Single-cell-based dissection of stemness programs identifies an RPL39L-centered prognostic model and immunosuppressive features in glioblastoma

Glioblastoma (GBM) exhibits profound intratumoral heterogeneity and cellular plasticity, with stemness-related programs driving therapeutic resistance and poor prognosis. However, stemness-associated molecular signatures derived from bulk transcriptomes often suffer from microenvironmental confounding and limited reproducibility. We established a multi‑layer integrative framework using single‑cell RNA‑seq data (GSE182109). Malignant cells were identified via CNV inference, and stemness gradients were characterized by CytoTRACE. Univariate Cox regression was first used to identify prognostically relevant genes, followed by unified five-fold cross-validation of random survival forest, LASSO-Cox, Elastic Net-Cox, and survival SVM-RFE. The best-performing random survival forest was selected for subsequent interpretation. Spatial transcriptomics, The Human Protein Atlas (HPA) immunohistochemistry, immune infiltration inference, and drug sensitivity databases (PRISM, CTRP, GDSC) were used for multi‑dimensional validation. A high‑stemness subpopulation (cluster 3) was identified. Among 55 candidate genes, RPL39L, SEC11C, COX6C, BCAP31, RPL10, and NUDT5 were prioritized. The risk model significantly stratified patients into high‑ and low‑risk groups ( P < 0.0001). RPL39L showed the highest importance score and was significantly upregulated in tumor tissues, especially in malignant regions. High RPL39L expression was consistently associated with worse overall and progression‑free survival across multiple independent cohorts (GSE16011, GSE61335, GSE74187). RPL39L expression positively correlated with ESTIMATE scores and regulatory T cell (Treg) infiltration, suggesting a role in shaping an immunosuppressive microenvironment. Additionally, RPL39L expression was associated with both sensitivity- and resistance-related drug-response patterns, while molecular docking prioritized lapatinib as a candidate for further experimental validation. This single‑cell dissection identifies RPL39L as a core stemness‑associated gene in GBM. RPL39L is closely tied to adverse prognosis, immune suppression, and drug response, highlighting its potential as a prognostic biomarker and a candidate for therapeutic targeting.