Abstract IA002: Spatial microenvironment alterations across the clinical spectrum and genomic landscape of human PDAC

Abstract

Human pancreatic cancer has a relatively homogenous genome and established classical and basal transcriptional subtypes, but despite this, there exist many intermediate state tumours, and a complex, spatially heterogeneous tumour microenvironment. To identify inter-omic dependencies, we investigated its genotype to phenotype molecular hierarchy and mapped the relationships between cellular compartments to identify tumour features associated with clinical outcomes. We used imaging mass cytometry to profile the in situ multi-cellular organization of 81 cell types in 221 resected cases with paired whole genome sequencing. We further confirmed and deeply profiled phenotypes using single cell RNA sequencing, visium spatial transcriptomics, and laser capture – mass spectrometry. Cells were organized in highly reproducible microenvironments with complex intra-tumour spatial heterogeneity with distinct relationships observed between tumour phenotype and the physically associated stroma or immune response. We show that tumour phenotypes, vascularization, immune response, and stromal biophysical state are reinforced by genomic aberrations and that sub-tumoural clonal heterogeneity is associated with phenotypic shifts and spatial heterogeneity. Investigating how these tumour microenvironments shift across the disease spectrum, we see enrichments and alterations in spatial environments from resected to advanced disease, between metastatic sites, and upon specific treatments. When predicting survival, spatial single cell data outperformed genomic or clinical features highlighting the importance of multi-cellular content, but machine-learning models which integrated multi-omics with clinical features provided the best prediction of patient survival requiring only 10 non-redundant robust molecular measures. Our findings define a phenotypic and molecular framework of PDAC from genome to tumour–microenvironment offering a refined basis for patient stratification.