Abstract A046: Spatial transcriptomics reveals heterogeneity of epithelial and stromal compartments in healthy and tumor pancreas

Abstract Pancreatic Ductal Adenocarcinoma (PDAC) is one of the most lethal cancers, with a 5-year survival rate of only 13%. PDAC develops from precursor lesions, most commonly Pancreatic Intraepithelial Neoplasia (PanIN). Studying PanIN in human patients is difficult due to its microscopic nature and the lack of indication for sampling the pancreas in the absence of disease. In collaboration with Gift of Life - Michigan, a local organ and tissue donor organization, we obtained over 80 pancreata for research. Surprisingly, we found PanIN in the vast majority of the organs, regardless of the donor’s age; the prevalence of PanIN compared to the relatively low incidence of PDAC indicates that most lesions are destined not to progress. To gather insight into factors that might distinguish lesions designed to progress from lesions that are likely to remain dormant, we sought to characterize both epithelial cells and their supporting microenvironment. Using spatial transcriptomics, we determined that PanIN epithelial cells are transcriptionaally similar to pancreatic cancer; further, limited genetic characterization revealed that they contain similar oncogenic KRAS mutations as pancreatic malignancy. We then sought to determine whether factors in the microenvironment distinguish sporadic PanIN from those associated with overt malignancy. We used 10x Genomic CytAssist Visium technology and profiled an initial batch of 7 tissue sections from donor pancreata and an equal number from tumors. We employed the BayesPrism algorithm to deconvolute the Visium data (which lacked single-cell resolution), using our single-cell RNAseq atlas to define stroma- and epithelial-specific expression profiles. We then segmented the tissue into epithelial and stromal compartments. We further identified various stromal subpopulations, with macrophages and fibroblasts being the most abundant cell types. We defined subpopulations of epithelial cells, fibroblasts, and macrophages through unsupervised clustering. In both donor samples and tumors, we identified heterogeneous populations of stromal cells present in specific geographic patterns, with PanIN-adjacent stroma clearly distinct from stroma surrounding healthy ducts. Cell-cell neighborhood analysis revealed a distinct geographic arrangement of epithelial and stromal compartments, indicating potential communication patterns. Independent unsupervised statistical analysis using Non-negative Matrix Factorization (NMF) using Coordinated Gene Association in Pattern Sets (CoGAPS) R package also revealed similar patterns. Our next steps involve validating gene expression signatures that define each subpopulation and identifying tumor-specific or PanIN-specific genes. Ultimately, we aim to investigate interactions between epithelial cells, fibroblasts, and macrophages to identify potential tumor-promoting or tumor-restraining pathways. Citation Format: Ahmed M. Elhossiny, Jude Okoye, Imade Williams, Yaqing Zhang, Hannah Watkoske, Atul Deshpande, Elana Fertig, Arvind Rao, Eileen S. Carpenter, Timothy L. Frankel, Marina Pasca di Magliano. Spatial transcriptomics reveals heterogeneity of epithelial and stromal compartments in healthy and tumor pancreas [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Advances in Pancreatic Cancer Research; 2024 Sep 15-18; Boston, MA. Philadelphia (PA): AACR; Cancer Res 2024;84(17 Suppl_2):Abstract nr A046.