Circular RNAs and Hepatocellular Carcinoma: New Epigenetic Players With Diagnostic and Prognostic Roles

Hepatocellular carcinoma (HCC) is one of the leading causes of cancer-related death worldwide. Due to the lack of potent diagnosis and prognosis biomarkers and effective therapeutic targets, the overall prognosis of survival is poor in HCC patients. Circular RNAs (circRNAs) are a class of novel endogenous non-coding RNAs with covalently closed loop structures and implicated in diverse physiological processes and pathological diseases. Recent studies have demonstrated the involvement of circRNAs in HCC diagnosis, prognosis, development, and drug resistance, suggesting that circRNAs may be a class of novel targets for improving HCC diagnosis, prognosis, and treatments. In fact, some artificial circRNAs have been engineered and showed their therapeutic potential in treating HCV infection and gastric cancer. In this review, we introduce the potential of circRNAs as biomarkers for HCC diagnosis and prognosis, as therapeutic targets for HCC treatments and discuss the challenges in circRNA research and chances of circRNA application.


INTRODUCTION
Hepatocellular carcinoma (HCC) is ranked as the sixth most common neoplasm and the third leading cause of cancer-related death worldwide (1). Curative treatments, including liver transplantation, liver resection, and ablation, are only available for early stage HCCs. But due to the absence of specific symptoms at early stages and the lack of early diagnostic biomarkers, most HCC patients are diagnosed at an advanced stage and not eligible for curative treatments (2). So far, the tyrosine kinase inhibitors (sorafenib and lenvatinib) and the combination of atezolizumab (an anti-PD-L1 antibody) with bevacizumab (a vascular endothelial growth factor inhibitor) are the only approved first-line systemic treatment for advanced HCCs (3)(4)(5). Moreover, although 5-year overall survival (OS) rate reaches up to 50%, recurrence occurs in more than 70% HCC patients after curative surgery (6), which severely impairs the prognosis of HCCs. Therefore, development of early diagnostic and/or prognostic biomarkers and identification of novel therapeutic targets are urgently required for improving HCC outcomes.

CircRNAs AS NON-INVASIVE DIAGNOSIS BIOMARKERS OF HCCs
Early diagnosis is critical for the improvement of HCC prognosis and outcomes. The serum alpha-fetoprotein (AFP) (43), alphafetoprotein-L3 (AFP-L3) (44) and Des-gamma-carboxy prothrombin (DCP) (45) are the most commonly used noninvasive circulating biomarkers for HCC diagnosis in clinical practice, but limited evidence supports their contribution to HCC early detection or improvement in HCC outcomes (46). Therefore, identifying novel biomarkers for HCC early diagnosis is urgently required. CircRNAs are dysregulated in HCCs (47), and more resistant to nuclease degradation compared to linear RNAs due to their covalently closed loops (38), which make them potential biomarkers for HCC diagnosis. Here, we mainly introduce some recent studies indicating that circRNAs are promising non-invasive circulating diagnosis biomarkers of HCCs.
By analyzing differentially expressed circRNAs (DECs) in plasma between 40 healthy controls and 71 HCC patients, Sun et al. recently identified a three-circRNA signature, including hsa_circ_0004001, hsa_circ_0004123, and hsa_circ_0075792, which are highly expressed in HCC patients (50). This 3-circRNA signature showed a potent diagnosis value (AUC: 0.89, sensitivity: 0.905; specificity: 0.781) (50). In this study, the authors did not compared the diagnosis value of the 3-circRNA signature with that of AFP, but the AUC, sensitivity and specificity values of the 3-circRNA signature may slightly better than those of AFP from other studies (48,49).
Down-regulation of hsa_circ_0001445 (circSMARCA5) in HCCs was identified in 3 cohorts of 208 pairs of HCC and corresponding adjacent noncancerous liver (ANL) tissues, and circSMARCA5 overexpression inhibited HCC growth and metastasis in vitro and in vivo (52). By comparing plasma circSMARCA5 levels in 103 healthy controls, 117 hepatitis (hepatitis B and C), 143 cirrhosis and 135 HCC patients, Li et al. found that plasma circSMARCA5 decreases gradually but significantly from healthy controls, hepatitis, cirrhosis to HCC patients (53). It worth noting that, in patients whose AFP levels FIGURE 1 | Schematic representation of circRNA biogenesis and functions. CircRNAs are covalently closed loop structures generated by back-splicing of premature transcripts mediated by either Alu repeats, inverted complementary repeats (ICR), or RNA binding proteins (RBP). CircRNAs could function as miRNA sponges to block miRNA-mediated inhibition on target genes; or as scaffolds, recruiters or decoys of some RNA binding proteins to regulate the function of the associated proteins. Some circRNAs containing IRES and ORFs serve as RNA template for translation.
Recently, a large-scale, multicenter study identified a plasma circRNA panel (CircPanel) containing three circRNAs (hsa_circ_0000976, hsa_circ_0007750, and hsa_circ_0139897), which had greater accuracy than AFP in distinguishing individuals with HCC from Non-HCC (

CircRNAs AS PROGNOSIS BIOMARKERS OF HCCs
Recurrence occurs in more than 70% HCC patients post curative surgery (6). Even for the most effective liver transplantation (LT), the recurrence rate is near 20% (55). Although the TNM staging system of the American Joint Committee on Cancer (AJCC), the Barcelona Clinic Liver Cancer (BCLC) classification, and the Cancer of the Liver Italian Program (CLIP) staging system, have been employed to evaluate the prognosis of HCC patients (56), their prognostic predictive performance was unsatisfactory partly because those assessments do not take the critical and complicated molecular pathogenesis of HCCs into account (57). Recent studies demonstrated that some circRNAs could serve as prognostic biomarkers for HCCs.
Hsa_circ_0000267 was up-regulated in HCC tumor tissues and positively associated aggressive clinicopathological features such as tumor size and TNM stage (58). Multivariable analysis identified enhanced hsa_circ_0000267 as an independent prognostic factor for the OS in HCC patients (58).
Chromosomal amplification at 7q21-7q31 was closely related to tumor recurrence in various types of cancers including HCCs (59)(60)(61)(62). By assessing the expression of 43 putative circRNAs in this chromosomal region in 4 pairs of matched HCC and normal tissues, Huang et al. identified that hsa_circ_0082002 (circMET) was the most significantly and consistently up-regulated circRNA in HCCs (63). By evaluating circMET expression in 90 paired HCC and normal tissues and in a tissue microarray (TMA) consisting 209 HCC tissues, levels of circMET were significantly and positively correlated with microvascular invasion, absent tumor encapsulation, multiple tumors and late HCC stage. Multivariate analysis identified circMET as an independent predictor for both OS and postoperative recurrence in HCC patients (63).
Sun et al. have demonstrated that hsa_circ_0027345 (circLRIG3), a nucleus-enriched circRNA, was highly expressed in HCCs and positively associated with aggressive clinicopathological features, such as tumor size, vascular invasion, Edmondson's grade and later TNM stage, and was identified as an independent risk prognostic factor for OS in HCC patients (64).
The down-regulation of hsa_circ_0001727 (circZKSCAN1) in HCCs were determined by two independent groups (65,66). Expression of circZKSCAN1 was associated with multiple clinicopathologic factors, such as tumor numbers, cirrhosis, vascular invasion, microscopic vascular invasion, or tumor grade (65), and was determined as an independent and significant factor affecting both OS and RFS in HCCs (66).
In addition to serving as a non-invasive biomarker, decreased expression of circSMARCA5 in HCC tumor tissues was significantly correlated with poorer tumor differentiation, more advanced tumor stage, larger tumor size and presence of microvascular invasion. Multivariate analysis demonstrated that circSMARCA5 expression level was an independent risk factor for OS and RFS in HCCs (52).
A recent study has demonstrated that an up-regulated plasma circRNA, hsa_cic_0005397, was positively correlated with tumor size (p = 0.020) and TNM stage (p = 0.006) (67). Meanwhile, a dynamic monitoring of plasma hsa_cic_0005397 in HCC patients who had undergone surgical resection revealed that plasma hsa_cic_0005397 level was drastically dropped in HCC patients after operation, but prominently elevated in recurrent or m e t a s t a t i c H C C p a t i e n t s , s u g g e s t i n g t h a t p l a s m a hsa_cic_0005397 could serve as a prognostic biomarker for post-operational recurrence and metastasis in HCC patients (67). Moreover, the plasma hsa_cic_0005397 level was negatively associated with OS in HCC patients (67).
However, due to the limited sample size and the lack of multicenter validation, the diagnosis and prognosis potential of those circRNAs should be further determined in the future.

ONCOGENIC AND TUMOR SUPPRESSIVE CircRNAs IN HCCs
Like many other non-coding RNAs, such as lncRNAs and miRNAs, circRNAs have been demonstrated to function as either oncogenes or tumor suppressors in HCCs.
By comparing circRNA expression profiles in primary HCCs with or without post-surgery pulmonary metastasis, Hu et al. identified that hsa_circ_0085616 (circASAP1) is highly expressed in HCCs with greater metastatic potential (31). Mechanistically, circASAP1 functions as a sponge for miR-326 and miR-532-5p and enhances the expression of CSF-1 and MAPK1, resulting in an increase in intra-tumor infiltration of tumor associated macrophages and activation of the MAPK-ERK signaling pathway (31). Consistent with this study, Li et al. also demonstrated that circASAP1 plays an oncogenic role in HCCs possibly by activating the b-catenin, ERK and AKT signaling pathways (68).
Through bioinformatics analysis and experimental validation, Li et al. demonstrated that up-regulation of hsa_circ_0074854 (circMAT2B) in HCCs was correlated with poor prognosis and could serve as an independent prognostic factor (69). By functioning as a sponge of miR-338-3p, circMAT2B released miR-338-3p-mediated inhibition on PKM2, resulting in enhanced glycolysis, proliferation and metastasis in HCCs (69).
In addition to acting as a miRNA sponge, some circRNAs have been reported to function as protein decoys, scaffolds and recruiters to promote HCC progression. Wang et al. demonstrated that hsa_circ_102034 (circRHOT1) was highly expressed in HCCs and promoted proliferation, migration and invasion in HCC cells by acting as a protein recruiter to recruit TIP60 onto the promoter region of NR2F6 to initiate NR2F6 expression (70).
By performing a circRNA microarray assay specifically targeting human circRNA splicing sites in seven paired HCC tumor and normal samples, Han et al. demonstrated that hsa_circ_0007874 (circMTO1) was significantly decreased in HCCs and correlated with poor prognosis of HCC patients (71). CircMTO1 could serve as a miR-9 sponge to release miR-9-mediated inhibition on p21, resulting in inhibition on HCC growth both in vitro and in vivo (71).
Compared to oncogenic circRNAs, the tumor suppressive circRNAs are less studied probably due to their low expression in HCCs. Only countable circRNAs have been identified as tumor suppressors in HCCs. For example, circSMARCA5, a downregulated circRNA in HCC, has been identified as a tumor suppressor for HCC progression. Over-expression of circSMARCA5 retained miR-17-3p and miR-181b-5p to release their common target, TIMP3, a well-known tumor suppressor in HCCs, resulting in enhanced HCC growth and metastasis both in vitro and in vivo (52,72). By functioning as a decoy of fragile X mental retardation protein (FMBP), an RNA binding protein, circZKSCAN1 prevented the binding of FMBP with cell cycle and apoptosis regulator 1 (CCAR1) mRNA and in turn inhibited CCAR1 expression and CCAR1-mediated activation of the Wnt/ b-catenin signaling pathway, resulting in compromised stemness in HCC cells (66).

CircRNAs RELATED TO DRUG RESISTANCE IN HCCs
Advanced HCCs are not legible for curative treatments. Although chemotherapy, targeted therapy and immunotherapy are commonly employed for those advanced HCCs, the therapeutic efficacy is unsatisfied due to low objective response rate and acquired resistance (73)(74)(75). In addition to playing oncogenic or tumor suppressive roles in HCC progression, some circRNAs have been recently reported to be related to drug resistance in HCC treatments.
For PD1 antibody mediated immunotherapy, circMET has been demonstrated to confer HCC cells resistance to anti-PD1 treatment by enhancing the immunosuppressive tumor microenvironment (63). Mechanistically, circMET functions as a sponge of miR-30-5p to induce SNAIL-mediated Dipeptidyl 4 (DPP4) expression, resulting in degradation of CXCL10, an important chemokine promoting intra-tumor infiltration of effector T cells, and leading to subsequent resistance to anti-PD1 treatment (63).

EXOSOMAL CircRNAs IN HCC PROGRESSION
Exosome is a type of nano-sized secreted vesicles, which are present in all body fluids under both physiological and pathological conditions (81). By carrying various biomolecules, including proteins, nucleic acids, lipids, and transferring from one cell to another, exomes play an important role in mediating intercellular communication (81). In 2015, Li, et al. for the first time demonstrated the presence of abundant circRNAs in exosomes (82). By RNA-seq analysis, they demonstrated enrichment of circRNAs in exosome compared to producer cells (82). A number of exosomal circRNAs have been identified to function as either diagnosis/prognosis biomarkers or oncogenic/tumor suppressive factors in HCCs (83,84). For example, an exosomal circRNA, hsa_circ_0070396, was identified as a better diagnostic biomarker than AFP in distinguishing HCCs from healthy controls (AUC: 0.8574 vs. 0.781), and differentiating early HCC patients (with BCLC are 0 or A) from advanced ones (with BCLC are B+C) (AUC: 0.7132 vs. 0.6535) (85). Higher circulating exosomal circAKT3 was identified in HCC patients compared to healthy controls, and positively correlated with tumor recurrence rates and mortality, suggesting that circAKT3 could serve as a prognostic biomarker of HCCs (86). A recent study has demonstrated that three exosomal circPTGR1 isoforms secreted from HCC cells with higher metastasis potential to promote the metastasis of HCC cells with lower metastasis potential via regulating the miR449a-MET pathway (87). Su et al. have demonstrated that circRNA Cdr1as competitively bound to miR-1270 to upregulate AFP level, thereafter accelerating proliferative and migratory abilities of HCC cells (88). Meanwhile, exosome-transmitted circRNA Cdr1as stimulated malignant behaviors of surrounding normal cells and finally contributed to the progression of HCC (88). An interesting study showed that a pro-invasive exosomal circRNA-100338 secreted from HCC cells could transfer to HUVECs to affect cell proliferation, angiogenesis, permeability, and vasculogenic mimicry (VM) formation ability of HUVECs, and in turn promote HCC metastasis (89).
Instead of transporting exosomal circRNAs from HCC cells to surrounding normal cells, the exosomal circRNA can also be generated and transported from surrounding normal tissues to HCC cells, resulting in altered HCC progression. For example, an exosomal circRNA, circ-0051443, which was generated in normal cells and transferred to HCC cells, could inhibit HCC progression by competitive binding to miR-331-3p, resulting in enhanced apoptosis and cell cycle arrest in HCC cells (90).
It is clear that exosomal circular RNAs play a significant role in HCC progression. Therefore, identifying novel exosomal circRNAs and understanding their biological functions can not only help us better understand the new mechanisms of potential HCC development, but also improve the clinical diagnosis, prognosis and treatment of HCCs.
We have summarized all the circRNAs mentioned in this review in Table 1.

CHALLENGES AND CHANCES OF CircRNAs
With the advent of high-throughput sequencing and highefficiency big data analysis, circular RNAs have been emerged as a novel class of non-coding RNAs in eukaryotes. Accumulating evidence indicates that circRNAs are novel non-coding players the involved in various biological processes and diseases. However, some fundamental questions regarding the function and regulation of circRNAs remain unknown.
First, the biological role of circRNAs is less defined. Only less than 3% (> 7000 circRNAs) of recorded circRNAs (295,526 circRNAs integrated from circBase, circNet, and circRNAdb) (91) were curated functional (92), which at least in part suggests that the most circRNAs may be the non-functional by-products of RNA splicing events. In a previous study, Guo et al. claimed that compared to mRNAs, most circRNAs are less abundant, less cell-type specific and less conserved (93). Moreover, ribosome profiling provides no evidence for their translation (93).
Second, serving as miRNA sponges may not be the general role of circRNAs. Although abundant studies support the role of circRNAs as miRNA sponges in regulating gene expression and executing their biological functions, the existence of numerous circRNAs in P. falciparum and S. cerevisiae (94,95), who lack known siRNA and miRNA pathways, suggests that as miRNA sponges may not be a general role for most circRNAs. This is consistent with a previous study showing that there is no particular enrichment of AGO2 binding on exons involved in circRNAs compared to their neighboring linear exons (93).
Third, the cis and trans regulators for circRNAs remain illuminated. Some studies demonstrated that the significantly longer introns and the inverted repeat sequence bracketing the regions that produce circRNAs are essential cis regulators for circRNA formation in humans and flies (38,(96)(97)(98). Meanwhile, canonical splicing signals are required for the circularization of most circRNAs (93,99), suggesting that canonical spliceosomal machinery may serve as trans regulators for circRNA formation. However, accumulating strong evidence indicates that the expression of circRNAs does not simply correlate with their linear host genes (100,101), suggesting that circRNA formation is a complicated process which may be regulated by additional circRNA-specific cis and trans regulators.
Although it remains controversial on the general roles of circRNAs, most functional circRNAs identified so far have been demonstrated to serve as miRNA sponges in regulating gene expression. By employing this feature, some artificial circRNAs have been engineered to sequester miRNAs relevant in human disease and shown promising potential for application in molecular medicine and biology. For example, an artificial circRNA carrying an array of miRNA-122 binding sites could sequester liver specific miR-122 from HCV RNA, resulting in inhibition on HCV viral protein production and HCV replication (102). Liu et al. generated an artificial circRNA sponge for miR-21, an oncomiR in many types of cancers, to block miR-21-mediated proliferation in gastric carcinoma cells (103). Therefore, in companion of growing understanding of circRNAs, expended application of circRNAs could be established for both research and clinic purposes.

CONCLUSION AND PROSPECT
In summary, circRNAs are emerging as a class of novel noncoding RNAs participating diverse physiological processes and pathological conditions. Although it remains controversial on the general roles of circRNAs, accumulating evidence indicates that circRNAs are functional by serving as either miRNA sponges, protein decoys, or translational templates. Like other types of non-coding RNAs, deregulation of circRNAs has been demonstrated in HCCs. A number of deregulated circRNAs have been identified as either oncogenic or tumor suppressive regulators in HCC progression. Moreover, some deregulated circRNAs could serve as non-invasive circulating biomarkers for HCC early diagnosis with great specificity and sensitivity superior to clinically used serum AFP. Meanwhile, some circRNAs have been involved in drug resistance in HCC treatments. In this review, we summarized the recent findings of circRNAs in HCC diagnosis, prognosis, progression and drug resistance and proposed that circRNAs may have great potential to serve as novel biomarkers for HCC early diagnosis, prognosis and therapeutic targets for HCC treatments.
Despite the advancements achieved in the circRNA field, several fundamental questions remain elusive. Given the fact that only a few percentage functional circRNAs (< 3%) were identified from the great circRNA popularity, whether circRNAs are a group of functional non-coding RNAs is still an open question. The existence of numerous circRNAs in species deficient with RNAi and miRNA pathways raises the question whether as miRNA sponges is the general role of circRNAs or just is an exception. Although some studies have demonstrated that the formation of circRNAs depends on the cis elements, such as the inverted repeats, and the trans regulators, such as spliceosomes, of linear host genes, the lack of correlation at expression levels and cell-type specificity between circRNAs and their linear host genes strongly suggests the existence of circRNA specific regulators. Therefore, novel techniques and concepts regarding those fundamental questions are warranted in the future to expand our understanding of circRNAs and improve our investigation on circRNAs.

AUTHOR CONTRIBUTIONS
KA, X-DW, YF, XL, YM, JL, XW, and QH performed extensive literature search and discussion. KA and X-DW drafted the manuscript. XK, JG, and HW edited the manuscript. All authors contributed to the article and approved the submitted version.   (90) for SMC-morning Star Youth Scholars Program, the Clinical