Circular RNAs and Cardiovascular Regeneration

circular RNAs (circRNAs) are a type of non-coding RNAs that are widely present in eukaryotic cells. They have the characteristics of stable structure, high abundance, and cell or tissue specific expression. circRNAs are single-stranded RNAs that are covalently back spliced to form closed circular loops. They may participate in gene expression and regulation through a variety of action modes. circRNAs can encode proteins or function by acting as miRNA sponges for protein translation. Since 2016, a growing number of research studies have shown that circRNAs play important role in the pathogenesis of cardiovascular disease. With the construction of circRNA database, the differential expression of circRNAs in the heart tissue samples from different species and the gradual elucidation of its mode of action in disease may become an ideal diagnosis biomarker and an effective therapeutic target. What can be expected surely has a broader application prospect. In this review, we summarize recent publications on circRNA biogenesis, expression profiles, functions, and the most recent studies of circRNAs in the field of cardiovascular diseases with special emphasis on cardiac regeneration.


INTRODUCTION
Circular RNAs (circRNAs) are single-stranded RNAs that, unlike linear RNA, form a covalently closed continuous loop without 5 ′ end caps or 3 ′ Poly (A) tails. The concept of "circular RNA" was introduced by Sanger et al. when the team found that viroids are single-stranded covalently closed circRNA molecules (1). The cytoplasmic localization of circular RNA in eukaryotic cells was discovered by Hsu et al. through the electron microscope in 1979 (2). These pilot studies established the foundation of this research field.
Transcription of circRNAs had been a mystery for many years. The circular transcription of the Sry gene was discovered in mice in the early 1990s (3). In 2012, Salzman et al. (4) discovered that circRNA is a transformed transcript produced by reverse splicing of mRNA precursor and found that it is abundantly present in different types of human cells. As the field advances rapidly, a large number of circRNAs were discovered with the utilization of high-throughput sequencing technology, and their biological functions were intensively investigated. In 2016, Hansen et al. (5) found that circular RNA can act as a sponge of microRNA (miRNA) to regulate the growth and development of cells. This study shed new light on the circRNA field. Most recently, Li et al. developed a quickly screening and discovering tool for functional circular RNAs based on the CRISPR-Cas13d system, and discovered a set of functionalities that are important for cell growth and embryonic development (6). This technology provided a new research tool to the circRNA field. Cardiovascular diseases (CVDs) are the leading cause of mortality worldwide. Several lines of evidence showed that circRNAs play important roles in regulating cardiovascular function. Jakobi et al. were the first group to provide a comprehensive catalog of RNase R-resistant circRNA species for the adult murine heart and explored the circRNA landscape of heart tissue (7). Over the next years, studies had reported that circRNAs are involved in the regulation of the physiology and pathology of the cardiovascular system. In particular, it is noted that circRNAs are involved in the pathogenesis of CVDs, such as myocardial infarction (MI) (8-14), heart failure (15, 16) and coronary artery disease (CAD) (17-23). Some circRNAs served as potential biomarkers for the diagnosis of CVDs (24-26). These findings suggest that circRNAs may be the new target molecules for the diagnosis and treatment of CVDs. In this review, we summarize circRNA classification, biogenesis, properties, functions, and some new research progress in the field of CVDs.

CLASSIFICATION OF CIRCRNA
circRNAs can be divided into three types according to the different sources of the sequences: ecRNAs (exonic circRNAs) which are derived from single or multiple exons (4,27), ciRNAs (circular intronic RNAs) which are derived from introns (28), EIciRNAs (exon-intron circRNAs) which are composed of exons and introns (29) and tricRNAs (tRNA intronic circRNAs) which are formed by splicing tRNA introns (30).
Lariat-driven circularization (exon skipping) is formed by connecting the splice site of 30 nucleotides upstream of the exon to the site of 50 nucleotides downstream ( Figure 1A). This connection leads to exon-skipping and the formation of an RNA lariat consisting of several exons and introns. The introns are then removed to generate circRNAs (27,31).
Intron-pairing driven circularization (direct back-splicing) is formed when pre-mRNA flanking introns contain inverted complementary sequences ( Figure 1B). The complementary pairing on both sides of the intron can lead to alternative cyclization and then a generation of various circRNAs, including ecircRNAs and EIciRNAs (27,32). Furthermore, longer introns can be found in the flanking sequences of circRNAs, and reverse complementary sequences in longer introns can aid the formation of circRNAs (29,33).
Circular intronic RNAs are produced by eukaryotic spliceosome-mediated splicing ( Figure 1C). The lariat intron generated from the splicing reaction evades normal debranching and degradation, and the 3 ′ "tail" downstream from the branchpoint is trimmed leading to the formation of a stable circRNA. Conserved motifs at both ends, including the 7-nt GU-rich element near the 5 ′ splice site and the 11-nt C-rich element near the branch point site, are combined to prevent introns form circular branches, which promote the formation of loop structures (28,34).
Reverse complementary sequences, such as Alu repeats, are located in upstream and downstream introns. RBP-driven circularization is formed when certain transactivator RNA binding proteins that bind to each flanking intron trigger the splicing of the donor and acceptor sites close enough to form circRNA ( Figure 1D) (32,(34)(35)(36)(37)(38).

FUNCTION OF CIRCRNA
Despite the rapid growth in the field, the biological functions of circRNAs in eukaryotic cells have not been fully understood. circRNAs share some common characters. First, circRNAs are widely distributed and abundantly expressed in a diverse of cells. circRNAs can be found in a large amount in the cytoplasm of eukaryotic cells derived from animals and plants (27,39). In humans, more than 30,000 circRNAs have been discovered and are still increasing year by year (40,41). Second, circRNAs are stable. Due to the covalently closed structure, circRNA is resistant to degradation by ribonuclease (RNase) or exonuclease and is more stable than linear RNA (42). The expression of circRNA differs according to time, tissues, or species (39,43). circRNA profiles change at different stages of cardiac differentiation or during cardiogenic differentiation of induced pluripotent stem cells (44,45). Moreover, circRNAs are evolutionarily conserved (43,46). In 2016, Werfel et al. (47) reported high homology of 1288 circRNAs across humans, mice, and rats. However, many studies have also illustrated that circRNAs are species-specific (48,49). circRNAs show different expression profiles between normal and diseased tissues (45,47,50). Increasing evidence suggests some circRNAs are derived from genomic loci associated with human diseases, and contribute to transcriptional, post-transcriptional, and translational regulations (51). To summarize, there are four main modes for circRNA function (Figure 2). 1) circRNAs can act as competitive endogenous RNAs (ceRNAs) to regulate gene expression by microRNAs (miRNAs) sponge effects (Figure 2A). miRNAs are important post-transcriptional regulators of gene expression that act by direct base pairing to target sites within untranslated regions of messenger RNAs (mRNAs) (52,53). circRNAs contain miRNA response elements (MREs) that promote the binding between circRNAs and miRNAs. This binding can decrease the level of functional miRNAs and increase the expression of miRNA targets (53,54). It has been reported that circRNAs regulate cell function by acting as miRNA sponges. For example, circFOXK2 promotes cell growth, migration, invasion, and apoptosis by binding to multiple sites and functioning as a sponge for miR-942 (55). Similarly, circRNA_100876 regulates the progression of triple-negative breast cancer by functioning as a sponge for miR-136 (56). Other circRNAs, such as circSLC26A4, circRNA_0000253, and circRNA_ANKIB1 can also function  as the sponge of miR-1287-5p (57), miRNA-141-5p (58), and miR-195a-5p (59), respectively. circALMS1_6 may participate in the regulation of cardiac remodeling by functioning as a sponge for miR-133 (60).
2) circRNAs can function as RBP sponges and RBPs can also participate in back-splicing ( Figure 2B) (61)(62)(63)(64)(65)(66). RBPs are a group of proteins involved in gene transcription and translation. circRNAs can interact with RBP and inhibit their activities (66)(67)(68). circMbl absorbs MBL proteins and regulates the subsequent physiological processes (32). circPABPN1 can bind to HuR to suppress the translation of PABPN1 mRNA (69). circANRIL competitively recruits PES1 to inhibit ribosome biogenesis (70). circFoxo3 interacts with different RBPs to participate in the processes of cardiomyocyte senescence and cell cycle progression (71). circAmotl1 can protect cardiomyocytes and promote cell proliferation and wound healing by binding to PDK1, AKT1, and STAT3 (72,73). The studies confirm the involvement of circRNAs in post-transcriptional regulation by chelating RBP. In 2020, Okholm et al. (74) conducted an extensive screen of circRNA-RBP interactions and analyzed circRNA-RBP interactions using a large set of eCLIP data with binding sites of 150 RBPs in the ENCODE cell lines HepG2 and K562 with deep-sequenced total RNA samples. Through this study, they confirmed the interactions between circCDYL and RBPs in bladder cancer cells.
3) circRNAs mediated regulation of the transcription of parental genes ( Figure 2C). For example, circβ-catenin can produce a novel 370-amino acid β-catenin isoform using the start codon as the linear β-catenin mRNA transcript and terminates translation at a new stop codon created by circularization (75). ci-ankrd52 is a circular intronic RNA that is abundant in the nucleus and has little enrichment at microRNA target sites. ci-ankrd52 can bind to the transcription sites and acts as a positive regulator of Pol II transcription (28). For the circRNAs that interact with RNA polymerase II, exons are typically circularized with introns which are 'retained' between exons. These circRNAs are termed exon-intron circRNAs or EIciRNAs. They are mainly localized in the nucleus and interact with U1 snRNP to promote transcription of their parental genes (29). 4) circRNAs can be translated into proteins via some modification ( Figure 2D). As we know, the translation is performed by ribosomes and involves initiation, elongation, termination and ribosome recycling. Base-modification N6methyladenosine (m6A) is a common form of base modification in RNAs. It can promote efficient initiation of protein translation from circRNAs in human cells. Legnini et al. revealed that m6Adriven translation of circRNAs is widespread with hundreds of endogenous circRNAs carrying the translation potential (76). circ-ZNF609 is an example of a protein-coding circRNA in eukaryotes. It is related to heavy polysomes and can be translated into a protein in a splicing-dependent and cap-independent manner (77). circRNAs play biological functions through the formation of complexes with proteins; otherwise, a novel protein circFAM188B-103aa encoded by circFAM188B that promotes the proliferation but inhibits the differentiation of chicken SMSCs was identified (78). Moreover, artificial (79) and endogenous circRNAs containing an internal ribosome entry site (IRES) that directly recruits ribosomes (80) can also be translated into protein. Additionally, circRNA with an infinite ORF has hundred-fold higher productivity than linear transcript by rolling circle amplification in an IRES-independent manner (81). 5) circRNAs may bind, store, sort, and sequester proteins to particular subcellular locations, can and act as dynamic scaffolding molecules that modulate protein-protein interactions ( Figure 2E). circRNAs can bind to RNAs and can also bind, store, sort or sequester selected proteins such as RBPs to modulate their activity or localization (82). RNAbinding protein 3 (RBM3) dynamically adjusts the proliferation of hepatocellular carcinoma cells by regulating the production of SCD-circRNA2 encoded by the 3 ′ -UTR of the stearoyl-CoA desaturase (SCD) gene (83). Recent studies have shown that RBP quaking could also modify the formation of circRNA through forming RNA-protein complexes (RPCs) (36). In addition to interacting with RBPs, circRNAs can function as protein sponges by adsorbing one or more proteins in binding sites, thereby acting as protein scaffolding by the mediating interaction between proteins. For example, CircFOXO3 could mediate the formation of circFOXO3-p21-CDK2 ternary complex and then serve as scaffolding, affecting the cell cycle progression of cancer (84).

SEQUENCING OF CIRCRNA
RNA-seq emerges as a powerful research tool to study the expression and function of non-coding RNAs including circRNAs (85). The technology of circRNA-seq generally includes library construction, computer sequencing, data analysis and processing, and function prediction (86). Either full transcriptome or circRNA profiling may be used to sequence circRNAs. The full transcriptome profiling is aimed to explore the expression patterns of both coding and non-coding RNA. This approach is suitable for the study of the biological function of circRNA. The circRNA profiling is focused on enriched circRNAs and this approach is appropriate to discover unknown circRNAs. Technically, the main difference between the two approaches is the construction of the sequencing library. The circRNA sequencing library not only requires the removal of most rRNA and poly (A), but also requires the use of ribonuclease RNase R to remove the interference of linear RNA. It has been reported that the abundance of circRNAs decreases after delinear RNA because some circRNAs are sensitive to RNase R mediated digestion (87,88). It is worthy to note that the alternative splicing of circRNA requires distinguishing the source of sense and antisense chains in the sequencing results. Therefore, constructing a chain-specific library is ideal as it may improve the accuracy of circRNA sequencing (87). Thus far, more than 100,000 unique human circRNAs have been discovered (89,90).
Microarray chip is another efficient tool for circRNA analysis, and it is commonly used in clinics for disease diagnosis. Compared to RNA-seq, microarray chip analysis is different in the following aspects: (2) microarray analysis of circRNA requires a known reference sequence, while RNA-seq can be utilized to analyze unknown circRNAs; (3) microarray chip analysis can be used to quantify circRNA expression when comparing with RNA-seq (98); and (4) microarray chip analysis can efficiently detect reverse splice site sequences and obtain a larger number of circRNAs than RNA-seq (99). However, some limitations of microarray chip analysis include: (2) high total RNA input is required during sample pretreatment; and (3) unlike full transcriptome sequencing, microarray chip does not give the linear RNA data (100). If the reference sequence is unknown, many studies usually use RNA-seq to determine the full transcriptional sequence, then analyze the circRNAs by microarray.

CIRCRNA AND CVDs
With the development of deep sequencing technology, we can now understand the types and differential expression of circRNAs and their associated miRNAs in cardiovascular tissues (101)(102)(103). Many circRNAs have been reported to be associated with CVDs and their expression pattern are different between healthy and diseased human hearts (15, 19, 104, 105). Quantitative proteomics may be used to discover the regulatory networks of circRNAs in cardiovascular tissues (106). Here, we summarized recent publications on the roles of circRNAs in the development and treatment of CVDs ( Table 1).
1) Cardiac hypertrophy. Cardiac hypertrophy is the heart's response to a variety of extrinsic and intrinsic stimuli that impose increased biomechanical stress and can be caused by various cardiovascular diseases. circRNA wwp1 exerts inhibitory roles of cardiac hypertrophy via down-regulation of ANF and miR-23a in isoproterenol hydrochloride-induced cardiac hypertrophy (107). A circRNA HRCR functions as an endogenous miR-223 sponge to sequester and inhibit miR-223 activity, resulting in an increase of ARC expression and protection of the heart from pathological hypertrophy and heart failure (108). Modulation of circRNAs levels may provide a promising therapeutic target for the treatment of cardiac hypertrophy.
2) Cardiac fibrosis. Activation and phenotypical transition of cardiac fibroblasts contribute to cardiac fibrosis. It was reported that circ_BMP2K enhances the regulatory effects of miR-455-3p on its target gene SUMO1 which leads to the inhibition of TGF-β1 or Ang II to induce the activation of cardiac fibroblasts (109). circRNA_010567, circRNA_000203, and circHIPK3 were upregulated in Angiotensin-II (Ang-II)-induced activation of cardiac fibroblasts (111,112,114). circNFIB was downregulated in TGF-β induced activation of primary adult cardiac fibroblasts (113). Some studies showed that targeting circRNAs improve myocardial ischemic and reperfusion injuries by attenuating myocardial fibrosis. For example, circ_LAS1L is down-regulated in patients with acute myocardial infarction and regulates cardiac fibroblast activation, growth, and migration by inhibiting miR-125b/SFRP5 pathway (12). circPAN3 knockdown attenuated autophagy-mediated cardiac fibrosis after myocardial infarction via miR-221/FoxO3/ATG7 axis (110). The roles of circRNAs in cardiac fibrosis has been summarized in a recent review article (54).
MicroRNA-31-5p acts as a negative regulator of circPAN3 by directly suppressing QKI in doxorubicin-induced apoptosis of cardiomyocytes (117). circPAN3 also ameliorates myocardial ischemia and reperfusion injury by regulating miR-421/Pink1 axis-mediated suppression of autophagy (118). HECTD1 overexpression increases cell viability and decreases cell apoptosis and migration, and circDLPAG4/HECTD1 mediates ischemia/reperfusion injury in endothelial cells via ER stress (119). Down-regulation of circFndc3b was observed in mice with myocardial infarction, and overexpression of circFndc3b increases angiogenic activity and reduces cell apoptosis in cardiac endothelial cells and cardiomyocytes which led to improved left ventricular functions (8). Other studies showed that miR-133 was regulated by circMAT2B. CircMAT2B knockdown attenuates oxygen-glucose deprivation-induced injury through up-regulating miR-133 in H9c2 cells (120). circNFIX can serve as a pro-apoptosis factor in cardiomyocytes (121). The expression of circ ACAP2 is induced by myocardial infarction which leads to increased cardiomyocyte apoptosis by sponging miR-29 (109). Salidroside inhibits apoptosis and autophagy of cardiomyocytes by regulation of circular RNA hsa_circ_0000064 in cardiac ischemia-reperfusion injury (122). These data suggest that circRNAs may be the new targets for designing cardioprotective treatments against cardiomyocyte death. 4) Coronary heart disease. circRNAs are involved in the pathogenesis of atherosclerosis and coronary heart disease. It has been reported that suppression of circDHCR24 alleviates aortic smooth muscle cell proliferation and migration by targeting miR-149-5p/MMP9 axis in human aortic vascular smooth muscle cells after PDGF-BB treatment (20). Recent studies showed that the expression level of circZNF609 in peripheral blood leukocytes of patients with coronary artery disease was significantly decreased, and circZNF609 regulates the release of inflammatory cytokines such as IL-6, IL-10, and TNF-α by serving as sponges to different miRNAs that control the expression of these cytokines (19). circMAP3K5 was downregulated in patients with coronary heart disease and acted as a microRNA-22-3p sponge to promote resolution of intimal hyperplasia via TET2mediated smooth muscle cell differentiation (23). circRNA-100338 may induce angiogenesis after myocardial ischemiareperfusion injury by sponging miR-200a-3p in human coronary endothelial cells (17). The level of hsa_circ_0001445 in plasma was associated with the severity of coronary atherosclerosis. In vitro, hsa_circ_0001445 was downregulated in extracellular vesicles secreted by human coronary smooth muscle cells upon exposure to atherogenic conditions (22). The 3 circRNAs (hsa_circ_0016868, hsa_circ_0001364, hsa_circ_0006731) have been verified by the coronary artery segments Sanger sequencing obtained from an 81-year-old male patient with the sudden death of myocardial infarction (18). In addition, a recent study on the differential expression of circRNAs in plasma samples from patients with coronary heart disease identified 9 circRNAs that promote the expression of transient receptor potential cation channel subfamily M member 3 by inhibiting hsa-miR-130a-3p (21). These data demonstrate that circRNAs play important roles in the pathogenesis of coronary heart disease and may serve as diagnostic or therapeutic targets for coronary heart disease.
Frontiers in Cardiovascular Medicine | www.frontiersin.org 5) Heart failure. Despite the detailed roles of circRNAs in the progression of human heart failure remains elusive, recent highthroughput sequencing studies identified many circRNAs with changed expression in patients with heart failure. The top highly expressed EAT circRNAs corresponded to genes involved in cell proliferation and inflammatory responses. A recent study on circRNA expression profile in epicardial adipose tissue in patients with heart failure showed that EAT circRNAs may contribute to the pathogenesis of metabolic disorders (16). Another study showed that the upregulation of Hsa_circ_0097435 contributes to the pathogenesis of heart failure via sponging multiple microRNAs (15).

6) Cardiac regeneration and repair.
Recent studies indicated that modulating circRNA activity may have great therapeutic potential for myocardial regeneration and repair. Up-or downregulation of circRNAs and miRNAs and circRNA-miRNA coexpression had been shown to change the expression of the genes associated with myocardial ischemia and reperfusion injuries (101,103). It has been reported that circRNAs can regulate inflammatory factors to improve myocardial ischemia and reperfusion injury. The circTLK1 exacerbates myocardial ischemia and reperfusion injury via targeting miR-214/RIPK1 through TNF signaling pathway (11). circ003593 has also been shown to confer cardioprotection through NLRP3 inflammasome myocardial infarct rats (13). circRNA CDR1as was identified in pig hearts. Elevated circRNA CDR1as in the infarction region of the pig heart is negatively associated with infarct size and positively associated with improved heart function (10). circ-0001273 can remarkably inhibit myocardial cell apoptosis and promote repair in myocardial infarction hearts (9). circCDYL was downregulated in myocardial tissues and hypoxia myocardial cells after acute myocardial infarction. circCDYL overexpression and downregulation can promote and inhibit the proliferation of cardiomyocytes in vitro, respectively. Additionally, circCDYL can promote the proliferation of cardiomyocytes through the miR-4793-5p/APP pathway (14). The downregulation of circFASTKD1 induces angiogenesis and improves cardiac function and repair after myocardial infarction (123). Hsa_circ_0007623 can bind to miR-297 and acts as a sponge of microRNA-297 which promotes cardiac repair after acute myocardial ischemia and protects cardiac function (124). circHipk3 was overexpressed in the fetal or neonatal heart of mice and functioned to promote the proliferation of cardiomyocyte and endothelial cells which leads to angiogenesis. Further study showed that circHipk3 regulates cardiac regeneration in mice post myocardial infarction by interacting with Notch1 and miR-133a (104). These findings highlight the physiological role of circRNAs in cardiac repair and indicate that modulation of circRNA may represent a potential strategy to promote cardiac function and remodeling after myocardial injuries.

FUTURE PERSPECTIVES
The studies we discussed in the paper highlight the significance of circRNAs in the pathogenesis of CVDs. circRNAs are stable and abundantly present in the circulatory system which enables them to serve as biomarkers for the diagnosis and treatment of CVDs; however, there are some critical issues to be addressed. Firstly, there is no reliable methodology for the detection of circRNAs. In terms of circRNA detection, newer, simpler, and more reliable methods will be expected to appear. This will provide convenience for us to study circRNAs, facilitate the faster output of research results, and obtain more target circRNAs with diagnostic and therapeutic significance. Secondly, some circRNA biomarkers come from small samples and populations. This makes us question the reliability and representativeness of the research results. Thirdly, the mechanism underlying circRNA functions in the cardiovascular system remain largely elusive. For the function and mechanism studies, the current research methods are limited and difficult to operate. New research protocols need to be further explored. Rapid development can be achieved only by breaking through the technological bottleneck in the field of circRNA research. Lastly, there is a lack of efficient approaches for modulating circRNA expression in the cardiovascular system. It is supposed to be that in the future there will be new and more diverse methods in modulating the overexpression and inhibition of circRNAs. This will promote the development of the mechanism underlying circRNA functions. Then the research results can be quickly used for clinical diagnosis and treatment in the field of vascular diseases.

AUTHOR CONTRIBUTIONS
All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

FUNDING
This work is supported by NIH R01 grant HL142627 and American Heart Association Grant 20TPA35490001.