Establishment and validation of an immune-based prognostic score model in glioblastoma

Background: Immune escape is one of the landmark features of glioblastoma (GBM). Immunotherapy is undoubtedly a revolution in the field of tumor treatment, especially the application of immune checkpoint inhibitors and CAR-T cells, which have achieved amazing results in fighting against cancer. This study aimed to establish a TP53-related immune-based score model to improve the prognostic of GBM by investigating the gene mutations and the immune landscape of GBM. Methods: Data were obtained from The Cancer Genome Atlas (TCGA) and the Chinese Glioma Genome Atlas (CGGA) databases. Differentially expressed genes (DEGs) analysis between the TP53 mutated (TP53(MUT)) and wild-type (TP53(WT)) GBM patients was conducted. The CIBERSORT algorithm was applied to evaluate the proportion of immune cell types and RNA sequencing (RNA-seq) data from the TCGA and CGGA were used as discovery and validation cohorts, respectively, to build and validate an immune-related prognostic model (IPM). Genes in the IPM model were first screened by univariate Cox analysis, then filtered by the least absolute shrinkage and selection operator (LASSO) Cox regression method to eliminate collinearity among DEGs. A nomogram was finally established and evaluated by combining both the IPM and other clinical factors. Results: PTEN was the top most mutated gene in GBM patients (118/393), followed by TP53 (116/393). 332 immune-related genes were identified and the immune response in the TP53(WT) group was remarkably greater than in the TP53(MUT) group. The final IPM model composed three immune-related genes: IPM risk score = (0.392 x S100A8 expression) + (0.174 x CXCL1 expression) + (0.368 x IGLL5 expression), significantly correlated with the overall survival (OS) of GBM in the stratified TP53 status subgroups and was an independent prognostic variate for GBM. By integrating the IPM and clinical characteristics, a nomogram was generated to facilitate clinical utilization, with the results suggesting that it has better predictive performance for GBM prognosis than the IPM. Conclusions: The IPM model can identify patients at high-risk and can be combined with other clinical factors to estimate the OS of GBM patients, demonstrating that it is a promising biomarker to optimize the prognosis of GBM.