Protein-based classification reveals an immune-hot subtype in IDH mutant astrocytoma with worse prognosis

Intertumoral heterogeneity complicates treatment of IDH-mutant astrocytoma. We analyze spatiotemporal multi-omics data and discover four clusters: adipogenesis/fatty-acid-metabolism (AFM), proliferative/progenitor (PPR), immune/mesenchymal-enriched (IME), and neuronal (NEU). PPR and IME are associated with poorer prognosis, a result validated in The Cancer Genome Atlas (TCGA) and a Chinese cohort of 273 IDH-mutant astrocytomas. Longitudinal analysis of 189 initial-recurrent pairs shows an evolutionary shift toward PPR/IME subtypes. Mechanistically, PPR is enriched for CDKN2A/B deletions, whereas IME features increased gemistocytic differentiation (GD) and infiltration by exhausted T cells and plasma cells. Spatial multi-omics link GD morphology to mesenchymal-like (MES-like) tumor cell aggregates and lymphocyte-rich niches. MES-like tumor cells in IME overexpress interferon-stimulated genes such as GBP1, which we show promotes proliferation and migration. Finally, we develop an AI-powered classifier for patient stratification. Our work delineates protein-based clustering of IDH-mutant astrocytoma and reveals an immune-hot subgroup that may inform therapeutic development.