ID #606 Single-cell and spatial mapping reveal driver-dependent tumor states and immune niches in glioneuronal tumors
Dilan Savran, Peter Zheng, Jared Collins, Natalie Schoebe, Christian Thomas, David Capper, Stefan Pfister, Romain Appay, Luca Bertero, Czinzia Lavarino, Andreas von Deimling, Wolfgang Wick, Olaf Witt, David Jones, Sandro Krieg, Pratiti Bandopadhayay, Rameen Beroukhim, Felix Sahm, Philipp SieversAbstract
Glioneuronal tumors (GNTs) are a heterogeneous group of brain tumors that are often clinically indolent but can show variable outcomes, including occasional progression. Although recurrent RAS/MAPK alterations motivate targeted therapies, resistance and relapse suggest that driver-linked tumor programs and the cellular context contribute to heterogeneity. Since many GNT types are rare, previous studies have examined subsets, which limit class-wide comparisons.
We comprehensively profiled 67 FFPE GNTs across thirteen entities spanning glial- to neuronal-enriched phenotypes using single-nucleus RNA sequencing. Across the cohort, we identified recurrent astrocyte-like (AC-like) and oligodendrocyte progenitor–like (OPC-like) tumor programs and detected two additional programs less well characterized across GNTs: extracellular matrix–like (ECM-like) and neuronal-like (NEU-like). While most tumors contained mixtures of these states, their relative composition varied by entity and tracked with driver: FGFR1- and NTRK-driven tumors were biased toward NEU-like cells, whereas BRAF-driven tumors favored AC/OPC/ECM-like programs.
Beyond state composition, clinically aggressive entities showed the strongest MAPK and VEGF signaling signatures despite differing dominant states, and proliferative programs, including G2M checkpoint, increased with MAPK activity. Immune programs also diverged with clinical behavior: less aggressive tumors showed tumor-intrinsic inflammatory programs alongside inflammatory myeloid activation, whereas aggressive entities more often exhibited reduced myeloid content with residual suppressive programs, although some retained substantial inflammatory microglia. PDGFRA-altered tumors showed especially strong inflammatory myeloid signatures.
Together, these data establish a spectrum-spanning single-cell map of GNTs that enables cross-entity comparison and highlights driver-linked tumor states, aggressiveness-associated signaling, and myeloid inflammatory or suppressive phenotypes. Ongoing spatial transcriptomics analyses will localize these programs within tissue architecture and support state–immune colocalization analyses across clinical behavior.