Abstract B041: A High-Resolution Integrative Framework for Annotating Cell States Underlying Interpatient Heterogeneity and Tumor Plasticity in Pediatric and Adult Gliomas

Abstract In this study, we systematically map the cell states and molecular signaling networks associated with the mesenchymal-like (MES-like) transformation of tumor cells in both pediatric and adult gliomas. While previous work has correlated the transition of homeostatic macrophages into a GPNMB-high cell state with tumors being more mesenchymal, it is increasingly clear that the MES-like transformation involves changes in other non-neoplastic brain cell types. However, the specific cell state changes and molecular signals underlying this process remain poorly understood. Here, we developed and applied a computational framework utilizing ConDecon, a clustering-independent gene expression deconvolution method, to deconvolve a large-scale cohort of 1, 573 pediatric and adult glioma bulk samples using reference single-cell atlases from 4 glioma subtypes (glioblastoma, adult low-grade glioma, pediatric high-grade glioma, and pediatric ependymoma). We used the inferred cell probability distributions to partition cells in each reference single-cell atlas into transcriptomic states that co-vary across patients, representing cell states associated with interpatient heterogeneity. We then examined associations between the abundances of these cell states and clinical variables, as well as candidate paracrine interactions through integrated ligand-receptor analysis. By incorporating population-level information in the annotation of cell states, our analysis enabled the high-resolution annotation of cell states within the tumor microenvironment, including a distinct 'mesenchymal niche' defined by the co-occurrence of MES-like tumor cells, GPNMB-high macrophages, activated endothelial cells, and pericytes. The abundance of this niche inversely correlates with that of neuro-epithelial tumor cell states and is associated with poor survival across gliomas. Ligand-receptor profiling identified key paracrine signals associated with this niche, specifically involving extracellular matrix adhesion (FN1, Collagen), angiogenesis (VEGF, ANGPTL), biosynthesis (Visfatin), and tumor-immune communication (CD44, SPP1). Altogether, our study presents a novel computational framework for high-resolution mapping of cell-state plasticity within the tumor microenvironment by integrating bulk and single-cell transcriptomic data prior to clustering and annotating the single-cell data. By identifying the constituent cell states and paracrine signals associated with the mesenchymal niche, our approach provides insights into the molecular and cellular mechanisms underlying glioma progression and may inform novel therapeutic strategies targeting resistance. Citation Format: Annan Timon, Pablo Cámara. A High-Resolution Integrative Framework for Annotating Cell States Underlying Interpatient Heterogeneity and Tumor Plasticity in Pediatric and Adult Gliomas abstract. In: Proceedings of the AACR Special Conference in Cancer Research: Brain Cancer; 2026 Mar 23-25; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2026;86 (6Suppl): Abstract nr B041.