Loss of Caveolin-1 and caveolae leads to increased cardiac cell stiffness and functional decline of the adult zebrafish heart

Caveolin-1 is the main structural protein of caveolae, small membrane invaginations involved in signal transduction and mechanoprotection. Here, we generated cav1-KO zebrafish lacking Cav1 and caveolae, and investigated the impact of this loss on adult heart function and response to cryoinjury. We found that cardiac function was impaired in adult cav1-KO fish, which showed a significantly decreased ejection fraction and heart rate. Using atomic force microscopy, we detected an increase in the stiffness of epicardial cells and cells of the cortical zone lacking Cav1/caveolae. This loss of cardiac elasticity might explain the decreased cardiac contraction and function. Surprisingly, cav1-KO mutants were able to regenerate their heart after a cryoinjury but showed a transient decrease in cardiomyocyte proliferation.


Results
Caveolin-1 expression in the intact and regenerating zebrafish heart. We began our analysis by examining Cav1 expression in intact hearts. We used the Tg(fli1a:GFP) line 26 , which expresses GFP in the endocardium and endothelium, and stained with antibodies against Cav1 and tropomyosin (Fig. 1a). Robust Cav1 expression was detected in the vasculature (asterisks in Fig. 1b, b′, b′′) and in the endocardium (arrowheads in Fig. 1b, b′, b′′ and c, c′, c′′). Strong expression was also found in the epicardium (arrows in Fig. 1b, b′, b′′), in the bulbus arteriosus and in the valves (Fig. 1a). Additionally, Cav1 expression was detected in the area between the cortical and trabecular myocardium (dashed area in Fig. 1b, b′, b′′ and inset in Fig. 1b′). We then analysed Cav1 expression in the regenerating zebrafish heart after cryoinjury (Fig. 1d). We used the Tg(wt1b:GFP) line that expresses GFP in the epicardium upon injury 27 . Seven days post cryoinjury (dpci), Cav1 was strongly expressed in epicardial cells (Fig. 1e, f, f′ brackets) covering the injured site, overlapping with GFP. High expression was also detected in the endocardium within the injured area (Fig. 1f, arrows). To confirm these observations, we utilised the Tg(fli1a:GFP) line and found that Cav1 was expressed in GFP + endocardial cells invading the damaged tissue (Fig. 1g, h, h′, arrows). We also surveyed the expression of caveolae-related genes during heart regeneration by quantitative (q)PCR (Fig. 1i). cav1 and cavin1b were upregulated after injury, in contrast to cav2 and cav3 whose expression remained stable. These results show that Cav1 is expressed in the endocardium, endothelium and epicardium of the intact heart, three cell types that are activated during regeneration [28][29][30][31] . Also, upon injury, Cav1 expression is strongly increased in epicardial cells surrounding the injured site, and in the endocardium invading the injured area.
Generation of cav1-KO zebrafish by CRISPR/Cas9 editing. We generated cav1-KO zebrafish to study the role of Cav1 and caveolae in heart homeostasis and regeneration. We used CRISPR/Cas9 editing to target the third exon of cav1, which corresponds to the C-terminal region of the protein (Fig. 2a). The zebrafish cav1 gene generates two transcripts-cav1a and cav1b-sharing the majority of the coding sequences, with cav1a giving rise to 31 amino acids longer protein 23 . We were able to introduce and identify mutations in the cav1 locus, with the majority of them being small deletions (Fig. 2b). The predicted effect on the protein was an open reading frame shift that would lead to an amino acid change and the generation of a premature stop codon (Fig. 2c). We selected the cav1 cn100 mutation, which showed significantly decreased cav1 expression, increased cav2 expression, but with no effect on cav3, cavin1b or cavy transcription (Fig. 2d). Mutant embryos had no morphological abnormalities, developed normally and were fertile (data not shown). We next investigated Cav1 expression by labelling 7 dpci WT and mutant samples with an antibody against Cav1a (Fig. 2e-j and Supplementary Fig. S1a). The strong Cav1a signal in the epicardium, endocardium, endothelium, bulbus arteriosus and valves ( Fig. 2e-g) was lost in cav1 cn100 hearts ( Fig. 2h-j). We repeated this analysis with an antibody that recognises both Cav1 proteins, Cav1a and Cav1b (Fig. 2k-p and Supplementary Fig. S1b). Normal Cav1 expression in epicardium, endocardium and endothelium ( Fig. 2k-m) was absent in cav1 cn100 mutants ( Fig. 2n-p), indicating the lossof-function (LOF) nature of the mutation. Cavin1 is also an essential component of caveolae 32 and deletion of Cav1 diminishes Cavin1 expression in mice 33 . We therefore asked whether Cavin1 was affected by Cav1 loss. Examination of 7dpci cav1 +/+ cryoinjured hearts revealed strong Cavin1 expression in the epicardium, in cardiomyocytes adjacent to the injured area, in the bulbus arteriosus and in the valves ( Supplementary Fig. S2). Cavin1 expression was decreased in cav1 cn100 hearts ( Supplementary Fig. S2). Specifically, Cavin1 was absent in the valves, whereas its expression was greatly reduced in the bulbus arteriosus ( Supplementary Fig. S2e) and in cardiomyocytes within the proliferative zone ( Supplementary Fig. S2g, g′). Taken together, these results show that the cav1 cn100 mutation leads to the loss of Cav1 and to the reduced expression of Cavin1 in cryoinjured hearts.
Loss of caveolae in cav1 cn100 mutant hearts. Expression of Cav1 in a system without endogenous Cav1 expression leads to de novo caveolae formation 5 , whereas deletion of Cav1 results in loss of caveolae [6][7][8] . Hence, we asked whether cav1 cn100 mutants could form caveolae. To address this, we analysed cav1 +/+ and cav1 cn100 hearts by transmission electron microscopy (TEM) (Fig. 3). We found caveolae in abundance in cav1 +/+ hearts (Fig. 3a,  a′), and the membrane of the endothelial cells was packed with caveolae-invaginations (Fig. 3a′, arrowheads). By contrast, cav1 cn100 hearts were deprived of caveolae in the coronary vasculature of the cortical layer (Fig. 3b,b′,c). No membrane-bound caveolae were detected, indicating the complete loss of caveolae in cav1 cn100 mutant hearts.
Response of caveolae-depleted hearts to injury. We next examined the effects of caveolae loss in adult heart regeneration. We cryoinjured cav1 +/+ and cav1 cn100 hearts and allowed them to regenerate for 90 days. We then analysed the hearts by Acid Fuchsin Orange-G (AFOG) staining, which labels both the damaged area and the healthy myocardium. We found that homozygous cav1 cn100 and heterozygous cav1 +/cn100 hearts had regenerated similarly to cav1 +/+ hearts 90 dpci (Fig. 4g-j). Cryoinjury results in the formation of a scar tissue that progressively degrades during the course of 90 days [34][35][36] . Thus, we monitored the regeneration process by analysing injured hearts at 30 and 60 dpci (Fig. 4a-f). cav1 cn100 hearts had a similar scar size to that of cav1 +/+ controls, both at 30 and 60 dpci. www.nature.com/scientificreports/ www.nature.com/scientificreports/ www.nature.com/scientificreports/ To further investigate the effect of caveolae loss in heart regeneration, we examined another cav1 mutant, cav1 cn101 . Interestingly, expression of cav1 transcript in the cav1 cn101 mutant embryos was unchanged, whereas cav2 was upregulated ( Supplementary Fig. S3a), in accord with our observations in cav1 cn100 mutants. In addition, www.nature.com/scientificreports/ Heart regeneration is unaffected in cav1 cn100 mutants. (a-j) cav1 +/+ and cav1 cn100 hearts were cryoinjured and harvested 30 (a, b), 60 (d, e) or 90 dpci (g, h), and processed for AFOG staining, which labels collagen in blue, fibrin in red and healthy myocardium in brown. (i) heterozygous cav1 +/cn100 heart, 90 dpci. (c, f, j) Quantification of the injury site. Injuries were quantified as a percentage of the damaged tissue (collagen and fibrin) to the total area of the ventricle. 30 dpci n WT = n cn100 = 10; 60 dpci n WT = n cn100 = 10; 90 dpci n WT = 9, n cn100 = 12, t-test. Scale bars 250 μm. www.nature.com/scientificreports/ the cav1 cn101 mutation had the same effect on protein expression as the cav1 cn100 mutation, with a loss of Cav1 expression ( Supplementary Fig. S3). Likewise, caveolae were absent in cav1 cn101 hearts, similar to our observations for cav1 cn100 (Supplementary Fig. S4). We then cryoinjured cav1 +/+ and cav1 cn101 hearts and examined the regeneration process every 30 days for 90 days ( Supplementary Fig. S5). cav1 cn101 hearts regenerated normally and we did not detect any differences in the size of the injury at 30, 60 or 90 dpci. These results demonstrate that loss of Cav1 and caveolae do not affect heart regeneration.
Activation of TGFβ signalling and fibrosis are unaffected in cav1 cn100 hearts. Caveolae are involved via Cav1 in the regulation of the TGFβ pathway 12 , a major signalling pathway that controls extracellular matrix (ECM) deposition, and is activated during zebrafish heart regeneration 37 . As both cav1 cn100 and cav1 cn101 hearts regenerated normally, we focused only on cav1 cn100 . To address TGFβ activity in cav1 cn100 hearts upon cryoinjury, we quantified the nuclear localization of phospho-Smad3 (psmad3), a downstream effector of TGFβ ( Supplementary Fig. S6a, b). We used 14 dpci Tg(fli1a:GFP) hearts to calculate the proportion of psmad3 + nuclei in endocardial cells within the damaged tissue ( Supplementary Fig. S6a′, b′, c) and in the cardiomyocytes surrounding the injured site ( Supplementary Fig. S6a′′, b′′, d). Analysis revealed that TGFβ signalling was equally active in control and cav1 cn100 hearts, in both endocardial cells and cardiomyocytes ( Supplementary Fig. S6c, d).
Heart cryoinjury in the Cav1-KO mouse leads to extensive collagen deposition and cardiac remodelling 22 . Thus, we exploited the AFOG staining protocol to examine the collagen and fibrin content after injury, and we also measured ventricular size ( Supplementary Fig. S7a, b). We found no differences between cav1 +/+ and cav1 cn100 hearts, neither in collagen deposition nor in fibrin amount in the injury, nor in the size of the ventricle. Furthermore, because hearts of Cav1-KO mice show increased interstitial fibrosis 1,6,18,38 , we examined this parameter in intact cav1 cn100 hearts using Picrosirius Red to stain collagen fibres. Results showed no difference in interstitial fibrosis between cav1 +/+ and cav1 cn100 hearts (Supplementary Fig. S7c-e). Thus, loss of Cav1 and caveolae does not affect TGFβ activity or fibrosis in intact or cryoinjured hearts.
Epicardial, endocardial and cardiomyocyte function in cav1 cn100 hearts upon injury. We next examined the behaviour of the different cell types involved in the regeneration process. We first analysed the epicardium and endocardium, where Cav1 is highly expressed 7 dpci ( Supplementary Fig. S8). We crossed the Tg(wt1b:GFP) line to cav1 cn100 to study epicardial proliferation and we found no difference in epicardial proliferation between cav1 +/+ and cav1 cn100 hearts ( Supplementary Fig. S8a, b, c). We then evaluated the abundance of endocardial cells within the damaged tissue by crossing double transgenic fish Tg(fli1a:GFP)/Tg(myl7:mRFP) expressing GFP in endocardial/endothelial cells and RFP in the membrane of cardiomyocytes, with cav1 cn100 mutants ( Supplementary Fig. S8d, e). Three-dimensional volume rendering and analysis of the GFP + cells inside the RFParea showed that endocardial cells in cav1 cn100 hearts populated the injured area similarly to those of cav1 +/+ hearts ( Supplementary Fig. S8f). These data indicate that epicardial proliferation and the behaviour of endocardial cells are unchanged in cav1 cn100 injured hearts. Cardiomyocyte proliferation has been reported to be reduced upon ventricular resection in a cav1a-KO zebrafish model 24 . The cardiomyocytes adjacent to the injured area 7 dpci are highly proliferative 39 and we analysed BrdU incorporation in control and cav1 cn100 hearts (Fig. 5a, b, a′, b′). Cardiomyocyte proliferation was significantly lower in cav1 cn100 hearts than in cav1 +/+ hearts (Fig. 5c). We confirmed this result in the cav1 cn101 animals ( Supplementary Fig. S9). We then addressed the proliferation status of cardiomyocytes at 14 dpci (Fig. 5d, e). We found that at this time point, cav1 cn100 cardiomyocytes proliferated at the same rate as cav1 +/+ cardiomyocytes (Fig. 5d′, e′, f). These data indicate that loss of caveolae leads to the transient attenuation in cav1 cn100 cardiomyocyte proliferation at 7 dpci, which is normalised by 14 dpci, leading to normal cardiac regeneration at 90 dpci. Impaired heart function and cardiac elasticity in cav1 cn100 mutants. Cav1-deficient mice have been reported to show decreased systolic function 1,8 . Therefore, we analysed heart function in our mutants using ultrasound imaging. Echocardiography of eleven months-old adult cav1 +/+ and cav1 cn100 animals revealed that cav1 cn100 hearts were less efficient in pumping blood than control hearts as their ejection fraction was significantly lower (Fig. 6a). Additionally, the heart rate of cav1 cn100 animals was lower than in control siblings (Fig. 6b). These findings indicate that caveolae are essential for normal cardiac function.
Caveolae provide protection against mechanical stress 13,14 . Because the biomechanical properties of cells influence the behaviour of tissues 40 and caveolae are involved in mechanoprotection, we investigated the mechanical response of cav1 +/+ and cav1 cn100 cardiac tissue to gain insight into the differences observed in cardiac function. We used AFM to determine the response of adult cav1 cn100 hearts to force. Freshly isolated hearts were placed on an agarose gel, with the apex of the ventricle oriented towards the direction of the cantilever, and force measurements were taken (Fig. 6c-e). We found that cardiac tissue stiffness was significantly greater in cav1 cn100 animals than in control animals (Fig. 6f). Applying the same force, the extent of deformation was 1.5 times greater in control hearts than in cav1 cn100 hearts, without causing permanent deformation as the curves returned to zero (Fig. 6g). This result suggests that changes in epicardial and underlying cells, including cortical myocytes, were responsible for the observed difference in stiffness that we detected. That is because the experiments were performed by producing a compressive deformation (indentation) of 1.5 μmin cav1 cn100 mutant hearts. The force-distance curves (Fig. 6g) did not show evidence of tissue and/or cell rupture during the deformation. Therefore, the epicardium is a major contributor to the measured stiffness (550 Pa). However, the use of large probes (R = 30 µm) implies that deformation is also transmitted to surrounding regions beyond the 1.5 μm indentation depth. By using bottom-effect corrections 41 we estimate that the cells within 5-10 μm distance from the contact point will also contribute to the measured stiffness. Membrane invaginations such as caveolae, provide the necessary stretch capacity for cells to buffer the impact of mechanical forces 14  www.nature.com/scientificreports/ www.nature.com/scientificreports/ www.nature.com/scientificreports/ of membrane reservoirs, caveolae, due to the loss of the broad and robust expression of Cav1 in the ventricle, results in increased tissue stiffness.

Discussion
Caveolin-1 is the main structural protein of caveolae, small membrane invaginations involved in signal transduction and mechanoprotection 7,13,14 . Here, we investigated the importance of Cav1 and caveolae in heart homeostasis and regeneration after cryoinjury in zebrafish. We found that while Cav1 expression in the heart is strongly increased upon cryoinjury, its deletion, resulting in loss of caveolae, does not affect heart regeneration. However, loss of Cav1 and caveolae leads to increased stiffness of epicardial cell and cells of the cortical zone, including cardiomyocytes, and impaired cardiac function. We found that Cav1 is expressed in the endothelium, endocardium and epicardium of intact adult hearts. Also, cav1 transcription is upregulated following injury and Cav1 is strongly expressed 7 dpci in epicardial cells covering the damaged cardiac tissue, and in the endocardium invading the injured area. We generated cav1-KO zebrafish strains to examine the role of Cav1 in the heart. We used two cav1 mutant alleles, cav1 cn100 and cav1 cn101 . Both mutations are LOF alleles, as they led to the lack of both Cav1a and Cav1b proteins expression. TEM analysis showed that caveolae were completely absent in both cav1 cn100 and cav1 cn101 hearts, whereas they were abundant in cav1 +/+ hearts. Thus, cav1 cn100 and cav1 cn101 mutations lead to loss of Cav1 and caveolae. This loss also affected Cavin1 protein expression, although transcription was unaffected. The cav1-KO hearts lacked Cavin1 in the valves and the overall expression of the protein was decreased, which is in accord with results from the Cav1-KO mouse 32,33 .
We investigated the effect of Cav1 and caveolae loss in cardiac regeneration using the cryoinjury model, as it mimics the mammalian myocardial infarction in formation of necrotic myocardium and scar tissue, which in the zebrafish gradually resolves leading to complete tissue regeneration at 90 dpci [34][35][36] . We found that mutant hearts carrying the cav1 cn100 or cav1 cn101 LOF mutations in heterozygous or homozygous condition regenerate similar to the WT heart upon cryoinjury and we did not detect differences in injury size at 30 and 60 dpci. These findings were unexpected, as it has been reported that ventricular resection in heterozygous cav1-KO zebrafish results in enhanced fibrosis, leading to defects in heart regeneration 30 dpa 24 . The fibrotic response in ventricular resection is minimum, whereas cryoinjury leads to massive fibrosis and to the formation of a transient scar tissue [34][35][36]42 . However, our data indicate that the formation and resolution of the scar tissue is unaffected by Cav1 loss. Additionally, caveolin-1 negatively regulates TGFβ signalling 12,] and loss of murine CAV1 enhances collagen deposition after injury 21,22 . Notably, we found no changes in the quantity of collagen within the damaged tissue between cav1-KO and WT hearts, or in TGFβ pathway activation. Likewise, there were no signs of interstitial fibrosis in intact cav1-KO hearts, as has been reported in murine Cav1-KO models 1,6,18,38 . These results suggest activation of a compensatory mechanism 43 that buffers Cav1 loss, which in heterozygosity might be induced stronger upon cryoinjury than after resection, in which fibrosis is minimal. Hence, since heart regeneration after ventricular resection lasts up to 60 days 42 , it would be of interest to examine the process at the final time point. Intriguingly, cav1 cn100 mutation results in decreased expression of cav1 transcript, whereas, cav1 transcription is unchanged in cav1 cn101 . However, both of our mutations lead to loss of Cav1 protein expression. This could suggest that compensation might be triggered also after protein loss 43 , or that a Cav1-independent TGFβ regulatory pathway may exist in zebrafish heart regeneration.
We detected Cav1 in the epicardium covering the injured area and in endocardial cells invading the damaged tissue. Upon injury, the epicardium is also populated with fibroblasts and macrophages that contribute to fibrosis 44,45 , and loss of Cav1 in mice is associated with increased fibrosis and macrophage infiltration after cardiac injury 21,22 . The proliferation rate of epicardial cells was unchanged in cav1-KO hearts, indicating that these two cell types are unaffected in our model. In addition, endocardial cells have been associated with collagen deposition 31 and Cav1 was completely lost in these cells. However, endocardial cell abundance was similar between WT and cav1 cn100 hearts, suggesting that their proliferation and migration was unaffected. Collectively, these findings further indicate that loss of Cav1 and caveolae does not lead to deregulation of fibrosis in intact or in cryoinjured hearts. Moreover, loss of Cav1 and caveolae triggered a transient but significant decrease in cardiomyocyte proliferation 7 dpci. This was not reflected in defective or even delayed regeneration. Indeed, cardiomyocyte proliferation in mutant fish 14 dpci was similar to that of controls, suggesting that cardiomyocytes were overall able to increase in numbers, populate the damaged area and contribute to normal heart regeneration in cav1 mutants. In summary, considering that the heart of both of homozygous or heterozygous cav1 LOF mutants regenerates normally, we can only conclude that Cav1 and caveolae are not required for zebrafish heart regeneration.
Systolic and diastolic functions are also affected in Cav1-null mouse hearts 1,8,18,19 and, in line with this finding, the cav1 cn100 line showed impaired cardiac performance, as indicated by the reduced EF and a lower heart rate. As we did not detect cardiac fibrosis, the reduction in cardiac elasticity is a plausible explanation for the cardiac dysfunction in cav1-KO fish. Membrane invaginations such as caveolae provide the necessary stretch capacity for cells to buffer the impact of mechanical forces 13,14 . Our AFM-force spectroscopy analysis revealed that cav1-KO hearts were significantly stiffer than WT hearts, as cardiac tissue elasticity was decreased almost two-fold after loss of Cav1 and caveolae. The sensitivity of AFM allows us to conclude that the measurements represent the stiffness of both the epicardial cells and the underlying cortical cells, including cardiomyocytes. We have shown that Cav1 is strongly expressed in the epicardium of intact hearts and that its expression is lost in cav1 cn100 hearts. Additionally, RNA-seq data show that cav1 expression is moderate in cortical cardiomyocytes, but significantly higher than in trabecular cardiomyocytes 44 . Caveolae are required for smooth muscle cell contractility 46,47 and cell contraction results in changes to membrane tension that caveolae buffer 14 . In addition, Cav1 and caveolae regulate RhoA activation (a GTPase protein that regulates the cytoskeleton), which in turn regulates actomyosin Scientific RepoRtS | (2020) 10:12816 | https://doi.org/10.1038/s41598-020-68802-9 www.nature.com/scientificreports/ contractility [48][49][50] . Accordingly, these data suggest that loss of Cav1 and caveolae impairs cell elasticity and/or contractility and, consequently, pumping efficiency, which could explain the significant reduction in cardiac performance. Cav1 deficiency in the mouse leads to diverse cardiac phenotypes that are attributed to extensive fibrosis and endothelial loss of Cav1 and caveolae. Our results demonstrate that loss of Cav1 and caveolae result in cardiac stiffening accompanied by reduced cardiac function, suggesting that a global change in the mechanical properties of the heart leads to the cardiac phenotypes observed in Cav1-deficiency models. Adult heart cryoinjury. Fish were anesthetised by immersion in 0.04% tricaine (Sigma-Aldrich, St Louis, MO) in fish water and placed on a wet sponge under a stereoscope with the ventral side exposed. The cardiac cavity was opened using microscissors and microforceps, and the pericardium was removed. The ventricle of the heart was exposed and dried and was then touched by a copper-made probe previously immersed in liquid nitrogen 28 . The fish were immediately returned to water to recover.

Methods
Bromodeoxyuridine injection. Adult fish were anesthetised and placed on a wet sponge under a stereoscope. 5-bromo-2′-deoxyuridine (BrdU) was diluted in phosphate-buffered saline (PBS) to 2.5 mg/ml and 30 μl were injected intraperitoneally 24-h before dissection of the hearts.
Echocardiography. Analysis of cardiac function by echocardiography in eleven months-old adult fish was performed as described 53 . Briefly, the fish were anesthetised by immersion in 60 mM tricaine and 3 mM isoflurane in fish water and transferred to a sponge immersed in the same solution. Images were acquired using the VEVO 2100 system (VisualSonics Inc., Toronto, ON, Canada) with a 50-MHz ultrasound probe. The transductor was immersed in the medium dorsally to the cardiac cavity. The fish were immediately transferred to fresh water to recover after the procedure.
Histological stains. Acid Fuchsin Orange G-staining (AFOG) and Picrosirius Red staining were performed following standard protocols 42 . Immunofluorescence. Sections of paraffin-embedded tissue were permeabilised with PBT (PBS with 0.01% TritonX-100) and washed with PBS before incubation with blocking solution (2% bovine serum albumin, 10% goat serum and 2 mM MgCl 2 in PBS). Sections were then incubated overnight at 4ºC with the antibodies of choice. Caveolin 1a (Cav1a), this isoform is 31 amino acids longer in the N-terminus region than Cav1b. The Cav1a antibody (Cell Signalling Technology, catalogue #D46G3) was raised against the N-terminus region of Cav1a, including residues surrounding Glu20. www.nature.com/scientificreports/ Whole-mount confocal imaging. Analysis of endogenous fluorescence of whole-mount hearts was performed as described 31 . Briefly, hearts were fixed overnight with 2% paraformaldehyde and after several PBS washes the tissues were immersed in 3% agarose. Samples were then incubated in CUBIC I solution 54 at 37 °C for one week. Agarose blocks containing the hearts were mounted for imaging on a petri dish and approximately 700 μm of the injured ventricle was scanned on a Leica SP8 confocal microscope (Leica Microsystems, Wetzler, Germany) using a 10 × objective.
Quantitative RT-PCR. Three-to-five biological replicates with three technical replicates of each sample were used for the expression analysis of genes by qPCR using the power SYBR Green Master Mix (Applied Biosystems, Foster City, CA) and the ABI PRISM 7900HT FAST Real-Time PCR System. All measurements were normalised to the expression of 18s 55 . The following primers were used for the qPCR analysis: 18sFwd TCG CTA GTT GGC ATC GTT TATG, 18sRvs CGG AGG TTC GAA GAC GAT CA, cav1Fwd TGG GAT GGG GGA ATG GAA AC, cav1Rvs TAA ACG GCG AGT GAG CGT AT, cav2Fwd GCG TTT ATT GCA GGG ATT GT, cav2Rvs GGA TCA CTG GCA TCA CCA C, cav3Fwd CAA CGA AGA TGT CGT GAA GG, cav3Rvs GAG ACG GTG AAG GTG GTG TAA, and for cavin1b and cavy from 17 .
Samples were post-fixed in 1% osmium tetroxide for 60 min and dehydrated through a series of ethanol solutions (30%, 50%, 70%, 95%, 100%) and acetone. After the last dehydration step, samples were incubated in a 1:3, 1:1, 3:1 mixture of DURCUPAN resin and acetone and cured at 60 °C for 48 h. Ultrathin Sections (50-60 nm) were obtained using a diamond knife (Diatome AG, Biel, Switzerland) in an ultramicrotome (Leica Reichert ultracut S. Leica Microsystems) and collected in 200-mesh copper grids. The sections were counterstained with 2% uranyl acetate in water for 20 min followed by a lead citrate solution. Sections were examined with a JEOL JEM1010 electron microscope (Tokyo, Japan) equipped with an Orius SC200 digital camera (Gatan Inc., Pleasanton, CA).

Image analysis and quantification.
To analyse cardiomyocyte proliferation, MEF2-positive nuclei were counted in an area of 100 μm around the injury site using Fiji (ImageJ, NIH). BrdU-MEF2-positive cells were also counted and the % proliferation index was expressed as the MEF2:BrdU-MEF2 ratio. For TGFβ signalling activation, all phospho-smad3-positive cardiomyocyte (100 μm of the injury) or GFP-positive cells (inside the injured area) were counted and normalised to the total number of cardiomyocytes or GFP-positive cells. To quantify the regeneration process, at least three sections, in the middle of the ventricle that contain both of the valves, of each heart were used and the injured area (fibrotic tissue and collagen) was measured using Fiji and expressed as a percentage of the total ventricular area. The 3D analysis of the whole-mount hearts was carried out using Fiji and IMARIS programmes. The volume of the GFP signal inside the injured area (RFP negative) was also measured and presented in relation to the volume of the injury. Electron microscopy images of the plasma membrane of coronary endothelial cells were taken at 50,000 × magnification. Uncoated membrane invaginations of 40-90 nm size were counted 17 and expressed as density per μm 2 of the perinuclear area. Two endothelial cells per three sections of the same heart were examined. Fiji was used to calculate the perinuclear area and for caveolae identification.
Atomic force microscopy (AFM)-force spectroscopy. Adult zebrafish were sacrificed by immersion in 0.16% tricaine and the heart was dissected. The atrium was removed and the ventricle was placed horizontally atop a 4% agarose gel immersed in PBS with 0.1 M KCl to arrest the heartbeat uniformly. AFM-force spectroscopy experiments were performed with a JPK Nanowizard III microscope (JPK Instruments, Berlin, Germany) coupled with an inverted optical microscope (AXIO Observer D1; Carl Zeiss, Germany) and equipped with Plateau-CONT-SPL cantilevers (Nanosensors, Neuchatel, Switzerland), with a nominal spring constant of 0.02-0.77 N/m and a spherical tip shape (R = 30 µm). The actual spring constant of the cantilever was determined using the thermal noise method as implemented in the AFM software. Force-distance curves (FDC) were acquired to determine Young's modulus of the zebrafish heart apex. The tip-sample distance was modulated by applying a triangular waveform 16 . The tip velocity was set to 10 µm/s and the amplitude to 15 μm. The maximum force exerted on the heart apex during a single FDC was of 11 nN. For each zebrafish heart, a complete sequence of 375 FDC was performed. These FDC were distributed in three different areas of 100 × 100 µm 2 several hundreds of microns apart. In each area, 125 FDC were measured (all along the zebrafish apex). To determine the contact point, we used a ratio of variances protocol. Young's modulus was obtained by fitting a section of the force-distance curve (approach semi-cycle of the whole FDC) with a Hertz model for spherical indenters.
Statistical analysis. Sample sizes, statistical tests and P-values are specified in the figure legends and were determined with GraphPad Prism software (GraphPad Software Inc., San Diego, CA). Statistical tests were twotailed. P values below 0.05 were considered of statistical significance.