Abstract B029: SpatialGland: A novel gland-based spatial analysis of the tumor microenvironment to resolve immune-ductal cell interactions in primary PDAC

Abstract

Pancreatic ductal adenocarcinoma (PDAC) is a highly malignant disease with one of the highest mortality rates and worst patient prognoses amongst all forms of cancer. Despite recent improvements to standard of care treatments and the development of individualized therapies, PDAC remains especially challenging. A particular challenge for developing PDAC therapies is the diverse and immunosuppressive pancreatic stroma within which tumor onset and progression occurs. Different immune and stromal cells within this tumor microenvironment (TME) contribute distinctly to tumorigenesis, metastasis, and drug resistance. Understanding the diverse range of interactions between these immune cells within the TME and malignant PDAC cells is critical, and herein we utilize spatial transcriptomics to present a novel cell-level and a gland-level resolution view of these interactions within the PDAC microenvironment. In this study, we characterized the impact of the tumor microenvironment on malignant PDAC cells using a state-of-the-art approach for spatial transcriptional profiling. Spatial resolution of formalin-fixed, paraffin-embedded (FFPE) tissue biopsies was performed using 10X Genomics’ Xenium platform. A custom 462 gene panel was optimized for identification and functional analysis of relevant immune and pancreatic cells in human PDAC. Xenium data was segmented, annotated, and superimposed on H&E-stained tissue for downstream spatial analyses. Following clustering and annotation using a modified version of scType, we first validated our ability to distinguish and resolve malignant tumor and healthy ductal cell populations in our tissue. For this analysis, we constructed a novel algorithm, SpatialGland, which identifies distinct pancreatic glands in PDAC tissue. Using the output of this algorithm, individual glands were pathologist verified to validate regions of healthy and malignant tissue, and all ductal cells were scored for their malignancy. This pipeline generated a reliable, high confidence method for ductal cell characterization. Building upon our development of SpatialGland, we next utilized this gland-level resolution of our PDAC Xenium data to query both the malignant cells comprising the gland, as well as the different glands’ individual immune neighborhoods. We began by exploring the distinction between classical and basal PDAC cells within individual glands. Ultimately, we demonstrated that even in PDAC tumors that were overwhelming classical, individual basal cells were present and identifiable within glands. Furthermore, differential analysis of malignant cells within individual gland neighborhoods revealed distinct patterns of gene expression within tumor cells that varied based on the majority immune cell types (T cells, B cells, Tregs, APCs) present within 50µm of the gland, contributing to distinct metabolic, pro-metastatic, and immune-resistant tumor cell signatures. These findings demonstrate the utility of spatial transcriptomics to provide gland level insights into the PDAC TME that can inform novel therapeutic developments.