As one of the most common malignant tumors of digestive system, hepatocellular carcinoma (HCC) is characterized by diverse etiology, high incidence, and poor prognosis. According to statistics, both its incidence and mortality rates are the highest in Asia (Singal et al., 2020). Due to the lack of obvious symptoms, the majority of HCC patients are diagnosed at advanced stages, with limited treatment options. Despite progress has been made in diagnosis and therapy during the last decades, the recurrence and metastasis rates remain high (Ghouri et al., 2017). To confront the current situation, it is crucial to develop new strategies for screening and monitoring of HCC. Alpha-fetoprotein (AFP) is currently the most widely used biomarker for HCC diagnosis, but its sensitivity and specificity are still not satisfactory (Daniele et al., 2004). In recent years, the development of various sequencing platforms and bioinformatics technologies has facilitated the identification of many novel biomarkers. Public databases have provided us with rich and diverse data resources, and we can further improve our understanding of HCC by integrating data from different sources. Although several biomarkers have been reported relating to the diagnosis and prognosis of HCC, such as alpha-fetoprotein lens culinaris agglutin-3 (AFP-L3), des-γ-carboxy prothrombin (DCP), glypican-3 (GPC3), and so on, their practical applications are yet to be evaluated (Piñero et al., 2020). With the vigorous development of bioinformatics, many biomarkers have been identified to be associated with the development of HCC, but few of them have been proven to be of practical use. The number of reliable tumor biomarkers that can be used for the early detection and prognostic assessment is still small in clinical practice. Thus, identification of novel potential biomarkers is warranted, which may contribute to update of diagnostic techniques and improvement of therapeutic efficacy.

It is well known that non-coding RNAs (ncRNAs) account for the vast majority of the human transcriptome, mainly including microRNAs (miRNAs), long non-coding RNAs (lncRNAs), and circular RNAs (circRNAs) (Chan and Tay, 2018). MicroRNA (miRNA) is a class of evolutionarily conserved non-coding RNA (ncRNA) with 18–25 nucleotides in length (Kloosterman and Plasterk, 2006). It participates in a series of physiological and pathological processes via mediating the post-transcriptional regulation of target genes (Krol et al., 2010). Aberrant expression of numerous miRNAs has been linked to cancer initiation and progression (Lee and Dutta, 2009). Long non-coding RNA (lncRNA) is known as a type of ncRNA whose length exceeds 200 nucleotides. Studies have revealed that lncRNAs play an important role in cancer development through a variety of mechanisms (Bhan et al., 2017). Circular RNA (circRNA) is a newly discovered ncRNA with a closed-loop structure. Compared with the traditional linear RNA, circRNA is more resistant to RNA exonuclease, without terminal 5′ caps and 3′ polyadenylated tails, and thus more stable (Jeck and Sharpless, 2014). CircRNAs have been confirmed to exert effects on regulating cellular metabolism in cancer (Yu T et al., 2019). In recent years, increasing numbers of researchers have dedicated themselves to exploring the biological functions of ncRNAs. Various computational methods have also been developed for the prediction of potential associations between ncRNAs and disease, which is of critical importance for the identification of biomarkers (Lan et al., 2020; Chen et al., 2021).

The competing endogenous RNA (ceRNA) hypothesis was first put forward by Salmena et al., in 2011 (Salmena et al., 2011). CeRNA is a class of ncRNA that can competitively bind shared miRNAs and cross-regulate each other at the post-transcription level. In the cytoplasm, lncRNA and circRNA can serve as miRNA sponges by common miRNA response elements (MREs) and indirectly regulate the downstream target genes (Taulli et al., 2013; Yao et al., 2019). This ceRNA-based regulatory mechanism has been discovered in multiple cancers. For example, lncRNA HOTAIR regulated HER2 expression through competition for miR-331-3p, thereby facilitating tumor development (Liu et al., 2014). LncRNA H19 was reported to exerted oncogenic functions in gallbladder cancer via modulating miR-342-3p and FOXM1 (Wang et al., 2016). In addition, circRNA ciRS-7 could act as the “super sponge” of microRNA-7 (miR-7) and inhibits the activity of miR-7 (Peng et al., 2015). Accumulating evidence has indicated that ceRNA regulation network serves a role in biological processes of HCC development, such as proliferation, metastasis, epithelial to mesenchymal transition (EMT), and chemotherapy resistance (Wu et al., 2018; Liu Z et al., 2019; Huang et al., 2020; Song et al., 2020).

The ceRNA network provides new perspectives for improving diagnosis and treatment for HCC. Even though several ceRNAs have been found associated with HCC progression (Bai et al., 2019; Guo et al., 2019; Wang et al., 2019; Yu J et al., 2019), our current understanding of ceRNA regulatory network in HCC is still very limited. Further exploration is needed to unravel unknown functions and mechanisms of related ceRNA. Sequencing data used in this study were collected from public databases. Starting from the differentially expressed messenger RNAs (mRNAs), we inversely predicted the targeted miRNAs and their relevant lncRNAs and circRNAs, and then constructed a comprehensive ceRNA network. Bioinformatics tools were applied to analyze and discuss the crucial pathways as well as the diagnostic performance and prognostic value of the key genes. LncRNAs, circRNAs, miRNAs and targeted mRNAs engaged in the ceRNA network may become potential diagnostic biomarkers and therapeutic targets for HCC. The research process is shown in Figure 1.

Amounting evidence has shown that ceRNA might play a role in cancer initiation and progression (Qi et al., 2015). In the ceRNA hypothesis (Salmena et al., 2011), the ability of ceRNA to competitively bind to miRNA can influence tumorigenesis and cancer progression via regulating mRNA expression. To date, several ceRNAs have been identified to have a role in HCC (Xu G et al., 2020). However, there are still many ceRNAs of potential significance that have yet to be identified and require further exploration. Through bioinformatics analysis, this study attempted to establish a lncRNA/circRNA-miRNA-mRNA network holding biological functions in HCC. Integration of various databases would help achieve more reliable results. Based on the ceRNA hypothesis, a ceRNA network was successfully constructed via stepwise reverse prediction from mRNA to lncRNA/circRNA. Our results are expected to provide valuable guidance for HCC management.

The minichromosome maintenance (MCM) family is mainly known for their involvement in DNA replication (Maiorano et al., 2006). Given that DNA replication is a crucial pathway in tumor development, members of the MCM family are implied to be closely related to cancer development as well (Neves and Kwok, 2017; Yu et al., 2020). The overexpression of MCMs in various cancer tissues has been demonstrated by multiple studies, and is generally connected with poor prognostic features (Giaginis et al., 2011; Peng et al., 2016; Liu et al., 2017; Issac et al., 2019). Through functional enrichment analysis, we found that the hub genes with MCMs predominating were mainly involved in cell cycle and DNA replication pathways. There is evidence suggesting that high expression of MCM4 is correlated with clinicopathological variables and prognosis of HCC and silencing MCM4 can suppress the tumorigenicity of hepatoma cells (Xu et al., 2021). Liu et al. found that knockdown of MCM6 in Huh7 cells could cause a delay in S/G2-phase progression through down-regulating the cell cycle checkpoint (Liu et al., 2018). In addition, it has been demonstrated that MCM7 promotes cancer progression through cyclin D1-dependent signaling (Qu et al., 2017). The above studies have displayed the potential of MCMs as biomarkers to engage in HCC management, which is consistent with the results observed in this study. Known as a molecular switch, CDC6 is considered to have a transcriptional effect on E-Cadherin and subsequently affect EMT (Sideridou et al., 2011; Petrakis et al., 2012). In cancer cells, aberrant expression of CDC6 is involved in proliferation and tumor growth by modulating cell cycle (Lim and Townsend, 2020). Xu et al. revealed that CDC6 was regulated by miR-215-5p to involve in the proliferation of HCC (Xu H et al., 2020). They also found that CDC6 was negatively associated with overall and disease-free survival in HCC patients. In this study, CDC6 was modified by hsa-miR-26a-5p, which has been reported to have an effect on cell proliferation, migration, and invasion in digestive malignancies (Tian et al., 2019; Li H. H et al., 2020; Li Y et al., 2020; Zhang et al., 2020). The GINS complex, a component of the DNA replication machinery, usually participates in DNA replication through interactions with MCMs and other components (Labib and Gambus, 2007; Yoshimochi et al., 2008). GINS1 is a subunit of the GINS complex. Previous studies have demonstrated that GINS1 is upregulated in tumor samples and correlated with poor prognosis (Bu et al., 2020a; Bu et al., 2020b; Li H et al., 2021). These studies are consistent with our findings. The five mRNAs (CDC6, GINS1, MCM4, MCM6, and MCM7) in the ceRNA network were upregulated in HCC tissues and of great value in HCC diagnosis and prognosis.

MicroRNA (miRNA), a type of single-stranded nonencoding RNA, is widely involved in tumor generation and development (Krol et al., 2010). MiRNAs regulate gene expression primarily by degrading mRNA or inhibiting its translation (Bartel, 2009). In this ceRNA network, eight miRNAs (hsa-miR-150-5p, hsa-miR-19b-3p, hsa-miR-23b-3p, hsa-miR-26a-5p, hsa-miR-651-5p, hsa-miR-10a-5p, hsa-miR-214-5p and hsa-miR-486-5p) were found underexpressed in HCC samples. They linked the ceRNAs to the target mRNAs. Several miRNAs of them have been reported to have a place in proliferation, migration, and invasion of HCC cells (Li T. et al., 2014; Pang et al., 2018; He et al., 2019; Wu et al., 2019; Zhu et al., 2020; Hayashi et al., 2021). Similarly, there have been researches that explored the functions of the five lncRNAs (DLGAP1-AS1, GAS5, LINC00665, TYMSOS, and ZFAS1) on tumorigenesis and progression of HCC or other tumors (Zhang et al., 2019; Zhou et al., 2019; Ding et al., 2020; Lin et al., 2020; Gu et al., 2021). Their expression levels were highly relevant to tumor growth, invasion, and metastasis. With the development of high-throughput sequencing and emergence of bioinformatic methods, studies have increasingly revealed the important roles of circRNAs in various tumors (Lei et al., 2020; Li R et al., 2020). Their unique covalent closed-loop structures make them advantageous for clinical application. Due to the short development period, abundant circRNAs and their functions in HCC need further excavation. Our study reversely predicted six circRNAs (hsa_circ_0003209, hsa_circ_0008128, hsa_circ_0020396, hsa_circ_0030051, hsa_circ_0034049, and hsa_circ_0082333) and all of them had reliable diagnostic value in HCC.

Notably, our study discovered a lncRNA-miRNA-mRNA axis of great clinical significance, namely ZFAS1/hsa-miR-150-5p/GINS1 ceRNA axis. Their expression levels and predicted interactions in HCC are in line with the ceRNA hypothesis. LncRNA ZFAS1 may contribute to tumorigenesis of HCC by sponging hsa-miR-150-5p and regulating the expression of the target mRNA GINS1. Moreover, the diagnosis and prognosis analysis of each gene showed favourable outcomes (p < 0.05). Our GSEA analysis found that the target mRNA GINS1 was significantly associated with cell cycle, DNA replication, p53 signaling pathway, mTOR signaling pathway, Notch signaling pathway, Wnt signaling pathway, etc. The majority of these pathways have been confirmed to be involved in HCC development (Greenbaum, 2004; Giovannini et al., 2016; Dimri and Satyanarayana, 2020). The previous study revealed that ZFAS1 acted as an oncogene in HCC progression by binding miR-150 and abrogating its tumor-suppressive function (Li et al., 2015). Besides, in vitro experiments have verified that miR-150-5p inhibition significantly promotes hepatoma cell migration and invasion (Li T et al., 2014). GINS1 was reported to be associated with tumor grades and poor survival of HCC patients (Li S et al., 2021). Furthermore, cell cycle, cell proliferation assay, and in vivo animal model experiment indicated that knocking down GINS1 induced in G1/S phase cell cycle arrest and decreased HCC cells proliferation (Li S et al., 2021). These studies provide excellent support for our studies.

In summary, through integrating data from a variety of databases, we successfully constructed a ceRNA network containing five lncRNAs, six circRNAs, eight miRNAs, and five mRNAs. The ceRNA network can promote HCC progression via cell cycle, DNA replication, and other pathways. Clinical diagnostic and survival analyses demonstrated that the ZFAS1/hsa-miR-150-5p/GINS1 ceRNA regulatory axis had a high diagnostic and prognostic value. Our findings are expected to provide potential biomarkers and therapeutic targets for HCC management. Nevertheless, our study also presents some limitations. Results in this study are based on bioinformatic predictions and further experiments and clinical practice are warranted. Additionally, since circRNAs has not been fully studied in HCC, data of circRNAs from the available datasets are limited and insufficient for prognostic analysis. We expect more research with larger sample sizes to expand and refine our conclusions.