The Immunological Contribution of a Novel Metabolism-Related Signature to the Prognosis and Anti-Tumor Immunity in Cervical Cancer

Abstract

Cervical cancer is the most frequently diagnosed malignancy in the female reproductive system. Conventional stratification of patients based on clinicopathological characters has gradually been outpaced by a molecular profiling strategy. Our study aimed to identify a reliable metabolism-related predictive signature for the prognosis and anti-tumor immunity in cervical cancer. In this study, we extracted five metabolism-related hub genes, including ALOX12B, CA9, FAR2, F5 and TDO2, for the establishment of the risk score model. The Kaplan-Meier curve suggested that patients with a high-risk score apparently had a worse prognosis in the cervical cancer training cohort (TCGA, n = 304, p < 0.0001), validation cohort (GSE44001, n = 300, p = 0.0059) and pan-cancer cohorts (including nine TCGA tumors). Using a gene set enrichment analysis (GSEA), we observed that the model was correlated with various immune-regulation-related pathways. Furthermore, pan-cancer cohorts and immunohistochemical analysis showed that the infiltration of tumor infiltrating lymphocytes (TILs) was lower in the high-score group. Additionally, the model could also predict the prognosis of patients with cervical cancer based on the expression of immune checkpoints (ICPs) in both the discovery and validation cohorts. Our study established and validated a metabolism-related prognostic model, which might improve the accuracy of predicting the clinical outcome of patients with cervical cancer and provide guidance for personalized treatment.

Keywords: cervical cancer; immune infiltration; metabolism-related genes; prognostic model; risk score signature.

Figure 1

Construction of a metabolism-related risk score signature in cervical cancer. (A) Schematic diagram of this study design. (B,C) Identification of differentially expressed metabolism-related genes (DEGs) between tumor and normal samples using GEO dataset (GSE63514) and annotation of GPL570 platform. The volcano plot (B) and Venn diagram (C) are shown. (D) The uni- and multi-variate Cox regression analysis results of the five metabolism-related hub genes in TCGA-CESC cohort. (E,F) The association between risk score and clinicopathological characters as well as enrichment scores of specific gene sets, including EMT, angiogenesis and hypoxia, of patients in TCGA cohort. The statistical difference of two groups was compared through the Wilcoxon test. * p < 0.05; **** p < 0.0001. (G) Representative immunostaining pictures of the five hub genes (ALOX12B, FAR2, F5, CA9 and TDO2) in tumor and normal tissues. Scale bar = 50 μm. The protein levels were plotted as a boxplot. * p < 0.05. MRGs, metabolism-related genes; ANT, adjacent non-tumor tissue; SCC, squamous cell carcinoma.

Figure 2

Verification of the metabolism-related risk score signature. (A) The left panel shows the risk curve and scatter plot of each sample in TCGA-CESC project reordered by the risk score, and the heatmap of expression profiles of the five hub genes. The middle panel displays the results of Kaplan-Meier analysis. The 3- and 5-year ROC curves curves of this optimal model (the right panel) revealed the AUC values. (B) The risk curve, heatmap of expression profiles (the left panel), results of Kaplan-Meier analysis (the middle panel) and ROC curves at 3 and 5 years (the right panel) of patients in GSE44001 cohort. (C) Forest plot of the univariate Cox regression analyses results of this risk score signature in all 33 types of cancer from TCGA database (the left panel). The Kaplan-Meier survival analyses and ROC curves of TCGA-HNSC and TCGA-LGG cohorts were plotted on the right panel. AUC, area under curve; ROC, receiver operating characteristic. * p < 0.05; ** p < 0.01; *** p < 0.001, **** p < 0.0001.

Figure 3

Functional enrichment analysis of the metabolism-related risk score signature. (A) Immune network of the 24 ImmuCellAI cell types in the TCGA cohort. The size of each cell was calculated by the formula log10 (p-values of univariate Cox regression analyses). The color of each cell was used to represent the different survival impact of these cell types. The thickness of the lines estimated by Spearman correlation analyses depicted the strength of correlation between diverse cell types. Red represents positive correlation whereas negative is in blue. (B) Correlation between these cell types and our established risk score in the TCGA cohort. Spearman analyses were applied to calculate the correlation coefficients and p-value < 0.05 was enrolled. (C) Gene ontology (GO) enrichment analysis of the differentially expressed genes between high- and low-risk groups in the TCGA-CESC cohort. Adjusted p-value < 0.05 was considered statistically significant. (D) GSVA analysis of hallmark pathways in the TCGA cohort was performed. Differential analysis of GSVA score between high- and low-risk groups was displayed. (E) Patients in the high-risk group were associated with lower infiltrating density of most cell types according to Charoentong’s research. * p < 0.05; ** p < 0.01; *** p < 0.001.

Figure 4

Correlation between the metabolism-related risk score and immune landscape. (A) Representative immunostaining pictures of the five hub genes and four cell types (CD4+, CD8+, CD57+ and CD68+ cells). The upper panel comprises images of five hub genes, images of four cell types were in the middle panel. Scale bar: 50 μm. The lower panel illustrates the infiltration scores of tumor-infiltrating CD4+, CD8+, CD57+ and CD68+ cells in the epithelial or stromal cell compartments. * p < 0.05. (B) Differences in the infiltration levels of 28 immune cells between high- and low-score groups in the pan-cancer validation cohorts. * p < 0.05; ** p < 0.01; *** p < 0.001.

Figure 5

Different expression patterns of immunostimulators (A), immunoinhibitors (B), receptors (C) and chemokines (D) between high- and low-score groups in TCGA training cohort and GSE44001 validation cohort. * p < 0.05; ** p < 0.01; *** p < 0.001.

Figure 6

The potential of the established signature in predicting the immunotherapeutic benefits of cervical cancer patients. (A) Kaplan-Meier curves for patients in the TCGA-CESC cohort stratified by the risk score and expressions of immune checkpoints, such as PD1, PD-L1, CTLA-4, CD28, CD38 and CD47. * p < 0.05; ** p < 0.01; *** p < 0.001; **** p < 0.0001. (B) Kaplan-Meier curves for patients in the GSE44001 validation cohort stratified by the risk score and expressions of immune checkpoints, such as PD1, PD-L1, CTLA-4, CD28, CD38 and CD47. * p < 0.05; ** p < 0.01.