Abstract PR007: Integrated spatial proteomics of human PDAC uncovers an expanded tumour-immune-stroma spectrum with genomic associations

Abstract

Pancreatic ductal adenocarcinoma (PDAC) exhibits extensive cellular and molecular heterogeneity. How unique malignant states interact with and organize local microenvironments to drive therapy resistance and progression remains poorly understood, limiting tumour profiling and patient stratification. Existing binary transcriptional classifications divide PDAC into classical and basal subtypes, yet many tumors occupy intermediate states within diverse immune and stromal microenvironments. To resolve these states, we performed imaging mass cytometry (IMC) on three serial tissue microarray sections from 218 resected PDACs, targeting epithelial, immune, and stromal compartments, respectively. This generated 2457 multiplexed images profiling >2.6 million cells into 83 cell types. These data were integrated with matched whole-genome sequencing (WGS), single-cell transcriptomics, lesion-level deep spatial proteomics, and clinical outcomes. We identified six reproducible tumour phenotypes spanning the classical-to-basal continuum including classical-like states lacking epithelial differentiation transcription factors, S100A4+ non-basal states, and hybrid intermediate states with distinct clinical associations. Trajectory analyses supported progressive transitions across this continuum and revealed mixtures of epithelial states in most tumours. In parallel, we identified eight recurrent immune and stromal microenvironments with strong tumour-subtype specific associations. To define the molecular programs underlying these spatial states, whole-slide IMC-guided laser capture microdissection mass spectrometry (LCM-MS) of 262 tumour and microenvironment regions identified unique functional programs involving differentiation, ECM organization, metabolic activity, stress response and immune regulation. Matched WGS further revealed genotype-phenotype relationships linking recurrent genomic alterations to specific tumour subtypes and spatial microenvironments. Across patients, both epithelial phenotype and microenvironment composition were associated with overall survival, with poorer outcomes observed along the tumour spectrum. The same epithelial continuum was detected in advanced pancreatic and metastatic liver and lung lesions, supporting its conservation across disease sites. Finally, cross-omic survival modelling showed that combining spatial imaging and genomic features outperformed clinical variables and single-modality models, prioritizing tumour states, fibroblast phenotypes, and copy-number alterations with the greatest prognostic value. Our findings define a spatially and molecularly resolved view of PDAC where epithelial states, microenvironments, and genomic alterations converge to shape disease progression, offering a basis for therapeutic targeting and stratification.