Identification of two heterogeneous subtypes of hepatocellular carcinoma with distinct pathway activities and clinical outcomes based on gene set variation analysis.

Background: High heterogeneity is an essential feature of malignant tumors. This study aims to reveal the drivers of hepatocellular carcinoma heterogeneity for prognostic stratification and to guide individualized treatment.

Methods: Omics data and clinical data for two HCC cohorts were derived from the Cancer Genome Atlas (TCGA) and the International Cancer Genome Atlas (ICGC), respectively. CNV data and methylation data were downloaded from the GSCA database. GSVA was used to estimate the transcriptional activity of KEGG pathways, and consensus clustering was used to categorize the HCC samples. The pRRophetic package was used to predict the sensitivity of samples to anticancer drugs. TIMER, MCPcounter, quanTIseq, and TIDE algorithms were used to assess the components of TME. LASSO and COX analyses were used to establish a prognostic gene signature. The biological role played by genes in HCC cells was confirmed by experiments.

Results: We classified HCC tissues into two categories based on the activity of prognostic pathways. Among them, the transcriptional profile of cluster A HCC is similar to that of normal tissue, dominated by cancer-suppressive metabolic pathways, and has a better prognosis. In contrast, cluster B HCC is dominated by high proliferative activity and has significant genetic heterogeneity. Meanwhile, cluster B HCC is often poorly differentiated, has a high rate of serum AFP positivity, is prone to microvascular invasion, and has shorter overall survival. In addition, we found that mutations, copy number variations, and aberrant methylation were also crucial drivers of the differences in heterogeneity between the two HCC subtypes. Meanwhile, the TME of the two HCC subtypes is also significantly different, which offers the possibility of precision immunotherapy for HCC patients. Finally, based on the prognostic value of molecular subtypes, we developed a gene signature that could accurately predict patients' OS. The riskscore quantified by the signature could evaluate the heterogeneity of HCC and guide clinical treatment. Finally, we confirmed through experiments that RFPL4B could promote the progression of Huh7 cells.

Conclusion: The molecular subtypes we identified effectively exposed the heterogeneity of HCC, which is important for discovering new effective therapeutic targets.