Abstract
BACKGROUND: Glioblastoma (GBM) remains the most prevalent and aggressive primary central nervous system (CNS) malignancy; however, the clinical efficacy of the preferred chemotherapeutic agent, Temozolomide (TMZ), is severely compromised by innate and acquired resistance. Sphingolipid metabolism acts as a pivotal regulator of GBM cell fate, and the imbalance of the "sphingolipid rheostat" is intimately linked to TMZ resistance. This provides potential targets for developing novel prognostic models to inform stratified treatment risk strategies, while offering a promising entry point for TMZ chemosensitization and stratified drug combinations. METHODS: We integrated single-cell and bulk transcriptomics from TCGA and GEO. Through a multi-dimensional framework combining Weighted Gene Co-expression Network Analysis (WGCNA), differential expression profiling, Cox regression, and machine learning, we identified candidate genes associated with the molecular landscape coupled with sphingolipid dysregulation and TMZ sensitivity in GBM to construct a reliable prognostic model. We verified mRNA expression of model genes and protein expression of TIMP1 in clinical specimens via RT-qPCR and tissue microarrays (TMA), respectively. Furthermore, we functionally characterized the core target, TIMP1, via lentiviral knockdown in U87 cells, employing Transwell, CCK-8, and IC50 assays to evaluate its impact on malignancy and, crucially, its capacity to modulate TMZ chemosensitization. RESULTS: Single-cell analysis stratified GBM samples into distinct metabolic subclasses, revealing significant metabolic heterogeneity. Integrating TCGA and GEO profiles with WGCNA-based multi-dimensional intersection, we identified 95 candidate genes, refined via Cox regression and machine learning into a potent six-gene model (MXRA8, TIMP1, TREM1, S100A4, RMI2, IRF7) reflecting critical axes of extracellular matrix (ECM) remodeling, inflammation, and DNA repair. We delineated the model's role in shaping an immune-excluded tumor microenvironment (TME) characterized by stromal remodeling, T-cell exhaustion and functional impairment of natural killer (NK) cell subsets, while uncovering specific therapeutic vulnerabilities for distinct risk subgroups. Experimental validation confirmed widespread upregulation of core targets in clinical specimens. Functionally, TIMP1 knockdown significantly suppressed proliferation and invasion. Most importantly, silencing TIMP1 effectively restored sensitivity to TMZ (chemosensitization). CONCLUSIONS: This study establishes and validates a robust GBM prognostic model integrating the sphingolipid-associated molecular landscape with chemotherapy resistance. It provides a comprehensive perspective on the interplay among sphingolipid dysregulation, immune evasion, TMZ resistance, and the critical functional role of TIMP1. Beyond enabling precise patient stratification, this model highlights specific therapeutic vulnerabilities, offering a translational framework for developing combinatorial strategies to target the sphingolipid regulatory network and overcome GBM chemoresistance.