In the present study, we focused on the HBV-related HCC because it is not only one of the common subtypes of liver cancer, but also a tricky public health problem. With the development of high-throughput sequencing and microarray technology, identifying DEGs and potent pathways in hepatocellular carcinoma is becoming a routine task, which provides an effective way to explore potential targets in diagnosing and prognosing HCC in the early stages[18, 19]. However, only a small part of studies focused on the establishment of signatures for HBC-related HCC[20] which makes our study meaningful. Here, we excavated DEGs between HCC and adjacent normal tissues based on two sets of gene expression profiling data. 457 genes were identified as DEGs which underwent GO analysis, KEGG analysis. PPI network were constructed and 10 hub genes were selected to undergo survival analysis. Innovatively, A diagnostic signature composed of 2 gene pairs and a prognostic signature composed of 4 pairs were built with high efficiency. Moreover, high risk of HBV-related HCC was primarily elucidated based on cellular composition analysis and enrichment.
Firstly, combined KEGG and GO analysis, it was clearly implicated that the identified DEGs were significantly involved in both cell cycle and metabolism process. Interestingly, up-regulated genes were mainly enriched in pathways ways associated with cell cycle and down-regulated genes mainly involved in metabolism pathway. For the cell cycle, there have been some studies unveiled the relationship of Hepatitis B virus, mitosis and hepatocarcinogenesis. HBV is usually involved in the development of liver cancer through HBx, immune imbalance and viral DNA integration into the host genome[21, 22]. Higher HBV load may contribute chronic liver inflammation by mediating immune response of host cytotoxic T lymphocyte (CTL) to promote the destruction and regeneration of HBV infected hepatocytes, and then increase the opportunity of mitotic replication errors[23, 24]. The accumulation of genetic and epigenetic alterations on cell growth advantage ultimately promotes the hepatocarcinogenesis[25]. HBx protein is an oncoprotein encoded by HBV, which may act as a transcription activator of oncogene or host gene to participate in cell growth regulation, DNA repair and epigenetic modification[26]. It may also affect cell cycle, cell transformation and other signal transduction pathways, and then contribute to HCC development[27, 28]. HBx can inhibit the expression of Wnt/catenin, or activate the expression of HURP through p38 / MAPK pathway and SATB1 protein by interacting with p53, TGF β, Fas and TNF, leading to the accumulation of anti-apoptotic proteins, and eventually inhibit the apoptosis of hepatoma cells[22, 29–32]. On the other hand, genomic integration of HBV leads to host genome dysfunction, affecting cell proliferation, cell cycle progression, apoptosis, and even chromosome stability[33]. Studies have found that recurrent integration of TERT and MLL4 is often observed in HCC[34]; HBx-LINE integration, as a function of long non-coding RNA, has been proved to promote the carcinogenesis[35].
Secondly, for the phenomenon that downregulated genes mainly participated in the metabolism pathways, it has been found that HBV infection is widely involved in the metabolic process of liver cells, thus participating in the hepatocarcinogenesis[22, 36]. Studies found that HBV infection leads to up-regulation of glucose metabolism (such as gluconeogenesis, glucose aerobic oxidation, and pentose phosphate pathway) and lipid metabolism (such as fatty acids, phospholipids and cholesterol biosynthesis)[37, 38] in liver. Recent studies on the mechanism of metabolic recombination provide a new perspective for HBV induced hepatocarcinogenesis. Studies have shown that HBx overexpression can participate in the transcriptional activation of steroid regulatory element binding protein-1 and peroxisome proliferator activated receptor (PPAR γ), leading to hepatocyte steatosis[39]. On the other hand, HBx also participate in the process of glucose metabolism recombination by regulating mitochondrial autophagy through BNIP3L, which can increase the stemness of hepatoma cells[40]. Some of the above studies are consistent with our study, but why most metabolic-related genes are down-regulated deserves further study. Furthermore, for the oxidation-reduction process, it is recognized that the formation of ROS is a common feature of many cellular biological functions and an important target for cytotoxic effects in the progression of HCC[41]. Important ROS-related liver cancer growth signal transduction pathways include nuclear translocation of nuclear factor-Kb (NF-kB)[42]. PI3K/AKT/mTOR signaling cascade is also involved in oxidative press of HCC. Its activation is regarded as a stimulator for cell growth and proliferation[43]. Unlike NF-kB, the increased level of ROS suppresses the phosphorylation of AKT and mTOR, thereby enhances the apoptosis of HCC cells[44]. Recent studies have reported various genes and drugs targeting the PI3K/AKT/mTOR cascade. A combination of sorafenib and C2-ceramide therapy was thought to stimulate the production of ROS, activating caspases-dependent cell apoptosis via PI3K/AKT/mTOR pathway[45]. CYP3A5, a member of CYP450, can induce ROS accumulation, inhibiting AKT phosphorylation at Ser473[44]. CYP450 is considered as another major source of ROS production except from NADPH, especially in liver[41]. However, there are few studies associating CYP450 with the oxidation-reduction process of HCC patients. In this study, a total of 57 genes enrich in oxidation-reduction process, 16 of them belong to CYP450 family, indicating a potential target for the HCC therapy.
The diagnostic model, which was established after cross-validation Lasso regression and best subset selection regression, presented to have a good diagnostic performance both in the training and validation cohort. We thought highly of the diagnostic model constructed in the present study firstly because of the application of gene pairs, which was a good way to eliminate overall sequencing differences between datasets[46], and we used a quite serious variates selection pipeline to determine the gene pairs included in the model. Here, RGS5|FAHD2A and CXCL14|SAMD5 are two gene pairs that composed the signature. It could be conceived that clinically, the expression levels of these two pairs of genes can be compared to identify whether patients with hepatitis B have already acquired HCC. This can solve the problem that different high-throughput methods may not be suitable for model coefficients but the pity is that the number of samples in our training set is not very large. Secondly, because all possible variate combinations have been traversed in the Best subset selection, the selected features can be optimal, and additionally, cross-validation Lasso avoided the overfitting. This model has shown very high efficacy in repeated validations, and we believe it can be a good method for assisting pathological diagnosis of HBV induced liver cancer.
As to the hub genes, including all hub genes were up-regulated in cancer tissues, and most of them are related to cell cycle and mitosis. CDK1 was the one with highest degree. According to previous studies, CDK1, also called cell division cycle protein 2 (CDC2), is an important regulator of the cell cycle, which is required for the transition from the G2 phase into mitosis[47, 48]. For HCC, several studies have suggested that CDK1 related to HCC cell proliferation[49] and regulated apoptin-related cell apoptosis[50]. Interestingly, FOSTAMATINIB was predicted to have therapeutic effect for HBV-related HCC patients. Fostamatinib, a FDA-proved tyrosine kinase (Syk) inhibitor, has been proved to be able to treat chronic immune thrombocytopenia and rheumatoid arthritis[51, 52]. Recent studies even hinted that it has effect on COVID-19[53]. The effect of fostamatinib on HCC was hardly reported and lacked large-scale clinical trials[54]. Here, we further revealed the potential mechanism of fostamatinib treatment of HBV-reduced HCC, that is, targeting CDK1 and AURKA and breaking the disease-causing PPI network.
Furthermore, a prognostic model composed 4 gene pairs was built for HBV-HCC patients, which presented an excellent performance (AUC = 93.1%). The advantages of pairs discussed above also made sense in the prognostic model. However, we hope to have a larger cohort to establish and validate models in the future, for there are still few high-quality datasets for hepatitis B-related liver cancer, especially those with clinical follow-up data.
With xCell, we primarily demonstrated the underlying mechanism related to the rising risk of death from HBV-related HCC, which mainly points to the loss of stromal cells in the tissues. It had been reported that the score of stromal and immune cells of tumor tissue have association with its clinical characteristics[55]. Tissue-based approaches were confirmed to be able to characterize the tumor infiltrates in clear cell renal cell carcinoma and primary glioblastoma multiforme, according to previous studies[56, 57]. A recent research showed that high‐risk cohorts in HCC have higher immune, and stromal scores than that in low‐risk cohorts, which was consistent with our results[58].
Further, result of ssGSEA hinted that the apart from WNT/ß−Catenin Signaling, DNA Repair and G2M Checkpoint which had been fully discussed above, Myc and E2F targets also related with the higher risk of HBV-related HCC. And metabolism pathways exhibited down-regulation in the high risk group. For it was wildly accepted that HBx and c-Myc were respectively oncogenic factors in the HBV virus and host hepatocyte, for critical interaction was confirmed between the two proteins[59]. HBx-mediated Myc stabilization greatly promoted the carcinogenic effect of the virus, on account that SCF (Skp2) ubiquitin E3 ligase could intervene Myc ubiquitination that leaded to Myc oncoprotein, upon which HBx has an effect of stabilization[60, 61]. At the same time, by activating PIK3CA/Akt/mTOR and c-Myb/COX-2 pathways, c-Myc drove apoptosis, for which E2F1 was found acting as a weakening factor[62, 63]. In RB1-altered tumors, the E2F pathway was found activated which was already reported in HCC. While E2F pathway could also activate in CCN-HCC without RB1 inactivation event which might be partly contributed by the ability of cyclin E/Cdk2 complexes to phosphorylate Rb[63]. Briefly, we demonstrated that some pathways not only involved in the occurrence of HCC but also involved in the subsequent progression of the tumor.