Loss of Super-Enhancer-Regulated circRNA Nfix Induces Cardiac Regeneration After Myocardial Infarction in Adult Mice

1Silencing circRNA (circular RNA) Nfix promotes cardiomyocyte proliferation and angiogenesis by suppressing Ybx1 (Y-box binding protein 1) ubiquitin-dependent degradation and rescuing miR-214.

2Superenhancers mark abundant, tissue-specific, and developmental stage-specific circRNAs.

3circNfix-associated superenhancer regulates circNfix expression by recruiting transcriptional factor Meis1 (Meis homeobox 1).

4The findings that circRNA Nfix negatively regulates cardiac regeneration provides a novel therapeutic target to improve the prognosis after myocardial infarction.

5Understanding the role of superenhancers on controlling mouse cardiac regeneration may uncover the epigenetic modifications restricting the regenerative ability of adult human heart, enabling the development of potential therapeutic avenues.

6Activation of cardiomyocyte proliferation in situ is a promising approach for replacing lost cardiomyocytes after injury or disease, such as myocardial infarction (MI) and heart failure. ncRNAs (noncoding RNAs) play vital roles in cell-cycle regulation, and incorporating them is a powerful strategy for promoting cardiomyocyte proliferation., circRNAs (circular RNAs) are a novel class of ncRNAs that are circularized by joining the 3′ end of the RNA to the 5′ end. The circular structure maintains their stability and strengthens their miRNA (microRNA)/protein-binding capacity., Thus, circRNAs may be more effective than noncircular RNAs in regulating the cardiac gene regulator network and inducing cardiomyocyte proliferation. Moreover, emerging evidence indicates that circRNAs may play fundamental roles in the regulation of cardiac regeneration. Certain kinds of circRNAs have been found to control the regulation of cell proliferation, differentiation, development, and even tissue regeneration. circRNAs also participate in biological processes and pathological conditions of the cardiovascular system, including the development of cardiomyopathy, cardiomyocyte hypertrophy, senescence, and apoptosis.– In addition, high-throughput RNA sequencing has recently revealed circRNAs that are frequently differentially expressed between neonatal and adult heart, and these deregulated circRNAs are likely involved in heart regeneration. Therefore, targeting circRNAs might be an attractive strategy for promoting a cardiac regenerative response because of their important roles in cell proliferation and cardiomyocyte function. However, the role of circRNAs in cardiac regeneration has not been previously investigated.

7Identification and validation of the key circRNA function in the acquisition of regenerative capability is also critical for seeking possible therapeutic targets. Recently, circRNAs that are differentially expressed between neonatal and adult hearts have been uncovered by RNA sequencing, but further recognizing the subgroups essential for cardiac regeneration from a fairly large number of candidates is difficult. Superenhancers are large clusters of active enhancers that are enriched for the binding of key master transcription factors. These clusters of enhancers are developmental and cell-type specific and act as causal mechanisms of cell differentiation, specification, and development by controlling cell-identity genes or ncRNA expression., Because superenhancers are frequently identified near genes or ncRNAs that are important for controlling cell identity and differentiation, such superenhancers may be used to quickly identify key nodes regulating cell differentiation and development.,, Importantly, superenhancer-associated ncRNAs have been found to drive cardiogenesis and cardiomyocyte differentiation,, both of which are key links during cardiomyocyte proliferation. Here, we postulated that superenhancers are likely to mark circRNAs that control cardiomyocyte differentiation and development, thereby helping to identify key circRNAs regulating cardiac regeneration.

8In line with this hypothesis, we first performed a comprehensive analysis of the relationship between superenhancers and circRNA networks in hearts. We further identified circRNA Nfix (circNfix) as a cardiomyocyte-enriched circRNA that was regulated by the binding of Meis1 (Meis homeobox 1) to its associated superenhancer in the adult heart. Of crucial importance, downregulation of circNfix promoted adult cardiomyocyte proliferation and cardiac regeneration by rescuing Ybx1 (Y-box binding protein 1) and miR-214 (microRNA-214), significantly decreasing the fibrotic area and promoting functional recovery after MI.

9The data, methods and materials related to this study are available to other researchers on reasonable request.

10C57BL/6 mice and SD rats were purchased from Guangdong Medical Laboratory Animal Center. The Cas9 knock-in mouse model was obtained from Shanghai Model Organisms Center, Inc. Mouse MI was performed as described. All animal experiments were approved by the Animal Research Committee of Southern Medical University and performed in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals.

11Neonatal cardiomyocytes were isolated from 1 day-old (P1) and 7-day-old (P7) C57BL/6 mice or P1 SD rats, and adult cardiomyocytes were isolated from ≈8 week-old adult mouse as previously described.

12Total RNA from cell or tissue lysates was isolated using TRIzol reagent (R6830-01E.Z.N.A, OMEGA). Quantitative reverse transcription–polymerase chain reaction (QRT-PCR) was performed by using SYBR Green PCR Master Mix (Takara, Dalian, China) in a LightCycler 480 System (Roche, Germany). The primers are listed in Table I in the online-only Data Supplement.

13After collection, hearts were fixed with 10% formalin or embedded in optimum cutting temperature embedding medium (Sakura) as appropriate. Immunohistochemical and immunofluorescent analyses were performed as previously described.

14Results were analyzed with SPSS 18.0 software. All data are presented as the mean±SD. Unless mentioned otherwise, after normality testing using the Shapiro-Wilk test, Student t test (2-group comparison) or 1-way/2-way repeated-measurements ANOVA, followed by Bonferroni post hoc testing (multi-group comparison), were used to examine the difference as appropriate. The survival rate was determined using the Kaplan-Meier method. The difference between survival curves was determined using the log-rank (Mantel-Cox) test. A value of P<0.05 was considered statistically significant.

15More detailed materials and methods used in this study are provided in the Methods in the online-only Data Supplement.

16Using an enhancer catalog generated in 23 human tissues, defined by the presence of H3K27ac (acetylation of histone H3 lysine 27) marks, we assigned enhancers to the most proximal promoters of circRNAs. A variety of circRNAs were spatially associated with superenhancers and typical enhancers (TEs), which were named superenhancer-associated circRNAs (SE-circRNAs) and typical enhancer–associated circRNAs (TE-circRNAs), respectively. We characterized SE-circRNAs in diverse human tissues, including heart, lung, brain, colon and gastric tissue. SE-circRNAs showed higher expression levels than TE-circRNAs or other circRNAs in these tissues (Figure IA in the online-only Data Supplement). Moreover, SE-circRNAs were more likely to be tissue-type specific than TE-circRNAs (Figure IB and IC in the online-only Data Supplement), and tissue-specific superenhancers are likely to be associated with tissue-specific circRNAs (Figure ID in the online-only Data Supplement).

17We next investigated the role of SE-circRNAs and their conservation at different stages of development. The catalog of H3K27Ac superenhancers in 4 mouse organs covering the embryonic and adult stages revealed widespread changes in superenhancer distribution across different developmental stages (Figure IE in the online-only Data Supplement). A gene ontology analysis of proximal genes around embryonic-specific superenhancers in the heart showed that they are related to biological processes characteristic of cardiogenesis and angiogenesis, whereas adult-specific superenhancers are linked to the downregulation of signaling pathways essential for cardiogenesis (Figure IF in the online-only Data Supplement). These results suggested that superenhancers might be critical drivers or inhibitors of cardiac regeneration.

18A previous study identified circRNAs that are differentially expressed between adult and neonatal (P1) rat hearts. Using this transcriptomic data set, we filtered circRNAs for those that are conserved among humans, rats, and mice (Figure IIA in the online-only Data Supplement). We mapped these murine orthologs to stage-specific heart superenhancers to identify SE-circRNAs (Figures IA and IIA in the online-only Data Supplement). circRNA_Nfix (circNfix) and circRNA_Slc8a1 (circSlc8al) were found to be the predominant SE-circRNAs that are differentially expressed in neonatal and adult hearts (Figure 1A). circNfix was more enriched in cardiomyocytes than was circSlc8a1 (Figure 1B). Moreover, Nfix mRNA expression levels tended to be decreased in adult mouse hearts and cardiomyocytes compared with neonatal mouse hearts and cardiomyocytes, respectively (Figure IIIA in the online-only Data Supplement). The inverse trend for circular and linear transcript expression might indicate an important biological function of circRNA. Next, we focused on circNfix for detailed expression profiling and functional characterization.

19Divergent primers amplified circNfix in cDNA but not in genomic DNA, indicating that this RNA species is circular in form (Figure 1C). The results of the Sanger sequencing analysis of PCR products were in accordance with the circNfix sequence shown in circBase (a database for circRNAs) where the circRNA ID numbers were hsa_circ_0005660 and mmu_circ_0001704. The circNfix nucleotide sequence is strongly conserved, with >95% homology among humans, rats, and mice (Figure 1C; Figure IVA and IVB in the online-only Data Supplement). We then demonstrated that circNfix had greater stability than the linear transcript (Figure 1E and 1F).

20QRT-PCR assays showed that circNfix was expressed in several tissues in adult mice and was enriched in the heart (Figure 1G). Moreover, we detected circNfix expression in several cardiac cell types isolated from adult mouse hearts. CircNfix was highly expressed in cardiomyocytes compared with other cardiac cell types (Figure 1H). Importantly, circNfix expression was significantly increased in cardiomyocytes during heart development (Figure 1L through 1M). Furthermore, MI in P0 mice, which induces robust cardiomyocyte proliferation at day 7 after MI, was associated with significant downregulation of circNfix in the border zone (Figure 1K, 1M; Figure VA in the online-only Data Supplement). By contrast, circNfix expression was modestly increased following MI at P56, a time point at which cardiomyocytes have a limited proliferation capacity (Figure 1K, 1M; Figure VA in the online-only Data Supplement). In addition, circNfix was observed to be located mainly in the cytoplasm of P7 cardiomyocytes (Figure 1N, 1O).

21We used siRNAs (small interfering RNAs) targeting back-splicing of circNfix to specifically knock down circNfix expression in vitro (Figures IIIA through IIID, Table II in the online-only Data Supplement). Notably, knockdown of circNfix significantly increased the percentage of P7 cardiomyocytes expressing cell-cycle activity markers (Figure 2A through 2D; Figure VIA in the online-only Data Supplement). The time-lapse imaging of P7 cardiomyocytes labeled with the fluorescent mitochondrial dye tetramethylrhodamine ethyl ester showed that P7 cardiomyocytes in the si-circNfix group underwent cell division, whereas no cardiomyocyte in the control group was observed to go through cell division (Figure 2E; Movie I and Movie II in the online-only Data Supplement). Moreover, we investigated the effects of circNfix overexpression on P0 cardiomyocyte proliferation. Overexpression of circNfix was found to reduce the ratio of P0 cardiomyocytes expressing proliferation markers (Figure 2F through 2H). We observed that silencing circNfix had no effect on P7 cardiomyocyte proliferation (Figure IIIB through IIIE in the online-only Data Supplement).

22Moreover, we observed more rounded cells, with disorganization and a reduced number of sarcomere structures in P7 cardiomyocytes when circNfix was knocked down (Figure VIB in the online-only Data Supplement). This observation was in accordance with an increase in dedifferentiation markers’ expression (Figure 2I). Flow cytometry assays further revealed that more P7 cardiomyocytes were accumulated at the synthesis and gap 2/mitotic phases of the cell cycle. (Figure 2J). In addition, the number of apoptotic cardiomyocytes after H2O2 treatment was significantly reduced by circNfix silencing, whereas circNfix overexpression increased the number, suggesting that circNfix expression might be associated with cardiomyocyte survival (Figure VIIA and VIIB in the online-only Data Supplement).

23To assess the in vivo effect of circNfix downregulation, we delivered AAV9-shRNA (adeno-associated virus-9 short hairpin RNA; circNfix) to adult mice and detected the proliferation of cardiomyocytes. The transfection efficiency of AAV9-shcircNfix-GFP (green fluorescent protein) was ≈80% by GFP and cardiomyocyte marker costaining, and circNfix expression was significantly reduced (Figure VIIIA through VIIIF in the online-only Data Supplement). We also demonstrated the specificity of AAV9-mediated circNfix knockdown for cardiomyocytes (Figure IXA through IXF in the online-only Data Supplement). In addition, knockdown of circNfix was observed to markedly elevate the ratio of proliferative cardiomyocytes and increase the total cardiomyocyte number (Figure IIIA through IIID, Figure XA in the online-only Data Supplement). Moreover, QRT-PCR assays showed that circNfix downregulation significantly increased the expression of the dedifferentiation markers in vivo (Figure XIA in the online-only Data Supplement). Western blotting and immunofluorescence analysis further confirmed that circNfix knockdown increased RUNX1 (runt-related transcription factor 1) expression in adult hearts (Figure 3E; Figure XIB in the online-only Data Supplement). Additionally, circNfix downregulation resulted in transient sarcomere disassembly and reduction in vivo (Figure XIC in the online-only Data Supplement). Disrupted cardiomyocyte adhesion, suggested by the disordered arrangement of cardiomyocytes and enlarged intracellular space, further confirmed the transient dedifferentiation of adult cardiomyocytes (Figure XID in the online-only Data Supplement). All the above results indicated that loss of circNfix was able to induce cardiomyocyte dedifferentiation and proliferation followed by redifferentiation.

24Moreover, we isolated cardiomyocytes from adult mouse hearts injected with AAV9-shcircNfix or AAV9-shNC. Ex vivo studies showed that total cardiomyocyte number and the percentages of pH3+ (phosphorylated histone H3 positive), Ki67+ (marker of proliferation Ki-67 positive), and mononuclear cardiomyocytes were significantly higher in the AAV9-shcircNfix group than in the AAV9-shNC group (Figure 3K, 3L; Figure XIIA in the online-only Data Supplement). The expression of dedifferentiation markers was also significantly higher in the AAV9-shcircNfix group than in the AAV9-shNC group (Figure XIIA through XIIE in the online-only Data Supplement). We also found that circNfix loss resulted in transient sarcomere disassembly and reduction in isolated adult cardiomyocytes (Figure XIIB through XIIE in the online-only Data Supplement). Additionally, downregulation of circNfix had no effect on cardiac fibroblast proliferation (Figure XIIIA in the online-only Data Supplement).

25CRISPR/Cas9 technology was used to knock out circNfix within cardiomyocytes both in vitro and in vivo. We successfully constructed a circNfix knockdown cell line using HL-1, a mouse cardiomyocyte cell line, with the CRISPR/Cas9 method as previously described (Figure XIVA and XIVB,; Table III in the online-only Data Supplement). Deletion of circNfix was observed to increase the proliferation of HL-1 cardiomyocytes (Figure XIVC and XIVD in the online-only Data Supplement). Moreover, we successfully decreased circNfix expression in vivo using CRISPR/Cas9 knock-in mice, inducing a significant increase in cardiomyocyte proliferation in these circNfix-deficient mice (Figure XVA through XVK, Table III in the online-only Data Supplement).

26Next we studied whether knockdown of circNfix could elevate myocardial repair and preserve cardiac function after MI. Knockdown of circNfix was performed in an MI model by injection of AAV9-shcircNfix or AAV9-shNC in the peri-infarcted area. As expected, knockdown of circNfix markedly increased the ratio of proliferative cardiomyocytes in the infarct border zone at 14 days post-MI (Figure 3F through 3H; Figures XB and XVIA in the online-only Data Supplement).

27We observed a 2.4-fold decrease in apoptotic cardiomyocytes in the AAV9-shcircNfix group at 7 days post-MI compared with the control group (Figure 3I). We found circNfix knockdown increased IB4− (isolectin B4 negative), CD31− (cluster of differentiation 31 negative), and vWF (von Willebrand Factor)-stained vessel density in the infarcted or border area (Figure 3I; Figure XVIIA through XVIID, and XVIIF in the online-only Data Supplement). We also observed a markedly increased proliferative rate of cardiac endothelial cells after circNfix knockdown in vivo, using cTnT (cardiac troponin)/CD31/Ki67 triple-staining on heart sections (Figure XVIIE and XVIIG in the online-only Data Supplement). Immunofluorescence analysis of α-SMA (α-smooth muscle actin) and CD31 revealed that arteriolar density was significantly greater in the AAV9-shcircNfix group than in the AAV9-shNC group (Figure XVIIIA and XVIIIB in the online-only Data Supplement). These results suggested increased angiogenesis during circNfix knockdown, inducing cardiac regenerative processes.

28We further explored the effect of circNfix silencing on functional recovery after MI. Echocardiography revealed that ejection fraction and fractional shortening of cardiac functions were significantly preserved in the AAV9-shcircNfix groups compared with the control group (Figure 4A through 4C). Positron emission tomography–computed tomography, TTC (triphenyltetrazolium chloride) staining, and Masson trichrome staining revealed that the infarcted area and cardiac fibrosis were significantly reduced in the AAV9-shcircNfix group compared with the control group (Figure 4D through 4G). These responses are associated with a significant increase in survival rate after MI in the circNfix knockdown group compared with that in the control group (Figure 4H).

29To determine whether forced circNfix overexpression impairs neonatal cardiomyocyte proliferation and cardiac regeneration, we injected adenovirus-expressing circNfix or control vector into P1 mouse hearts. We observed a high transfection efficiency and specificity of adenovirus -circNfix for cardiomyocytes (Figures XIXA through XIXE and XXA through XXI in the online-only Data Supplement). Furthermore, circNfix overexpression significantly reduced the proportion of proliferative cardiomyocytes and the total cardiomyocyte number (Figure 5A through 5I). In addition, the percentage of TUNEL+ cardiomyocytes in the infarct border zone was significantly higher in mice overexpressing circNfix than in the control mice (Figure 5F). IB4, α-SMA, and CD31 immunostaining showed that circNfix overexpression inhibited angiogenesis after P1 mouse MI (Figure 5G and 5H; Figure XXIA in the online-only Data Supplement). Accordingly, echocardiographic ejection fraction at 7 days after MI was significantly lowered in the adenovirus -circNfix group compared with the control group (Figure 5L and 5M). Histological analysis showed a significant increase in the infarct size in circNfix-overexpressing mice compared with that in the control mice (Figure 5J and 5K). Collectively, these findings demonstrated that repression of cardiomyocyte proliferation by circNfix overexpression could impair cardiac regenerative repair in a neonatal mouse model.

30We also assessed the effect of circNfix on the regulation of cardiomyocyte hypertrophy. Neither circNfix silencing nor circNfix overexpression was found to significantly alter cardiomyocyte size in vivo or cardiomyocyte volume ex vivo (Figure XXIIA through XXIID in the online-only Data Supplement).

31The cardiac superenhancer at the Nfix locus is present in human and mouse hearts (Figure 6A and 6B). We validated that this superenhancer region is enriched in other superenhancer marks, including H3K4me1 (monomethylation of histone H3 at lysine 4), H3K27me3 (trymethylization of histone H3 at lysine 27), P300, and DNase hypersensitive site, using ChIP-sequence (ChIP-seq) data (Figure 6A and 6B). Consistent with the increase in circNfix expression in adult hearts, active histone marks at the Nfix locus were markedly increased, whereas inactive histone marks were reduced in adult hearts (Figure 6A and 6B). We also found that the superenhancer from which circNfix is derived is uniquely active in adult heart tissue compared with other tissues (Figure XXIIIA and XXIIIB in the online-only Data Supplement). Importantly, the enhancer H3K27ac histone marks were also noticeably increased in adult cardiomyocytes compared with embryonic cardiomyocytes (Figure 6B). Superenhancers often recruit transcription factors to fulfill their enhancer activity. We therefore used ChIP-seq to map the genome-wide occupancy of Meis1, a transcription factor controlling the cell-cycle arrest of adult cardiomyocytes, inside the circNfix-associated superenhancer. The ChIP-seq profile of Meis1 revealed a prominent peak that overlapped with circNfix-SE (Figure 6B). Moreover, a Meis1 motif was observed in the circNfix-associated superenhancer sequence, which is highly conserved among species (Figure 6B). Chromosome 3C assays showed chromatin looping between the circNfix promoter and superenhancer, fully supporting the superenhancer as a regulatory element for circNfix (Figure 6C; Figure XXIIIC and Table IV in the online-only Data Supplement). We next cloned individual constituent enhancers (E1, E2, E3, E4) of circNfix-SE into enhancer-reporter vectors and measured their activities by using a luciferase reporter system. E2 was shown to be strongly active in cardiomyocytes (Figure 6D). Using luciferase, electrophoretic mobility shift assay and ChIP–quantitative PCR assays, we confirmed that Meis1 bound to E2 in P7 cardiomyocytes (Figure 6F and 6G). Moreover, Meis1 silencing significantly reduced circNfix expression, although it increased Nfix mRNA abundance (Figure 6H and 6I). We further observed that Meis1 silencing increased the percentage of P7 cardiomyocytes exhibiting proliferation kinetics indexes, and these effects could be abrogated by circNfix overexpression (Figure 6J).

32We used CRISPR-Cas9 knock-in mice to generate loss-of-function mutations in the Meis1 gene as previously described (Figure XXIVA through XXIVD, Table III in the online-only Data Supplement)., These Meis1 mutations significantly decreased circNfix expression and increased cardiomyocyte proliferation in vivo, which could be reversed by AAV9-mediated circNfix overexpression (Figure XXVA through XXVG in the online-only Data Supplement). We also demonstrated that circNfix knockdown could reverse the inhibitory effect of Meis1 on cardiomyocyte proliferation (Figure XXVIA through XXVIF in the online-only Data Supplement). Additionally, we failed to observe any synergistic effect between Meis1 and circNfix knockdown (Figures XXVA through XXVG, and XXVIIA through XXVIID in the online-only Data Supplement). The above results indicated that circNfix expression is tightly regulated by the interaction between Meis1 and circNfix_SE.

33To better explore the function of circNfix in cardiomyocytes, we performed RNA pulldown assays of circNfix using biotinylated probes targeting the circNfix back-spliced sequence (Figure 7A; Figure XXVIIIA in the online-only Data Supplement), and the precipitates were subjected to mass spectrometry analysis. Several RNA-binding proteins were identified by mass spectrometry analysis, and the most abundant among them was Ybx1 (Table V in the online-only Data Supplement). Bioinformatics analysis revealed that Ybx1 and circNfix are likely to interact (Figure XXVIIIB in the online-only Data Supplement), and several Ybx1 motifs were observed in the circNfix sequence (Figure XXVIIIC in the online-only Data Supplement). We then confirmed the interaction between circNfix and Ybx1 using RNA pulldown followed by Western blotting and radioimmune precipitation (Figure 7A and 7B).

34We next investigated the molecular consequence of the interaction between circNfix and Ybx1. CircNfix knockdown significantly decreased Ybx1 protein expression in cardiomyocytes, which could be attenuated by proteasome inhibitor MG132 (Figure 7C and 7D,; Figure XXVIIIE in the online-only Data Supplement). Ybx1 mRNA expression levels were not altered when circNfix was overexpressed (Figure 7E). Moreover, circNfix inhibited nuclear translocation of Ybx1 in P0 cardiomyocytes (Figure 7F and 7G).

35Nuclear Ybx1 has been reported to bind to the promoters of Ccna2 (cyclin A2) and Ccnb1 (cyclin B1) and activate their transcription. We also found that Ybx1 knockdown could markedly decrease Ccna2 and Ccnb1 expression, which could abrogate circNfix knockdown-induced overexpression (Figure 7G). Subsequent ChIP–quantitative PCR assay confirmed the interactions of Ybx1 with the promoters of Ccna2 and Ccnb1 (Figure 7I). Moreover, Ybx1 overexpression significantly increased cardiomyocyte proliferation (Figure XXVIIID in the online-only Data Supplement), whereas Ybx1 silencing could block circNfix knockdown-induced cardiomyocyte proliferation (Figure 8N; Figure XXIXA through XXIXC in the online-only Data Supplement).

36Furthermore, circNfix knockdown and overexpression could reduce and increase the ubiquitination of Ybx1 in cardiomyocytes, respectively (Figure 7J). Moreover, Nedd4l, an E3 ubiquitin ligase that interacts with its substrates for ubiquitination, was also among the proteins identified by mass spectrometry analysis following circNfix RNA pulldown (Table VI in the online-only Data Supplement). Notably, a protein docking program predicted that Ybx1 and Nedd4l were likely to interact (Figure XXVIIIF in the online-only Data Supplement), and coimmunoprecipitation assays confirmed the interaction between Ybx1 and Nedd4l (Figure 7J). Next, we observed the association of circNfix with Nedd4l by using RNA pulldown and radioimmune precipitation assays (Figure 7A and 7B). The altered expression of circNfix did not affect Nedd4l expression (Figure XXVIIIG in the online-only Data Supplement) but significantly affected the interaction of Ybx1 with Nedd4l (Figure 7J). We further confirmed that silencing of Nedd4l increased Ybx1 protein stability by decreasing ubiquitination, which could abrogate the effect of circNfix overexpression (Figure 7K and 7L). Collectively, these results indicated that the inhibitory effect of circNfix on cardiomyocyte proliferation may be achieved by enhancing Ybx1 ubiquitin-dependent degradation.

37We then explored whether circNfix functions as a miRNA sponge to regulate gene expression. Bioinformatics prediction analysis using the Segal Laboratory program showed that several miRNAs may interact with circNfix. Among them, miR-214/761 has the most binding sites on circNfix, reaching 3 conserved binding sites (Figure 8A; Figure XXXA in the online-only Data Supplement). QRT-PCR assays showed that miR-214 had much higher expression levels in the myocardium than miR-761 had (Figure 8B). CircNfix downregulation significantly increased miR-214 gene expression levels (Figure 8B). The results of luciferase, RNA pulldown, and RNA fluorescense in situ hybridization assays confirmed the binding of miR-214 on circNfix (Figure 8C through 8E).

38Next, we explored the effects of miR-214 mimic transfection on several candidates that have been associated with cardiac regeneration (Figure XXXB in the online-only Data Supplement). QRT-PCR revealed that the miR-214 mimic significantly decreased Gsk3β (glycogen synthase kinase 3 β) expression (Figure XXXB in the online-only Data Supplement). Furthermore, the binding site of miR-214 in the 3′UTR of Gsk3β is highly conserved across species (Figure XXXC in the online-only Data Supplement), suggesting Gsk3β as a common target gene of miR-214. Consistently, RNA fluorescense in situ hybridization experiments in cardiomyocytes showed overlap between miR-214 and Gsk3β 3′UTR (Figure 8F). The 3′UTR of the Gsk3β gene was also cloned into a luciferase vector and transfected into cardiomyocytes. Transfection with miR-214 mimic significantly reduced luciferase activity, although mutation of the miR-214 binding site eliminated this reduction (Figure 8G; Figure XXXC in the online-only Data Supplement). In addition, circNfix overexpression increased luciferase activity, which could by abrogated by miR-214 mimic transfection (Figure 8G). Similarly, Gsk3β expression was significantly decreased in hearts transfected with AAV9-shcircNfix compared with that in hearts transfected with AAV9-shNC (Figure 8H). Gsk3β expression was significantly reduced by miR-214 mimics but increased by miR-214 inhibitors (Figure 8I). A miR-214 inhibitor abolished the repressive effect of circNfix downregulation on Gsk3β (Figure 8J). Gsk3β is known to regulate cardiomyocyte proliferation via degrading β-catenin. Consistently, we also found that β-catenin expression was significantly reduced in cardiomyocytes transfected with siRNA-circNfix (Figure 8K). In addition, we did not find an obvious interaction between Ybx1 and miR-214 (Figure XXXD and XXXE in the online-only Data Supplement). Moreover, Meis1 knockdown significantly repressed Gsk3β expression (Figure 8L), which was reversed by circNfix interference (Figure 8M), suggesting that the Meis1-circNfix-miR-214-Gsk3β regulatory axis is active in cardiomyocytes. Then, we investigated whether such a regulatory axis was involved in regulating cardiomyocyte proliferation. As expected, the miR-214 mimics significantly increased the percentage of proliferating cardiomyocytes (Figure XXXF in the online-only Data Supplement). Additionally, miR-214 inhibition counteracted the effect of circNfix downregulation on cardiomyocyte proliferation, which was reversed by Gsk3β knockdown (Figure 8N; Figure XXIXA through XXIXC in the online-only Data Supplement). Furthermore, the inhibitory effect of losing either Ybx1 or miR-214 is similar to the loss of both, abolishing circNfix-induced cardiomyocyte proliferation (Figure 8N; Figure XXIXA through XXIXC in the online-only Data Supplement). Our results demonstrated that circNfix requires both Ybx1 and miR-214 to regulate cardiomyocyte proliferation.

39Considering the important role of Gsk3β on angiogenesis, we also detected VEGF (vascular endothelial growth factor) secretion via ELISA after altering this regulatory axis. Inhibiting miR-214 was found to abolish the facilitating effect of circNfix knockdown on VEGF secretion, which could be restored by Gsk3β downregulation (Figure XXXIA in the online-only Data Supplement). To clarify the link between circNfix, Gsk3β, and angiogenesis, we performed rescue experiments in adult and neonatal mice after MI. Coinjection of AAV9-Gsk3β and AAV9-shcircNfix substantially abolished the angiogenic effects of circNfix downregulation in adult mice after MI (Figure XXXIB through XXXID in the online-only Data Supplement). Moreover, the Gsk3β inhibitor SB216763 was used to perform rescue experiments in neonatal mice after MI as previously described. Interestingly, SB216763 treatments significantly restored the IB4+ capillary density in the peri-infarcted zone of circNfix-overexpressing mice (Figure XXXIE in the online-only Data Supplement). These results indicated Gsk3β as the main downstream effector mediating the function of circNfix in angiogenesis.

40In the current study, we identified a cardiomyocyte-specific circRNA, circNfix, whose expression level is controlled by the interaction of its proximal superenhancer and Meis1. Downregulation of circNfix promotes cardiac regenerative repair and functional recovery after MI. Mechanistically, circNfix blocks cardiomyocyte proliferation by promoting Ybx1 ubiquitination degradation and acting as a miR-214 sponge.

41In this study, we demonstrated that circRNAs are involved in cardiac regeneration, as superenhancer-associated circNfix negatively regulates cardiomyocyte proliferation both in vitro and in vivo by affecting the cardiomyocyte cell-cycle reentry and dedifferentiation processes. We further revealed that circNfix downregulation induced 2% to 4% cardiomyocyte proliferation in the peri-infarcted myocardium, which was sufficient to facilitate functional recovery after MI.,,, Along with cardiomyocyte proliferation, circNfix downregulation promoted angiogenesis and inhibited cardiomyocyte apoptosis, which are also essential for cardiac repair and improvement of cardiac function after MI. These circNfix-induced beneficial effects decreased the fibrotic scar area by nearly 60% and restored cardiac function to almost normal levels after MI, indicating a potent therapeutic effect of attenuating postinfarction cardiac remodeling and subsequent heart failure. Recently, studies have indicated that ncRNAs, including miRNAs and lncRNAs (long noncoding RNAs), have important roles in regulating cardiomyocyte proliferation and cardiac regeneration., However, the use of these ncRNAs are limited by rapid degradation, poor conservatism and uncertain effects on humans. By contrast, we discovered that circNfix has structural robustness and a highly conservative nature. More importantly, circNfix expression in adult human myocardium was remarkably higher than that in fetal human myocardium. These characteristics suggest circNfix might overcome the stated ncRNA limitations and have clinical translational value. Notably, circNfix is clearly enriched in cardiomyocytes under basal conditions. Targeting circNfix might be less likely to have a significant negative effect on other organs and thus represents a promising strategy for improving prognosis after MI.

42In this study, we proposed and confirmed a novel concept in which superenhancers could be used to identify key circRNAs during cardiac regeneration by marking the cell-identity subgroups. During the cardiac development process, the embryonic and adult stages of mouse hearts are quite different regarding cardiogenic potency. We compared their differences in the superenhancer catalog and found that most of their superenhancers are developmental specific, indicating the important role of superenhancers in cardiogenic potency. Similarly, we revealed that superenhancer-associated circRNAs have unique characteristics, including higher expression levels and developmental stage and tissue specificity, which are linked to regulators essential for cell differentiation and development. Based on these findings, we conducted an integrated analysis of superenhancer-associated circRNAs and their expression profiles in embryonic and adult hearts and further filtered out adult cardiomyocyte-specific circNfix. Functional assays confirmed that circNfix downregulation was required and sufficient for cardiac regeneration. Moreover, we observed that circNfix-associated superenhancers regulated circNfix expression by recruiting Meis1, a key transcription factor controlling cardiomyocyte cell-cycle arrest. Our results are similar to previous studies showing that superenhancers are involved in causal mechanisms underlying the induction of cell differentiation and proliferation by recruiting master transcription factors. Collectively, superenhancer-associated circRNAs might be the key regulators of cardiac regeneration because of their specific expression profiles and causal functions in inducing cell differentiation and proliferation. The identification of ncRNAs with these features from traditional genome-wide profile analyses is time and labor intensive, and, more importantly, susceptible to false-positive discovery. By contrast, integrated analysis of superenhancers and their associated circRNA expression profiles might serve as a powerful filter to quickly identify the key regulators in heart regeneration processes according to RNA expression and genetic alteration.

43In this study, we investigated the mechanisms through which circNfix blocks cardiomyocyte proliferation and found that they are related to Ybx1. circRNAs often interact with RNA-binding proteins to fulfill their biological functions. The RNA pulldown, radioimmune precipitation, and fluorescense in situ hybridization assays confirmed that circNfix interacts with Ybx1. As an important transcription factor associated with cell proliferation, Ybx1 is involved in embryogenesis, muscle regeneration, oncogenic cell transformation, and cardiomyocyte differentiation/dedifferentiation., In the current study, Ybx1 induced cardiomyocyte proliferation by increasing cyclin A and cyclin B1 transcription, suggesting Ybx1 as an essential regulator of cardiac regeneration. Moreover, our studies showed that the interaction between circNfix and Ybx1 prevented nuclear translocation of Ybx1, causing cytoplasmic retention and degradation. The E3 ubiquitin ligases have been demonstrated to induce polyubiquitination and proteasomal degradation of Ybx1. Nedd4l is a member of the E3 ubiquitin ligases that are required for heart development. Our study revealed that Ybx1 is a substrate for E3 ubiquitin ligase Nedd4l in the presence of circNfix. The formation of the circNfix/Ybx1/Nedd4l ternary complex arrests Ybx1 in the cytoplasm and promotes Ybx1 degradation through ubiquitination-proteasome pathways. Although the exact molecular mechanism for the associations among circNfix, Ybx1, and Nedd4l remains to be clarified, we speculate that the interaction of Ybx1 with circNfix might change the conformation of Ybx1 and thus exposes its concealed ubiquitination sites to Nedd4l.

44To understand the critical role of circNfix in cardiac regeneration, we further explored another mechanism by which circNfix exerts its functions. Bioinformatic tools predicted that circNfix contains 3 binding sites for miR-214, which has been demonstrated to be essential for cardiomyocyte proliferation. In the current study, we found that miR-214 was involved in circNfix knockdown, promoting cardiomyocyte proliferation by inhibiting Gsk3β expression. Moreover, our study revealed that the circNfix/miR-214/Gsk3β pathway regulates VEGF release from cardiomyocytes. We also observed that circNfix downregulation significantly promoted angiogenesis after MI. Previous studies have indicated an association between miR-214 and Gsk3β and their important roles in endothelial cell angiogenesis.– Thus, the circNfix/miR-214/Gsk3β pathway might play an important role in regulating angiogenesis after MI. Compared with other ceRNAs (competing endogenous RNAs) shown to sponge miR-214, circNfix might be more efficient in modulating miR-214 activity attributable to its abundance, stability, and the numbers of miR-214 binding sites. These mechanistic studies further confirmed that circNfix is a crucial regulator of cardiac regeneration by regulating cardiomyocyte proliferation and angiogenesis.

45Our study has certain limitations. First, although the effect of circNfix on Ybx1 and miR-214 functions are rather evident, more detailed experiments in vivo might help to further verify their roles. Second, we used CRISPR-Cas9 knock-in mice to generate Meis1-deficient and circNfix-deficient mice. Although Meis1 and circNfix were not completely removed, our results from these deficient mice are quite reliable to determine the function of Meis1 and circNfix. Confirming our findings in complementary conventional genetic models would be interesting.

46The cardiac-specific circRNA Nfix, which is regulated by an superenhancer, inhibits cardiomyocyte proliferation by promoting Ybx1 ubiquitin-dependent degradation and repressing miR-214 activity. Loss of circNfix induced cardiac regeneration and angiogenesis and inhibited cardiomyocyte apoptosis after MI, which significantly restored cardiac function and improved the prognosis. Therefore, the newly discovered circNfix might be a potent therapeutic target for restoring cardiac function and preventing heart failure after MI.

47This work was supported by grants to Dr Bin from the National Natural Science Foundation of China (No. 81771857, No. 81571698, and No. 81271640).

48None.