Abstract B055: Single-cell meta-analysis reveals a neural-progenitor to mesenchymal transition driving immune evasion and resistance in glioblastoma

Abstract Background: Glioblastoma (GBM) remains the most aggressive primary brain tumor, with median survival under 18 months despite multimodal therapy. Intratumoral heterogeneity and adaptive transcriptional states contribute to relapse and immunotherapy resistance, yet a unified single-cell map linking developmental state, immune microenvironment, and clinical outcome is lacking. Methods: We performed a systematic single-cell RNA-seq meta-analysis across 42 public GBM datasets (≈220 000 cells; 1 123 patients). Data were quality-filtered, normalized, and batch-integrated using Seurat v5 + Harmony. Cluster annotation was guided by canonical lineage markers. RNA-velocity and pseudotime analyses defined transcriptional trajectories. Ligand-receptor interactions were inferred with CellChat, and spatial localization validated with the IVY Glioblastoma Atlas. Survival associations were tested in TCGA-GBM using prespecified mesenchymal-signature thresholds. False-discovery correction applied. Results: A recurrent neural-progenitor to mesenchymal (NP to MES) transition was identified in 27 % of tumors and enriched in recurrent disease. Transitioning cells upregulated SOX9, RUNX2, and STAT3, downregulated MHC-I/II, and secreted LGALS9, SPP1, and IL1B, promoting microglial M2 polarization. Spatial transcriptomics confirmed co-localization of MES-state clusters with exhausted T-cell niches (r = 0.71, p < 0.001). Tumors with >20 % MES-signature fraction exhibited shorter overall survival (median 8.6 vs 16.4 months; HR 1.92, p < 0.001) and lower PD-1 response (ORR 12 % vs 34 %, p = 0.02). Cross-species comparison revealed a conserved FGF2–STAT3 autocrine loop maintaining MES persistence. In silico inhibition of this axis reduced predicted MES expansion by ≈47 % and restored antigen-presentation pathways. Conclusions: This integrative analysis defines a conserved NP to MES transition as a driver of immune evasion and therapy resistance in GBM. Targeting the FGF2–STAT3 axis could re-sensitize MES-high tumors to immunotherapy and guide precision, state-adapted combination trials in glioblastoma. Disclosures: Nothing to declare. Citation Format: Randa Elzein, Safa Elzein. Single-cell meta-analysis reveals a neural-progenitor to mesenchymal transition driving immune evasion and resistance in glioblastoma [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(6_Suppl):Abstract nr B055.