Metabolic reprogramming of glioma-associated macrophages identifies detoxification and energetic macrophages as drivers of immunosuppression and therapeutic vulnerability.

Background: Glioma, a highly heterogeneous primary intracranial malignancy, features an immunosuppressive tumor microenvironment (TME) dominated by tumor-associated macrophages (TAMs). These glioma-associated macrophages (GAMs) critically drive disease progression, yet their metabolic reprogramming and clinical prognostic potential remain incompletely characterized. This study aimed to stratify GAMs by metabolic profiles and elucidate their clinical relevance, providing a framework for novel therapeutic strategies. Methods: We performed integrated multi-omics analysis of glioma single-cell RNA sequencing (scRNA-seq), bulk transcriptome sequencing, and clinical data. GAMs were stratified using metabolic pathway enrichment scores, and their abundance was correlated with patient prognosis. Supervised machine learning algorithms identified prognostic signature genes to construct a metabolic risk prediction model. Patients were stratified into high- and low-risk groups based on model-derived risk scores. Comprehensive profiling compared these groups across three dimensions: (i) dysregulated signaling pathways, (ii) tumor microenvironment characteristics, and (iii) genomic aberrations. Western blot (WB) analysis validated core gene expression in glioblastoma tumor tissues versus adjacent normal brain tissues. Results: This study reclassified GAMs into four metabolic subtypes-Glycolipid-Signaling (GSM), Detoxification and Energic (DEM), Polymetabolic (PmM), and Glycolipid Metabolism/Immunoregulatory (GMIM)-with DEMs exhibiting terminal differentiation, enrichment in detoxification/energy pathways, and significant correlation with advanced tumor grades and poor survival (p < 0.05). Machine learning leveraging DEM signature genes identified six core prognostic markers (CLIC1, FABP5, FCER1G, S100A8, S100A9, SPP1) and optimized a Stepwise Cox + Random Survival Forest model (C-index = 0.71). Applying this model, we identified high-risk gliomas exhibiting a paradoxical tumor microenvironment characterized by elevated immune cell infiltration and enhanced immunogenicity, yet impaired T-cell cytotoxicity. Concurrently, high-risk gliomas demonstrated hyperactivation of pro-tumorigenic pathways (e.g., mTOR, MAPK) and frequent EGFR amplification. Integration with EGFR amplification and IDH1 mutation status enhanced clinical prognostication. Western blot validation confirmed significant upregulation of all six core proteins in glioblastoma versus adjacent normal brain tissues. Conclusions: Metabolic subtyping identifies DEMs as critical drivers of glioma progression. The DEM-derived risk model, combined with EGFR/IDH status, provides a clinically actionable tool for prognosis and targeted therapy development.