Tbx20 Induction Promotes Zebrafish Heart Regeneration by Inducing Cardiomyocyte Dedifferentiation and Endocardial Expansion

Heart regeneration requires replenishment of lost cardiomyocytes (CMs) and cells of the endocardial lining. However, the signaling regulation and transcriptional control of myocardial dedifferentiation and endocardial activation are incompletely understood during cardiac regeneration. Here, we report that T-Box Transcription Factor 20 (Tbx20) is induced rapidly in the myocardial wound edge in response to various sources of cardiac damages in zebrafish. Inducing Tbx20 specifically in the adult myocardium promotes injury-induced CM proliferation through CM dedifferentiation, leading to loss of CM cellular contacts and re-expression of cardiac embryonic or fetal gene programs. Unexpectedly, we identify that myocardial Tbx20 induction activates the endocardium at the injury site with enhanced endocardial cell extension and proliferation, where it induces the endocardial Bone morphogenetic protein 6 (Bmp6) signaling. Pharmacologically inactivating endocardial Bmp6 signaling reduces expression of its targets, Id1 and Id2b, attenuating the increased endocardial regeneration in tbx20-overexpressing hearts. Altogether, our study demonstrates that Tbx20 induction promotes adult heart regeneration by inducing cardiomyocyte dedifferentiation as well as non-cell-autonomously enhancing endocardial cell regeneration.


INTRODUCTION
Adult mammalian hearts have limited regeneration capacity in response to cardiac damage. Injured hearts lose cardiac muscle and replace with fibrotic scar tissue, ultimately leading to arrhythmia and heart dysfunction (Xin et al., 2013;Tzahor and Poss, 2017). However, zebrafish and neonatal murine hearts exhibit increased regeneration capacity after various insults (Kikuchi and Poss, 2012;Li et al., 2015). Heart regeneration occurs through diverse mechanisms including activation of epicardial, myocardial, or endocardial tissues (Kikuchi and Poss, 2012). Although various mitogenic factors and signaling pathways have been identified to enhance heart regeneration (Kim et al., 2010;Heallen et al., 2013;D'Uva et al., 2015;Gemberling et al., 2015;Wang et al., 2015;Liu and Zhong, 2017;Mohamed et al., 2018;Singh et al., 2018), the signaling and transcriptional control of heart regeneration by myocardial dedifferentiation and endocardial activation are largely unknown. Understanding injury-induced heart regeneration will provide therapeutic strategies to empower regenerative capacity to the diseased human heart.
T-box transcription factor 20 (Tbx20), a key cardiac transcriptional factor, is required for heart development and homeostasis (Greulich et al., 2011). In humans, the heterozygous nonsense or missense mutations of TBX20 are associated with diverse cardiac pathologies such as dilated cardiomyopathy, atrial septal defect, cardiac valve defects and tetralogy of Fallot (Kirk et al., 2007;Huang et al., 2017). Both tbx20-null zebrafish and mice are embryonic lethal and exhibit deleterious cardiovascular malformations with defects of CM proliferation and heart tube looping (Cai et al., 2005;Singh et al., 2005;Lu et al., 2017). Tbx20-deficient mice also display defects in cardiac chamber differentiation, endocardial cushion formation, and atrioventricular canal (AVC) patterning (Singh et al., 2005;Stennard et al., 2005;Shelton and Yutzey, 2007;Cai et al., 2013). Myocardial-specific Tbx20 ablation in adult mice leads to thinner ventricle wall and cardiomyopathy accompanied with arrhythmias (Shen et al., 2011;Sakabe et al., 2012). Conversely, inducible tbx20 overexpression in embryonic cardiomyocytes leads to increased CM proliferation and thickening of the myocardium in adult hearts (Chakraborty et al., 2013). Myocardial-specific tbx20 overexpression in zebrafish embryos also results in enlarged heart with both increased cardiac progenitor cell formation and the proliferation of differentiated CMs (Lu et al., 2017). Recent studies report that Tbx20 overexpression in adult mouse hearts after myocardial infarction increases CM proliferation in the injury border zone and improves cardiac function recovery (Xiang et al., 2016). Despite that previous studies demonstrate essential roles of Tbx20 transcription factor during heart development, injury repair and congenital heart disease, it is currently not understood whether and how endocardial cells respond to Tbx20 induction in the myocardium after cardiac damage, and the extent to which Tbx20 regulates CM dedifferentiation and proliferation during heart regeneration.
The heart develops through generation of CMs and tightly associated endocardial cells (Staudt and Stainier, 2012). Endocardial cells represent a subset of a larger endothelial cell pool (Harris and Black, 2010). During development, the endocardium is organized into arterial and venous subpopulations with comparable gene expression profiles (Staudt and Stainier, 2012). Endocardial differentiation and growth occur without an accretion of external cells, in a manner independent of vascular endothelial growth factor (VEGF) signaling (Dietrich et al., 2014). After cardiac injury, activated endocardium coincides with changes in cell morphology and gene expression (Kikuchi et al., 2011b). The activation and maturation of the endocardium require Notch signaling, which supports myocardial regeneration (Munch et al., 2017;Zhao et al., 2019). However, no signaling factor or molecular program has been shown to be essential for endocardial cell proliferation during zebrafish heart regeneration.
In this report, we have determined Bmp6 signaling as an early endocardial injury-response to myocardial Tbx20 induction, which promotes endocardial cell regeneration, a previously unrecognized mechanism. Tbx20 is also sufficient to induce injury-induced CM dedifferentiation, thus stimulate CM proliferation. Overall, our findings reveal novel roles and distinct mechanisms of myocardial Tbx20-mediated network in governing cardiac muscle production and endocardial cell proliferation during heart regeneration.

Tbx20 Is Induced in the Regenerating Zebrafish Heart Following Injury
To define the spatiotemporal expression pattern of T-Box Transcription Factor 20 (Tbx20) during adult heart regeneration, we first evaluated tbx20 expression during the window of zebrafish cardiac regeneration by in situ hybridization (ISH). In the uninjured adult heart, we detected faint tbx20 expression in the ventricles and atriums (Figures 1A,B). Within 1 day post amputation (dpa), tbx20 expression was induced in both the myocardium and atrial epicardium (Figures 1C,D). By 3 dpa, tbx20 was strongly upregulated in the ventricular and the atrial myocardium (Figures 1E,F). Specifically, more tbx20 + cells were accumulated in the injury border zone of the ventricle compared to the remote (uninjured) zone at 3 dpa and 7 dpa (Figures 1E-H). Furthermore, qPCR analyses validated the marked upregulation of tbx20 at the border zone of injured ventricles and the atrium at 1 dpa, 3 dpa and 7 dpa (Supplementary Figure S1A). The induction of tbx20 at the injury border zone and the atrial epicardium was also detectable at 5 days post cardiac cryoinjury (dpci), an independent injury approach (Supplementary Figures S1B-E). However, tbx20 induction at the atrial epicardium was hardly detectable at uninjured hearts ( Figure 1B and Supplementary Figure S1D). As controls, tbx20 transcripts were not detectable in uninjured and injured hearts from 1 dpa to 7 dpa using tbx20 sense probes (Supplementary Figures S1F-M), confirming the specificity of tbx20 upregulation in the ventricle and the atrial epicardium after cardiac damage.
To unambiguously define the myocardial expression of tbx20, we generated a Tg(tbx20: moxGFP) transgenic zebrafish expressing monomeric oxidizing GFP (moxGFP) driven by a tbx20 promoter-based upstream region. We were able to detect CMs that expressed weak moxGFP in the ventricle of uninjured hearts after co-immunostaining for GFP and α-actinin (Z-disk marker) ( Figure 1I and Supplementary Figure S1N). In the injured adult heart, extensive and stronger moxGFP signals were observed in the ventricles, as well as at the injury border zone by 7 dpa and 5 dpci, respectively, ( Figure 1J and Supplementary Figure S1O). While moxGFP signals were hardly detectable in the epicardium in uninjured hearts (Supplementary Figures  S1P, S2A), strong GFP signals colocalized with epicardial marker pan-cytokeratin (PCK) were detectable in the atrial epicardium of injured hearts (Supplementary Figures S1Q, S2B). These FIGURE 1 | Cardiac injury triggers a localized increase in tbx20 expression. (A-H) Representative images of ISH for tbx20 on the heart sections of uninjured (A,B) and injured ventricles and atriums at different time points as indicated (C-H). Red arrowheads indicate atrial epicardium. (I,J) Confocal images of uninjured (I) and 7 dpa (J) heart sections from Tg(tbx20:moxGFP) zebrafish immunostained for GFP and α-actinin (red). DAPI marks nuclei. Boxed areas are magnified on the right panels. Dashed lines in (J) demarcate amputation planes. Scale bar: 100 µm (A-J); 20 µm (inset images).
results indicate the tbx20 expression is induced in the ventricular myocardium and the atrial epicardium after cardiac injury, consistent with ISH analyses.
During zebrafish heart regeneration, newly formed CMs primarily come from the proliferation of pre-existing CMs (Jopling et al., 2010;Kikuchi and Poss, 2012). To assess whether tbx20 promotes CM proliferation during heart regeneration, TRE3G:tbx20 CMOE and control TRE3G:tbx20 zebrafish were treated with DOX ( Figure 2B), and subjected to ventricular apex resection next day. Heart sections at 7 dpa, a time point when CM proliferation peaks, (Wills et al., 2008;Wang et al., 2011), were immunostained with antibodies against proliferation marker PCNA and the CM nuclear marker Mef2. The results revealed that CM proliferation in injured TRE3G:tbx20 CMOE hearts markedly increased by ∼61%, compared with that in control hearts (25.1 ± 1.4% versus 15.5 ± 1.3%) (Figures 2D,E). By contrast, tbx20 overexpression had no discernible effects on CM proliferation in uninjured adult hearts after 7 days of DOX treatment in which PCNA + Mef2 + CMs were not detectable in tbx20 overexpressing hearts FIGURE 2 | Myocardial tbx20 overexpression in adult hearts promotes CM proliferation and reduces fibrotic scars. (A) Schematic diagram of transgenes of the control line Tg(TRE3G:tbx20) and the inducible myocardial tbx20 overexpression line Tg(TRE3G:tbx20 CMOE ) with DOX treatment. (B) Experimental strategy employed to induce myocardial tbx20 overexpression during heart regeneration. (C) Representative images of ISH with tbx20-E2A-mCherry probe on 7 dpa heart sections from DOX treated Tg(TRE3G:tbx20) and Tg(TRE3G:tbx20 CMOE ) fish. Scale bar: 100 µm. (D) Upper panels: Representative confocal fluorescence images of heart sections from 7 dpa zebrafish immunostained for PCNA (green) and Mef2 (red). Insets showing high-magnification images of proliferating cardiomyocytes, arrowheads indicate PCNA + Mef2 + cells. Lower panels: Representative images of heart sections from 21 dpa fish stained with AFOG, muscle stained brown, collagen is blue and fibrin is red. The genotypes of fish were indicated above the images. Scale bar: 100 µm. (E) Quantification of CM proliferation at border zone and injury site on 7 dpa heart sections (n = 7 in each group). (F) Quantification of scar area at 21 dpa (n = 9 in each group). The fish in (C-F) were treated with DOX as indicated in (B). Each value in (E,F) represents mean ± SEM, **p < 0.01, ***p < 0.001. and control hearts (Supplementary Figure S4). Collectively, these results demonstrate that myocardial overexpressing tbx20 stimulates injury-induced CM proliferation.
Given that increased tbx20 expression is able to enhance CM proliferation at injured sites at an early regeneration stage, we reasoned that long-term high transcription level of tbx20 reduced fibrotic scars and hastened wound healing. To test this possibility, heart regeneration of DOX-treated TRE3G:tbx20 CMOE and TRE3G:tbx20 fish at 21 dpa were evaluated using Acid Fuchsin-Orange G (AFOG) staining of heart cryosections. Our study showed that hearts from TRE3G:tbx20 CMOE fish were evidenced by contiguous cardiac muscle formation and reduced fibrotic scars at the injured ventricle apex, whereas hearts from TRE3G:tbx20 fish remained variable of prominent scar tissues (Figures 2D,F).

Myocardial-Specific tbx20
Overexpression Enhances Injury-Induced CM Dedifferentiation Cardiomyocyte dedifferentiation, a transition from mature state to immature state, is a mechanism to ensure subsequent CM proliferation that naturally occurs in response to cardiac injury in neonatal mouse and adult zebrafish (Jopling et al., 2010;Porrello et al., 2011;D'Uva et al., 2015). This process is characterized by disassembly of sarcomeric structure, loss of cell-cell adhesion and re-expression of cardiac embryonic, fetal or progenitor genes (Kubin et al., 2011;D'Uva et al., 2015). Since increased tbx20 expression enhances CM proliferation and cardiac regeneration (Figure 2), we asked whether tbx20 cardiac overexpression was capable of boosting CM dedifferentiation following cardiac injury. To test this idea, we determined morphological and molecular changes of CMs in DOX-treated control and TRE3G:tbx20 CMOE hearts. We observed a marked reduction of a cell tight junction marker ZO-1 in α-actininmarked CMs in the injury border zone in TRE3G:tbx20 CMOE hearts compared to that in control hearts (Figures 3A-B

-I).
N-cadherin is a marker of cell-cell adhesion junction localized in the intercalated disks between neighboring CMs (Luo and Radice, 2003;Vite and Radice, 2014;Li et al., 2019). In injured hearts overexpressing tbx20 in CMs, we observed a reduction of N-cadherin in the border zone of the injured myocardium, indicating a loss of cell-cell contact between CMs (Figures 3C' ,D' and Supplementary Figure S5). Notably, we observed CMs marked by cardiac troponin T (cTnT) in the wound edge exhibited greater extent of sarcomere disassembly that were devoid of myofibril striations in tbx20-overexpressing hearts than that in control hearts (Figures 3C-E).
Concomitantly, q-PCR analyses showed that expression of CM dedifferentiation markers, including cardiac fetal markers, alpha-smooth muscle actin (α-SMA), natriuretic peptide a (nppa) and natriuretic peptide b (nppb) (Dirkx et al., 2013;Man et al., 2018), as well as a progenitor cell marker runx1 (Poling et al., 2012;D'Uva et al., 2015;Wang et al., 2017) were significantly increased in DOX-treated TRE3G:tbx20 CMOE hearts compared with control hearts ( We observed that α-SMA was markedly induced in the αactinin-marked myocardial compact layer ( Figure 3H and Supplementary Figures S6A, A-I-A-II"), and some of α-SMA signal was detectable in the trabecular layer adjacent to the injury site ( Figure 3H and Supplementary Figures S6A, A-II-A-II") in DOX-treated TRE3G:tbx20 CMOE injured hearts. In contrast, α-SMA re-expression was restricted to a small injury region in control hearts ( Figure 3G). Furthermore, Runx1 was upregulated in more regenerating CMs in TRE3G:tbx20 CMOE injured hearts than that in control wounded hearts ( Figures 3J,K,M). Similarly, ISH analyses revealed the increased expression of α-SMA and nppb, as well as cardiac progenitor markers gata4, gata5 and hand2 in injured tbx20-overexpressing hearts in comparison to control hearts ( Figure 3I and Supplementary  Figures S7A-H). qPCR analyses validated the upregulation of gata4, gata5 and hand2 in TRE3G:tbx20 CMOE hearts following ventricular apex resection (Supplementary Figure S7I). Taken together, our data demonstrated that enhanced cardiac tbx20 expression favors induction of cardiac fetal and progenitor gene programs, resulting in CM dedifferentiation and proliferation during regeneration.

Tbx20 Mediates Various Genetic Circuits Regulating Zebrafish Heart Regeneration
To decipher the molecular basis in response to cardiac injury with enhanced cardiac tbx20 expression, we analyzed gene expression profiles of the apical halves of the resected ventricles from DOX-treated TRE3G:tbx20 CMOE and TRE3G:tbx20 zebrafish at 7 dpa ( Figure 4A). We found that 1880 genes were differentially expressed in TRE3G:tbx20 CMOE hearts (Log FC > 0.5, pvalue < 0.05) compared to TRE3G:tbx20 hearts, in which 747 of them were upregulated and 1133 were downregulated (Supplementary Table S1).
The injured hearts were immunostained with cTnT antibody that labels CMs and GFP antibody that recognizes flk:GFPmarked endocardial cells ( Figure 5A). Previous studies report that endocardial expansion within the injury site is mediated through extension and migration of existing endocardial cells from the uninjured site (Munch et al., 2017). In the remote (uninjured) region of the tbx20-overexpressing and control heart, a coherent network of flk:GFP-marked endocardial cells surrounded cardiac muscles labeled by cTnT (Figures 5B,B-I,C,C-I). Within the injury site, we observed that a population of disorganized endocardial cells extended from the uninjured site and displayed cell protrusions, suggestive of migration (arrows in Figures 5B-II,C-II; Munch et al., 2017). Remarkably, Tbx20-overexprssion hearts in DOX-treated Tg(TRE3G:tbx20 CMOE ; flk:GFP) animals exhibited a marked increase in endocardial cells labeled by flk:GFP extending the injury site (Figures 5C,C-II). Quantification analyses indicated the proportion of the injured area occupied by flk:GFP + cells was increased in tbx20-overexpressing hearts (35.2 ± 2.5%) compared to that in controls (21.2 ± 2.2%) (Figure 5F). These findings suggest that myocardial tbx20 promotes extension and migration of endocardial cells into the injury site. We also appraised the endocardial cell proliferation adjacent to and within the cryoinjury site from DOX-treated TRE3G:tbx20 CMOE hearts at 5 dpci by co-immunostaining for endothelial/endocardial cell nuclear marker Fli1 and cell proliferation marker PCNA ( Figure 5A). We observed approximately threefold increase of proliferating endocardial/endothelial cells (Fli1 + PCNA + ) inTRE3G:tbx20 CMOE hearts (16.4 ± 1.0%) in comparison with that of control hearts (4.9 ± 1.5%) (Figures 5D,E,G). Taken together, these findings indicate that myocardial-specific tbx20 overexpression promotes endocardial cell migration and proliferation during heart regeneration.
Next, we assessed whether the endocardium is activated in myocardial tbx20 overexpressing hearts. Nfatc1 is required for endocardial development in the heart and serves as a specific marker for endocardial activation after cardiac injury (Wong et al., 2012;Munch et al., 2017). ISH analyses revealed a striking upregulation of nfatc1 at the injury site at 5 dpci in DOX-treated TRE3G:tbx20 CMOE hearts (Figures 5H,I). In consistent, re-survey of our RNA-seq data revealed an increase of two other endocardial markers, aldehyde dehydrogenase 1 family member a2 (aldh1a2) and leptin b (lepb), in DOXtreated TRE3G:tbx20 CMOE hearts (Supplementary Table S3). Previous studies report that the Retinoic Acid synthesis enzyme Aldh1a2 and the secreted regulator of energy homeostasis protein Lepb are induced in the endocardium after cardiac injury, indicative of endocardial activation (Kikuchi et al., 2011b;Kang et al., 2016). We also validated the upregulation of aldh1a2 and lepb at endocardial cells in the injured region in myocardial tbx20 overexpressing hearts by ISH analyses (Figures 5J-M). Altogether, these results suggest that inducible tbx20 overexpression in the adult myocardium contributes to endocardial cell migration and regeneration through endocardial cell activation, revealing crosstalk between the myocardium and endocardial cells during regeneration.

Tbx20 Induction Augments Endocardial Bmp6 Signaling During Heart Regeneration
To identify the molecular signaling involved in activating the endocardial regeneration program in the injured hearts with CM-specific tbx20 overexpression, we searched differentially expressed genes participating in endocardial/endothelial activation in our RNA-seq/GO analysis data generated from control and tbx20-overexpressing hearts. We found a profound upregulation of BMP signaling, including bmp6 ligand and its downstream targets, inhibitor of DNA-binding proteins (id1, id2a and id2b) in myocardial tbx20 overexpressing hearts following ventricular injury (Figures 6A,B and Supplementary Table S3). ISH analyses revealed that the expression of bmp6, id1 and id2b was marked induced in the injury border zone and the inside of the wound endocardium at 5 dpci hearts from DOX-treated TRE3G: tbx20 CMOE fish in comparison to TRE3G: tbx20 control fish (Figures 6C-I'). To determine the identity of bmp6 + cells in the injury site of hearts, we performed fluorescence in situ hybridization (FISH) experiments using bmp6 or id2b antisense-mRNA probes on DOX-treated Tg(TRE3G:tbx20 CMOE ;flk:GFP) hearts and Tg(TRE3G:tbx20;flk:GFP) control hearts following cryoinjury ( Figure 6J). These fluorescence hybridization hearts were subjected to double immunostaining of GFP antibody recognizing the flk:GFP transgenic endocardium and α-actinin antibody labeling cardiac muscle at 5dpi (Figure 6J). We found that bmp6 transcripts were located predominantly in flk:GFP-marked endocardial cells in tbx20-overexpressing and control hearts (arrowheads in Figures 6K,K-I,K-I' ,L,L-I,L-I'); however, only a few number of non-endocardial cells expressed bmp6 in tbx20-overexpressing hearts (arrows in Figures 6L-I,L-I'). Quantification analyses indicated a significant increase of bmp6 + Flk-GFP + cells in tbx20-overexpressing hearts (55.7 ± 2.7%), in comparison with control hearts (33.6 ± 2.9%) (Figure 6M). Similarly, expression of id2b, a Bmp6 downstream target, overlapped with flk:GFP in endocardial cells in both DOX-treated TRE3G:tbx20 CMOE hearts (arrowheads in Figures 6O,O-I,O-I') and TRE3G:tbx20 control hearts (arrowheads in Figures 6N,N-I,N-I'). id2b transcripts were only detectable in a few non-endocardial cells in tbx20-overexpressing hearts (arrows in Figures 6O-I,O-I'). Importantly, myocardial tbx20 overexpression resulted in a significant increase in id2b + Flk-GFP + endocardial cells at the cardiac injury region (52.5 ± 3.5%) compared to that in TRE3G:tbx20 control hearts (30.2 ± 2.5%) ( Figure 6P). In contrast, expression of bmp3 or bmp10 was not significantly upregulated in tbx20 overexpressing wound hearts (Supplementary Figure S8). Collectively, these findings indicated that Bmp6/Id2b signaling is activated mostly in the endocardium in response to Tbx20 induction following cardiac injury.

DISCUSSION
Based on our findings, we propose a working model for Tbx20-dependent transcriptional network in governing heart regeneration using zebrafish heart injury model ( Figure 7N). Following heart injury, tbx20 is strongly induced in the myocardial wound region and the atrial epicardium. Myocardialinduced tbx20 upregulates expression of embryonic or fetal genes as well as cell-cycle regulators, promoting CM proliferation via dedifferentiation. We find that Tbx20 induction also plays a key role in endocardial cell migration and regeneration through upregulating endocardial Bmp6 signaling after cardiac damage. Thus, this putative Tbx20-mediated transcriptional program would coordinate two distinct mechanisms of zebrafish heart regeneration. Because adult hearts are least likely to exist cardiac stem cells (Kretzschmar et al., 2018;He et al., 2019), myocardial regeneration occurs via stem-cell independent mechanisms, in which resident CMs in the injured heart undergo dedifferentiation prior to their proliferation (Jopling et al., 2010;Kikuchi et al., 2010;Senyo et al., 2013). Our study indicates that inducible tbx20 overexpression in adult zebrafish hearts enhance CM proliferation and myocardial regeneration, similar to that observed in mice by Xiang et al. (2016). Mechanistically, we have demonstrated that Tbx20 promotes injury-induced CM proliferation via dedifferentiation through mediating cellular changes and molecular dedifferentiation circuits. Specifically, CMs at the wound border zone express a myocardial fetal marker α-SMA and a stem cell marker Runx1 in injured zebrafish hearts. Both α-SMA and Runx1 were discovered originally as CM dedifferentiation markers in human patients suffered from myocardial infarction (Kubin et al., 2011). Thus, CM dedifferentiation following cardiac injury is a conserved mechanism, defining zebrafish as a model system to study injury-induced CM dedifferentiation during regeneration. Interestingly, we find that tbx20 induction is not only localized to the injured myocardium, but also detectable in the atrial epicardium following ventricle injury. During development, tbx20 is expressed in epicardial cells surrounding the atrioventricular canal groove (Yamagishi et al., 2004;Boogerd et al., 2018), but the function of epicardial tbx20 is still unclear. Previous studies indicate that epicardium is required for heart regeneration in zebrafish (Kikuchi et al., 2011a;Wang et al., 2015). It will be important to test whether atrial tbx20 + epicardial cells contribute to injury-induced CM proliferation, and how atrial epicardial activation relates to ventricle regeneration.
BMP signaling plays pivotal functions via diverse mechanisms involving vascular, myocardial, endocardial and mesenchymal tissues during cardiovascular development (Morrell et al., 2016). We determined that Bmp6 signaling in the endocardium is an early injury-response to myocardial Tbx20 induction, which promotes endocardial cell regeneration. Precedent for this type of crosstalk between myocardium and endocardium exists in the context during heart development. For example, Bmp2 and Bmp4 function in the myocardium is required for the epithelialmesenchymal transformation (EMT) within the endocardium, leading to the formation of endocardial cushions and valves (Ma et al., 2005;Choi et al., 2007). During zebrafish heart regeneration, our RNA-seq analyses revealed upregulation of multiple bmp ligands in tbx20-overexpressing hearts, consistent with part of BMP regulation following cardiac injury (Wu et al., 2016). However, the cellular sources of secreted bmp ligands remain unclear in injured hearts. In this study, we determined that bmp6 and id2b are primarily upregulated in the endocardium following diverse cardiac damages, while their expression was not detectable in uninjured WT hearts, revealing Bmp6 as a specific BMP signal in the endocardium in response to cardiac injury. Previous studies report that Notch signaling and Retinoic Acid (RA) pathways are activated in the wound endocardium (Kikuchi et al., 2011b;Munch et al., 2017). However, both Notch and RA pathways in the endocardium signal to the myocardium to stimulate CM regeneration (Kikuchi et al., 2011b;Munch et al., 2017;Zhao et al., 2019). Our data suggest that Bmp6 signaling in the endocardium promotes endocardial cell regeneration in response to myocardial Tbx20 induction. In uninjured mouse hearts, Tbx20-mediated increases in CM proliferation are due to activation of Bmp2 and PI3K signaling in the myocardium (Chakraborty et al., 2013). Thus, the influence of Tbx20 on distinct Bmp ligands in myocardial cells or the endocardium appears to be context dependent. We have also observed the increased immune response genes in Tbx20-overexpressing hearts in RNA-seq analyses. Immune response participates in heart injury and repair (Lai et al., 2019). Previous studies report that immune cells, such as macrophages, play important roles in the process of inflammation, scar resolution and wound healing, which in turn stimulates myocardial regeneration and repair (Lai et al., 2019). It will be interesting to test whether Tbx20 directly or indirectly regulates immune response genes during cardiac regeneration, and whether Bmp6/Id2b signaling might be activated in immune cells such as macrophages following cardiac injury.   . (A,B) Myocardial tbx20 overexpression upregulates BMP target genes id1 and id2b as well as the BMP ligand bmp6 after heart resection (A) or cryoinjury (B), respectively. Data from three biological replicates. Mean ± SEM, **p < 0.01, ***p < 0.001. (C-E) Representative ISH images of heart sections from DOX-treated uninjured Tg(TRE3G:tbx20) (C), 5 dpci Tg(TRE3G:tbx20) (D) and Tg(TRE3G:tbx20 CMOE ) zebrafish (E,E') with bmp6 probe. Red arrowheads indicate endocardium with bmp6 signal. (F-I) Representative ISH images of heart sections from 5 dpci DOX treated Tg(TRE3G:tbx20) and Tg(TRE3G:tbx20 CMOE ) fish with id1 (F,G) and id2b (H,I) probe. Red arrowheads indicate endocardium with id1 (G') and id2b (I') signal in the injury site. (J) Experimental procedures for DOX treatment, cryoinjury, FISH combined with immunostaining experiments. (K,L) Representative images of FISH analysis of bmp6 (red) combined with immunostaining for GFP and α-actinin (blue) on heart sections from 5 dpci Tg(TRE3G:tbx20;flk:GFP) (K) and Tg(TRE3G:tbx20 CMOE ;flk:GFP) (L) zebrafish. Boxed areas indicate locations of the magnified and channel-separated panels below. White arrowheads point to GFP + cells with bmp6 transcripts in border zone and injury site (K-I,K-I',L-I,L-I'), yellow arrows in (L-I) and (L-I') indicate non-endocardial cells with bmp6 transcripts. (M) Dotted diagram indicates the percentage of bmp6 + endocardial cells in injury site from 5 dpci DOX treated Tg(TRE3G:tbx20) (K, n = 5) and Tg(TRE3G:tbx20 CMOE ) (L, n = 6) fish. Mean ± SEM, ***p < 0.001. (N,O) Representative images of FISH analysis of id2b (red) combined with immunostaining for GFP and α-actinin (blue) on heart sections from 5 dpci Tg(TRE3G:tbx20;flk:GFP) (N) and Tg(TRE3G:tbx20 CMOE ;flk:GFP) zebrafish  . The values are mean ± S.E.M. Two-way ANOVA followed with Tukey's multiple comparison test. n.s.: none significance, ***p < 0.001. (N) Model of myocardial tbx20 function during zebrafish heart regeneration: In CMs, myocardial tbx20 overexpression promotes CM dedifferentiation and proliferation following injury. In addition, myocardial tbx20 promotes endocardial regenerative progress by enhancing its proliferation via partially through Bmp6 signaling. Dashed lines delineate the injured area. Scale bar: 100 µm.
Although adult zebrafish can efficiently regenerate their hearts in response to different forms of damage, there are some intrinsic disadvantages to the model system. For instances, proliferative CMs in zebrafish are mononuclear and diploid, whereas most adult human CMs become polyploidy that restrains CM proliferation (Xin et al., 2013;Tzahor and Poss, 2017). Although CM renewal is extremely low in uninjured mammalian hearts, a small number of proliferative CMs can be detectable in the injury border zone of murine hearts after myocardial infarction (Malliaras et al., 2013;Xiang et al., 2016;Xie et al., 2020). We believe that specific factors and mechanisms obtained from zebrafish regeneration study should be confirmed in mammal system such as mouse before exploring potential therapeutics in humans. Tbx20 plays a conserved role in promoting injury-induced CM regeneration in zebrafish and mouse (Xiang et al., 2016), raising a possibility that Tbx20 might be a target to be explored in the medical context. For instance, Tbx20 administration at local injury regions might be potential strategy to improve repair capacity in the wound human heart. Overall, understanding of the mechanisms underlying heart regeneration in model systems will provide inspiration for the regeneration intervention in humans.

Zebrafish Cardiac Injury
All animal work used in this study were approved by the Animal Care Committee of Fudan University. We bred wild-type zebrafish, or zebrafish carrying Tg(flk:GFP) and Tg(cmlc2:GFP) (Burns et al., 2005;Jin et al., 2005). In addition, we bred zebrafish carrying the newly generated transgenes described below. 6-18 months aged wild type or transgenic lines of the AB strain were used for ventricular resection or cryoinjury as previously described (Poss et al., 2002;Gonzalez-Rosa et al., 2011). Tricaine (ethyl-3-aminobenzoate methane sulfonate salt, Sigma) with concentration of 0.16 mg/mL were used to anesthetize zebrafish before use.
The cmlc2-TetON-3G cassette was first constructed by cloning TetON-3G from pCMV-Tet3G vector (Clontech, #z1164n) downstream of a 3.0-kb cmlc2 promoter. In order to identify the transgenic animals by lens fluorescence, the Cryaa-mCherry cassette from plasmid pT2-hsp70l-dnMEK1 (Liu and Zhong, 2017) was then cloned downstream of the cmlc2-TetON-3G cassette in the opposite orientation. The entire construct was franked by two copies of the chicken β-globin insulators at upstream and one human 5 β-globin insulator at downstream, and one human 3 β-globin insulator was also included between the two cassettes.

Immunostaining, ISH, AFOG and Histology Analyses
Zebrafish hearts were dissected and fixed in 4% paraformaldehyde (PFA) solution at 4% overnight before embedded in OCT (Thermo Fisher Scientific). 10 µm cryosections were used in all histological analyses, in situ hybridization (ISH) and immunostaining. The primers used to generate RNA probes were listed in Supplementary Table S4. ISH and Acid Fushin-Orange G (AFOG) analyses were performed as described, the quantification of the scar area was performed using Fiji software (Poss et al., 2002). The TSA -Plus fluorescence system (Perkin Elmer, #NEL752001KT) were used for fluorescence in situ hybridization (FISH) analyses according to the manufacturer's instructions. Primary and secondary antibody staining were performed according to the standard protocol, except for PCNA staining which required heat-induced antigen retrieval (Lepilina et al., 2006). All primary antibodies were incubated at 4% overnight and secondary antibodies were incubated at room temperature for 2 hrs. Primary antibodies used in this study include GFP (Invitrogen, #A21311, 1:200), GFP (Aves Labs, #GFP-1010, 1:500), α-actinin AFOG images were taken using an Olympus DP80 microscope. ISH images were taken using a Nikon microscope. FISH and immunofluorescence images were taken using a ZEISS LSM880 confocal microscope. For quantitative analyses of immunofluorescence images, sections containing the largest wounds were selected and manually counted. A defined region within and adjacent (150 µm and 100 µm away from the wound edge for Mef2 + PCNA + CMs and Fli1 + PCNA + cells, respectively) to the injury site were selected to quantify the proliferation ratio of CMs and endothelial cells.

Gene Expression Analysis
For RNA-seq analysis, 7 dpa hearts were dissected, and the apical portions of 6-8 ventricles were collected in each group. RNA samples were extracted using TRIzol Reagent (Invitrogen, #15596018). Next generation sequencing library construction and sequencing was performed by GENEWIZ Company, Suzhou on an Illumina HiSeq sequencer. Reads from the sequenced samples were qualified and aligned to zebrafish transcriptome (Ensembl genebuild GRCz10.86) using Hisat2 (v2.0.1). Differentially expressed genes were analyzed using the DESeq Bioconductor package, genes identified with altered expression levels with a Benjamini and Hochberg adjusted p-value < 0.05 were retained. GO-TermFinder was used to define Gene Ontology terms. For quantitative RT-PCR analysis (qPCR), RNA samples were extracted from ventricular apices, border zone and atrium at specific time points, respectively, 6-8 samples were pooled together for each group, cDNA was synthesized using the SuperScript TM III First-Strand Synthesis System (Invitrogen, #18080051) following the manufacturer's instructions. q-PCR was performed on a Q7 Real-Time PCR System (Applied Biosystems) using SYBR Green ROX dye (Applied Biosystems, #A25742). Primers used are listed in Supplementary Table S4.

Statistical Analysis
All quantitative analyses of immunofluorescence images and assessment of stained images were performed in blinded fashion. GraphPad Prism 8 was used for data analyzing. Statistical comparisons between two groups were analyzed by Student's t-test. For samples with more than two independent groups, one-way ANOVA with Dunnett's multiple comparisons test were performed. For samples with two experimental factors, two-way ANOVA followed by Tukey's multiple comparisons tests were performed. All statistical tests were calculated when normality test using D'Agostino-Pearson omnibus test and Shapiro-Wilk test was passed.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: https://www.ncbi.nlm. nih.gov/geo/, GSE144831.

ETHICS STATEMENT
The animal study was reviewed and approved by the Animal Care Committee of Fudan University.

AUTHOR CONTRIBUTIONS
YF, KL, and TZ designed the study. YF and KL performed the experiments. YF, KL, PS, JS, WT, and TZ analyzed the experimental data. YF and TZ wrote the manuscript. WT and TZ edited the manuscript. All authors contributed to the article and approved the submitted version.

ACKNOWLEDGMENTS
We acknowledge Haitao Zhou for assistance in fish care. We are grateful to laboratory members for helpful discussions.
FIGURE S3 | No leakiness of tbx20 expression is observed with DOX treatment using the TetON-3G system. (A-D) Adult transgenic zebrafish images indicating the selectable marker of lens, Tg(TRE3G:tbx20) showed green eyes in (B), Tg(TRE3G:tbx20 CMOE ) showed yellow eyes in (D). No mCherry signal was observed in the zebrafish fin, skin or muscle after DOX treatment (D). Scale bar: 250 µm. (E,F) Images of whole mount hearts of the transgenic fish at 7 dpa treated with DOX in the brightfield (E) or mCherry channel (F), mCherry signal was not detected in the outflow tract (asterisk) and the injured site (arrow) from Tg(TRE3G:tbx20 CMOE ) hearts. Scale bar: 500 µm. (G-I) ISH analysis of tbx20 overexpression using tbx20-E2A-mCherry probe on 7 dpa heart sections from Tg(TRE3G:tbx20) zebrafish treated with DOX (G), Tg(TRE3G:tbx20 CMOE ) zebrafish treated with (I) or without DOX solution (H). Scale bar: 100 µm.