A metabolism-associated gene signature for prognosis prediction of hepatocellular carcinoma

Abstract

Hepatocellular carcinoma (HCC), the most frequently occurring type of cancer, is strongly associated with metabolic disorders. In this study, we aimed to characterize the metabolic features of HCC and normal tissue adjacent to the tumor (NAT). By using samples from The Cancer Genome Atlas (TCGA) liver cancer cohort and comparing 85 well-defined metabolic pathways obtained from the Kyoto Encyclopedia of Genes and Genomes (KEGG), 70 and 7 pathways were found to be significantly downregulated and upregulated, respectively, in HCC, revealing that tumor tissue lacks the ability to maintain normal metabolic levels. Through unsupervised hierarchical clustering of metabolic pathways, we found that metabolic heterogeneity correlated with prognosis in HCC samples. Thus, using the least absolute shrinkage and selection operator (LASSO) and filtering independent prognostic genes by the Cox proportional hazards model, a six-gene-based metabolic score model was constructed to enable HCC classification. This model showed that high expression of LDHA and CHAC2 was associated with an unfavorable prognosis but that high ADPGK, GOT2, MTHFS, and FTCD expression was associated with a favorable prognosis. Patients with higher metabolic scores had poor prognoses (p value = 2.19e-11, hazard ratio = 3.767, 95% CI = 2.555–5.555). By associating the score level with clinical features and genomic alterations, it was found that NAT had the lowest metabolic score and HCC with tumor stage III/IV the highest. qRT‒PCR results for HCC patients also revealed that tumor samples had higher score levels than NAT. Regarding genetic alterations, patients with higher metabolic scores had more TP53 gene mutations than those with lower metabolic scores (p value = 8.383e-05). Validation of this metabolic score model was performed using another two independent HCC cohorts from the Gene Expression Omnibus (GEO) repository and other TCGA datasets and achieved good performance, suggesting that this model may be used as a reliable tool for predicting the prognosis of HCC patients.