Four circadian rhythm-related genes predict incidence and prognosis in hepatocellular carcinoma

Abstract

Circadian dysregulation can be involved in the development of malignant tumors, though its relationship with the progression of hepatocellular carcinoma is not yet fully understood. We identified genes related to circadian rhythms from the Cancer Genome Atlas (TCGA), measured gene expression, and conducted genomic difference analysis to construct a circadian rhythm-related signature. The resulting prognosis model proved to be an effective biomarker, as demonstrated by Kaplan-Meier survival analysis for both the training (n = 370, P = 2.687e-10) and external validation cohorts (n = 230, P = 1.45e-02). Further, we found that patients considered 'high risk', with an associated poor prognosis, displayed elevated levels of immune checkpoint genes and immune filtration. We also conducted functional enrichment, which indicated that the risk model showed a significant positive correlation with certain malignant phenotypes, including G2M checkpoint, MYC targets, and the MTORC1 signaling pathway. In summary, we identified a novel circadian rhythm-related signature allowing assessment of prognosis for hepatocellular carcinoma patients, and further can be used to predict immune infiltration sensitivity.

Keywords: Circadian clock; gene signature; immune; liver cancer; overall survival.

Figure 1

Genomic alterations to circadian-related genes in HCC tissues. (A) Oncoprint of 14 circadian-related genes in the TCGA LIH dataset generated by maftools. (B) Comparison of the expression of circadian-related genes among amplification, deletion, or normal in the LIHC dataset. (C, D) Heatmap of the circadian-related gene methylation types (C) and expression (D) in the LIHC dataset.

Figure 2

Survival analysis based on the risk score model. Survival analysis of the LIHC cohort in the TCGA. (A) Venn diagram of DEGs and prognostic genes that correlate with OS in tumor and tumor-adjacent normal tissue. (B) Lasso regression identified 4 genes correlated with prognosis. (C) Forest plots of the four genes that overlap between DEGs or prognostic genes that relate to OS based on univariate Cox regression analysis. (D) Kaplan Meier curves for the OS of patients in the high-risk and low-risk groups in the TCGA cohort (P-value <.0001). * p value < 0.05, ** p value < 0.01.

Figure 3

Validation of the risk score based on four signature genes from the ICGC dataset. (A) Distribution and expression of circadian-related genes based on the risk scores in the ICGC dataset. (B) Kaplan Meier curves for the OS of patients in the high-risk and low-risk groups in the ICGC dataset. (C, D) Multivariate Cox regression analyses of factors affecting OS in the TCGA LIHC cohort (C) and in the ICGC LIRI dataset (D). (E, F) Time-dependent ROC analyses indicated that risk score had greater predictive value than other clinical features for 1-year OS in TCGA LIHC and ICGC LIRI. ROC, receiver operating curve; *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Figure 4

Identification of differentially expressed genes (DEGs) between the high-risk and low-risk score groups in TCGA-LIHC dataset with the cut-off criteria of |logFC|>1 and adj.P <0.05. (A) Heatmap of upregulated and downregulated DEGs in the TCGA-LIHC dataset. (B) Functional enrichment analysis of differentially expressed genes (DEGs) between the high-risk and low-risk groups. (C, D) GSEA analysis in the TCGA dataset according to risk score, (C) hallmark pathway, and (D) KEGG pathways.

Figure 5

Identification of modules related to circadian genes in the TCGA-LIHC and ICGC datasets. (A, B) Module-trait relationships, where rows represent color modules and columns represent clinical traits (normal and tumor). Cells represent the corresponding correlation and P-value. (C, D) Sankey plot indicates the association between the circadian gene, hallmark pathway, and WGCNA modules.

Figure 6

Comparison of drug response and signature score between high-risk and low-risk groups. (A, B) Drug response of the TCGA (A) and ICGC (B) datasets. (C, D) Signature scores of TCGA (C) and ICGC (D).