Abstract PR016: High-plex spatial phenotyping resolves immune organization in pancreatic ductal adenocarcinoma

Abstract

Pancreatic ductal adenocarcinoma (PDAC) is profoundly immunosuppressive and characteristically exhibits sparse lymphocytic infiltration, yet many tumors harbor focal immune aggregates. The phenotypic diversity and patterns of multicellular organization of these immune aggregates remain incompletely defined. Here, we integrate spatial proteomics and spatial transcriptomics to map PDAC immune aggregates at single-cell resolution and cohort scale. We first optimized high-dimensional immunophenotyping on whole tumor sections using Akoya PhenoCycler-Fusion (CODEX) and performed serial-section integration with Bruker CosMx spatial molecular imaging to relate protein-defined immune and stromal phenotypes to transcriptional programs in nearby malignant glands. Immune aggregates spanned diverse lymphoid and myeloid phenotypes beyond canonical tertiary lymphoid structures and were embedded in distinct malignant and stromal contexts. To determine whether this heterogeneity reflected recurrent patterns of immune-cell organization, we next developed a custom 38-plex panel on Lunaphore COMET and profiled 2,429,896 cells across 264 tumor cores from 147 patients with untreated, resected primary PDAC and linked clinicopathologic data including stage, grade, margin and nodal status, recurrence, and survival. Across this cohort, we characterized immune context at multiple scales by defining multicellular immune communities and developing an interpretable immunoscore that captures coordinated variation in immune composition and spatial architecture. Overall immune infiltration and specific community abundances were associated with clinical outcomes, including a myeloid-rich community with the lowest proportion of T cells near antigen-presenting cells that was associated with worse overall survival. We linked microenvironmental features, including FAP+ fibroblasts, 4HNE-associated oxidative stress, and malignant cell states (classical, basal-like, neural-like progenitor), to variation in multicellular immune communities, their core-level immune ecotypes, and the immunoscore. Further spatial analyses resolved two distinct microenvironments surrounding intratumoral nerves, one characterized by CD163− myeloid cells and the other by fibroblasts and lymphocytes. This study establishes a multiscale framework for resolving immune organization in PDAC and provides a deeply annotated cohort-scale resource for investigating tumor–immune–stroma interactions. ChatGPT was used to proofread and refine the wording of this abstract; all text was reviewed and edited by the authors.