified two major transcriptional subtypes, classical and basal, the latter of which is associated with worse overall survival and limited sensitivity to chemotherapy. However, spatial localization and classical-basal heterogeneity are incompletely understood at the individual cell level within a bulk tumor. Design: We built and validated a multiplex immunofluorescence assay that employs three classical and two basal subtype markers to measure subtype composition at spatially resolved, single cell resolution in a multi-institutional cohort of formalin-fixed paraffin-embedded primary pancreatic cancer resection specimens. Using digital image analysis, supervised machine learning, and Cox proportional hazards regression, we analyzed protein-level classical-basal subtype landscape, cell localization, and outcome associations. Results: Subtyping data were successfully generated for 1.2 million tumor cells across 290 resected tumor specimens (median 3579 cells/tumor). Tumor level analysis of cellular composition revealed that 88% of tumors are “mixed”, harboring any number of both classical and basal cells, whereas tumors composed only of classical or basal cells represented 11% and 1.4% of cases, respectively. In addition to pure classical and basal cells, we identified “hybrid cells” co-expressing basal and classical markers in 79% of cases, ranging in abundance from <1% to 41% of cells within a given tumor. Spatial analysis at an individual gland level revealed that 62% of hybrid cells are located within glands harboring a mixture of both classical and basal cells, even though such mixed glands are substantially less common than purely basal or classical glands. As shown with previous bulk RNAseq datasets, multivariable Cox proportional hazard regression demonstrated that increasing basal cell composition was associated with worse disease-free survival (DFS) (Q4 vs Q1: hazard ratio [HR] 2.04, 95% CI: 1.27-3.30, PTREND 0.0007) and overall survival (OS) (Q4 vs. Q1: HR 1.83, 95% CI: 1.14-2.94, PTREND 0.01), whereas increasing classical cell composition was associated with improved DFS (Q4 vs Q1: HR 0.63, 95% CI: 0.39-1.01, PTREND 0.01). Compared to tumors with few or no hybrid cells, tumors with high hybrid cell composition exhibited worse DFS (High vs Low: HR 1.43, 95% CI: 1.04-1.95, PTREND 0.03). Conclusion: Multiplexed, protein-based expression analysis using a panel of classical and basal subtype markers reveals that most PDACs are composed of a mix of classical and basal cells. Furthermore, a substantial proportion of cases contain “hybrid” cells that harbor prognostic significance and may represent a transitional or plastic cell state based upon their gland-level distribution. Citation Format: Hannah L. Williams, Jinming Zhang, Srivatsan Raghavan, Peter S. Winter, Kevin Kapner, Sara Vayrynen, Andressa Dias Costa, Chen Yuan, Mai Chan Lau, Vicente Morales-Oyarvide, Douglas Rubinson, Lauren Brais, Emma Reilly, Margaret Kozak, David Linehan, Richard Dunne, Daniel Chang, Albert Koong, Aram Hezel, William C. Hahn, Alek K. Shalek, Andrew J. Aguirre, Jonathan A. Nowak, Brian M Wolpin. Spatially resolved, single cell assessment of pancreatic ductal adenocarcinoma expression subtypes reveals mixed and hybrid basal-classical marker expression with prognostic significance and discrete spatial localization [abstract]. In: Proceedings of the AACR Virtual Special Conference on Pancreatic Cancer; 2020 Sep 29-30. Philadelphia (PA): AACR; Cancer Res 2020;80(22 Suppl):Abstract nr PR-006.