The Gridlock transcriptional repressor impedes vertebrate heart regeneration by restricting expression of lysine methyltransferase

ABSTRACT Teleost zebrafish and neonatal mammalian hearts exhibit the remarkable capacity to regenerate through dedifferentiation and proliferation of pre-existing cardiomyocytes (CMs). Although many mitogenic signals that stimulate zebrafish heart regeneration have been identified, transcriptional programs that restrain injury-induced CM renewal are incompletely understood. Here, we report that mutations in gridlock (grl; also known as hey2), encoding a Hairy-related basic helix-loop-helix transcriptional repressor, enhance CM proliferation and reduce fibrosis following damage. In contrast, myocardial grl induction blunts CM dedifferentiation and regenerative responses to heart injury. RNA sequencing analyses uncover Smyd2 lysine methyltransferase (KMT) as a key transcriptional target repressed by Grl. Reduction in Grl protein levels triggered by injury induces smyd2 expression at the wound myocardium, enhancing CM proliferation. We show that Smyd2 functions as a methyltransferase and modulates the Stat3 methylation and phosphorylation activity. Inhibition of the KMT activity of Smyd2 reduces phosphorylated Stat3 at cardiac wounds, suppressing the elevated CM proliferation in injured grl mutant hearts. Our findings establish an injury-specific transcriptional repression program in governing CM renewal during heart regeneration, providing a potential strategy whereby silencing Grl repression at local regions might empower regeneration capacity to the injured mammalian heart.


INTRODUCTION
The human heart has limited capacity to regenerate new cardiac muscle after myocardial injury. Instead, impaired myocardium is replaced by fibrotic scar, eventually leading to heart failure (Tzahor and Poss, 2017;Xin et al., 2013). By contrast, zebrafish heart possesses natural regeneration capacity by dedifferentiation and proliferation of pre-existing cardiomyocytes (CMs) after injury (Jopling et al., 2010;Kikuchi et al., 2010;Li et al., 2015). Hearts of mice can regenerate if injured in the few days after birth, which coincides with the transient capacity for CM proliferation (Porrello et al., 2011). The limited regenerative potential in the adult mammalian heart can be ascribed to cell-intrinsic barriers or molecular blocks that prevent CMs from entering proliferation and completing cytokinesis after damage (González-Rosa et al., 2018;Li et al., 2015;Tzahor and Poss, 2017). Various mitogenic factors and signaling pathways have been identified as initiating and achieving heart regeneration Han et al., 2019;Harrison et al., 2019;Kikuchi et al., 2011;Marin-Juez et al., 2016;Mohamed et al., 2018;Monroe et al., 2019;Wu et al., 2016;Zhao et al., 2019). However, transcriptional programs and epigenetic signaling that serve as roadblocks to limit injury-induced CM renewal are incompletely understood.
Gridlock (Grl; also known as Hey2) is a basic helix-loop-helix (bHLH) transcriptional repressor belonging to the Hesr (Hairy/ Enhancer-of-split related) family (also known as Hey or Hrt family) containing Hey1, Grl and HeyL, which play important functions via diverse mechanisms involving vascular, myocardial, endocardial and neuronal tissues during development (Fischer et al., 2002;Nakagawa et al., 1999;Sakamoto et al., 2003;Zhong et al., 2000). Hey proteins mediate transcriptional repression by directly binding an E-box DNA motif, and can repress GATA-mediated transcription via physically interacting with GATA factors (Heisig et al., 2012;Jia et al., 2007;Kathiriya et al., 2004). We and others previously reported that Grl and Hey1 regulate arterial-venous differentiation in zebrafish and mice (Fischer et al., 2004;Zhong et al., 2001). Grl also contributes to zebrafish heart development by limiting expansion of cardiac progenitor cells (CPC) and embryonic CMs (Gibb et al., 2018;Jia et al., 2007). During cardiovascular development, Hey family genes function in response to Notch signaling, including atrioventricular boundary formation, arterial-venous differentiation and possible trabecular specification (Fischer et al., 2004;Koibuchi and Chin, 2007;Kokubo et al., 2007;Miao et al., 2018;Rutenberg et al., 2006;Tian et al., 2017;Xin et al., 2007;Zhong et al., 2001). In humans, HEY2 variants are associated with Brugada syndrome, a rare disease with high risk of sudden cardiac death (Bezzina et al., 2013). Although Hey proteins play pivotal roles during cardiovascular development and potential human disease, their functions in heart regeneration have not been investigated and remain unknown.
In this study, we reveal cell-autonomous effects of Grl on regulation of myocardial regeneration using loss-of-function and conditional gain-of-function studies. Grl functions as a negative regulator by restraining injury-induced CM dedifferentiation and proliferation. We identify KMT Smyd2 upregulation at the injured myocardium as a direct target of Grl. Smyd2 induction, triggered by injury-induced Grl reduction, is required for CM proliferation during regeneration. Our findings reveal novel roles and mechanisms of the Grl-Smyd2 network in governing vertebrate CM renewal and heart regeneration.

Cardiac injury triggers a grl dynamic expression pattern
To define the spatiotemporal expression pattern of grl during adult heart regeneration, we generated grl reporter lines Tg(grl:EGFP) and Tg(grl:mCherry) in zebrafish, where EGFP and mCherry were driven by grl promoter-based upstream region. In the adult heart, grl:EGFP expression was detectable throughout the myocardium labeled by myocyte enhance factor 2 (Mef2) and enriched in the primordial myocardial layer (PML) (Fig. 1A,A1; Fig. S1A,B). grl: EGFP was also overlapped with myosin heavy chain (MHC) that marked the CM cytoplasm ( Fig. 1B-D). In situ hybridization (ISH) analyses validated grl expression throughout the myocardium, with enrichment in the PML (Fig. 1E). Although grl was expressed in the arterial endothelium of aortae during embryogenesis (Jia et al., 2007;Li et al., 2020;Rowlinson and Gering, 2010;Satow et al., 2001;Zhong et al., 2000), grl:EGFP was absent in the endocardium or coronary vascular endothelium lining with flk1:mCherry expression in the adult heart ( Fig. 1F-H). Furthermore, Tg(grl: mCherry) transgenic hearts displayed no colocalization of mCherry with the epicardial marker tcf21:nucEGFP (Fig. S1C). grl:EGFP was also absent in the epicardium marked by injury-induced Raldh2 protein (Fig. S1D). These findings indicate that the myocardium is the primary source of Grl under homeostatic conditions.
To explore the injury response of grl during zebrafish heart regeneration, we performed ventricular apex resection using Tg(grl: EGFP) animals and investigated grl temporal expression profiles from injury onset until 30 days post amputation (dpa). We observed that EGFP fluorescence was reduced at the apical edge of the injured myocardium labeled by Mef2 at 1 dpa (Fig. 1J,J1; Fig. S1I,J) and declined to the lowest level at 3 dpa (Fig. 1K,K1), compared with uninjured hearts (Fig. 1I,I1). Thereafter, grl:EGFP increased to some extent at the injured myocardial cell edge at 7 dpa (Fig. 1L,L1; M,M1) and gradually returned to the uninjured level with formation of the PML from 14 dpa to 30 dpa (Fig. 1N,N1). Consistent with grl: EGFP expression, quantitative PCR (qPCR) analyses indicated a great reduction in grl expression at 1 dpa, which reduced to the lowest level at 3 dpa (Fig. 1O). Although grl expression increased to some extent at 7 dpa, it was still reduced by 75% compared with the uninjured level (Fig. 1O). Reduced expression of other Hey family genes, hey1 and heyL, was also detectable by qPCR. Expression of heyL was much lower than that of hey1, despite extremely low expression of both genes (Fig. 1O). We noticed that the increased level of grl:EGFP transgene at 7 dpa seemed to be much higher than that in grl transcripts measured by qPCR (Fig. 1L,O). These differences might be because the 7.8 kb upstream region in the grl: EGFP transgene does not fully reflect the endogenous grl transcripts during regeneration. Indeed, our ISH analyses revealed a discernible increase in grl transcripts at the injury site at 14 dpa rather than 7 dpa (Fig. S1G,H), consistent with qPCR analyses. Collectively, our findings indicate that the reduced expression of myocardial grl correlates with the regenerative responses of the zebrafish heart to injury.
Mutations in grl augment heart regeneration by reducing fibrotic scars and enhancing CM proliferation To investigate the effects of grl loss of function on heart regeneration, we generated zebrafish nonsense mutations in grl using the CRISPR/Cas9 technique, as a previous grl m145 mutant caused a point mutation of the terminator codon that is predicted to produce an extended protein with some residue activity (Zhong et al., 2000). Single guide RNA (sgRNA) was designed to target the first exon of grl ( Fig. 2A). Two grl deletion mutations, grl 5nt−/− mutants with a 5-nuleotide deletion and grl 19nt−/− mutants with a 19-nuleotide deletion, were identified (Fig. 2B). These mutants were predicted to produce a premature stop codon and encode a truncated peptide containing 4 amino acids in grl 5nt−/− mutants or 29 amino acids in grl 19nt−/− mutants, both of which lack the bHLH, Orange and YRPW domains (Fig. S2A). Some mutants survived to adulthood and were fertile. Considering that grl 5nt−/− is almost a null mutation, we chose grl 5nt−/− mutants to test the effects of grl loss of function on heart regeneration (Fig. S2B). Hearts from 5month-old grl 5nt−/− mutants were subjected to ventricular resection and then assayed for fibrin and fibrotic scar tissue using acid fuchsin-orange G (AFOG) staining at 30 dpa (Fig. 2C). We observed that more grl 5nt−/− mutant hearts contained large cardiac myofiber deposits and minimal fibrin or collagen deposits than injured wild-type (WT) sibling hearts in wound regions at 30 dpa ( Fig. 2D-G). Therefore, the grl mutant enhances cardiac muscle regeneration and reduces fibrotic scarring.
We next assessed injury-induced CM proliferation in grl 5nt−/− mutant hearts by conducting immunostaining of the DNA replication marker proliferating cell nuclear antigen (PCNA) and CM marker Mef2 following ventricular resection (Fig. 2C). We observed that PCNA-positive CMs were markedly increased in grl 5nt−/− hearts in comparison to injured WT sibling hearts (Fig. 2H,I). Quantification revealed that the CM proliferation index (number of PCNA + Mef2 + cells/number of Mef2 + cells) was significantly increased by 105% in grl 5nt−/− mutant hearts compared with control hearts (Fig. 2L). Similarly, in grl 19nt−/− mutants the CM proliferation index was elevated by 70% over control hearts (Fig. 2J,K,M). In contrast, the coronary vasculature was not impaired in grl 5nt−/− mutant hearts (Fig. S2C,D). Collectively, these findings demonstrate that mutations in grl enhance injury-induced CM proliferation.
Regenerating CMs normally undergo dedifferentiation, which is characterized by reactivation of cardiac embryonic or fetal genes as well as less organized sarcomeres in the myocardial injury region (D'Uva et al., 2015;Jopling et al., 2010;Kikuchi et al., 2010). We tested whether CM dedifferentiation might be regulated by Grl. In the region remote from injury sites, myofibrils organize in regular sarcomere units exhibiting cross-striations revealed by α-actinin in both conditional grl-overexpressing and control hearts (Fig. 3N,O).
Grl repressor negatively regulates transcripts encoding lysine methyltransferase Smyd2 in the regenerating myocardium Grl mediates transcriptional repression by binding E-box motifs (Fischer et al., 2005(Fischer et al., , 2002Heisig et al., 2012). We reasoned that potential Grl downstream genes should be downregulated in injured hearts overexpressing grl, but simultaneously upregulated in grldeficient hearts following damage. Therefore, we performed RNA sequencing (RNA-seq) on ventricular wound tissues collected from myocardial grl-overexpressing hearts, grl 5nt−/− mutant hearts and their respective control groups. The 4-HT-treated Tg(cmlc2:creER; cmlc2:nRSGG) and Tg(cmlc2:nRSGG) control animals, as well as grl 5nt−/− mutant and WT sibling animals, were subjected to ventricle resections. Total RNAs were isolated from injured ventricles at 7 dpa, which corresponds to when CM proliferation actively occurs (Fig. 4A), and sequenced. We generated and analyzed transcriptional profiles in grl-overexpressing hearts, grl 5nt−/− mutant hearts and their control groups (Fig. S4A,B, Tables S1 and S2). As expected, gene ontology (GO) analyses identified categories in 'regulation of cell proliferation', 'cardiac muscle cell differentiation' and 'heart contraction' that were enriched in the downregulated gene set in grl-overexpressing RNA profiles and in the increased gene set in grl 5nt−/− mutant RNA profiles ( Fig. 4B,C). However, we also obtained the enrichment of genes related to metabolic and biosynthetic processes, and to the immune response. We surveyed the list to identify potential Grl target genes that should be upregulated in grl 5nt−/− mutant hearts and simultaneously reduced in grl-overexpressing hearts following injury. About 50 genes were identified as fitting the criteria and being present in overlapped RNA profile datasets (Fig. S4C, Table S2). After filtering out genes with low expression levels (FPKM<1), we obtained 17 differential expression genes containing epigenetic regulatory genes (such as smyd2b), metabolism regulatory genes (such as znf326) and other unknown factors (Fig. 4D).
To assess whether smyd2 expression is normally upregulated by injury, we analyzed and compared smyd2b expression in uninjured and amputated WT hearts, considering smyd2a expression is very low (data not shown). Using ISH analyses, we observed a marked induction of smyd2b expression at the wound edge and a slight increase in the remote myocardial area in WT hearts at 7 dpa (Fig. 4H), whereas smyd2b transcripts were hardly detectable in uninjured WT hearts (Fig. 4F). Next, we determined whether smyd2 induction was controlled by Grl during regeneration. We noticed that grl 5nt−/− mutant hearts following resection displayed much stronger smyd2b upregulation at wound edges and remote areas compared with that in injured WT sibling hearts (Fig. 4I,H). Similar to uninjured WT hearts, smdy2b transcripts were hardly detectable in unresected grl 5nt−/− mutant hearts (Fig. 4G), suggesting that grl reduction is not sufficient to induce smyd2b expression in uninjured hearts. Remarkably, myocardial-specific grl overexpression abolished smyd2b induction at the wound region and remote areas, compared with that in injured Tg(cmlc2:nRSGG) control hearts (Fig. 4J,K). Collectively, these findings demonstrate that Grl negatively regulates smyd2b expression throughout the myocardium in response to injury.
Smyd2, acting as a key Grl transcriptional target, is required for heart regeneration by enhancing CM proliferation To assess whether Grl directly mediates transcriptional repression of smyd2, we performed luciferase reporter assays in HEK 293T cells. We generated a pGL3-smyd2-Fluc vector that links the smyd2 promoter containing four E-box motifs with firefly luciferase (Fluc) (Fig. 5A), as well as a pGL3-smyd2-Δ4Ebox-Fluc plasmid containing the Fluc-fused smyd2 promoter deleting four E-boxes (Fig. 5A). Grl-truncated mutants harboring deletions of the bHLH domain (Grl-ΔbHLH) and the YRPW motif (Grl-ΔYRPW) were also constructed (Fig. 5B). We co-transfected dual luciferase expression plasmid pGL3-smyd2-Fluc with pcDNA3-Grl, pcDNA3-Grl-ΔbHLH or pcDNA3-Grl-ΔYRPW. Grl transfection efficiently reduced smyd2 luciferase activity by 60% in comparison with pcDNA3 controls (Fig. 5C). By contrast, Grl-ΔbHLH mutant failed to reduce smyd2-luciferase activity, whereas the Grl-ΔYRPW mutant maintained the repression of smyd2-luciferase activity (Fig. 5C), indicating that Grl represses smyd2 transcription via its bHLH domain. Remarkably, deletion of four E-box motifs in the smyd2 promoter abrogated repression responsiveness to Grl when pcDNA3-Grl and pGL3-smyd2-Δ4Ebox-Fluc were co-transfected (Fig. 5C). To assess whether Grl binds the smyd2 promoter region, chromatin-immunoprecipitation (ChIP) was conducted by transfection of pEGFP-Flag-Grl or pEGFP-Flag control plasmids into cells. ChIP-PCR analyses revealed enrichment of the E-boxcontaining promoter region of smyd2 in the chromatinimmunoprecipitates using anti-Flag antibody, whereas the control IgG antibody did not enrich smyd2 promoter region (Fig. 5D,E). Whole genome ChIP-seq analyses for human HEY2 also identified enrichment of the SMYD2 promoter region (Heisig et al., 2012). Collectively, these results demonstrate that Grl mediates transcriptional repression of smyd2 by binding to the E-boxcontaining promoter, indicating that smyd2 is a direct transcriptional target of Grl, consistent with RNA profiling analyses in the setting of heart regeneration.
Because Smyd2 is induced mainly in the myocardium following injury, we tested whether Smyd2 functions as an essential regulator to promote CM proliferation during regeneration. LLY-507 and AZ505 are chemical inhibitors that specifically block the KMT activity of Smyd2 (Gao et al., 2017;Li et al., 2017;Nguyen et al., 2015). We administrated resected WT animals with LLY-507 or AZ505, and performed immunostaining with PCNA and Mef2 (Fig. 5F). Inhibition of Smyd2 KMT activity by LLY-507 treatment caused a marked reduction in CM proliferation, leading to a 25% reduction in the CM proliferation index in comparison with vehicle treatment (Fig. 5G,H,J). Similarly, administration of injured WT hearts with AZ505 inhibitor reduced the CM proliferation index by 34% (Fig. 5G,I,J). These findings indicate that the KMT activity of Smyd2 is required for CM proliferation during regeneration. Next, we assessed whether the elevated CM proliferation in grl mutant hearts might be due to Smyd2 induction following damage. We administrated grl 5nt−/− mutant animals with LLY-507 or AZ505 treatment following resections and analyzed the CM proliferation indices (Fig. 5F). Inactivation of Smyd2 activity with LLY-507 or AZ505 treatment resulted a significant reduction in the CM proliferation indices in grl 5nt−/− hearts compared with vehicletreated grl 5nt−/− hearts ( Fig. 5K; Fig. S5A-C). Considering that Smyd2 is a transcriptional repression target of Grl, we assessed CM proliferation in myocardial grl induction hearts with Smyd2 inactivation (Fig. 5F). LLY-507-treated grl induction hearts displayed almost the same percentage reduction in CM proliferation indices as in vehicle-treated hearts following resection ( Fig. 5L; Fig.  S5D-F), suggesting that smyd2 functions in the grl pathway during regeneration.
Smyd2 lysine methyltransferase mediates Stat3 activation during heart regeneration As a KMT, Smyd2 conducts its function by methylation of histones or non-histone substrates, including Stat3, during cell proliferation, survival and other events (Donlin et al., 2012;Gao et al., 2017;Li et al., 2017). We first assessed whether Smyd2 regulates histone methylation during Grl-mediated heart regeneration. Methylation of H3K4 and H3K9 were examined using cardiac wound tissues extracted from grl 5nt−/− mutant hearts or grl-overexpressing hearts. We learned that various methylation patterns of H3K4 and H3K9 appeared not to be affected in grl loss-or gain-of-function hearts (Fig. S6). On the contrary, levels of phosphorylated Stat3 (P-Stat3) and Smyd2 were both increased or reduced in grl 5nt−/− mutant hearts or grl-overexpressing hearts, respectively, whereas Stat3 levels failed to be affected (Fig. 6A,B), revealing correlation of the phosphorylated Stat3 level with the Smyd2 level during regeneration. Previous studies report that, during renal cystic tissue growth, Stat3 phosphorylation is induced by Smyd2 through methylation of its own protein (Li et al., , 2018a, leading to increased renal epithelial cell proliferation. Phosphorylated active Stat3 is also mitogenic and required for heart regeneration in zebrafish . We hypothesized that Smyd2 regulates myocardial regeneration by modulating the Stat3 methylation and phosphorylation activity. Anti-methylated lysine (methyl-lysine) antibody, a pan-methyl antibody available in the field, was used to recognize lysine-methylated proteins. We found that methylated proteins corresponding to the size of Stat3 were correlated with the P-Stat3 level or the Smyd2 level in grl 5nt−/− mutant hearts or grl-overexpressing hearts, respectively (Fig. 6A,B). We tested whether zebrafish Smyd2a or Smyd2b directly regulates Stat3 methylation and phosphorylation. Plasmids containing pcDNA3-Flag-Stat3 and pcDNA3-Myc-Smyd2a or pcDNA3-HA-Smyd2b were co-transfected into HEK 293T cells. We found that coexpression of Stat3 with Smyd2a or Smyd2b in HEK 293T cells resulted in an increase in levels of both Stat3 methylation and phosphorylation (Fig. 6C,E). In contrast, inactivating Smyd2 KMT activity with inhibitor AZ505 or (B) Deletion (Δ) fragments of Grl created for luciferase assays. (C) Luciferase activity in cells after co-transfection of pGL3-smyd2-Fluc with control plasmid vector, Grl, Grl-ΔbHLH or Grl-ΔYRPW expression plasmids, as well as transfection of pGL3-smyd2-Δ4Ebox-Fluc with Grl expression plasmids. The relative flyfire luciferase activity was normalized by Renilla luciferase and calculated as the ratio of each experimental group to the control group (n=3). (D,E) ChIP-PCR (D) and ChIP-qPCR (E) analyses of enrichment of Grl at its predicted binding sites in smyd2 promoter in HEK 293T cell. Enrichment levels of smyd2 promoter fragment were examined in pEGFP-Flag-grl and pEGFP-Flag transfected groups immunoprecipitated with anti-IgG or anti-Flag antibody. Genomic DNA isolated before IP was analyzed as the input control. Enrichment levels of qPCR analysis were normalized to that in the pEGFP-Flag-grl transfected group immunoprecipitated with IgG. (F) Experimental design for inhibitor treatment and CM proliferation analyses after ventricular resection in WT, grl mutant hearts, as well as in 4-HT-treated Tg(cmlc2;CreER;cmlc2:nRSGG) hearts and 4-HT-treated Tg(cmlc2;nRSGG) control hearts. Treatment at 6-7 dpa was with 5 µM LLY-507, 10 µM AZ505 or 0.5‰ DMSO. Red arrows represent experimental steps for Tg(cmlc2:creER;cmlc2:nRSGG) animals and Tg(cmlc2: nRSGG) control fish. Blue arrows represent experimental steps for grl 5nt−/− mutant fish and WT sibling fish. (G-I) Immunofluorescent section images of injured ventricles from DMSO-(G), LLY-507-(H) and AZ505-treated (I) WT fish at 7 dpa, stained with anti-PCNA (green) and anti-Mef2 (red) antibodies. Insets show higher-magnification images of the dashed boxes. Arrowheads indicate proliferating CMs. (J) Quantification of CM proliferation indices in 7 dpa ventricles derived from DMSO-, LLY-507-or AZ505-treated WT fish (n=5). (K) Quantification of CM proliferation indices in 7 dpa ventricles derived from DMSO-(n=11), LLY-507-(n=8) or AZ505-treated grl 5nt−/− fish (n=10). (L) Quantification of CM proliferation indices in 7 dpa ventricles derived from DMSOtreated Tg(cmlc2:nRSGG) control fish (n=5), DMSO-treated Tg(cmlc2:creER;cmlc2:nRSGG) fish (n=4) and LLY-507-treated Tg(cmlc2:creER;cmlc2: nRSGG) fish (n=5). Data presents as mean±s.e.m. *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001, Student's t-test (unpaired, two-tailed). Scale bars: 100 µm. LLY-507 not only caused a reduction in methylated Stat3, but also led to a decrease in phosphorylated Stat3 (Fig. 6D,F). Furthermore, co-immunoprecipitation (Co-IP) analyses revealed physical interactions between Flag-Stat3 with Myc-Smyd2a or HA-Smyd2b (Fig. 6G,H). Together, these findings indicate that zebrafish Smyd2 interacts with Stat3 to regulate the methylation and phosphorylation of Stat3 in cultured cells, in agreement with previous murine studies (Li et al., 2018a).

DISCUSSION
In this study, we describe the transcriptional control of methylation modulation in governing CM renewal and heart regeneration. We demonstrate that Grl functions as a negative regulator to restrain heart regeneration in zebrafish, and identify KMT Smyd2 as the Grlrepressing target in this context. Following cardiac injury, Grl reduction triggers Smyd2 induction at the injured myocardial cell edge, leading to CM proliferation and heart regeneration. We show that zebrafish Smyd2 functions as a KMT and regulates CM renewal in part by modulating the activity of Stat3. These findings suggest that manipulation of the Grl-Smyd2 network might be instrumental in potential strategies to relieve cardiac barriers and accelerate heart regeneration in mammalian hearts.
The limited regenerative potential of adult mammalian hearts can be attributed to cell-intrinsic barriers that prevent CMs from entering the cell cycle (González-Rosa et al., 2018;Li et al., 2015;Tzahor and Poss, 2017). Recent studies have uncovered a few molecular blocks that impede zebrafish heart regeneration (Dogra et al., 2017;Jopling et al., 2012;Missinato et al., 2018). We demonstrate that Grl acts as a transcriptional repressor to limit injury-induced CM proliferation. This is consistent with its roles in negative regulation of embryonic heart development and growth in zebrafish (Jia et al., 2007). Hey family genes play multiple roles during embryogenesis, with expression in various cell types (Fischer et al., 2002;Nakagawa et al., 1999;Zhong et al., 2000). In the embryonic heart, grl is enriched in the myocardial compact zone, as well as weakly expressed in the trabecular region and ventricular endocardium (Jia et al., 2007;Miao et al., 2018). During murine heart development, hey2 expression in myocardial compact zones and ventricular endocardium is regulated by Notch signaling (Miao et al., 2018), but its functional mechanisms remain unknown. In the adult zebrafish heart, we observed that grl expression is restricted to myocardial cells but not the endocardium or coronary vasculature. Notably, myocardial grl expression in the adult heart is reduced in response to heart damage. By contrast, Notch signaling is upregulated in endocardial and epicardial cells following injury, stimulating CM renewal in zebrafish (Zhao et al., 2019). Thus, the role of myocardial Grl in limiting heart regeneration is independent of endocardial Notch signaling. We observed that grl expression declines to the lowest level from 1 dpa to 3 dpa, increases to some extent at 7 dpa, and gradually restores from 14 dpa to 30 dpa. During regeneration, many factors that stimulate CM proliferation are induced after injury, which can counteract inhibitory factors such as Grl for CM proliferation. We believe that the spatiotemporal regulation of genome-wide networks enables CMs to regenerate after injury. Overall, the temporal expression patterns of grl are consistent with the functional model in which Grl negatively regulates CM proliferation during zebrafish heart regeneration. It is important to test whether murine Hey2 expression is reduced following cardiac injury, and whether conditional Hey2 deletion in mice can promote CM regeneration.
Smyd2 represents a novel class of KMTs and has been reported to methylate histone and non-histone substrates (Donlin et al., 2012;Gao et al., 2017;Li et al., 2017). Our studies identify a crucial role for Smyd2 during heart regeneration. However, Smyd2 is dispensable for cardiac development in mice (Diehl et al., 2010), suggesting that Smyd2 is required for injury-induced heart regeneration or functions in response to other cardiac stresses, consistent with our observations that zebrafish smyd2 expression was not detectable in uninjured adult hearts. In addition to Stat3, Smyd2 methylates other non-histone substrates, including BMP receptor 2 (BMPR2), NF-κB and HSP90 (Donlin et al., 2012;Gao et al., 2017;Li et al., 2017). It is likely that Smyd2 has multiple methylation substrates during cardiac regeneration. It is important to assess whether Smyd2 regulates methylation of NF-κB or BMPR2 following injury. Alternatively, Smyd2 might methylate chaperone Hsp90 to promote interaction of the methyl-Hsp90sarcomeric protein complex and myofilament organization during heart regeneration (Donlin et al., 2012). Future studies are warranted to investigate various methylation substrates of Smyd2 during cardiac regeneration, and to assess the extent to which methylation modulations play crucial roles during Grlmediated transcriptional controls. The findings of such studies will reveal novel transcriptional regulations and various methylation controls that direct vertebrate heart regeneration, which will be relevant in developing strategies for regeneration interventions in humans.

Generation of grl mutants
The grl mutants were generated using CRISPR/Cas9-mediated mutagenesis. Cas9 capped mRNA was produced by in vitro transcription from a pGH-T7-zCas9 vector (Liu et al., 2014) using mMESSAGE mMACHINE T7 Transcription Kit (Ambion). The sgRNA target site of grl was designed via the CRISPRscan website. The sgRNA transcription template was PCR amplified and then transcribed using MAXIscript T7 Kit (Ambion). Cas9 capped mRNA and sgRNA were co-injected into zebrafish embryos at the one-cell stage. The final concentrations of Cas9 mRNA and sgRNA were 400 ng/µl and 30 ng/µl, respectively.
Generation of Tg(grl:EGFP), Tg(grl:mCherry) and Tg(cmlc2:loxP-nlsmCherry-STOP-loxP-grl-EGFP) zebrafish To generate Tg(grl:EGFP) zebrafish, a 7.8 kb fragment containing the grl promoter sequence was amplified from BAC14 (Zhong et al., 2000) and cloned into the pCR-Blunt II-TOPO vector. A 2888 bp tol2 3′end-tol2 5′ end-EGFP cassette was amplified from pT2KXIGΔin vector (Urasaki et al., 2006), discarding the original EF1p promoter. After verification of the grl promoter sequence, a 7.6 kb grl promoter fragment flanked by a 24-nucleotide overlap of tol2 3′end-tol2 5′end-EGFP cassette was amplified and assembled with this cassette using Gibson Assembly Master Mix (NEB). DNA construct (40 ng/µl) and Tol2 transposase mRNA (50 ng/µl) prepared from pCS-TP were co-injected into zebrafish embryos at the one-cell stage. To generate the Tg(grl:mCherry) line, the EGFP sequence in pT2KXIGΔin vector was replaced with the mCherry fragment. The methods used were the same as described above.
To generate the Tg(cmlc2:nRSGG) strain, full name as Tg(cmlc2:loxP-nlsmCherry-STOP-loxP-grl-EGFP), the MultiSite Gateway Kit (Invitrogen) was used to simultaneously clone a cmlc2:loxP-nlsmCherry-STOP-loxP cassette, grl CDS fragment and EGFP fragment into the pDESTol2pA2 destination vector. The cmlc2:loxp-nlsmCherry-STOP-loxp cassette was flanked by attB4 and attB1r. The grl CDS fragment was flanked by attB1 and attB2. The EGFP fragment was flanked by attB2r and attB3. Three attP-containing donor vectors ( pDONR P4-P1R, pDONR 221 and pDONR P2R-P3) were used in separate BP recombination reactions with the above-mentioned attB-containing PCR fragments to generate three entry clones. The three entry clones and the pDESTol2pA2 destination vector were used together in an LR recombination reaction between attL-and attRflanked regions to create the cmlc2:loxP-nlsmCherry-STOP-loxP-grl-EGFP construct. The construct was co-injected with Tol2 transposase mRNA into one-cell stage embryos to generate transgenic lines, which were screened for nlsmCherry and then EGFP upon Cre-mediated recombination.

Ventricular apex resection and chemical treatment
Zebrafish at 5-6 months of age were used for ventricular resection surgery as previously described (Poss et al., 2002). About 20% of ventricular muscle was excised from the ventricle apex after anesthesia as previously described (Poss et al., 2002). 4-HT (Sigma) was dissolved in ethanol to give a 20 mM stock solution. To induce expression of grl in adult zebrafish heart, Tg(cmlc2:creER;cmlc2:nRSGG) fish were incubated with 5 µM 4-HT for bath treatment for 2 days. After 24 hours, fresh 4-HT was prepared and added for an additional 24 h treatment. AZ505 ditrifluoroacetate (MCE) was dissolved in DMSO to give a 20 mM stock solution and diluted in aquarium water to 10 µM for bath treatment from 6 to 7 dpa. LLY-507 (MCE) was dissolved in DMSO to give a 10 mM stock solution and diluted in aquarium water to 5 µM for bath treatment from 6 to 7 dpa. DMSO (0.5‰) was used as a vehicle control for AZ505 and LLY-507 treatment.
To quantify regenerative status shown by AFOG staining at 30 dpa, we scored heart sections for regeneration, with 'class 1' indicating complete regeneration showing contiguous ventricle wall, 'class 2' indicating partial regeneration showing some collagen deposition and 'class 3' indicating a block in regeneration showing fibrin and collagen deposition. Multiple sections of each heart were scored according to these criteria.
To quantify CM proliferation index, four to six sections showing the largest injury area were selected from each heart, and images were taken using a 10× objective. The number of Mef2 + and Mef2 + PCNA + cells were manually counted using ImageJ software within a defined region including almost all Mef2 + PCNA + cells around the injury site (1344×384 pixel region of the 1388×1040 pixel images). The percentages of Mef2 + PCNA + cells from all sections were averaged to determine a proliferation index for each heart.

qPCR analysis
For cDNA synthesis using PrimeScript II 1st Strand cDNA Synthesis Kit (TaKaRa), 1 µg of total RNA extracted from ventricles was used. RT-qPCR was carried out on biological triplicates with SYBR Premix Ex Taq II (Takara) using a Roche LightCycler 480 II system. The relative expression levels of target genes were normalized to β-actin and quantified by the 2 −ΔΔCT method (Livak and Schmittgen, 2001).

RNA-seq and data analysis
Ventricles at 7 dpa were collected from 4-HT-treated Tg(cmlc2:creER; cmlc2:nRSGG) and Tg(cmlc2:nRSGG) animals, as well as grl 5nt−/− and WT sibling zebrafish. Six ventricles were pooled per biological replicate. Each group contained three biological replicates. Total RNAs of the ventricles were isolated using TRIzol Reagents (Life Technologies) and checked for a RIN number to inspect RNA integrity using the Agilent 2100 Bioanalyzer system (Agilent Technologies). Qualified total RNAs were further purified by RNAClean XP Kit (Beckman Coulter) and RNase-Free DNase Set (Qiagen). Libraries were constructed using the TruSeq RNA Sample Preparation Kit (Illumina) followed by a quality check with Qubit 2.0 Fluorometer (Invitrogen) and the Agilent 2100 Bioanalyzer system (Agilent Technologies), Sequencing was carried out on the Illumina HiSeq X Ten (Illumina) at Shanghai Biotechnology Corporation.
Sequenced raw reads were preprocessed using Seqtk to filter out rRNA reads, sequencing adapters, short-fragment reads and other low-quality reads to obtain clean reads for data analyses. Hisat2 (version 2.0.4) was used to map the clean reads to the zebrafish GRCz10 reference genome. After genome mapping, the mapped reads of genes were converted to FPKM (fragments per kilobase of exon model per million mapped reads) using Stringtie (version 1.3.0) (Pertea et al., 2016(Pertea et al., , 2015 for standardization of gene expression. The fold-changes of genes were calculated according to the FPKM in each sample. Genes with differential expression levels between samples were analyzed using edgeR (Robinson et al., 2010), with GO enrichment and KEGG pathway enrichment analyses.

Transfection, ChIP and Luciferase assay
The Flag tag sequence fused with zebrafish grl CDS was ligated into the pEGFP-N1 vector. The plasmids were transfected into HEK 293T cells using Lipo 3000 Transfection Reagent (Invitrogen) as previously described (Ni et al., 2011). At 48 h after transfection, ChIP was performed using EZ-ChIP Kit (Millipore). HEK 293T cells were crosslinked with 1% formaldehyde for 10 min. After incubation with glycine to stop crosslinking, cells were rinsed in PBS and homogenized in SDS lysis buffer. Cell lysates were sonicated with a FS-300 Ultrasonic Processing apparatus. ChIP was then performed by using an antibody against Flag (mouse; Sigma) or a normal mouse IgG antibody (Millipore) as negative control, according to the manufacturer's instructions.
HEK 293T cells were cultured overnight at 8×10 4 cells per well in 24-well plates. Firefly luciferase pGL3-smyd2-Fluc plasmid vector (0.5 µg) and Renilla luciferase pRL-TK plasmid vector (5 ng, as a normalization control) were co-transfected with different Grl mutant plasmids (0.5 µg) using 1 µl of Lipofectamine 2000 (Life Technologies) according to the manufacturer's protocol. Approximately 24 h later, cell lysates were collected and the firefly luciferase and Renilla luciferase activities measured using a Dual-Luciferase Reporter Assay System (Promega). The experiment was repeated three times.

Methylation and phosphorylation assay
For detection of methylation and phosphorylation of Stat3, HEK 293T cells were transfected with pcDNA3-Myc-Smyd2a, pcDNA3-HA-Smyd2b and pcDNA3-Flag-Stat3 as indicated. When inhibitors were used, cells were treated with AZ505 at different concentrations for 6 h before sample collection, whereas LLY-507 treatment was for 28 h. At 48 h post transfection, cells were washed with PBS twice and lysed in denaturing IP lysis buffer (150 mM NaCl, 0.1% NP40 or Triton X-100, 25 mM Tris-HCl, pH 8.0, 5 mM EDTA, 10% glycerol, 1% SDS, 1× protease inhibitor cocktail, 1× phosphatase inhibitor cocktail) and treated at 100°C for 20 min. The lysates were cleared by centrifugation at 13,600 g for 20 min. The supernatant was diluted with denaturing binding buffer without SDS to give a final SDS concentration of 0.1%, and then directly incubated with anti-Flag-M2-affinity beads (Bimake, Houston, TX) for 4 h. After extensive washing with IP washing buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.1% Triton X-100, 1 mM EDTA), the immunoprecipitated proteins were boiled in 1× SDS loading buffer and analyzed by western blot.

Imaging
AFOG and ISH images were taken using a Nikon Eclipse Ni microscope with a Nikon Digital Sight DS-Ri1 camera. Immunostaining images were taken using a Zeiss Axio Observer.Z1 microscope, Zeiss LSM 710 confocal microscope and Andor Dragonfly 500 High Speed confocal microscope.

Statistical analysis
GraphPad software was used to perform statistical analysis. Data were analyzed by the two-tailed Student's t-test and Chi-square tests. For the Student's t-test, data are represented as mean±s.e.m. and considered significant at P<0.05.