Identification and validation in a novel quantification system of the glutamine metabolism patterns for the prediction of prognosis and therapy response in hepatocellular carcinoma

BACKGROUND: Hepatocellular carcinoma (HCC) has one of the highest mortality rates worldwide. Abnormal glutamine metabolism (GM) has been reported to be involved in HCC progression. The current study sought to examine the predictive value of GM in HCC patient’s prognosis and therapy response. METHODS: The RNA-sequencing data and clinical information of HCC samples were obtained from The Cancer Genome Atlas (TCGA) database (N=377) and Gene Expression Omnibus (GEO) database (N=242). By analyzing a data set from TCGA, we showed that the GM landscape of HCC patients was developed based on the non-negative matrix factorization (NMF) algorithm. Univariate Cox regression and least absolute shrinkage and selection operator (LASSO)–penalized Cox regression analyses were used to construct a risk model. The accuracy of the model, which was based on the GM-related genes (GMRGs), was verified by Kaplan-Meier (K-M) and receiver operating characteristic (ROC) curves. We also verified the reliability of the model based on GEO data. Finally, the immune infiltration analysis, pathway enrichment analysis, and treatment response prediction results were compared to each other in the 2 risk groups. RESULTS: In our study, the HCC samples were divided into 2 GM-related patterns; that is, C1 and C2. The multi-analysis revealed that the GM-related patterns were associated with the pathologic stage, T stages, N stages, histologic grade, and the tumor immune microenvironment (TIME). Next, the prognostic model containing 5 GMRGs (i.e., aldehyde dehydrogenase 5 family member A1, ASNSD1, carbamoyl-phosphate synthetase 1, GMPS, and PPAT) was constructed to calculate the risk score. The high-risk group of HCC patients had significantly worse overall survival (OS) than the low-risk group in both datasets (P<0.001). Multivariate Cox regression uncover the riskScores may serve as an independent prognostic marker for HCC patients [TCGA: hazard ratio (HR) =2.909 (1.940−4.362), P<0.001; GEO: HR =2.911 (1.753−5.848), P=0.043]. Finally, we found that the prognostic model was significantly correlated with the pathologic stage and TIME of the HCC patients in both databases. Moreover, the prognostic model may guide the immunotherapy, chemotherapy, and targeted drugs choice. CONCLUSIONS: In summary, we developed a GM-related 5-gene risk-score model, which may be a useful tool for predicting prognosis and guiding the treatment of HCC patients.