Integrated tissue transcriptome deconvolution of grade-associated tumor microenvironment patterns in CNS tumors.

Abstract

e14039 Background: Central nervous system (CNS) tumors exhibit marked tumor microenvironment (TME) heterogeneity, including immune exclusion and immunosuppression that may limit benefit from immunotherapy. We applied an integrated tissue transcriptome-based framework to define clinically relevant TME patterns associated with tumor grade. Methods: Tissue RNA sequencing from 39 CNS tumor tissue transcriptome samples comprising Low-Grade (LG, n=12) and High-Grade (HG, n=27) tumors was analyzed using complementary deconvolution approaches (quanTIseq, MCP-counter, EPIC, and xCell). Outputs were harmonized into a consensus TME profile summarizing overall immune infiltration, myeloid compartment activity, stromal/fibrotic programs, and immunosuppressive balance, and cross-method concordance was assessed to support robustness. Gene-set scoring (GSVA) and enrichment analysis (GSEA) quantified pathway-level programs including endothelial/angiogenesis, cancer-associated fibroblast (CAF), interferon-gamma (IFNγ) T-cell–inflamed, and antigen presentation (MHC class I/II). LG vs HG comparisons used nonparametric testing. Results: Consensus profiling demonstrated a sharp grade-associated separation in overall immune infiltration, which was higher in LG than HG (median 0.151 vs −0.108; p=0.010) with directionally concordant differences across deconvolution methods. HG tumors exhibited immunoregulatory skewing, including higher inferred regulatory T-cell abundance (Treg fraction median 0.035 vs 0.007; p=0.018) and higher Treg:CD8 ratio (median 0.147 vs 0.031; p=0.013). HG tumors also showed increased vascular/endothelial program activity (endothelial signature median −0.120 vs −0.260; p=0.032) with a trend toward increased CAF program activity (p=0.097). IFNγ T-cell–inflamed and MHC class I/II programs did not differ significantly by grade (all p≥0.37). Conclusions: Multi-algorithm consensus transcriptomic profiling identifies immune-enriched LG tumors and immune-excluded HG tumors characterized by Treg-associated immunosuppression and vascular/endothelial enrichment. These findings support TME-guided stratification of CNS tumors and motivate evaluation of combination strategies pairing immune modulation with anti-angiogenic and/or stromal-targeting approaches in HG disease.