Model-Dependent Autophagy–Apoptosis Gene Expression in Glioblastoma: Integration of U87-MG, Stem-like, and Spheroid Models with TCGA-GBM

Background/Objectives: Glioblastoma (GBM) is a highly aggressive brain tumor characterized by marked heterogeneity, therapy resistance, and poor prognosis. Dysregulation of autophagy and apoptosis may contribute to tumor adaptation, yet experimental GBM models differ in their translational relevance. This study compared model-dependent expression patterns of selected autophagy- and apoptosis-related genes and evaluated their patient-level relevance in TCGA-GBM. Methods: Relative expression of p62/SQSTM1, ATG7, BAX, and BCL2 was assessed descriptively across SVG p12, U87-MG, U87-MG spheroids, and U87-MG CD133+ using RT-qPCR-derived fold-change values. Experimental profiles were integrated with TCGA-GBM transcriptomic and clinical data using Cox proportional hazards regression and Spearman correlation analysis. Results: U87-MG showed upregulation of p62/SQSTM1, BAX, and BCL2 with modest ATG7 downregulation. U87-MG CD133+ showed reduced BAX, ATG7, and BCL2 expression despite retained p62/SQSTM1 upregulation, whereas spheroids displayed marked BAX upregulation together with reduced ATG7 and BCL2 expression. In the final TCGA-GBM cohort (n = 281), none of the four genes was significantly associated with overall survival in univariable, age- and sex-adjusted, or radiotherapy-stratified Cox models. Exploratory treatment-complete sensitivity analyses did not materially alter these findings. Gene–gene correlations were weak to moderate. Conclusions: GBM model selection substantially influences observed autophagy- and apoptosis-related transcriptional patterns. Limited concordance between experimental expression changes and patient-level survival associations indicates that these mRNA-level findings are hypothesis-generating rather than evidence of validated prognostic biomarkers and require a broader pathway, protein-level, and functional validation.