TMET-30. Uncovering the metabolic programs underlying malignant cell state heterogeneity in glioblastoma

Abstract

Abstract BACKGROUND Glioblastoma (GBM) is an incurable and aggressive brain cancer marked by profound intra-tumoral heterogeneity. Malignant cells exist in four core transcriptional states: stem-like (OPC-like and NPC-like) and more differentiated astrocytic (AC-like) or mesenchymal (MES-like) states. Using spatial transcriptomics, we previously identified a layered tumor structure of these states, with hypoxia emerging as an organizing driver, implicating oxygen availability as a central factor in cell state dynamics. However, how these cellular states interact with each other and with the tumor microenvironment remains incompletely understood. METHODS To uncover state-specific vulnerabilities for reducing GBM cell state diversity, we conducted a small-molecule screen in gliomasphere models using over 1,600 cysteine-reactive covalent inhibitors, assessing their effects on cell states via RNA-seq. We then employed the Sonar metabolic reporter to visualize NAD+ and NADH levels in gliomaspheres. Through experiments with orthotopic xenografts in mice and co-culture with human cortical organoids, we characterized distinct redox profiles across GBM cell states by single-cell and bulk RNA-seq. Spatial single-cell transcriptomics and mass spectrometry imaging (MSI) metabolomics on adjacent sections of frozen patient samples were used to identify preferential metabolic programs in specific cell states. RESULTS Our small-molecule screen identified compounds that promote differentiation and are enriched for NRF2 activation, a master regulator of the antioxidant response. Spatial single-cell transcriptomics and mass spectrometry metabolomics experiments confirmed that MES-like cells exhibit a heightened glycolytic signature, whereas stem-like states preferentially depend on oxidative phosphorylation (OXPHOS) for their energetic needs. Finally, pharmacologic perturbations revealed that mitochondrial and glycolysis inhibition selectively deplete different cell state populations in gliomaspheres. CONCLUSION Our findings provide deeper insight into the metabolic underpinnings of GBM cell states and support the rationale for combinatorial therapies aimed at reducing intratumoral heterogeneity that exploit their divergent dependencies.