Identification of the Tumor Immune Microenvironment and Therapeutic Biomarkers by a Novel Molecular Subtype Based on Aging-Related Genes in Hepatocellular Carcinoma

Abstract

Background: Hepatocellular carcinoma (HCC) is one of the most prevalent malignant tumors with poor prognosis. Increasing evidence has revealed that immune cells and checkpoints in the tumor microenvironment (TME) and aging are associated with the prognosis of HCC. However, the association between aging and the tumor immune microenvironment (TIME) in HCC is still unclear. Methods: RNA expression profiles and clinical data concerning HCC were downloaded from The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) databases. Based on differentially expressed aging-related genes (DEAGs), unsupervised clustering was used to identify a novel molecular subtype in HCC. The features of immune cell infiltration and checkpoints were further explored through CIBERSORTx. Enrichment analysis and both univariate and multivariate Cox analyses were conducted to construct a 3-gene model for predicting prognosis and chemosensitivity. Finally, the mRNA and protein expression levels of the 3 genes were verified in HCC and other cancers through database searches and experiments. Results: Eleven differentially expressed AGs (GHR, APOC3, FOXM1, PON1, TOP2A, FEN1, HELLS, BUB1B, PPARGC1A, PRKDC, and H2AFX) correlated with the prognosis of HCC were used to divide HCC into two subtypes in which the prognosis was different. In cluster 2, which had a poorer prognosis, the infiltration of naive B cells and monocytes was lower in the TCGA and GEO cohorts, while the infiltration of M0 macrophages was higher. In addition, the TCGA cohort indicated that the microenvironment of cluster 2 had more immunosuppression through immune checkpoints. Enrichment analysis suggested that the MYC and E2F targets were positively associated with cluster 2 in the TCGA and GEO cohorts. Additionally, 3 genes (HMGCS2, SLC22A1, and G6PD) were screened to construct the prognostic model through univariate/multivariate Cox analysis. Then, the model was validated through the TCGA validation set and GEO dataset (GSE54236). Cox analysis indicated that the risk score was an independent prognostic factor and that patients in the high-risk group were sensitive to multiple targeted drugs (sorafenib, gemcitabine, rapamycin, etc.). Finally, significantly differential expression of the 3 genes was detected across cancers. Conclusion: We systematically described the immune differences in the TME between the molecular subtypes based on AGs and constructed a novel three-gene signature to predict prognosis and chemosensitivity in patients with HCC.