Abstract
BACKGROUND: Glioblastoma (GBM) exhibits profound cellular heterogeneity and a highly immunosuppressive microenvironment in which tumor-associated macrophages represent a dominant immune component. However, how macrophage state-specific transcriptional programs and co-expression networks integrate to shape patient outcome and immunotherapy-related phenotypes remains insufficiently defined. METHOD: Single-cell RNA-seq data were integrated to resolve GBM cellular architecture and macrophage subpopulations. Macrophage regulatory and co-expression programs were inferred using R-SCENIC and hdWGCNA. Prognostically relevant genes were selected by integrating macrophage subcluster markers with key co-expression modules and survival screening in TCGA. A multi-algorithm machine learning framework was used to construct the Macrophage Co-expression-derived Risk Score (MCRS), which was validated in multiple independent cohorts. Immune landscapes, immunotherapy-related metrics, and pan-cancer analyses of SPP1 were systematically evaluated. In addition, an in silico virtual knockout analysis of SPP1 was performed to assess its potential downstream transcriptional effects in the GBM microenvironment. Finally, SPP1 was functionally validated in GBM cell lines. RESULTS: Macrophages segregated into distinct functional subpopulations with differential regulatory programs and prognostic relevance. hdWGCNA identified key macrophage modules linked to these states, enabling construction of the MCRS, which robustly stratified patient survival across TCGA, CGGA, and GEO cohorts. High MCRS was associated with coordinated immune-metabolic pathway activation, altered tumor purity, reduced immunogenicity, and increased immune escape potential. Pan-cancer analyses revealed widespread overexpression and adverse prognostic associations of SPP1. Virtual knockout of SPP1 induced distinct transcriptional changes enriched in immune-related biological processes and pathways, further supporting its involvement in macrophage-associated immune regulation. Experimental assays further showed that SPP1 promoted glioma cell proliferation, invasion, and clonogenicity. CONCLUSION: By integrating single-cell macrophage heterogeneity with co-expression network modeling, this study establishes MCRS as a robust prognostic and immunological stratifier in GBM and identifies SPP1 as a key macrophage-associated effector. Combined with in silico perturbation and experimental validation, these findings provide a biologically grounded framework for risk assessment and immunomodulatory targeting in GBM.