Development and validation of metabolic models for predicting survival and immune status of hepatocellular carcinoma patients.

BACKGROUND

Metabolic reprogramming is associated with the carcinogenesis of hepatocellular carcinoma (HCC). The effects of metabolism-related genes on predicting survival and immune status in HCC remain unclear.

OBJECTIVES

To develop and validate metabolic models for predicting the survival and immune status of HCC patients.

MATERIAL AND METHODS

The metabolic core genes for overall survival (OS) and disease-free survival (DFS) were retrieved. Then, glycolysis and fatty acid metabolism prognostic models were constructed and validated using The Cancer Genome Atlas (TCGA) and the International Cancer Genome Consortium (ICGC) data. Decision trees based on machine learning were developed for classifying the prognostic risks of HCC patients. The associations between the metabolic signatures, immunotherapy and immune cell infiltration were investigated. Experimental validations were performed using reverse transcription-quantitative polymerase chain reaction (RT-qPCR) and immunohistochemistry (IHC).

RESULTS

We identified 30 prognostic core genes for glycolysis metabolism and 12 prognostic core genes for fatty acid metabolism. Subsequently, 2 glycolysis models and 2 fatty acid metabolism models were developed to predict the OS and DFS of HCC patients, respectively. Two decision trees were constructed to classify the low-, intermediateand high-risk groups of HCC patients for OS and DFS. Moreover, the patients in the high-risk groups of glycolysis and fatty acid metabolic models tended to have higher expression of programmed cell death ligand-1 (PD-L1 or CD274), programmed cell death 1 (PDCD1), cytotoxic T-lymphocyte-associated protein-4 (CTLA-4), and lymphocyte activating 3 (LAG3). Most of the metabolic core genes were significantly associated with immune cell infiltration. In addition, ATP-binding cassette subfamily B member 6 (ABCB6), peptidylprolyl isomerase A (PPIA), uroporphyrinogen decarboxylase (UROD), and non-SMC condensin II complex subunit H2 (NCAPH2) were positively correlated with both tumor mutational burden (TMB) and microsatellite instability (MSI) scores. The expression of ABCB6, PPIA, UROD, and NCAPH2 was validated using RT-qPCR and IHC.

CONCLUSIONS

We established novel prognostic models based on metabolism-related genes to better predict the outcome and immune status of HCC patients.