Accounting for 75-85% of all primary liver cancer, hepatocellular carcinoma (HCC) is the main histological classification of liver cancer, which is the fourth most frequent cause of cancer-related death globally (Harris et al., 2019; Yang J.D. et al., 2019). The liver is the second most common cancer-prone organ, after the lungs, as was shown by the recent cancer study in China (Fu and Wang, 2018). On the whole, the estimated morbidity of HCC per 100,000 world standard population is 40.0 in males and 15.3 in females (Zhu et al., 2016). Major risk factors for HCC include genetic predisposition, epigenetic variation, chronic hepatitis B infection, hepatitis C virus infection, smoking, obesity, aflatoxin exposure, and diabetes (Puszyk et al., 2013; Baecker et al., 2018). Transplantation is the most useful way to treat HCC; however, after the transplantation process, the tumor recurrence and metastasis rates are high (Au and Chok, 2018; Aufhauser et al., 2018). More than 70% of patients at advanced stage are not suitable for transplantation, whether due to the tumor burden or liver dysfunction (Wang et al., 2019). Therefore, it is urgent to recognize new biomarkers that can act as molecular targets for therapy, and predictors of the prognosis of HCC. With the development of times and technological progress, microarray and high-throughput sequencing technologies have matured and become more reliable, and public databases are improving, such as the Gene Expression Omnibus (GEO)¹ and the Cancer Genome Atlas (TCGA)². The advancement of microarray (Yang X. et al., 2018) and high throughput sequencing technologies (Weinstein et al., 2013) has provided a highly efficient tools to explore key genetic or epigenetic changes in disease to identify biological markers that can be applied to disease diagnosis, therapy, and prognosis (Weinstein et al., 2013; Wang et al., 2018; Yang X. et al., 2018; Li et al., 2019). Additionally, the application of integrated bioinformatics methods in cancer research can solve the problem of different results due to errors caused by different technical platforms or small sample size, thus finding much valuable biological information (Liu X. et al., 2018; Deng et al., 2019; Yan et al., 2019; Yang K. et al., 2019).

In this research, by analyzing and distinguishing genes in human HCC samples and normal hepatocyte samples using TCGA and GEO datasets, we firstly screened out differentially expressed genes (DEGs). Then, GO and KEGG pathway enrichment analyses were applied in the further exploration of the main biological functions, which regulated by the DEGs. After that, the final step is to utilize a protein–protein interaction (PPI) network, survival analyses and drug-gene interaction network analyses to ascertain crucial genes and pathways which affecting the pathogenic mechanism and prognosis of HCC patients.

Hepatocellular carcinoma is a type of primary liver tumor with poor prognosis and high mortality, and the progress in its diagnosis and treatment has always attracted widespread attention from researchers around the world. Because the high recurrence and metastasis rate of HCC remains a challenge, identifying new molecules as biological markers is urgently needed. Integrated bioinformatics analysis, which focuses on screening of DEGs, discovering hub node of network-based and doing survival analysis, which has been diffusely used to recognize latent biological markers related to cancer diagnosis, therapy, and prognosis estimation. In recent years, increasing researches have demonstrated that abnormal gene expression is a factor in the tumorigenesis and development, so it is feasible to screen differential genes as biomarkers to assist diagnosis and treatment. In 2017, by developing an integrated approach including GO and KEGG analysis, PPI network creation, hub gene identification, and overall survival analysis, Li L. et al. (2017) picked out 16 hub genes for HCC from three GEO datasets, five of which may be playing a part in the occurrence, development, invasion, metastasis or recurrence of HCC. In 2018, Zhang L. et al. (2018) used bioinformatics methods to select 10 genes from the GEO dataset GSE64041 for the identification of hub genes and pathways of HCC. Gu et al. (2020) recognized 13 crucial genes correlated with progression and prognosis of HCC from the TCGA-LIHC dataset by weighted gene coexpression network analysis. Compared with previous similar studies, our study not only integrated a large sample size of mRNA expression data from the GEO database but also analyzed RNA sequencing result and clinical data from the TCGA-LIHC database to screen out potential hub genes in HCC. And in the second place, this study validated the DEGs through multiple databases. Finally, we explored the relationship between seven hub genes and existing drugs for cancer therapy, which may provide some guidance for the molecular targeting therapy of HCC in the future.

In our research, DEGs in HCC based on the GEO expression profile of GSE121248 (70 HCC samples and 37 normal samples) and TCGA-LIHC RNA sequencing data (374 HCC samples and 50 normal samples) were identified by bioinformatics analysis. In total, 763 significantly robust DEGs, including 247 upregulated DEGs and 516 downregulated DEGs, were identified. The enrichment analysis results of GO indicated that the DEGs were mostly relevant to “oxidoreductase activity, acting on paired donors,” “monooxygenase activity,” “arachidonic acid epoxygenase activity,” “oxidation-reduction process,” “epoxygenase P450 pathway,” “response to drug,” “extracellular exosome,” “extracellular region,” and “cytosol.” The analysis of KEGG pathway showed that the DEGs were mainly concentrated in the following: “fatty acid degradation pathway,” “metabolic pathways,” “chemical carcinogenesis pathway,” “cell cycle pathway,” and “biosynthesis of antibiotics pathway.” Previous studies have reported that the arachidonic acid-derived metabolites and cytochrome P450 epoxygenase CYP2J2 possibly play vital roles in regulating malignant tumor, stimulating tumor cell growth, and inhibiting tumor cell apoptosis (Liu L. et al., 2011; Xu et al., 2011; Yarla et al., 2016). Additionally, metabolic pathways are important for cancer cell survival because the metabolic demands of cancer cells are often expressed as increased, and HCC shows a significant alteration in lipid metabolism (Pope et al., 2019). Moreover, dysregulation of the cell cycle processes and mitotic cell cycle plays a vital role in the tumorigenesis and progression (Williams and Stoeber, 2012; Wlodarchak and Xing, 2016). These theories are consistent with our results in GO and KEGG enrichment analysis.

Through building PPI network and analyzing it, we identified crucial hub genes in the PPI network, including CDC20, CDK1, MAD2L1, BUB1, BUB1B CCNB1, CCNB2, and CCNA2. Using Oncomine and GEPIA validation, the mRNA expression of these eight hub genes in HCC samples was higher than normal liver samples, the finding that was in accord with the microarray results. Subsequently, HPA database data displayed that the protein and mRNA expression of hub genes were consistent, and most genes were overexpressed in HCC tissue. To inquire prognostic biological markers of HCC, we applied the GEPIA to analyze the influence of hub genes expression level on survival of HCC patients and found that, except CCNB2, the high level gene expression of CDC20, CDK1, MAD2L1, BUB1, BUB1B, CCNB1, and CCNA2 were related to HCC patients poor overall survival. Therefore, these seven genes may be functional in HCC occurrence and development.

It was reported that high expression of CDC20 (cell division cycle protein 20) is associated with poor survival in astrocytoma (Ding et al., 2017), cutaneous squamous cell carcinoma (Chu et al., 2019) and pancreatic ductal adenocarcinoma (Dong et al., 2019). CDC20 promotes the progression of prostate cancer by stabilizing hypo-catenin in tumor-like dry cells (Zhang et al., 2019). However, the expression of cell division cycle protein 20 in HCC still lacks accurate experimental data. As a part of the Ser/Thr protein kinase family, CDK1 (cyclin-dependent kinase 1) is a key molecule that controls the eukaryotic cell cycle. By phosphorylating Bora, Cyclin A/cdk1 could facilitate the phosphorylation, activation and mitotic entry of Aurora A-dependent Plk1 (Vigneron et al., 2018). It is reported that CDK1 overexpression has been found in colorectal cancer, pancreatic ductal adenocarcinoma and thyroid cancer (Zhang P. et al., 2018; Piao et al., 2019; Zheng et al., 2019). It was also reported that CDK1 amplification rate in HCC tissues was usually up to 46% (18/39), which was meaningfully related to poor overall survival (p = 0.008) (Wu et al., 2018). These results were in accord with our study findings.

As a pro-oncogene upregulated in gastric cancer, MAD2L1 (mitotic arrest deficient 2-like protein 1) can be downregulated expression by miR-30a-3p, resulting in inhibition of the proliferation of gastric cancer cells (Wang et al., 2019). Besides, by restraining MAD2L1, miR-200c-5p can inhibit HCC cells proliferation, migration and invasion (Li Y. et al., 2017), suggesting that MAD2L1 can be used in HCC patients prognostic evaluation and targeted therapy. As a cyclin controlling the G1/S and G2/M phases in the cell cycle, CCNA2 (cyclin-A2) is more expressed in CRC samples than in normal samples. The reduction of CCNA2 gene expression would disrupt cell cycle progression and induce apoptosis, thus significantly inhibiting the growth of CRC cells (Gan et al., 2018). By maintaining the expression of CCNA2 protein and the production of arginine, arginine metabolic enzyme argininosuccinate lyase (ASL) can promote the production of nitric oxide synthase, thus promoting the formation of HCC (Hung et al., 2017).

As a mitotic checkpoint serine/threonine kinase, BUB1 is related to tumorigenesis in many cancers. shRNA silencing inhibits the expression of BUB1 gene in glioblastoma tumor cells, thereby reducing the proliferation and tumorigenicity of tumor cells in vivo and in vitro (Yu et al., 2019). Increased BUB1 expression signally facilitates cell proliferation, while decreased BUB1 expression restrains liver cancer cells proliferation (Zhu et al., 2020). The proliferation, migration, and invasion of PCa cell lines can be enhanced via BUB1B overexpression (Fu et al., 2016). Worse OS and DFS of HCC patients can be predicted by the high expression of BUB1B (Zhuang et al., 2018). CCNB1, an important protein regulating the G2/M (mitotic) cell cycle, is activated by Chk1, exerting its oncogenic role in colorectal cancer cells growth in vivo and in vitro (Fang et al., 2014). Abnormal FOXM1 expression can transcriptionally activate CCNB1 expression, thereby promoting the proliferation of HCC cells (Chai et al., 2018).

After exploring the potential information about the interactions between the seven hub genes and existing chemotherapeutic drugs, we found that various drugs could influence the expression levels of these hub genes. However, only topotecan, oxaliplatin and azathioprine could simultaneously reduce the expression level all seven hub genes. And it should be noted that further experiments are needed to support whether HCC patients with hub gene overexpression can benefit from hub gene inhibition or whether these key genes may be targets of drug treatment of tumor need ulteriorly biological experiments support.

In the current study, we have discussed that the development of HCC is associated with the overexpression of seven hub genes, which lead to poor overall survival, indicating that they may be considered as potential prognostic biomarkers for HCC. However, our study has several limitations: (1) some important clinical information (for example, different age, tumor size, TNM stage and grade) were not considered; (2) biological experiments must be carried out in the future to verify the results of our research; (3) the molecular mechanism of hub gene upregulation remains unclear. Therefore, the verification of hub genes will be the focus of our next work.

Adopting a series of bioinformatics analysis methods, the current study identified 763 DEGs and seven hub genes (CDC20, CDK1, MAD2L1, BUB1, BUB1B, CCNB1, and CCNA2) that may be involved in hepatocellular carcinoma tumorigenesis and progression. Additionally, multiple database analysis and survival analysis demonstrated that these seven hub genes may regard as a latent prognostic biomarker and the overexpression of these seven hub genes might lead to reduced overall survival in HCC patients. These results provide a theoretical basis for the molecularly targeted therapy and prognosis evaluation of HCC.

The data in this study were obtained from the GEO (GSE121248) and TCGA public databases, and the acquisition and application methods complied with the corresponding database guidelines and policies.

ZL conceived the study, analyzed the raw data, and wrote the manuscript. YL prepared the diagrams in the manuscript. BC and QZ provided the useful suggestions in methodology. YC designed and instructed the research. All the authors have read and approved the final manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.