Chromatin-remodelling factor Brg1 regulates myocardial proliferation and regeneration in zebrafish

The zebrafish possesses a remarkable capacity of adult heart regeneration, but the underlying mechanisms are not well understood. Here we report that chromatin remodelling factor Brg1 is essential for adult heart regeneration. Brg1 mRNA and protein are induced during heart regeneration. Transgenic over-expression of dominant-negative Xenopus Brg1 inhibits the formation of BrdU+/Mef2C+ and Tg(gata4:EGFP) cardiomyocytes, leading to severe cardiac fibrosis and compromised myocardial regeneration. RNA-seq and RNAscope analyses reveal that inhibition of Brg1 increases the expression of cyclin-dependent kinase inhibitors such as cdkn1a and cdkn1c in the myocardium after ventricular resection; and accordingly, myocardial-specific expression of dn-xBrg1 blunts myocardial proliferation and regeneration. Mechanistically, injury-induced Brg1, via its interaction with Dnmt3ab, suppresses the expression of cdkn1c by increasing the methylation level of CpG sites at the cdkn1c promoter. Taken together, our results suggest that Brg1 promotes heart regeneration by repressing cyclin-dependent kinase inhibitors partly through Dnmt3ab-dependent DNA methylation.

T he high mortality and morbidity following myocardial infarction is a public health problem worldwide. Myocardial infarction results in the loss of billions of cardiomyocytes in heart failure patients while myocardial regeneration is severely limited. Various cell-based and cell-free strategies are being explored for promoting heart regeneration in animal models and human patients [1][2][3] . However, the efficacy of cardiac cell-based therapy is still uncertain, with frequent occurrence of engraftment-induced arrhythmia, so the clinical implications remain unclear 4 . In contrast, lower vertebrates such as zebrafish can perfectly regenerate the injured heart by cardiomyocyte dedifferentiation and proliferation [5][6][7][8] . Although cardiac regeneration after ventricular resection occurs in mouse neonatal heart at 1 day after birth, this regenerative capacity is lost within 7 days after birth 9 , suggesting that regenerative potential is gradually lost during mouse heart development and maturation. In spite of the very limited regenerative capacity, mammalian cardiomyocytes are able to divide and renew in adulthood [10][11][12] . Therefore, harnessing the mechanisms underlying zebrafish heart regeneration may provide insights into mammalian heart regeneration and have therapeutic applications.
ATP-dependent chromatin remodelling is involved in controlling chromatin structure that in turn regulates many physiological and pathological processes. Instead of covalently modifying DNA or histones, the SWI/SNF (SWI/sucrose non-fermentable)-like complex, a member of the family of ATP-dependent chromatinremodelling complexes, uses energy from ATP hydrolysis, and regulates gene transcription by rearranging nucleosome positions and histone-DNA interactions, and thus facilitates the transcriptional activation or repression of targeted genes 13 . The SWI/SNF complex contains 410 components, of which brahma-related gene 1 (BRG1, or SMARCA4) is one of the central ATPase catalytic subunits. This complex plays an important role in the development of the central nervous system, thymocytes, heart and other organs. Brg1 is essential for zygote genome activation 14 , erythropoiesis 15 , cardiac development 16,17 and neuronal development 18,19 . Other members of the mammalian SWI/SNF complex are also required for heart morphogenesis, including Baf60c (ref. 20), Baf180 (ref. 21) and Baf250a (ref. 22). In particular, Brg1 controls cardiovascular development in a timeand tissue-specific manner. Brg1 deletion in mice results in embryonic lethality before implantation 23 . Endothelial and endocardial depletion of Brg1 results in embryonic death and failure of myocardial trabeculation around E10.5 in mice 16 . Mice with myocardial depletion of Brg1 die around E11.5 due to thin compact myocardium and the absence of the interventricular septum 17 . In embryos, Brg1 promotes cardiomyocyte proliferation by maintaining Bmp10 and suppressing p57 kip2 (cdkn1c) expression 17 . Brg1 suppresses Ask1 and Cdkn1a to inhibit apoptosis and promote proliferation of neural crest cells 24 . Besides its effects on cardiomyocyte proliferation, Brg1 also controls a and b myosin heavy-chain switching in the embryonic and adult hearts under hypertrophic stimulations 17 . The function of Brg1 in heart development is evolutionarily conserved between zebrafish and mammals. Mutation of brg1 in zebrafish causes cardiac hypoplasia and severe arrhythmia with abnormal expression patterns of several heart-specific genes 25 . Besides its functions in organ development, Brg1 is also required for hair regeneration and epidermal repair. Brg1 knockdown impairs bulge cell proliferation partly through elevating the cyclindependent kinase inhibitor p27 Kip1 (cdkn1b) 26 .
Although several subunits of the SWI/SNF complex are essential for cardiac development, little is known about how this complex orchestrates zebrafish heart regeneration at the chromatin level. To address this important question, we set out to determine whether and how the disruption of Brg1 affects zebrafish heart regeneration. Here we find that brg1 mRNA and protein are induced during the course of cardiac regeneration, and inhibition of Brg1 leads to severe cardiac fibrosis and compromised myocardial regeneration. Myocardial-specific expression of dn-xBrg1 blunts myocardial proliferation and regeneration by increasing cell-cycle-dependent inhibitors in the myocardium. Furthermore, injury-induced Brg1 interacts with Dnmt3ab to suppress the expression of cdkn1c by increasing the methylation level of CpG sites at the cdkn1c promoter. This study has gained molecular insights of Brg1 into zebrafish heart regeneration and has shed light on potential intervention of this complex for promoting heart repair and regeneration in humans.

Results
Brg1 is upregulated after ventricular apex amputation. In spite of great efforts in many laboratories, it remains challenging to induce mammalian cardiomyocytes to re-enter mitosis by either activating a single cyclin-dependent kinase or inactivating a single cyclin-dependent kinase inhibitor [27][28][29] . We hypothesized that a global epigenetic change might occur during zebrafish heart regeneration and so manipulating epigenetic programmes might be an efficient means of inducing mammalian cardiomyocytes to re-enter mitosis. To evaluate the functions of the SWI/SNF complex during zebrafish cardiac regeneration, we performed in situ hybridization screens to identify expression patterns of the complex components after ventricular apex amputation. Interestingly, several members of this complex (brg1, baf60c and baf180) were induced during regeneration ( Fig. 1 and Supplementary Fig. 1). brg1 transcripts were upregulated as early as 2 days post amputation (d.p.a.), peaked and concentrated proximal to the injury site at 3, 7 and 14 d.p.a., and became undetectable at 30 d.p.a., when regeneration was nearly complete (Fig. 1). These data support our hypothesis that the chromatinremodelling BAF complex is associated with zebrafish heart regeneration.
Inhibition of Brg1 blocks heart regeneration in zebrafish. Elevated expression of brg1 mRNA and protein following ventricular apex amputation suggested that brg1 might participate in regeneration. To determine the function of Brg1, we applied a dominant-negative Xenopus Brg1 (dn-xBrg1) that carries a K770T771-to-A770A771 mutation in the ATP-binding pocket 19 . This mutant Brg1 protein can bind to the other components of the SWI/SNF complex but its ATPase domain is disrupted and thus plays a dominant-negative role. Brg1 is highly conserved with 84% identity in amino acids between zebrafish and frog, and the ATP-binding pocket, which is mutated in the Xenopus dominant-negative Brg1, is identical between the two proteins. We generated a Tg(hsp70:dn-xbrg1) transgenic strain in which the dn-xBrg1 was driven by the zebrafish heat shock promoter 31 Supplementary Fig. 3). We then performed ventricular amputation in Tg(hsp70:dn-xBrg1) zebrafish and their wild-type siblings followed by heating at 37°C for 30 min daily from 5 to 30 d.p.a. Acid fuchsin orange G (AFOG) stain for extracellular matrix showed increased fibrosis and lack of sealing of the wound in dn-xBrg1 transgenic hearts ( Fig. 3b and Supplementary Fig. 4c) compared with wild-type siblings ( Fig. 3a and Supplementary Fig. 4a) at 30 d.p.a. Myocardial regeneration, visualized by MF20 immunostaining, was defective in dn-xBrg1 transgenic hearts ( Fig. 3d and Supplementary  Fig. 4d,e) compared with the fully regenerated myocardium in wild-type siblings ( Fig. 3c and Supplementary Fig. 4b,e) at 30 d.p.a. After heat shock from 5 to 30 d.p.a., we stopped heat shock treatment and examined the hearts at 60 d.p.a. by AFOG and MF20 staining. Heat shock caused comparable lethality between dn-xBrg1 transgenic zebrafish and their wild-type siblings ( Supplementary Fig. 4f). We still found cardiac fibrosis and compromised myocardial regeneration in dn-xBrg1 transgenic hearts ( Supplementary Fig. 5c,d) compared with wild-type sibling hearts ( Supplementary Fig. 5a,b), suggesting that inhibiting Brg1 caused permanent defects in heart regeneration. On the other hand, conditional over-expression of wild-type brg1 had no effects on myocardial proliferation ( Supplementary Fig. 6), suggesting that other members of SWI/ SNF complex are also required for this process. Taken together, these data demonstrated that inhibition of Brg1 caused a regenerative defect in the zebrafish heart and so Brg1 is required for cardiac regeneration.
Previous studies suggest that regenerated cardiomyocytes originate from the de-differentiation and proliferation of cardiomyocytes near injury site 7,32 . Thus, we evaluated whether inhibition of Brg1 had any effect on the de-differentiation of cardiomyocytes after injury. We found comparable dedifferentiation of cardiomyocytes in the injury site, as measured disassembly of cardiac sarcomeres by transmission electron microscopy ( Supplementary Fig. 7a-d), or disrupted Z-disks labelled by cypher-EGFP fusion protein 33 ( Supplementary  Fig. 7e-l) in both wild-type sibling and Tg(hsp70:dn-xbrg1) transgenic hearts at 14 d.p.a. We next compared the index of cardiomyocyte proliferation in Tg(hsp70:dn-xBrg1) and wild-type sibling hearts at 14 d.p.a. after heat shock by quantifying the percentage of BrdU and myocardial marker Mef2C doublepositive nuclei near the injury area. Compared with B14% BrdU þ /Mef2C þ cardiomyocytes in wild-type sibling hearts at 14 d.p.a. (Fig. 3e,g), we found only 2.4% proliferating cardiomyocytes in Tg(hsp70:dn-xBrg1) hearts ( Fig. 3f,g), which was further confirmed by measuring PCNA þ /Mef2C þ proliferating cardiomyocytes (Supplementary Fig. 8d-f). In addition, both Tg(hsp70:dn-xBrg1) transgenic and wild-type sibling hearts, without heat shock treatments, had similar BrdU þ /Mef2C þ proliferating cardiomyocytes (Supplementary Fig. 8a-c), suggesting the minimal effects of heat shock on the index of proliferating cardiomyocytes. As previously reported 7,34 , Gata4-positive cardiomyocytes appeared near the injured area in wild-type Tg(gata4:EGFP) hearts at 14 d.p.a. (Fig. 3h). However, there were markedly fewer Gata4:EGFP-positive cardiomyocytes in Tg(hsp70:dn-xBrg1; gata4:EGFP) hearts at 14 d.p.a. (Fig. 3i,j). These data suggested that Brg1 is essential for cardiomyocyte proliferation and regeneration after ventricular apex amputation in zebrafish.
The endocardium and epicardium are also known to be activated after ventricular amputation 35 . Immunostaining revealed that epicardial and endocardial marker Raldh2 (Supplementary Fig. 9a-c) and epicardial reporter Tg(tcf21:DsRed) (Supplementary Fig. 9d-f) were similarly expressed in wild-type sibling and Tg(hsp70:dn-xBrg1) hearts, suggesting that the organwide activation of endocardium and epicardium was not affected after inhibition of Brg1. In addition, coronary vessel regeneration accompanied myocardial regeneration after ventricular resection 36 . By labelling coronary vessels and endocardium with the Tg(flk1:EGFP) transgene, we found fewer flk1 þ endothelium and endocardium in the injured area of Tg(hsp70:dn-xBrg1) hearts than in wild-type sibling hearts (Supplementary Fig. 9g-i), suggesting that Brg1 is also critical for the formation of newly regenerated coronary vessels since the Raldh2 þ endocardium is less affected during heart regeneration.
Brg1 represses cyclin-dependent kinase inhibitor genes. To decipher the molecular mechanisms underlying the regulation of myocardial proliferation by Brg1, we performed RNA-seq to compare the transcriptomes in Tg(hsp70:dn-xBrg1) and wild-type sibling hearts after heat shock daily from 5 to 14 d.p.a. We obtained B4 million 100 bp pair-end reads for each sample. Further bioinformatics analyses revealed that 1,204 genes were upregulated while 1,092 genes were downregulated after the inhibition of Brg1 (Supplementary Data Sets 1 and 2). The genes with a twofold difference were selected for further analysis. cdkn1a was one of the upregulated genes in the Tg(hsp70:dn-xBrg1) hearts (Fig. 4a). Since Brg1 inhibition had a pronounced effect on myocardial proliferation and fibrosis in Tg(hsp70:dn-xBrg1) hearts, we focused on the genes regulating cell-cycle progression and proliferation as well as cardiac fibrosis. Using reverse transcription-PCR (RT-PCR), we found that other CDK inhibitors cdkn1a, cdkn1ba, cdkn1bb, cdkn1c and cdkn1d were also increased in the dn-xBrg1 transgenic heart at 14 d.p.a. (Fig. 4b). A previous study reported that Meis1 is required for the transcriptional activation of CDK inhibitors (cdkn2a, cdkn2b and cdkn1a) in mice 37 . Consistently, meis1a, meis2a and meis2b were induced in the dn-xBrg1 transgenic heart (Fig. 4b). In addition, several fibrotic genes col1a1a, col1a2, tgfb1a, tgfb3 and vimentin were also upregulated in the dn-xBrg1 transgenic heart (Fig. 4b). Importantly, brg1 was induced while cdkn1c was reciprocally repressed in the early phases of regeneration (Fig. 4c). Taken together, our data suggested that Brg1 normally represses the expression of CDK inhibitors for priming heart regeneration in zebrafish.  Since Brg1 is induced in injured cardiomyocyte (Fig. 2h) and is essential for cardiomyocyte proliferation (Fig. 3e-g and Supplementary Fig. 8d-f), we tested whether the Brg1-cdkn axis acted in cardiomyocytes during heart regeneration. RNAscope in situ hybridization analysis revealed that both cdkn1a (Fig. 5a,b,i,j) and cdkn1c (Fig. 5e,f,m,n) were very lowly expressed in wild-type heart with or without injury, but both cdkn1a and cdkn1c was markedly induced in Tg(hsp70:dn-xBrg1) transgenic hearts (Fig. 5d,h,l,p) compared with wild-type sibling hearts (Fig. 5c,g,k,o). Interestingly, co-staining with MF20 showed that both cdkn1a and cdkn1c were enriched in the myocardium (Fig. 5l,p). Furthermore, by generating Tg(myl7:CreER;ubi: loxP-DsRed-STOP-loxP-dn-xBrg1) transgenic zebrafish, tamoxifen-induced myocardial-specific inhibition of Brg1 resulted in decreased PCNA þ /Mef2C þ proliferating cardiomyocytes at 7 d.p.a. (Fig. 6a-c), as well as increased cardiac fibrosis (Fig. 6d,e) and compromised myocardial regeneration at 30 d.p.a. (Fig. 6f,g). Consistently, we also found that both cdkn1a and cdkn1c were upregulated in the MF20 þ myocardium in Tg(myl7:CreER;ubi:-loxP-DsRed-STOP-loxP-dn-xBrg1) transgenic hearts (Supplementary Fig. 10j,l,n,p) compared with Tg(ubi:loxP-DsRed-STOP-loxP-dn-xBrg1) control hearts ( Supplementary Fig. 10i,k,m,o) after 4-HT induction by RNAscope. Together, these data support our hypothesis that Brg1 acts to suppress cdkn1a and cdkn1c in the myocardium to regulate heart regeneration in zebrafish.
We then asked how Brg1 regulates the transcriptional activation of CDK inhibitors such as cdkn1c during regeneration. Since DNA methylation is an important mechanism for regulating gene expression, we determined the DNA methylation pattern of the promoters of cdkn1c, meis1a and tgfb1a by performing bisulfate sequencing of 8-10 individual CpG sites. Bisulfate sequencing showed less methylation in these promoters of Tg(hsp70:dn-xBrg1) transgenic hearts than that in wild-type sibling hearts (Fig. 7a,b and Supplementary Fig. 11). Furthermore, methylation of the cdkn1c promoter increased in injured hearts at 3 and 5 d.p.a. while brg1 was reciprocally induced compared with mock controls (Fig. 7c,d), confirming a potential endogenous role of Brg1 in repressing the expression of cdkn1c. To investigate the mechanism underlying the effects of Brg1 inhibition on cdkn1c promoter methylation, we performed chromatin immunoprecipitation (ChIP) and quantitative ChIP assays, and found that Brg1 bound with the cdkn1c promoter region, which was demethylated after inhibition of Brg1 during cardiac regeneration (Fig. 7a,e). Consistent with Brg1-cdkn1a/1c function in the myocardium, we found that cdkn1c promoter was hypomethylated in Tg(myl7:CreER;ubi:loxP-DsRed-STOP-loxP-dn-xBrg1) transgenic hearts compared with Tg(ubi:loxP-DsRed-STOP-loxP-dn-xBrg1) control hearts after 4-HT induction by bisulfite sequencing (Supplementary Fig. 12a); and that myc-tagged dn-Brg1 directly bound to the cdkn1c promoter in Tg(myl7:CreER;ubi:loxP-DsRed-STOP-loxP-dn-xBrg1) transgenic hearts by ChIP assay (Supplementary Fig. 12d).
Methyltransferases are known to maintain the patterns of methylated cytosine residues in the mammalian genome and are the key molecules in regulating the level of DNA methylation. Seven DNA methyltransferases were annotated in the zebrafish genome website (Zv9.0). By RT-PCR, we found that dnmt1, dnmt3aa and dnmt3ab were expressed in adult zebrafish heart,  suggesting that both de novo and maintenance DNA methylation might occur during heart regeneration. By over-expressing Brg1 and Dnmt3ab in 293T cells, or H9C2 cardiac cells, we demonstrated that either Brg1 or dn-xBrg1 and Dnmt3ab form a protein complex by co-immunoprecipitation experiments (Fig. 7f,g). These data are consistent with previous studies in cancer cells 38 and hypertrophic hearts 39 . An additional reporter system was further used to examine whether this interaction affected the cdkn1c promoter activity. Indeed, over-expression of either brg1 or dnmt3ab inhibited the cdkn1c promoter activity, which was synergistically enhanced by co-expression of brg1 and dnmt3ab in 293T cells (Fig. 7h) or P4 rat neonatal cardiomyocytes ( Supplementary Fig. 12c). Furthermore, we also found that baf60c increased while dnmt3ab decreased in dn-xBrg1 transgenic hearts compared with their sibling hearts at 14 d.p.a. (Supplementary Fig. 13). These data suggest a feedback effect of SWI/SNF complex on baf60c, as well as synergistic interaction of Brg1 and Dnmt3ab in regulating DNA methylation of cdkn inhibitors on heart regeneration by over-expression of dn-xBrg1. Importantly, dnmt3ab was induced from 3 to 14 d.p.a. and peaked at 7 d.p.a. compared with that at mock hearts ( Supplementary Fig. 14a), and nanoparticle-mediated dnmt3ab siRNA decreased BrdU þ /Mef2C þ proliferating cardiomyocytes ( Supplementary Fig. 14b-d), supporting the role of dnmt3ab during heart regeneration. We have previously reported methodology that nanoparticle-delivered siRNA efficiently inhibits targeted gene expression in adult zebrafish hearts 40 . Together, our data support the notion that Brg1 and Dnmt3ab form a protein complex in 293T cells and H9C2 cardiomyocytes, and this complex might be utilized to increase DNA methylation of cdkn1c promoter and so repressing its transcription for promoting zebrafish heart regeneration.
Cdkn1a/1c mediate effects on myocardial proliferation. We then asked whether CDK inhibitors functionally act downstream  ARTICLE from Brg1 during heart regeneration. Using nanoparticle-mediated siRNA knockdown method 40 , we were able to decrease the expression of cdkn1a and cdkn1c in 2 d.p.a. hearts at 24 h after siRNA injection (Fig. 8a,b). We found comparable BrdU þ / Mef2C þ proliferating cardiomyocytes in wild-type sibling hearts without (Fig. 8c) or with control siRNA injection (Fig. 8d), showing that siRNA injections daily from 5 to 14 d.p.a. had no effect on injury-induced cardiomyocyte proliferation (Fig. 8h).   Importantly, we found more BrdU þ /Mef2C þ proliferating cardiomyocytes in Tg(hsp70:dn-xBrg1) hearts with either cdkn1a (Fig. 8f) or cdkn1c siRNA (Fig. 8g) than in control dn-xBrg1 transgenic hearts (Fig. 8e,h). Another independent siRNA for either cdkn1a or cdkn1c was also able to rescue the proliferation index in Tg(hsp70:dn-xBrg1) transgenic hearts ( Supplementary Fig. 15a-f). However, siRNA knockdown of either cdkn1a or cdkn1c had little or no effect on BrdU þ / Mef2C þ proliferating cardiomyocytes in wild-type sibling hearts at 14 d.p.a. (Supplementary Fig. 16), consistent with very low levels of these genes in wild-type hearts after ventricle resection (Fig. 5). These data further support the notion that Brg1 promotes heart regeneration by repressing CDK inhibitors such as cdkn1a and cdkn1c.

Discussion
Brg1 plays an essential role in embryonic development such as in zygote genome activation 14 , erythropoiesis 15 and the development of T cells 41,42 , heart 16,17 and neurons 18,19 . However, its function in adult heart regeneration has not been addressed.
Here we showed that Brg1 was activated during zebrafish heart regeneration; either global or myocardial-specific over-expression of dn-xBrg1 interfered with myocardial proliferation and regeneration; and mechanistically, Brg1 promoted regeneration by suppressing the CDK inhibitors cdkn1c. Although Brg1 was broadly expressed in cardiomyocytes, endothelial/endocardial cells, epicardium and inflammatory cells (macrophages and neutrophils), myocardial-specific enrichment and function of Brg1 and cdkn1c suggest that the Brg1-cdkn1c axis acts in the myocardium to regulate cardiomyocyte proliferation and regeneration. Future studies are warranted to parse Brg1 function in the endocardium, epicardium and inflammatory leukocytes in adult heart regeneration. Previous studies have shown that the components of the SWI/SNF complex (also called the BAF complex) are differentially expressed in embryonic stem cells, neuronal progenitors and differentiated neurons 43   ARTICLE complex. During zebrafish heart regeneration, we found that brg1, baf60c and baf180 were induced on heart sections. They had no or little expression in control hearts and were induced from 1 d.p.a. with peak expression around 7-14 d.p.a., and then declined from 14 to 30 d.p.a. for brg1 and baf60c while declined from 21 to 30 d.p.a. for baf180. Overall, brg1 and baf60c mRNA are more abundant while baf180 is less expressed during heart regeneration. Therefore, we propose that a similar cardiacregenerative BAF complex might contain at least brg1, baf60c and baf180 during heart regeneration in zebrafish. However, this hypothesis remains to be tested in future studies. Previous studies by Field and colleagues have shown that either deletion of a CDK inhibitor or activation of a CDK has limited effects on promoting mammalian cardiomyocyte proliferation 2,12,[27][28][29] . Our data support the notion that an increase of CDK inhibitors and related meis genes occurs after inhibition of Brg1 during heart regeneration in zebrafish; this was partly supported by RNA-seq analysis of Tg(hsp70:dn-xBrg1) and wildtype sibling hearts after daily heat shock from 5 to 14 d.p.a., as well as being further confirmed by quantitative RT-PCR and its myocardial enrichment by RNAscope in situ hybridization analysis. During regeneration, cdkn1c was downregulated in injured hearts at 3 and 5 d.p.a., while brg1 was reciprocally upregulated, suggesting that Brg1 normally represses cdkn1c expression during this process, consistent with previous reports that cell-cycle-dependent kinase inhibitors are downstream of Brg1 in cardiac development 17 , mammalian neural crest cell development 24 , bulge stem cells during tissue regeneration 26 and adult neural stem cells maintenance 46 , as well as Brg1 directly binds to the cdkn1c promoter as predicted by ChIP-seq analysis 47 . Furthermore, our data showed that Brg1 directly bound to the promoter of cdkn1c by ChIP assay, repressed the cdkn1cluciferase reporter, and siRNA knockdown of either cdkn1a or cdkn1c partially rescued the blunted myocardial proliferation with transgenic over-expression of dn-xBrg1. Taken together, we have shown, for the first time, that Brg1 promotes adult cardiomyocyte proliferation by repressing CDK inhibitors, specifically by direct repression of cdkn1c during heart regeneration.
It has been shown that Brg1 interacts directly or indirectly with other transcription factors or epigenetic components 17,[24][25][26]38,48 , and SWI/SNF chromatin-remodelling factors can induce changes in DNA methylation to regulate gene expression 49 . We then hypothesized that Brg1 might interact with other transcription and/or epigenetic factors to repress the transcription of cdkn1c. DNA methyltransferases are known to catalyse the reaction of transferring the methyl group to DNA from S-adenosyl methionine. Dnmt3a and Dnmt3b are de novo DNA methyltransferases that normally act as transcriptional repressors by DNA methylation or transcriptional co-repressors 38,[50][51][52][53][54][55][56] .
Here we showed that inhibition of Brg1 led to a decreased level of DNA methylation in the promoters of cdkn1c, meis1a and tgfb1a, and increased the expression of cdkn1c accordingly. Importantly, the level of DNA methylation in the cdkn1c promoter increased after ventricular resection at 3 and 5 d.p.a., which is consistent with the repression of cdkn1c during normal regeneration. Indeed, RT-PCR showed that zebrafish dnmt3ab was induced during heart regeneration and it is required for myocardial proliferation, and co-immunoprecipitation analysis showed that it directly interacted with Brg1 in 293T cells and H9C2 cardiomyocytes, consistent with the previous report that Brg1 and Dnmt3a interact in cancer cells 38 and in hypertrophic cardiomyocytes 39 . Our data reveal that dnmt1, dnmt3aa and dnmt3ab are expressed in adult zebrafish heart, and their respective role in heart regeneration need to be addressed in the future. Together, Brg1 suppresses expression of cdkn1c and possible other CDK inhibitors, at least, partly through its interacting with Dnmt3ab to increase the level of DNA methylation in the cdkn1c promoter, leading to an automatic regenerative capacity in the heart of adult zebrafish. Therefore, conditional activation of the BAF complex and related signalling pathways might shed light on improving mammalian myocardial regeneration.

Methods
Zebrafish lines. Zebrafish were raised and handled according to a zebrafish protocol (IMM-XiongJW-3) approved by the Institutional Animal Care and Use Committee at Peking University, which is fully accredited by AAALAC International. Tg(hsp70:dn-xBrg1), Tg(myl7:cypher-EGFP), Tg(myl7:CreER), Tg(ubi:loxP-DsRed-STOP-loxP-Brg1) and Tg(ubi:loxP-DsRed-STOP-loxP-dn-xBrg1) zebrafish lines were generated by using Tol2-based transgenesis 57 . The dominant-negative Xenopus Brg1 (dn-xBrg1) plasmid clone was kindly provided by Dr Kristen L. Kroll (Washington University at St Louis) 19,58 , and the Tg(myl7:CreER) and Tg(ubi:loxP-DsRed-STOP-loxP-EGFP) plasmid clones were kindly provided by C Geoffrey Burns (Massachusetts General Hospital, Boston, MA, USA) 59 . Tg(coronin1a:EGFP) 60  For heat shock experiments, we crossed heterozygous Tg(hsp70:dn-xBrg1) with wild-type TL zebrafish, and so expected to have 50% heterozygous transgenic fish and 50% wild-type siblings. Heterozygous Tg(hsp70:dn-xBrg1) transgenic and wild-type sibling adult zebrafish received a daily heat shock in 37°C water for 30 min from 5 to either 14 or 30 d.p.a. Each cycle of heat shock was carried out by transferring zebrafish to system water at 31°C, accompanying with gradually increasing temperature from 31 to 37°C for about 10 min and then remaining at 37°C for another 20 min. To induce the Cre recombination in adult zebrafish, Tg(myl7:CreER;ubi:loxP-DsRed-STOP-loxP-dn-xBrg1) or control Tg(ubi:loxP-DsRed-STOP-loxP-dn-xBrg1) transgenic zebrafish were bathed in 5 mM tamoxifen (Sigma, St Louis, MO) for 24 h, which was made from a 10 mM stock solution dissolved in 100% ethanol at room temperature. Zebrafish were treated with tamoxifen at a density of 3-4 per 150 ml of water, and then returned to circulating zebrafish system water. Ventricular resections were performed at 3 days after tamoxifen treatment. Zebrafish were confirmed for their genotyping and randomly picked for all experiments.
Adult zebrafish heart resection. The ventricular resection was performed according to a well-established procedure 5,34 . Briefly, adult zebrafish were anaesthetized with tricaine and the pericardial sac was exposed by removing surface scales and a small piece of skin. The apex of the ventricle was gently pulled up and removed with Vannas scissors. The zebrafish was then placed back into a water tank, and water was puffed over the gills with a plastic pipette until it breathed and swam regularly. The surface opening sealed automatically within a few days. In particular, we crossed heterozygous Tg(hsp70:dn-xBrg1) with wildtype TL zebrafish, and so expected to have 50% heterozygous transgenic fish and 50% wild-type siblings for performing ventricular resections.
Construction and sequencing of high-throughput RNA-seq libraries. Total RNA was isolated using an RNeasy Mini kit (QIAGEN). After confirming the quality and integrity of RNA on agarose gels, we used B1 mg of total RNA to construct the RNA-seq libraries by applying a TruSeq RNA Sample Prep kit (Illumina, San Diego, CA). We carried out the RNA-seq using Illumina HiSeq 2500 to generate B4 million 100 bp pair-end reads for each sample. Low-quality reads and sequencing adapters were removed from the raw sequencing data, and the clean reads were mapped onto the zebrafish transcriptome (danRer7) using Tophat 64 . The expression level of each gene was calculated using Cufflinks, and the genes differentially expressed between samples were calculated using Cuffdiff 64 . Genes with 42-fold difference between the two groups were selected for further analyses.
siRNA delivery into adult zebrafish heart. siRNAs were encapsulated in polyethylene glycol-polylactic acid nanoparticles using a double emulsion-solvent evaporation technique and then injected into the pericardial sac 40,65,66 . Briefly, zebrafish were allowed to recover for 1 day after ventricular resection. To evaluate the effect of siRNA on its target gene expression, the hearts were collected at 2 d.p.a., and total RNA was isolated to assess the expression of the respective genes by quantitative RT-PCR. To evaluate the effect of genes on cardiomyocyte proliferation, 50 ml polyethylene glycol-polylactic acid nanoparticle-encapsulated siRNAs was injected first, and B1 h later, 50 ml 2.5 mg ml À 1 BrdU (B5002; Sigma) was injected into the thoracic cavity daily from 7 to 14 d.p.a. The hearts at 14 d.p.a.
were collected for subsequent experiments. siRNA sequences for cdkn1a, cdkn1c and dnmt3ab are shown in Supplementary Table 2. mRNA or protein detection assays and AFOG staining. In situ hybridization and AFOG staining were performed on paraffin sections 34 . Adult zebrafish hearts were fixed in 4% paraformaldehyde at room temperature for 2 h, dehydrated and then embedded in paraffin and sectioned at 5 mm. Zebrafish brg1 cDNA was cloned from an embryonic cDNA library, of which primer sequences are shown as brg1-F and brg1-R in Supplementary Table 1. Digoxigenin-labelled brg1 probes were synthesized using T7 RNA polymerase (Roche).
RNAscope (Advanced Cell Diagnostics, Hayward, CA) was performed on 10 mm sections from freshly frozen hearts embedded in O.C.T. Compound (Embedding Medium for Frozen Tissue Specimens to ensure Optimal Cutting Temperature; SAKURA; 4583). Tissues were fixed in pre-chilled 10% neutral buffered formalin, followed by dehydration, then treated with Pretreat 1 for 10 min at room temperature. After Pretreat 1, slides were washed with water and incubated for 30 min at room temperature with Pretreat 4. Following Preteat 4, the RNAscope 2.0 HD Detection Kit Brown was applied for visualizing hybridization signals. Three injured and mock hearts were used for each RNAscope experiment. Immunostaining was performed with primary antibodies (1:50 diluted in PBT containing 10% FBS) incubated overnight at 4°C.
RT-PCR analysis. Total RNA was isolated and purified using an RNeasy mini kit (74106; Qiagen). About 1 mg RNA was used for reverse transcription with a Prime Script RT Reagent kit (RR037A; TakaRa), and quantitative RT-PCR was performed using a SYBR Premix DimerEraser kit (RR091A; TakaRa). Primer sequences are listed in Supplementary Table 1.
Chromatin immunoprecipitation. Chromatin was isolated from zebrafish hearts using Chromatin Prep Module (Catalogue# 26158; Thermo Scientific Pierce). ChIP assays were performed using Agarose ChIP Kit (Catalogue# 26156; Thermo Scientific Pierce) 34 . Chromatin was immunoprecipitated using anti-Brg1 (J1) antibody 30 , and Brg1-bound sequences were amplified with the respective gene primers. The primer sequences are listed in Supplementary Table 1. The original uncropped images of gels are shown in Supplementary Fig. 17.
DNA methylation and immunoprecipitation. Zebrafish genomic DNA was extracted, purified and resuspended in 1 Â TE buffer. For each sample, a total of 500 ng genomic DNA was treated using MethylCode bisulfite conversion kit (MECOV-50; Invitrogen). The CpG islands of the cdkn1c promoter region were primarily located between À 1,586 and À 1,354 bp. The bisulfite-treated DNA was subjected to PCR to amplify the cdkn1c promoter region. The primers were designed according to the website (http://www.urogene.org/methprimer/) and are listed in Supplementary Table 1. Alterations of promoter methylation were confirmed by Sanger sequencing. In addition, zebrafish dnmt3ab was isolated from an embryonic cDNA library and then subcloned into the pCDNA3.1 vector. Myc-tagged dnmt3ab and brg1 were co-transfected into 293T cells (CRL-1573, American Type Culture Collection; ATCC), which were then collected for immunoprecipitation after 24 h (ref. 34). For immunoprecipitation, H9C2 cells (CRL-1446, ATCC) were infected with either Ad-brg1/Ad-Myc-dnmt3ab or Ad-dn-xbrg1-Flag/Ad-Myc dnmt3ab, and infected cells were then collected for immunoprecipitation after 24 h. The recombinant adenovirus was constructed and amplified by SinoGenoMax Co., Ltd, Beijing, China. The antibody for immunoprecipitation were anti-Myc (2276S; Cell Signaling Technology), anti-Flag (AP1013a, ABGENT) and anti-Brg1 (J1) as described previously 30 . The original uncropped images of blots are shown in Supplementary Fig. 17.
Transmission electron microscopy. Tg(hsp70:dn-xBrg1) and wild-type sibling hearts were collected at 14 d.p.a. and fixed in 2% glutaraldehyde, 2% paraformaldehyde and 0.1 M PBS overnight at 4°C. Subsequent embedding, ultra-thin section preparation and staining were performed by the Electron Microscopy Core Facility of Peking University. A TecnaiT20 (LaB6, 200KV) transmission electron microscope (FEI, Hillsboro, OR, USA) was used to image stained sections 33 . Three transgenic and wild-type sibling hearts were used for transmission electron microscopy.
Statistical analysis. All statistics were calculated using Prism 5 Graphpad Software. The statistical significance between two groups was determined using paired Student's t-test, with tow-tailed P value, and the data were reported as mean±s.e.m. Among three or more groups, one-way analysis of variance followed by Bonferroni's multiple comparison test or Dunnett's multiple comparison test was used for comparisons.
Data availability. Data that support the findings of this study have been deposited in Gene Expression Omnibus with the accession code GSE81627. All other relevant data are available from the corresponding authors on reasonable request.