Novel biomarker genes for the prediction of post

We compared the microenvironments of tumor and adjacent normal samples from HCC patients with either NAFLD or HBV who had tumor resection and different overall survival based on the transcriptome data and follow-up information. We also used text-mining software and biomedical literature data to screen two potential key genes FABP4 and VWF from DEGs. Finally, we built and validated a multi-feature joint model to predict the prognosis of NAFLD-HCC after hepatectomy.

FABP4 is a lipid chaperone protein that binds with high affinity to hydrophobic ligands including saturated and unsaturated long-chain fatty acids. It is responsible for promoting lipid storage, distribution, transportation, decomposition and metabolism [22]. FABP4 is mainly secreted by adipocytes and macrophages. It was elucidated that FABP4, which is not normally expressed in the liver, could be synthesized and secreted by hepatocytes, peritumoral endothelial cells, intra-tumoral hepatic stellate cells, and HCC cells [23]. The up-regulated level of FABP4 in the systemic circulation of patients with NAFLD is associated with liver inflammation and fibrosis [24]. FABP4 could provide fatty acids to malignant cells to maintain cell proliferation and affect cancer progression [25]. A recent study suggested that targeted inhibition of LPL/FABP4/CPT1 fatty acid metabolic axis can effectively prevent the progression of nonalcoholic steatohepatitis to liver cancer [26]. LPL was also screened out in our results, but due to its correlation with FABP4, we did not include it in our prediction model (Fig. 2C). In addition, FABP4 overexpression in intra-tumoral hepatic stellate cells may contribute to hepatocarcinogenesis in patients with metabolic risk factors by modulation of inflammatory pathways [27]. In our study, we also observed an overexpression of FABP4 in NAFLD-HCC tumor tissue. FABP4 may promote liver cancer through endothelial cells in NAFLD-HCC [28]. In another previous study, it was demonstrated that FABP4 in peritumoral endothelial cells from human HCC samples with metabolic syndrome was overexpressed compared with those with other risk factors for chronic liver diseases [28]. This study confirmed the oncogenic role of FABP4 in liver carcinogenesis, highlighting the key role of tumor microenvironment via cross-talks between endothelial and tumors cells mainly through microvesicles release from endothelial cells. This also agreed with our finding that FABP4 is an unfavorable biomarker in the prognosis prediction model, and it is associated with the endothelial cell infiltration score (Fig. 2). However, the effect of FABP4 as the biomarker for HCC is context-dependent. In our study, the level of FABP4 was not different between HBV-HCC tumor and adjacent tissues (P > 0.05). In a previous study from our hospital where most HCC patients carried HBV, FABP4 was low-expressed in tumor tissues compared with the adjacent tissue, and its expression as a favorable biomarker was significantly associated with the tumor size, portal vein tumor thrombus, recurrence-free survival and overall survival [29].

VWF, a multimeric glycoprotein synthesized primarily by endothelial cells, is well known to be involved in angiogenesis and hemostatic mechanisms [30, 31]. The binding of the VWF to integrin avβ3 could repress the VEGFR-2 activity and the downstream pro-angiogenic signaling pathways. The development of HCC is dependent on the formation of new blood vessels, and the surrounding blood vessels play as one type of important tumor microenvironment in tumor initiation and progression in HCC. Growing research evidence suggested that VWF may function as the bivalent mediator of HCC [30]. For example, higher preoperative VWF appeared to be negatively associated with post-resection liver dysfunction in patients with HCC undergoing partial hepatectomy, whereas, high post-resection plasma VWF concentrations indicated the early HCC recurrence [32]. Another study indicated that VWF levels were higher in patients with severe liver fibrosis stage and/or HCC development than in those without [31]. In our study, we found VWF served as a favorable marker for prognosis of NAFLD-HCC after hepatectomy.

Takaya et al. proposed that one possible explanation for these different results was the various progression and underlying causes of liver diseases [31]. For example, cirrhotic patients frequently have hypercoagulability which is associated with elevated VWF and results in markedly increased risk for thromboembolism. Most of the patients in our study had not developed cirrhosis (Table S1), whereas in the study by Takaya et al., all the patients in the case group had developed cirrhosis [31]. This implies the importance to consider the progression and the underlying causes of liver diseases of the patients when using VWF as a biomarker.

A multi-feature prognostic model for NAFLD-related HCC OS was constructed based on transcriptome analysis, follow-up information and text-mining. The risk score which calculated by combining TNM stage, gender and the expression level of FABP4 and VWF, could be used to predict the OS. Previously, one single VWF gene was proposed to predict the stage of HCC [31] or prognosis of HCC after hepatectomy [32]. According to our results, the combination of two key genes, gender and TNM performed better than the other combinations, indicating the advantage of integrating omics data analysis and text-mining.

However, it should be admitted that our research inevitably has some limitations. Firstly, functional experiments are needed to further explore their potential roles and underlying molecular mechanisms. Secondly, due to strict sampling criteria, only 47 patients with NAFLD-HCC undergoing hepatectomy were included. Multi-center clinical studies with larger sample size are expected to verify the results. Thirdly, our model was based on reported NAFLD-HCC related genes, the other DEGs may also play important roles in NAFLD-HCC development and require further exploration.