Abstract
Cancer-associated fibroblasts (CAFs) critically regulate tumor progression, angiogenesis, metastasis, and therapeutic resistance. This study investigated the characteristics of CAFs in glioblastoma (GBM) and developed a CAF-based risk signature to predict patient prognosis. The single-cell RNA sequencing (scRNA-seq) data were sourced from the Gene Expression Omnibus (GEO) database, whereas the bulk RNA-seq datasets were retrieved from The Cancer Genome Atlas (TCGA) and Chinese Glioma Genome Atlas (CGGA), respectively. The Seurat R package processed scRNA-seq data to identify CAF clusters using established markers. Prognostic genes were screened through univariate Cox regression, with Lasso regression constructing the final risk model. A nomogram incorporating clinical parameters was subsequently developed. Immunohistochemical validation was performed using the Human Protein Atlas (HPA) for the signature genes. The qRT-PCR validation was conducted in U251 and HA cells. ScRNA-seq analysis revealed five CAF clusters in GBM, including three prognostically relevant subtypes. Three key genes were refined to construct a risk signature functionally enriched in the the IL6_JAK_STAT3 signaling, P53 pathway, and inflammatory response. The signature correlated strongly with stromal and immune scores. Multivariate analysis confirmed risk signature independent prognostic value (P < 0.0001), followed by age (P = 0.005). The CAF-derived nomogram demonstrated superior predictive accuracy for 1-/2-year survival compared to clinical factors alone. The signature genes were validated in the HPA database and experimental validation. This study proposes CAF-derived molecular signatures as potential predictors of glioblastoma prognosis worthy of clinical validation. Systematic characterization of CAF heterogeneity may offer insights for interpreting GBM immunotherapy responses, providing a foundation for future exploration of stroma-targeted therapeutic strategies.