The regenerative capacity of the zebrafish heart is dependent on TGFβ signaling

In adult mammals, a heart infarction heals by deposition of a permanent collagenous scar instead of restoration of functional cardiac tissue (Laflamme and Murry, 2011). By contrast, an impressive regenerative capacity has been demonstrated in the heart of zebrafish and newts (Ausoni and Sartore, 2009; Major and Poss, 2007; Singh et al., 2010). In contrast to mammals, these animals sustain a robust proliferative competence of cardiomyocytes in adult life, which provides a source of a new myocardium.

The zebrafish serves as a particularly useful model system with which to study the regenerative response to myocardial damage, as shown by experiments involving resection of the ventricular apex, cryoinjury and genetic cardiomyocyte depletion (Chablais et al., 2011; Gonzalez-Rosa et al., 2011; Schnabel et al., 2011; Wang et al., 2011). Cell lineage-tracing analyses have demonstrated that complete repair of the amputated myocardial apex predominantly involves cell cycle re-entry of existing cardiomyocytes, as opposed to recruitment of epicardial stem cells (Jopling et al., 2010; Kikuchi et al., 2011a; Kikuchi et al., 2010). Despite progress in the last decade, little is known about the molecular factors that control heart restoration. Studies applying transgenesis have revealed a requirement of the FGF and retinoic acid (RA) signaling pathways in epicardial activation and in stimulation of cardiomyocyte proliferation (Kikuchi et al., 2011b; Lepilina et al., 2006). Using a small-molecule inhibition approach it has been shown that PDGF signaling is required for epicardial function and blood vessel formation in regenerating zebrafish hearts (Kim et al., 2010; Lien et al., 2006). These findings support the hypothesis that organ regeneration is regulated by secreted factors that mediate communication between adjacent cardiac and non-cardiac tissues.

The advantage of the zebrafish model system in regenerative biology has recently been reinforced by establishing a myocardial infarction model using a cryoinjury procedure, which involves direct freezing of the ventricular wall (Chablais et al., 2011; Gonzalez-Rosa et al., 2011; Schnabel et al., 2011). Compared with ventricular resection or genetic cardiomyocyte depletion, this procedure is of particular value because it mimics different physiological responses observed in mammals, including cell death and scarring. After reparative events, zebrafish infarcted hearts progress to a unique regenerative phase. The initially deposited scar tissue becomes completely replaced with new cardiac muscle within 2 months following cryoinjury. Remarkably, proliferation of cardiomyocytes is triggered simultaneously with the scar deposition. The progressive cardiomyocyte propagation into the injury area is accompanied by local regression of fibrotic tissue. Thus, cardiac regeneration involves different responses of cardiac and non-cardiac cell types, which have to be spatially and temporally orchestrated to achieve a balance between the reparative and regenerative processes. Little is known about the molecules that coordinate the function of the various cellular components to achieve this successful regeneration.

In the present study, we identified that a molecular link between the reparative and regenerative processes after cryoinjury-induced myocardial infarction is achieved through Smad3-dependent TGFβ/Activin signaling. We first determined the involvement of the related TGFβ and Activin-β (Inhibin β) ligands using expression analysis at the transcript and protein levels. Next, we visualized the spatiotemporal dynamics of pathway activation during heart regeneration using immunostaining of a direct downstream signal transducer, phospho-Smad3. To determine the role of TGFβ/Activin signaling, we blocked the relevant type I receptor with the previously characterized specific inhibitor SB431542. Inhibition of TGFβ/Activin signaling suppressed phosphorylation of Smad3 and impaired heart regeneration, resulting in ventricular deformation. The reversible action of the inhibitor was beneficial in dissecting the effects of TGFβ signaling on several successive cellular processes. The temporal inhibition of TGFβ/Activin signaling during the subacute phase revealed an unexpected role of the transient scar in supporting the mechanical properties of the damaged ventricle. We identified several extracellular regulators of tissue organization and remodeling, represented by Collagen, Fibronectin and Tenascin C, as targets of TGFβ/Activin signaling. Furthermore, continuous and short-term inhibition of the pathway demonstrated the stimulatory role of TGFβ/Activins for cardiomyocyte proliferation, which is a key mechanism of new myocardium formation. Our study provides evidence that TGFβ signaling coordinates a wide spectrum of subsequent cellular steps that are required for efficient heart regeneration in zebrafish.

The following zebrafish strains were used: wild-type AB (Oregon), cmlc2::DsRed2-Nuc (Rottbauer et al., 2002), flk1::EGFP (Lawson and Weinstein, 2002) and hsp70:dn-fgfr1 (Lepilina et al., 2006). Fish at 6-18 months of age were anesthetized in 0.1% tricaine (Sigma-Aldrich) and cryoinjury was performed as previously described (Chablais et al., 2011). Ventricular apex resection was performed as described (Poss et al., 2002). Animals were allowed to regenerate for various times at 26.5°C. The Alk5/4 inhibitor SB431542 (Tocris) was used as previously described (Jazwinska et al., 2007). The control fish were in water with 0.1% DMSO. For BrdU incorporation analysis, the fish were incubated for 24 hours in water containing 50 μg/ml BrdU (Sigma-Aldrich). hsp70:dn-fgfr1 fish received daily heat shocks by raising the water temperature to 37°C for 1 hour. The experimental research on animals was approved by the cantonal veterinary office of Fribourg.

Acid Fuchsin Orange-G (AFOG) staining was performed as described previously (Poss et al., 2002). For immunohistochemistry, the primary antibodies used were: rabbit anti-TGFβ receptor I at 1:50 (AnaSpec, 53895), mouse anti-Tropomyosin at 1:100 [developed by Jung-Ching Lin, obtained from the Developmental Studies Hybridoma Bank (DSHB), CH1], mouse anti-GFP at 1:100 (Roche, 11814460001), rabbit anti-Raldh2 at 1:400 (GeneTex, GTX124302), rabbit anti-Tenascin C at 1:500 (USBiological, T2550-23), mouse anti-Vimentin at 1:70 (developed by Alvarez-Buylla, obtained from DSHB, 40E-C), rabbit anti-α-SMA at 1:200 (GeneTex, GTX1000345), rabbit anti-Mcm5 at 1:5000 (kindly provided by Soojin Ryu, MPI for Medical Research, Heidelberg, Germany), rabbit anti-Fibronectin at 1:400 (Sigma-Aldrich, F3648), rabbit anti-phospho-Smad3 at 1:400 (Abcam, ab52903) and rat anti-BrdU at 1:100 (Abcam, ab6326). The secondary antibodies used (at 1:500) were: goat anti-rabbit Alexa Fluor 488, goat anti-mouse Alexa Fluor 488 (Molecular Probes), goat anti-rabbit Cy3 or Cy5 conjugated, goat anti-mouse DyLight 549 or 649 conjugated (Jackson ImmunoResearch).

Hearts were fixed, cryosectioned at 16 μm and immunostained as previously described (Chablais et al., 2011). The specimens were analyzed by confocal microscopy (Leica TCS-SP5). For quantification of proliferating cardiac cells, Mcm5/DsRed2 or BrdU/DsRed2 double-positive nuclei were normalized as the percentage of the total number of DsRed2-positive nuclei using ImageJ 1.44p software. For quantification of proliferating non-cardiac cells, the number of Mcm5-positive DsRed-negative cells was normalized as the percentage of the total number of DAPI-positive cells of the ventricle.

In situ hybridization on cryosections was performed as described (Chablais and Jazwinska, 2010). Primers are listed in supplementary material Table S1.

After standard in situ hybridization staining, the sections were washed in PBS for 30 minutes and postfixed in 70% ethanol for 30 minutes. The slides were rinsed in PBST (PBS, 0.3% Triton X-100) and incubated with PBST plus 5% goat serum for 1 hour. The primary antibodies are listed above. The slides were processed using the Vectastain ABC Kit (Vector Labs) according to the manufacturer’s instructions. Secondary antibodies (Vector Labs) were biotinylated anti-mouse (BA-2000) and biotinylated anti-rabbit (BA-1000) diluted 1:200. The antibody staining was detected using DAB solution (Sigma-Aldrich, D5905).

All ECGs were recorded in triplicate before injury and at subsequent time points after surgery. To reduce biological variation, the same six animals were recorded before and after injury. ECG acquisitions were performed as previously described (Chablais et al., 2011).

A class of molecular signals that are known to be injury activated in various organs is represented by structurally related TGFβ and Activin cytokines (Dobaczewski et al., 2010a; Werner and Alzheimer, 2006). In vertebrates, several isoforms of these family members have been identified (Massague and Gomis, 2006). We tested the expression of Activin and TGFβ genes by in situ hybridization using control hearts from sham-injured zebrafish [4 days post sham (dps)] and regenerating hearts at the onset of infarct healing [4 days post cryoinjury (dpci)], and at 14 dpci when the transition between scar formation and scar resorption becomes evident. activin-βA (inhbaa – Zebrafish Information Network) was not significantly detected in control and injured hearts, whereas activin-βB (inhbb – Zebrafish Information Network) was ubiquitously expressed only in regenerating hearts (supplementary material Fig. S1). By contrast, three known TGFβ isoforms, tgfb1, tgfb2 and tgfb3, showed specific upregulation in the post-infarct zone of the injured hearts at 4 and 14 dpci (Fig. 1F-H and supplementary material Fig. S2A-D). The injury area was identified by the absence of the ventricular myosin heavy chain (vmhc) transcript (Fig. 1E). The control hearts did express TGFβ genes (Fig. 1A-D).

To further determine the expression domain of the TGFβ isoforms in the infarcted hearts, we combined in situ hybridization with antibody staining for marker proteins of different cell types in the heart. High-resolution imaging of double-stained sections did not reveal expression of TGFβ genes in the intact Tropomyosin-positive myocardium, except for a thin layer of cardiomyocytes abutting the infarct area (Fig. 1I and supplementary material Fig. S2H,I). Most of the Raldh2-expressing epicardial cells and Vimentin-expressing fibroblasts of the post-infarct appeared to be positive for the TGFβ transcripts (Fig. 1J,K and supplementary material Fig. S2J-M). By contrast, EGFP-labeled endothelial cells did not seem to display TGFβ expression (Fig. 1L and supplementary material Fig. S2N,O). We concluded that all three TGFβ genes appear to be expressed in overlapping domains that encompass several cell types, including fibroblasts and epithelial cells of the infarct area, as well as some cardiomyocytes at the infarct boundary.

The TGFβ/Activin cytokines activate the downstream pathway via ligand-stimulated interaction of type I and II receptors (Massague and Gomis, 2006). We analyzed the expression of three isoforms of type I receptors, alk5a/b and alk4 (tgfbr1a/b and acvr1b – Zebrafish Information Network), which are specific for TGFβs and Activins, respectively (Shi and Massague, 2003). alk5a and alk4 appeared to be present in the injured area, whereas alk5b was strongly expressed in the entire heart (supplementary material Fig. S3). To confirm this finding, we visualized the distribution of Alk5b protein using an antibody against a unique amino acid sequence of this receptor. Immunofluorescence analysis detected the presence of this transmembrane receptor in the whole heart, including cardiomyocytes and Vimentin-positive cells of the post-infarct (supplementary material Figs S4 and S5). We concluded that both cardiac and non-cardiac cells of the infarcted heart can be responsive to TGFβ ligands.

The activated type I receptors transmit signaling by phosphorylation of receptor-activated SMADs (R-SMADs), which, in a complex with common-mediator SMADs (co-SMADs), are translocated to the nucleus to control target gene transcription (Shi and Massague, 2003). To determine the spatiotemporal activation of the TGFβ/Activin signaling pathway during cardiac regeneration, we visualized phospho-Smad3 (pSmad3), a specific R-SMAD for Alk5/4 receptors. To distinguish between cardiac and non-cardiac nuclei, we used a transgenic strain of zebrafish that expresses nuclear DsRed2 under the control of the cardiac-specific cardiac myosin light chain (cmlc2; myl7 – Zebrafish Information Network) promoter (cmlc2::DsRed2-Nuc) (Rottbauer et al., 2002). To recognize healthy myocardium, the sections were immunostained for Tropomyosin. In hearts of uninjured fish, we did not detect significant pSmad3 immunolabeling (supplementary material Fig. S6A). However, the hearts at 7 and 14 dpci displayed numerous pSmad3-positive nuclei of non-cardiac cells in the post-infarct (Fig. 2A and supplementary material Fig. S6B). Importantly, pSmad3 was also enhanced in a subset of cardiomyocytes localized at the infarct boundary (Fig. 2A). The number of pSmad3-positive nuclei declined at 30 dpci, when cardiac regeneration was nearly complete (supplementary material Fig. S6C,D). The visualization of pSmad3 demonstrates that the TGFβ/Activin signaling pathway is activated in the fibrotic tissue and in adjacent cardiomyocytes during heart restoration.

To examine the requirement of the TGFβ/Activin pathway for heart regeneration, we applied a pharmacological strategy to block type I receptor activity. We used the low molecular weight inhibitor SB431542 (Alk5/4-i), which specifically blocks TGFβ/Activin pathway receptors and does not affect closely related receptors from the bone morphogenetic protein (BMP) pathway or other signal transduction pathways (Ho et al., 2006; Inman et al., 2002; Jazwinska et al., 2007). The mechanism of this selectivity has recently been determined by solving the crystal structure of the mammalian TGFβ type I receptor kinase domain in complex with SB431542 (Ogunjimi et al., 2012). Based on this study, we verified that the key determinant for the selectivity of SB431542, residue S280, as well as all the other gatekeeper residues are conserved in Alk5/4 receptors in zebrafish (supplementary material Fig. S7).

To determine the activity of the SB431542 inhibitor in the heart after cryoinjury, we compared pSmad3 immunostaining between control and inhibitor-treated animals at 14 dpci. We found a noticeable reduction of pSmad3-positive cells within the post-infarct and the myocardium (Fig. 2B,C). We concluded that the application of 10 μM Alk5/4-i markedly suppressed TGFβ/Activin signaling in the heart of adult zebrafish.

To examine the requirement of TGFβ/Activin signaling for heart regeneration, we analyzed the histology of control and Alk5/4-i-treated hearts at different time points after cryoinjury. Acid Fuchsin Orange-G (AFOG) staining of heart sections at 4 dpci revealed a comparable damage of ∼20% of the ventricle in control and inhibitor-treated animals with abundant fibrin accumulation (Fig. 3A,B,E,F). At 14 dpci, control hearts formed two domains of the scar: a layer of fibrin along the outer border of the post-infarct, and a network of collagen in the interior of the wound (Fig. 3C). By contrast, the post-infarct of the Alk5/4-i-treated fish failed to deposit a collagenous matrix in the lost myocardium (Fig. 3G). In control hearts at 30 dpci, the fibrotic tissue was largely resolved and was replaced by new muscle tissue, demonstrating substantial cardiac regeneration (Fig. 3D). We found that such a regenerative process was not accomplished after the inhibition of TGFβ/Activin signaling (Fig. 3H). The post-infarct tissue was devoid of muscle and remained surrounded by a layer of fibrin. Moreover, the damaged area conspicuously bulged out from the ventricular circumference, indicating mechanical deformation of the tissue.

To better understand this phenotype, we compared it with another case of defective heart regeneration that occurs upon inhibition of FGF signaling, which has been shown to be required for heart regeneration after partial amputation (Lepilina et al., 2006). As predicted, overexpression of a dominant-negative (dn) form of the FGF receptor resulted in a lack of cardiac regeneration (Fig. 3I-L). In contrast to the phenotype caused by TGFβ inhibition, the post-infarct of hearts expressing dnFgfr1 contained abundant collagen fibers, indicating excessive scarring. Importantly, the fibrotic infarct did not protrude beyond the ventricular circumference. The phenotypic comparison indicates that TGFβ signaling is required for two processes during heart regeneration: collagen deposition during infarct healing and subsequent cardiac muscle regeneration.

To characterize the nature of the post-infarct tissue of the hearts treated with Alk5/4-I at 30 dpci, we performed immunofluorescence analyses using several cellular and extracellular markers (Chablais et al., 2011). The visualization of Tropomyosin cardiac protein confirmed the absence of myocardial tissue within the protruding post-infarct (supplementary material Fig. S8D,E). Instead, the injury zone was labeled by connective tissue markers, such as the extracellular matrix (ECM) protein Fibronectin and the intermediate filament protein Vimentin. The infarct margin was surrounded by a layer of myofibroblasts/smooth muscle cells that express α-smooth muscle actin (α-SMA). The ECM of the post-infarct was devoid of Tenascin C, a de-adhesive extracellular protein (supplementary material Fig. S8F). By contrast, Tropomyosin-positive myocytes efficiently invaded the post-infarct area in the control hearts. The residual Fibronectin, Tenascin C, α-SMA and Vimentin demarcated the position of cryoinjury (supplementary material Fig. S8A-C). The immunofluorescence analyses demonstrate that the inhibition of TGFβ/Activin signaling impaired the replacement of connective tissue with new cardiac tissue in the post-infarct.

To assess the physiological consequences of this phenotype, we recorded heart conduction with an electrocardiogram (ECG). We have previously shown that cryoinjury results in a prolonged ventricular repolarization, displayed as QT intervals, but that the regenerated hearts at 30 dpci recover their original electrical properties (Chablais et al., 2011). Here, we found that the fish treated with Alk5/4-i for 30 days after cryoinjury developed prolonged QTc values (Fig. 4). Thus, a lack of heart regeneration after the inhibition of TGFβ/Activin signaling alters cardiac electrophysiological properties, indicating heart dysfunction.

The phenotype caused by the inhibition of TGFβ/Activin was associated with the absence of a collagen-containing scar and a complete lack of cardiac regeneration (Fig. 3H). Both concomitant defects suggest a correlation between collagenous scar formation observed at 14 dpci and the heart restoration at 30 dpci. To determine whether collagen deposition is beneficial for subsequent cardiac regeneration, we designed a drug-shift experiment. After myocardial infarction, animals were treated with Alk5/4-i during the first 14 days to block collagenous scar formation. Then, the fish were transferred to normal conditions for another period of 14 days to allow regeneration to resume. At the end of the experiment at 28 dpci, heart regeneration and scarring were assayed using AFOG staining (Fig. 5A,C).

Most of the control hearts of this experimental setting displayed advanced regeneration with little scarring (eight hearts per group; Fig. 5A). By contrast, the Alk5/4-i-treated animals failed to restore the myocardial tissue (among seven animals, four hearts had a bulging infarct, two hearts had impaired regeneration without a significant deformation, and one heart was comparable to the control; Fig. 5C). The post-infarct of these hearts contained collagen fibers, especially around the infarction wall. This demonstrates that the initially blocked deposition of collagen was partially compensated during the 14 days in normal conditions. Despite the resumed regeneration, the cardiac muscle failed to replace the damaged tissue. Thus, the initial suppression of scarring led to irreversibly impaired regeneration. We concluded that TGFβ/Activin-dependent collagen deposition in the subacute phase is essential for subsequent regenerative success.

Collagen fibers are known to undergo continuous turnover in the healing infarcts (Bowers et al., 2010; Chen and Frangogiannis, 2010). Thus, the stimulatory role of TGFβ for collagen accumulation might occur by either enhancing collagen production and/or decreasing its degradation. To distinguish between these possibilities, we designed an experiment in which the drug treatment was applied after scar formation. For this purpose, the animals were kept in normal conditions for 14 days after cryoinjury to allow collagen deposition. After this time, the fish were transferred to water with Alk5/4-i for the next 14 days, and the amount of scarring was assessed. Histological analysis revealed advanced regeneration with only residual collagen fibers within the post-infarcts of the control hearts (Fig. 5B). By contrast, abundant scar material and regenerative failure were observed after the drug shift (among seven fish, five hearts displayed enhanced scarring without infarct bulging, and two hearts regenerated myocardium with residual scarring comparable to the control; Fig. 5D). The presence of the persisting collagen fibers suggests that the inhibition of Alk5/4 receptors did not enhance degradation of the pre-existing collagenous matrix. We concluded that TGFβ/Activin signaling controls collagen accumulation predominantly by stimulating its synthesis. Moreover, we found no deformation of infarct geometry, indicating a mechanical role of collagen as supporting material in the damaged tissue of the injured heart.

The main components of the cardiac ECM include structural proteins such as collagen, adhesive proteins such as fibronectins, and de-adhesive tissue remodeling proteins such as tenascins (Bowers et al., 2010; Corda et al., 2000). We set out to systematically determine the effect of TGFβ/Activin signaling on the expression of these ECM components during the regenerative phase at 14 dpci, when the fibrotic tissue is abundantly detected in the post-infarct. Consistent with our previous observations (Fig. 3G), the accumulation of collagen matrix was strongly impaired in the post-infarcts of Alk5/4-i-treated animals as compared with the controls (Fig. 6A,E,I,L). Moreover, Alk5/4-i treatment resulted in the reduction of Fibronectin, a broadly expressed core matrix component that provides a framework for cell adhesion (Fig. 6C,G). Notably, the tissue remodeling protein Tenascin C (Imanaka-Yoshida et al., 2004), which is normally induced at the infarct-myocardium interface, was completely suppressed after antagonizing Alk5/4 receptors (Fig. 6K,N). Taken together, the deposition of crucial ECM components in the post-infarct zone, including structural, adhesive and de-adhesive proteins, requires TGFβ/Activin signaling.

The source of the ECM in mammalian myocardial infarcts is fibroblasts, myofibroblasts and smooth muscle cells (Camelliti et al., 2005; Dobaczewski et al., 2010b; Porter and Turner, 2009). We aimed to determine whether TGFβ/Activin signaling regulates the ECM content by acting on the recruitment/differentiation of these ECM-producing cells. To visualize the fibroblasts of the infarct we used antibodies against Vimentin, a cytoplasmic intermediate filament-associated protein. The post-infarct of control and Alk5/4-i-treated fish displayed a dense network of Vimentin-positive cells, which was even augmented in the hearts treated with Alk5/4-i (Fig. 6K,N). This accumulation of fibroblasts might represent a compensatory reaction to their reduced activity in the deposition of ECM proteins. To identify myofibroblasts and smooth muscle cells, we used antibodies against α-SMA, a marker of their specific contractile microfilamentous apparatus. In control regenerating hearts, α-SMA specifically labeled the outer border of the post-infarct (Fig. 6D). A similar peripheral rim of α-SMA-positive cells persisted in hearts treated with Alk5/4-i (Fig. 6H). This demonstrates that the inhibition of TGFβ/Activin signaling did not impair the accumulation of Vimentin-positive and α-SMA-positive cells of the fibrotic tissue. Taken together, despite the presence of ECM-producing cells, the synthesis of various ECM components was markedly affected by exposure to Alk5/4-i. We concluded that disruption of TGFβ/Activin signaling results in functionally defective fibroblast-like cells that exhibit impaired ECM deposition.

Epicardial and endocardial cells are important components in heart regeneration in zebrafish (Kikuchi et al., 2011a; Kikuchi et al., 2011b). They induce expression of the RA-synthesizing enzyme Raldh2 (Aldh1a2 – Zebrafish Information Network) to increase production of RA, a cardiogenesis-stimulating factor. To test the role of TGFβ/Activin signaling in the epicardium and the endocardium after myocardial infarction, we immunoassayed Raldh2 in control and Alk5/4-i-treated animals. In uninjured animals, Raldh2 specifically labeled the epicardium (supplementary material Fig. S9). At 14 dpci, Raldh2-positive epicardial cells invaded the injury area and surrounded the damaged myocardium. Consistent with previous studies (Kikuchi et al., 2011b), the expression of Raldh2 was induced in a subset of endocardial cells within the intact myocardium. In comparison with the control animals, inhibition of TGFβ/Activin signaling did not affect the distribution of Raldh2-positive epicardial and endocardial cells (supplementary material Fig. S9B,C). This suggests that the lack of cardiac regeneration in Alk5/4-i-treated fish is not caused by the defective injury responses of epicardium and endocardium.

The reparative and regenerative phases of heart regeneration are overlapping, as the deposition of scar tissue is concomitant with the enhanced cell cycle re-entry of cardiomyocytes, which will lead to restoration of the functional cardiac tissue (Chablais et al., 2011). Consistently, closer examination of the control hearts at 14 dpci revealed spreading of myocytes along the border with fibrotic tissue (Fig. 2A and Fig. 6K′). Inhibition of TGFβ signaling impaired this shift of cardiomyocytes into the damaged tissue (Fig. 2B and Fig. 6N′). Notably, the impaired invasion of the myocardium was associated with the suppression of Tenascin C, which is a cell migration-promoting factor at the infarct-myocardium boundary (Imanaka-Yoshida et al., 2004). This observation suggests that TGFβ/Activin signaling is required for the invasion of a new myocardium into the injury area.

To examine the mitogenic role of TGFβ/Activin in regenerating hearts, we assessed cell proliferation with antibodies against Mcm5 (a DNA replication licensing factor) at 14 dpci. To identify nuclei of cardiomyocytes, we used a transgenic strain that expresses nuclear DsRed2 under the control of the cmlc2 promoter (cmlc2::DsRed2-Nuc). In control heart sections with a representative myocardial injury, the proportion of Mcm5-positive cardiomyocytes was 8.6% (Fig. 7A,B). Treatment with Alk5/4-i resulted in a 50% reduction in the number of proliferating cardiomyocytes (Fig. 7B,C). Non-cardiac cell proliferation (Mcm5-positive and DsRed2-negative cells), by contrast, was enhanced by the Alk5/4-i treatment by ∼30% (Fig. 7D). This result is consistent with our previous observation of augmented Vimentin-positive tissue in the injury area (Fig. 6K,N). We concluded that the mitogenic role of TGFβ/Activin signaling is tissue dependent: it has a stimulating effect on cardiomyocytes but not on the other cell types of the post-infarct.

The reduced proliferation of myocytes in Alk5/4-i-treated fish can be explained either by the primary requirement of TGFβ/Activin signaling as a mitogenic factor or as the secondary consequence of impaired cell migration and tissue remodeling at the myocardial-infarct boundary. To distinguish between these possibilities, we designed another drug-shift experiment to determine the immediate effects caused by short-term inhibition of Alk5/4 receptors during normally progressing regeneration (Fig. 7E). Transgenic fish after myocardial infarction were first allowed to initiate scar deposition and cardiomyocyte proliferation under normal conditions for the first 7 days, and were then kept in the presence of Alk5/4-i for only 3 days. To identify DNA-replicating cells, the fish were in addition exposed to BrdU for the last 24 hours of the experiment, and cell proliferation was assessed. In sections with scar tissue of control hearts, BrdU labeled ∼4% of the cardiomyocytes, which were predominantly located in the vicinity of the injury (Fig. 7F,H). A short-term exposure to Alk5/4-i resulted in a ∼70% reduction of BrdU-positive cardiomyocytes (Fig. 7G,H). The rapid reduction of BrdU incorporation suggests that TGFβ/Activin is directly or indirectly required to maintain the proliferative capacity of cardiomyocytes.

To further test the stimulatory role of TGFβ/Activin in cardiac regeneration, we used another injury method that is based on partial amputation of the ventricular apex (Poss et al., 2002). The design of the experiment was the same as that described above, in which the fish after ventricular resection were allowed to initiate regeneration for 7 days under normal conditions, followed by 3 days of Alk5/4-i treatment including BrdU treatment for the final day (supplementary material Figs S10, S11). The pulse exposure to the inhibitor significantly reduced cardiomyocyte proliferation. This demonstrates that the mitogenic role of TGFβ/Activin for cardiomyocytes is independent of the type of heart injury.

Heart regeneration following cryoinjury-induced myocardial infarction is remarkably efficient in adult zebrafish (Chablais et al., 2011; Gonzalez-Rosa et al., 2011; Schnabel et al., 2011). The present study identifies the requirement of TGFβ/Activin signaling to coordinate the sequence of diverse cellular events during this process. Our data support a model in which zebrafish heart regeneration can be subdivided into three overlapping phases (Fig. 8). First, myocardial cell death triggers an inflammatory response that is characterized by the infiltration of the infarct with activated leukocytes and fibroblast-like cells, which express TGFβ ligands (Fig. 8A). Second, as the wound becomes cleared of dead cells and matrix debris, the reparative phase begins at ∼4 dpci (Fig. 8B). TGFβ/Activin signaling stimulates the recruited fibroblast-like cells to synthesize the collagen-rich ECM. Third, at the onset of scar formation at 7 dpci, the regenerative phase begins (Fig. 8C). While the fibrotic tissue is still maturing, TGFβ/Activin signaling promotes proliferation of cardiomyocytes located in the vicinity of the post-infarct. Tissue remodeling and invasion of new cardiac muscle to replace scar tissue are associated with the TGFβ-dependent expression of Tenascin C, an extracellular protein with contra-adhesive properties. Thus, the regenerative phase is characterized by two opposing processes: scar deposition in the damaged area and scar degradation at the position where new cardiomyocytes invade the post-infarct. It is not known how a balance between both cellular events is achieved. Here, we show that the equilibrium between the reparative and regenerative responses is coordinated by the TGFβ/Activin signaling pathway. The inhibition of this signaling impairs both collagenous scar formation and cardiomyocyte propagation.

The injury response of the nonregenerative mammalian heart is also described in three phases – inflammation, granulation tissue proliferation and scar maturation – all of which are regulated by multifunctional TGFβ signaling (Dobaczewski et al., 2010b). Although the initial two phases are similar between zebrafish and mammals, in zebrafish the collagenous matrix does not ultimately develop a mature scar but instead undergoes remodeling and is replaced with a new myocardium (Chablais et al., 2011; Gonzalez-Rosa et al., 2011; Schnabel et al., 2011). Thus, the unique feature of the regenerating zebrafish heart is a switch from the deposition to the degradation of fibrotic tissue to allow for the expansion of the new myocardium. Studies in rodents demonstrate that the modulation of TGFβ signaling using pharmacological approaches or gene/protein therapies can reduce fibrosis and deleterious ventricular remodeling (Alegria et al., 2005; Kuwahara et al., 2002; Okada et al., 2005; Tan et al., 2010). Finding the optimal time window for antagonizing TGFβ signaling after myocardial infarction is essential in order to precisely target the pathological processes that occur during the specific phase of heart repair (Leask, 2010).

Downregulation of TGFβ/Activin signaling caused a wide range of cellular phenotypes affecting heart regeneration in zebrafish. Specifically, we identified the following processes to be affected after Alk5/4-i treatment: (1) suppressed collagen accumulation; (2) attenuated Fibronectin deposition; (3) blocked Tenascin C expression at the interface between myocardium and fibroblasts; (4) stimulated fibroblast proliferation; (5) attenuated cardiomyocyte proliferation; and (6) impairment of cardiomyocyte spreading into the post-infarct. These diverse effects demonstrate that TGFβ/Activin signaling can be considered as a molecular link between different responses exerted by various cell subpopulations in the regenerating heart. The comparison between mammalian and zebrafish hearts reveals a common effect on fibroblasts to stimulate their proliferation and ECM protein deposition (Leask, 2007; Dobaczewski et al., 2010a). The most notable difference occurs at the level of the cardiomyocyte response. In mammals, TGFβ promotes pathologic cardiomyocyte hypertrophy (Rosenkranz, 2004), as opposed to the healthy myocyte proliferation observed in zebrafish. These diverse outcomes might be due to either differential intrinsic cardiomyocyte properties or additional environmental factors. Elucidation of these underlying mechanisms will be crucial to understand the differences in regenerative capacities among vertebrates.

One of the important findings of the present study is that the initial transient scar deposition is beneficial for heart regeneration in zebrafish. This discovery is paradoxical to the prevalent concept of ‘scarless regeneration’ in mammalian fetal wound healing, limb regeneration in newts and fin regeneration in teleost fish (Brockes and Kumar, 2008; Palatinus et al., 2010; Poss, 2010). Indeed, in the case of the skin and appendages, the regenerative process can be achieved directly after wound healing without going through a reparative phase of scar formation. We found that such a shortcut is deleterious for the heart as a force-generating contractile organ. Insufficient collagen deposition after Alk5/4 inhibition coincided with the alteration of infarct shape, which was characterized by bulging tissue beyond the ventricular circumference. It is conceivable that a lack of collagen, which normally provides tissue stiffness, affects robustness of the damaged ventricular wall that is challenged to withstand the internal blood pressure (Holmes et al., 2005). Because the wall tension is proportional to the cavity diameter and inversely proportional to the wall thickness (Laplace’s law) (Yin, 1981), the increased ventricular wall stress of the collagen-deficient infarct can lead to biomechanical deformations resulting in an augmented curvature of the infarct. From this physical law, we surmise that transient scar formation is an essential step to maintain ventricular wall stiffness during the subacute regenerative stage. This raises the idea of potential advantages of having a transient scar that can progressively be replaced by new invading functional tissue.

We thank V. Zimmermann and M. Kaczorowski for excellent technical assistance and for fish care and M. Celio and N. Panchaud for critical reading of the manuscript. Electrocardiograms were performed in collaboration with G. Rainer and J. Veit.