Regulatory Functions and Mechanisms of Circular RNAs in Hepatic Stellate Cell Activation and Liver Fibrosis

Chronic liver injury induces the activation of hepatic stellate cells (HSCs) into myofibroblasts, which produce excessive amounts of extracellular matrix (ECM), resulting in tissue fibrosis. If the injury persists, these fibrous scars could be permanent and disrupt liver architecture and function. Currently, effective anti-fibrotic therapies are lacking; hence, understanding molecular mechanisms that control HSC activation could hold a key to the development of new treatments. Recently, emerging studies have revealed roles of circular RNAs (circRNAs), a class of non-coding RNAs that was initially assumed to be the result of splicing errors, as new regulators in HSC activation. These circRNAs can modulate the activity of microRNAs (miRNAs) and their interacting protein partners involved in regulating fibrogenic signaling cascades. In this review, we will summarize the current knowledge of this class of non-coding RNAs for their molecular function in HSC activation and liver fibrosis progression.


Introduction
Chronic liver disease (CLD) represents a gradual deterioration of liver functions, which results from a spectrum of aetiologies including toxins, viral hepatitis, alcohol abuse, and non-alcoholic fatty liver disease (NAFLD) [1]. CLD involves a replacement of healthy liver tissues with fibrotic scars, and hepatic stellate cells (HSCs) play a major role in this process at a cellular level. HSCs are a liver mesenchymal cell type residing in Disse space. In normal livers, HSCs are quiescent and store retinol in cytoplasmic lipid droplets, but upon liver injury, HSCs become activated, lose retinol storage, and trans-differentiate into myofibroblasts that produce extracellular matrix (ECM) mainly type I collagen [2,3]. Excessive ECM accumulation is one of the critical hallmarks of liver fibrosis, disrupting liver architecture and forming fibrous scars [4]. For a short-term injury, fibrosis could be reversed by HSC inactivation and/or apoptosis [5]. However, persistence of injury promotes the continuous production of ECM and eventually leads to irreversible liver fibrosis and cirrhosis [6]. Cirrhosis is the end-stage CLD where the liver structures and functions are severely compromised [7,8].

Role of circRNAs
CircRNAs serve an essential non-coding function in various cellular processes. Other than competing with pre-mRNA splicing and lowering the levels of linear mRNA isoforms, leading to altered gene expression [43,47,50], some speculate that they act as miRNA sponges if they contain multiple miRNA binding sites [69][70][71][72]. Moreover, since circRNAs possess protein binding sites [39], circRNAs can affect cellular function by controlling activity of their protein partners [50, 73,74]. Despite the lack of a 5 cap structure and a 3 Poly(A) tail, circRNAs can be translated into proteins via the internal ribosome entry site (IRES) and N 6 -methyladenosines (m 6 A)-mediated cap-independent translation [75,76].

Acting as miRNA Sponges
MiRNAs are a class of small and highly conserved non-coding RNA molecules that can regulate gene expression [77]. Most miRNAs are transcribed from DNA into primary miRNAs (pri-miRNAs) by RNA polymerases II and III [78]. Then, a class 2 ribonuclease enzyme process these pri-miRNAs into precursor miRNAs (pre-miRNAs), following by a cascade of cleavage steps to generate mature miRNAs [78,79]. Then, a mature miRNA duplex is incorporated into the AGO protein family to form the miRNA-induced silencing complex (miRISC) that can target mRNAs and cause mRNA degradation and/or translational repression [77][78][79][80][81], resulting in altered gene expression. CircRNAs appear to play an important role in the miRNA-mediated regulation of gene expression, possibly by sponging specific miRNAs and thus increasing protein levels of miRNA target genes (Figure 1a) [82][83][84][85]. Nevertheless, a caution should be taken that several miRNA binding sites within a circRNA are required for efficient sequestration of a particular miRNA [86].
play an important role in the miRNA-mediated regulation of gene expression, possibly by sponging specific miRNAs and thus increasing protein levels of miRNA target genes ( Figure 1a) [82][83][84][85]. Nevertheless, a caution should be taken that several miRNA binding sites within a circRNA are required for efficient sequestration of a particular miRNA [86]. (b) CircRNAs with RNA-binding protein (RBP) binding sites have the potential to decoy these proteins and disrupt their functions. (c) Some circRNAs have been demonstrated to serve as protein scaffolds, allowing the colocalization of proteins (e.g., enzymes and their substrates). (d) CircRNAs containing an internal ribosome entry site (IRES) or N 6 -methyladenosine (m 6 A) and a start codon can be translated to proteins under certain conditions. (e) Additionally, circRNAs may attract certain proteins to a particular location in the cell. (f) CircRNAs can alter gene expression by interacting with RNA polymerase II (RNA Pol II). (g) Some circRNAs can affect the splicing of the host gene and influence the ratio between mRNAs and circRNAs. Abbreviation: Tet1: tet methylcytosine dioxygenase 1; U1 snRNP: U1 small nuclear ribonucleoprotein particle; EIcircRNA: exon-intron circRNAs; ecircRNA: exonic circRNA; ciRNA: circular intronic RNAs; UAP56: spliceosome RNA helicase DDX39B; URH49: ATP-dependent RNA helicase DDX39A.

Interacting with Proteins
CircRNAs may bind with RBPs via RBP-binding sites [50]. RBPs play essential roles in the control of gene expression by regulating splicing, mRNA maturation, transport, localization, translation, and decay [87][88][89]. Various circRNAs have been discovered to interact with RBPs by acting as protein sponges or decoys, thus activating, or disrupting their biological functions (Figure 1b) [90][91][92]. Some circRNAs serve as a protein scaffold to promote the colocalization of two or more proteins, such as enzymes and their substrates, enhancing their activity (Figure 1c) [74,93]. Some circRNAs can recruit certain (b) CircRNAs with RNA-binding protein (RBP) binding sites have the potential to decoy these proteins and disrupt their functions. (c) Some circRNAs have been demonstrated to serve as protein scaffolds, allowing the colocalization of proteins (e.g., enzymes and their substrates). (d) CircRNAs containing an internal ribosome entry site (IRES) or N 6 -methyladenosine (m 6 A) and a start codon can be translated to proteins under certain conditions. (e) Additionally, circRNAs may attract certain proteins to a particular location in the cell. (f) CircRNAs can alter gene expression by interacting with RNA polymerase II (RNA Pol II). (g) Some circRNAs can affect the splicing of the host gene and influence the ratio between mRNAs and circRNAs. Abbreviation: Tet1: tet methylcytosine dioxygenase 1; U1 snRNP: U1 small nuclear ribonucleoprotein particle; EIcircRNA: exon-intron circRNAs; ecircRNA: exonic circRNA; ciRNA: circular intronic RNAs; UAP56: spliceosome RNA helicase DDX39B; URH49: ATP-dependent RNA helicase DDX39A.

Interacting with Proteins
CircRNAs may bind with RBPs via RBP-binding sites [50]. RBPs play essential roles in the control of gene expression by regulating splicing, mRNA maturation, transport, localization, translation, and decay [87][88][89]. Various circRNAs have been discovered to interact with RBPs by acting as protein sponges or decoys, thus activating, or disrupting their biological functions (Figure 1b) [90][91][92]. Some circRNAs serve as a protein scaffold to promote the colocalization of two or more proteins, such as enzymes and their substrates, enhancing their activity ( Figure 1c) [74,93]. Some circRNAs can recruit certain proteins to specific cellular locations and spatially modulate their function ( Figure 1e) [94][95][96]. EIciRNAs and ciRNAs, which are located in the nucleus, are involved in transcriptional regulation by promoting RNA Pol II activity, which enhances target gene transcription  Figure 1f) [40,41]. Moreover, it has been revealed that some nuclear retained ecircRNAs influence splicing by exon-skipping variants, consequently affecting the balance between circRNAs and their linear isoforms (Figure 1g) [97].

Translated into Proteins
Linear mRNA translation is usually a cap-dependent process that requires a 7-methylgu anosine (m 7 G) cap at the 5 end and a poly(A) tail at the 3 end [98]. Recent findings suggest that circRNA translation can occur in a cap-independent manner through IRES or m 6 A RNA modification in the 5 untranslated region (UTR) (Figure 1d) [99,100]. CircMblencoded proteins are expressed in small amounts under normal conditions, but following starvation, their expression notably increases [101]. These results suggest that circRNAderived peptides may play a role in maintaining homeostasis.

Anti-Fibrotic circRNAs
TGF-β signaling is regarded as the primary fibrogenic pathway that stimulates HSC activation and ECM synthesis [103]. TGF-β is found in trace amounts in healthy livers [133]. Following liver damage, macrophages start producing TGF-β and PDGF, which can activate excessive ECM production from HSCs and lead to the development of liver fibrosis [134]. Most reported circRNAs involved in HSC activation target miRNAs and proteins in the TGF-β pathway. One of these circRNAs is circPSD3 (mmu_circ_0001682), which is downregulated in primary HSCs and liver tissues from mice with carbon tetrachloride (CCl 4 )induced liver fibrosis [113]. CircPSD3 can function as a miR-92b-3p sponge, consequently promoting Smad7 expression [113]. Smad7 can block the activation of receptor-regulated Smads (R-Smads) and thus inhibits the TGF-β signaling pathway [135,136]. Furthermore, circPSD3 can preclude HSC proliferation and forestall fibrosis progression in vivo [113].
Some circRNAs play a significant role in other liver cell types but can indirectly have an impact on HSC activation. For example, in normal hepatocytes, circBNC2 expression is high but drastically decreased upon liver injury [118]. Interestingly, expression of activation markers (α-SMA and type I collagen) is significantly increased in HSCs incubated with conditional medium from circBNC2 knockout hepatocytes, which appear to contain high levels of the pro-fibrotic cytokines including TGF-β [118]. In contrast, the overexpression of circBNC2 in hepatocytes can reduce expression levels of these cytokines upon liver injury, and conditional medium from these hepatocytes can suppress expression of α-SMA and type I collagen in cultured HSCs, indicating the anti-fibrotic roles of circBNC2 [118]. Mechanistically, circBNC2 contains an open reading frame (ORF) and an IRES, suggesting their function as a protein template [118]. The circBNC2-translated protein (ctBNC2) is a protein product derived from circBNC2 translation [118]. This 681-amino acid protein can bind to CDK1 and cyclin B1 and promote CDK/cyclin complex translocation into the nucleus, a critical step to initiate mitosis and prevent apoptosis [118]. As ctBNC2 levels decrease in a damaged liver, the translocation of the CDK/cyclin complex could be impaired and induce apoptosis, which triggers the secretion of TGF-β and DAMPs from hepatocytes [151,152]. As previously mentioned, these molecules can activate ECM production from HSCs and promote fibrosis progression. Therefore, by protecting hepatocytes from apoptosis upon injury, circBNC2 could play an anti-fibrotic role by suppressing production of inflammatory cytokines [118].
A mitochondrial circRNA called circSCAR is found to be downregulated in NASH patients [130]. Lipid accumulation can induce endoplasmic reticulum (ER) stress, thus increasing the expression of the ER stress mediator CCAAT-enhancer-binding protein homologous protein (CHOP) [130]. CHOP can inhibit the expression of PGC-1α, which positively regulates circSCAR transcription and thus decreases circSCAR levels [130]. CircSCAR can interact with ATP5B, a regulator of the mitochondrial permeability transition pore (mPTP) [130]. Binding between circSCAR and ATP5B hinders the opening of mPTP and consequently the efflux of reactive oxygen species (ROS) from mitochondria into the cytosol [130]. Previous studies suggest that ROS promotes NF-κB signaling pathway and HSC activation [164,165]. Therefore, circSCAR may play an anti-fibrotic role by preventing leakage of ROS from mitochondria [130]. Some circRNAs are found to suppress HSC activation, but their underlying mechanisms are not fully understood. One example is hsa_circ_0004018 or circSMYD4, which is proposed to suppress HSC activation and proliferation via miR-660-3p [132]. By using bioinformatics tools, telomerase-associated protein 1 (TEP1) is predicted to be a target of miR-660-3p and experimentally confirmed by a luciferase assay [132]. While functions of TEP1 in HSCs have not been reported, higher levels of TEP1 in hepatocytes can indirectly suppress the proliferation and activation of HSCs, potentially because TEP1 prevents formation of critically short telomeres and reduces DNA-damage response, which normally triggers hepatocyte senescence and sends apoptotic signals to activate HSCs [132,166]. Intriguingly, although TEP1 play a role in telomere length regulation in many cell types [167,168], a study shows that TEP1 is not essential for telomere length regulation in murine liver [169]. This finding suggests the roles of TEP1 in other cellular functions that could modulate HSC activation [169]. In addition to being a component of the telomerase complex, TEP1 has been found in vault ribonucleoprotein complexes (VRCs) [170]. Although the function of VRCs in HSC activation is not yet known, major vault protein knock-out mice can intensify hepatic steatosis and induce fibrotic responses [171]. Further investigations are needed to expand the knowledge of how the circSMYD4/miR-660-3p/TEP1 axis inhibits HSC activation. Other examples are hsa_circ_0089761 and hsa_circ_0089763, which are two mitochondrial-encoded circRNAs downregulated in HSCs during LPS stimulation [172]. Their downregulation indicates their potential roles as anti-fibrotic circRNAs, but their mechanisms and targets remain to be further studied. Lastly, in a hepatitis B virus (HBV)induced activation model of LX-2 cells, circMTM1 is upregulated along with interleukin 7 receptor (IL7R), potentially via absorbing miR-122-5p [173]. This study shows that cir-cMTM1 knockdown expression and miR-122-5p overexpression reduce the expression levels of activated HSC markers, while the upregulation of IL7R attenuates the anti-fibrotic function of miR-122-5p [173]. Normally, IL7R is expressed in T cells and plays a pivotal role in T cell homeostasis [174]. Despite interesting results, the exact roles of circMTM1 and IL7R in HSC functions need a further investigation.
Interleukin-6 (IL-6) can activate HSC through the MAPK and JAK/STAT signaling pathways [104]. CircCHD2 can enhance hepatic leukemia factor (HLF) expression by interacting with miR-200b-3p [123]. HLF can promote the expression of IL-6 that can bind to the IL-6 receptor and activate the JAK/STAT3 pathway [182]. The JAK/STAT pathway is also a part of a non-SMAD pathway of TGF-β signal transduction that is essential for HSC activation [183]. The JAK/STAT3 pathway also promotes HLF expression, thereby regulating signal transduction in a feed-forward circuit manner [182].
The Hedgehog signaling pathway plays a significant role in HSC activation [109,110]. A recent study shows that circRSF1 sponges miR-146a-5p. Subsequently, this sequestration promotes Ras-related C3 botulinum toxin substrate 1 (RAC1) expression and Hedgehog signal transduction, resulting in HSC activation and proliferation [129]. Rac1 has been shown to be involved in the Hedgehog signaling pathway [190]. Activated Rac1 induces gliomaassociated oncogene (Gli) nuclear translocation, which is necessary for the Hedgehog signaling pathway [191]. Rac1 can also influence NF-κB and JNK signaling by promoting their transduction [192,193]. JNK signal transduction can phosphorylate Smad2 at the C-terminal and linker regions [194]. Initial findings reveal that phosphorylation of Smad2 in the linker region hinders its nuclear translocation and cellular signaling, but recent evidence shows that phosphorylation of the Smad linker can stimulate expression of fibrotic genes [194,195]. One study discovers that the phosphorylated Smad2 linker region is associated with increased expression of glycosaminoglycans (GAG) [196], and hyaluronan (HA), a class of GAG, is found to be able to activate HSCs via Notch1 [197], so the phosphorylation of the Smad2 linker region may be associated with HSC activation by increasing the expression of HA and promoting the Notch1 fibrogenic signaling pathway. However, this proposed mechanism of circRSF1/miR-146a-5p/Rac1 with subsequent signal transductions remains to be further explored.
Moreover, the direct regulation of circRNAs on fibrotic genes have been reported. For instance, the circARID1A/miR-185-3p axis post-transcriptionally regulates expression of COL1A1, a key marker of HSC activation [131]. The pro-fibrotic circARID1A can also promote the proliferation and migration of HSCs as well as inhibit their apoptosis [131]. Mechanistically, increased type I collagen expression can lead to a positive feedback loop of HSC activation in which accumulation of collagen further activates HSCs by increasing ECM stiffness [198]. This mechanical tension in turn leads to YAP activation in the Hippo signaling pathway and Akt activation in the PI3K/Akt signaling pathway [198,199].
The above discussion mainly focuses on the intrinsic expression of circRNAs in HSCs as most in vitro studies rely on the single cell type for their analysis. However, apart from HSCs, liver is a complex organ comprising of several cell types including hepa-tocytes, Kupffer cells, liver sinusoidal endothelial cells, and cholangiocytes, which are known to communicate with one another to maintain liver functions [200]. Some of these cells were reported to modulate liver fibrosis by transferring circRNAs to HSCs such as hepatocyte-derived circBNC2 [118]. Alternatively, circRNAs that regulate the production of inflammatory cytokines in Kupffer cells can have an indirect impact on the state of HSC activation. These circRNAs include circMcph1 [201,202] and circ1639 [203][204][205][206]. Nevertheless, a study of cell-cell interaction in mediating liver fibrosis through circRNAs is still lacking and need to be further investigated.

Conclusions and Future Perspectives
HSC activation is the major contribution to liver fibrosis, which can trigger the development of cirrhosis and ultimately hepatocellular carcinoma (HCC). Molecular mechanisms underlying the activation process are currently being explored in the hope of identifying new therapeutic targets for liver fibrosis. CircRNAs have been reported to regulate HSC activation through modulating signal transduction in fibrogenic signaling pathways. Depending on which miRNAs or proteins they target, circRNAs can be either anti-or pro-fibrotic. Although most discovered circRNAs so far are mechanistically explained as miRNA sponges for their functions, some circRNAs linked to HSC activation can act as protein templates or protein sponges, but only a few studies have reported this mode of action. Given that most circRNAs do not carry multiple binding sites for a single miRNA, the sequestration of miRNAs may not be the primary function of circRNAs [207]. Future studies should explore new regulatory mechanisms of circRNAs other than miRNA sponges in HSC activation. Recent technology based on proximity labeling enzymes in couple with CRISPR/Cas allows the isolation of proteins interacting with specific RNA transcripts [208]. The identification of these proteins could yield novel functions of circRNAs in fibrogenic HSCs.
The number of newly discovered circRNAs has increased over the past years thanks to advances in nucleotide sequencing technologies. Oxford nanopore sequencing combined with circRNA identifiers using long-read sequencing data (CIRI-long) could detect additional splicing events and alternative circularization of circRNAs at a higher efficiency compared with that of Illumina RNA-sequencing methods [209]. This advanced approach might allow the identification of new circRNAs that regulate liver fibrosis. To study roles of the newly discovered circRNAs, loss-of-function and gain-of-function experiments are essential. However, current methods for such genetic manipulation have significant drawback when studying circRNAs in terms of efficiency and specificity. A recent study describes a new technique known as twister-optimized RNA for durable the overexpression (Tornado) expression system, which enables a higher rate of RNA circularization than that of traditional methods [210]. Conversely, utilizing base editors to alter single nucleotide sequences in back-splice sites could impair circRNA biogenesis and reduce its expression while avoiding off-target effects on the cognate linear mRNA [211]. These new techniques could be applied for functional analyses of circRNAs in HSC activation. Lastly, recent studies illustrate cell-to-cell communications via circRNA-containing exosomes from other liver cell types to HSCs, adding another layer of complexity in the modulation of HSC fates [124,125]. More mechanistic understanding of circRNAs in HSC activation could provide new knowledge in biology of non-coding RNA and potentially shed light on development of a new therapeutic strategy for liver fibrosis.  [CrossRef] 28. Eddy, S.R. Noncoding RNA genes. Curr. Opin. Genet. Dev. 1999, 9, 695-