Identification of EMT-related gene signatures in glioblastoma to aid immunotherapy and prognosis

Abstract

Abstract Objective: Glioblastoma (GBM) is the most aggressive malignancy of the central nervous system, with the lowest survival rate of malignant brain tumors at approximately 6%. Epithelial mesenchymal transition (EMT) is closely associated with tumor metastasis and drug resistance. Therefore, it is necessary to construct models consisting of EMT-related genes (ERGs) to forecast prognosis and benefit from immunotherapy in GBM patients. Methods: To identify differentially expressed genes (DEGs) for GBM by TCGA, CGGC, and GEO databases. To collect ERGs, databases called dbEMT2 and MSigDB were employed. Weighted gene co-expression network analysis (WGCNA) was used to find the core differentially expressed EMT-related genes (CDEERGs) at the junction of DEGs and ERGs. We developed the CDEERGs prognosis model (CPM) using a variety of bioinformatics analysis methods. The CPM was employed to determine risk scores for patients in TCGA-GBM dataset, and CGGC-325 and CGGC-693 datasets were utilized to externally validate the CPM's predictive ability. The differences in immunological traits and immunotherapeutic indicators between different groups were compared. Finally, it was evaluated if CPM may be helpful for determining whether immunotherapy would be appropriate for oncology patients. Results:IGFBP2, RGS4, AGTR1, CCL5, and LOXL1 were the five risk factors and ACTL6A and MTHFD2 were the two protective factors in the CPM. Patients with GBM were separated into high-risk and low-risk subgroups according on median risk scores. Low-risk subgroup in the TCGA-GBM dataset was significantly better than that in high-risk subgroup, and prognosis of patients in CGGC-325 and CGGC-693 datasets remained consistent with that in TCGA-GBM dataset. Risk scores and the expression of CD274 and PDCD1 were positively associated, and CD274 expression was higher in high-risk subgroup than in low-risk subgroup. The expression of numerous immunotherapy markers was different in high-risk and low-risk subgroups. Compared to the other prognostic models, the CPM has greater predictive power. We discovered that patients with low-risk scores may be better candidates for immunotherapy by calculating the risk scores of patients in the IMvigor210 dataset. Conclusion: The present study constructs CPMs that could be used to predict the prognosis of GBM patients as well as to screen for patients who can benefit from immunotherapy and to screen for CDEERGs that may provide new therapeutic targets for the treatment of GBM patients.