Abstract Background H3-altered gliomas, now recognized by WHO CNS5 as grade 4 neoplasms driven by histone H3 mutations (notably H3K27M and H3G34R/V), exhibit marked molecular heterogeneity, dismal survival, and resistance to standard therapies. Despite their classification as WHO grade 4 neoplasms, current stratification fails to capture their clinical and epigenetic diversity. Methods We performed genome-wide DNA methylation profiling on 49 representative cases from a clinical cohort of 375 H3-altered gliomas, and validated subtype classifications using TCGA data. Single-cell RNA sequencing, ChIP-seq, and spatial transcriptomic analyses were employed to elucidate tumor-intrinsic programs and microenvironmental niches across subtypes. Results Unsupervised clustering identified four robust DNA methylation-defined subtypes: DMGK27M, GBMRTK, DHGG34, and IDH/H3comut, a novel subtype harboring co-occurring IDH and H3 mutations. The IDH/H3comut subtype exhibited global DNA hypermethylation, enrichment in neurodevelopmental gene programs, and significantly improved survival. In contrast, GBMRTK tumors showed hypomethylation of cell-cycle enhancers and aggressive phenotypes, while DHGG34 and DMGK27M displayed subtype-specific epigenetic features. Single-cell transcriptomics revealed distinct lineage compositions and microenvironments, with IDH/H3comut tumors enriched in NPC-like cells, and GBMRTK in astrocyte-like and vascular mimicry programs. Clonal and spatial analyses uncovered a compartmentalized coexistence of IDH- and H3-mutant cells, suggesting cooperative rather than exclusive evolution. Conclusions Our study redefines H3-altered gliomas through methylation-based taxonomy, identifies a novel IDH/H3comut subtype with favorable prognosis, and reveals subtype-specific therapeutic vulnerabilities, including potential responsiveness to hypomethylating agents or CDK4/6 inhibitors.
Fu et al. (2026) studied this question.