Comprehensive analysis reveals cholesterol metabolism-related signature for predicting prognosis and guiding individualized treatment of glioma

Abstract

Objective: Gliomas are the most common intracranial tumors with the highest degree of malignancy. Disturbed cholesterol metabolism is one of the key features of many malignant tumors, including gliomas. This study aimed to investigate the significance of cholesterol metabolism-related genes in prognostic prediction and in guiding individualized treatment of patients with gliomas.

Methods: Transcriptional data and clinicopathological data were obtained from The Cancer Genome Atlas (TCGA) and Chinese Glioma Genome Atlas (CGGA) databases. Intraoperative glioma samples retained in our unit and the corresponding clinicopathological information were also collected with the patients' knowledge. Firstly, cholesterol metabolism-related gene signatures (CMRGS) were identified and constructed based on difference analysis, least absolute shrinkage and selection operator (LASSO) regression analysis, and univariate/multivariate COX analysis. Then, the role of CMRGS in predicting the prognosis of gliomas and distinguishing immune landscapes was evaluated by using nomograms, survival analysis, enrichment analysis, and immune-infiltration analysis. Finally, the drug sensitivity of gliomas in different risk groups was evaluated using the oncoPredict algorithm, and potentially sensitive chemotherapeutic and molecular-targeted drugs were identified.

Results: The prognostic CMRGS contained seven genes: APOE, SCD, CXCL16, FABP5, S100A11, TNFRSF12A, and ELOVL2. Patients were divided into high- and low-risk groups based on the median cholesterol metabolic index (CMI). There were significant differences in clinicopathological characteristics and overall survival between groups. COX analysis suggested that CMRGS was an independent risk factor for glioma prognosis and had a better predictive performance than several classical indicators. In addition, GSEA, immune infiltration analysis showed that CMRGS could differentiate the immune landscapes of patients in groups. The reliability of CMRGS was validated in the CGGA cohort and our Gusu cohort. Finally, 14 drugs sensitive to high-risk patients and 16 drugs sensitive to low-risk patients were identified.

Conclusion: The CMRGS reliably predicts glioma prognosis in multiple cohorts and may be useful in guiding individualized treatment.

Keywords: Cholesterol metabolism; Drug sensitivity; Glioma; Immunoscape; Prognostic prediction; Signature.

Fig. 1

Flowchart of this study.

Fig. 2

Identification of prognostic Cholesterol-Metabolism-related genes from TCGA cohort. (A) Heatmap of 46 DEGs in normal and glioma groups. The redder, the higher the gene expression; the bluer, the lower the gene expression. (B) Forest plot of 20 Cholesterol-Metabolism-related genes performed by univariate Cox regression analysis. (C) LASSO coefficient profiles of Cholesterol-Metabolism-related genes. (D) Partial likelihood deviance of different numbers of variables revealed by the LASSO regression model (The log(λ) sequence plot of Cholesterol-Metabolism-related genes using LASSO regression). (E) Forest plot of 7 Cholesterol-Metabolism-related genes by multi-Cox regression. (F) Correlation between Cholesterol-Metabolism-related genes. DEG, differentially expressed gene; TCGA, The Cancer Genome Atlas; LASSO, least absolute shrinkage and selection operator.

Fig. 3

Validation of CMRGS in TCGA cohort. (A) Overall survival of high- and low-risk groups according to CMI in glioma patients from TCGA. (B) Heatmap of expression of the seven modeling genes in high- and low-risk groups (upper), distribution and median value of CMI (middle), and the distribution of survival status and CMI (below). The redder, the higher the gene expression; the bluer, the lower the gene expression. (C) PCA analysis of CMRGS. CMRGS, Cholesterol-Metabolism-related genes signature; CMI, Cholesterol-Metabolism index; DEG, differentially expressed gene; PCA, principal component analysis.

Fig. 4

Correlation between CMI and clinicopathological characteristics in TCGA dataset. (A) Heatmap of correlation between risk groups, age, gender, race, histology, grade, IDH mutation status, 1p19q mutation status, MGMT promoter status, and expression of Cholesterol-Metabolism-related genes. The redder, the higher the gene expression; the bluer, the lower the gene expression. (B) ROC curves of CMI and clinicopathological characteristics. (C and D) Univariate and multivariate Cox regression analysis of the combination of CMI and clinicopathological characteristics.CMI: Cholesterol-Metabolism index; IDH, isocitrate dehydrogenase; MGMT, O6-methylguanine-DNA methyltransferase; ROC, receiver operating characteristic.

Fig. 5

Prediction of the survival of glioma patients by nomogram, and GSEA analysis between different risk groups in TCGA cohort. (A) Nomogram used for predicting glioma patients was constructed. (B) Calibration plots for predicting 1-, 3- and 5-year overall survival. (C and D) GSEA functional enrichment analysis base on GO pathway.GSEA, gene set enrichment analysis; TCGA, The Cancer Genome Atlas; GO, Gene Ontology. ∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05.

Fig. 6

Comparison of immune landscapes between high- and low-risk groups based on CMI. (A) Comparison of immune checkpoint expression levels in different groups. (B and C) Comparison of immune cells and immune function between groups.(D) Scores for glioma microenvironment in different groups.∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05.

Fig. 7

Validation of CMRGS in External Cohorts. (A) Survival curve of high- and low-risk groups in TCGA cohorts. (B) Heatmap of the modeling genes expression in groups (upper), the distribution and median value of CMI (middle), and the distributions of survival status, survival time, and CMI (below). The redder, the higher the gene expression; the bluer, the lower the gene expression. (C) Time-dependent ROC analysis of CMRGS. (D) Immunohistochemical staining of S100A11 in different WHO-grade glioma tissues. (E) CMI of different WHO grade gliomas calculated based on seven genes' expression. CMRGS, Cholesterol-Metabolism-related genes signature; CGGA, Chinese Glioma Genome Atlas; CMI, cholesterol metabolism index; ROC, receiver operating characteristic.

Fig. 8

Identification of chemotherapeutic and molecularly targeted drugs based on CMRGS. (A) Predicted IC₅₀ of 30 FDA-approved drugs in GDSC2 in groups from TCGA and CGGA cohorts. The redder, the higher the gene expression; the bluer, the lower the gene expression. (B) Venn diagram for various combinations of training and testing sets. (C) Correlation between predicted IC₅₀ of four drugs (Irinotecan, Topotecan, Crizotinib, and Lapatinib) and seven genes for CMRGS among different combinations of training and test sets. CMRGS, Cholesterol-Metabolism-related genes signature; FDA, Food and Drug Administration; IC₅₀: half-maximal inhibitory concentration; GDSC2: Drug Sensitivity in Cancer v2; TCGA, The Cancer Genome Atlas; CGGA, Chinese Glioma Genome Atlas. ∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05.