GATA4 regulates Fgf16 to promote heart repair after injury

Myocardial infarction is becoming the leading cause of death worldwide. During myocardial infarction, a substantial number of cardiomyocytes die and cardiac hypertrophy occurs subsequently as a result of a compensatory response. Lack of adequate numbers of functional cardiomyocytes coupled with chronic overload and subsequent dysfunction of the remaining cardiomyocytes eventually lead to heart failure and death. As cardiomyocyte replication is the key mechanism by which the neonatal heart regenerates, unraveling the transcriptional regulatory network responsible for cardiomyocyte replication could provide useful information for the development of novel therapies to drive cardiac repair after injury.

The GATA family members are zinc finger transcription factors. They bind specifically to the HGATAR-containing DNA motifs through their highly conserved zinc-finger DNA-binding domains (Patient and McGhee, 2002). Vertebrates have six GATA factors; GATA4 plays a key role during heart development. It regulates many cardiac-specific genes that are important for embryonic and neonatal heart development (Monzen et al., 1999), including differentiation (Boheler et al., 2002), migration (Molkentin et al., 1997), hypertrophy (Charron et al., 2001; Liang et al., 2001) and survival (Aries et al., 2004) of cardiomyocytes. GATA4 also coordinates with a large number of cardio-regulators via the C-terminal zinc finger domain to maintain proper cardiac function throughout life (Durocher et al., 1997; Lu et al., 1999; Belaguli et al., 2000; Morin et al., 2000; Bhalla et al., 2001). Previous studies showed that GATA4 is crucial for regulation of cardiac hypertrophy, cardiomyocyte viability and fibrosis during adulthood (Bisping et al., 2006; Oka et al., 2006). In spite of the important function of GATA4 in heart development and homeostasis, the role of GATA4 in mammalian heart regeneration remains unknown.

Fibroblast growth factors (FGFs) are pluripotent growth factors that are essential for cell proliferation and differentiation during organ development and tissue regeneration (Ornitz and Itoh, 2015). FGFs promote coronary angiogenesis during development and repair (Cross and Claesson-Welsh, 2001) and are also important for wound healing (Werner and Grose, 2003). FGFs play a pivotal role in development of multiple systems, including the limbs, lungs and nervous system (Min et al., 1998; Ford-Perriss et al., 2001). FGF16 belongs to the FGF9 subfamily and is the only cardiac-specific member of the FGF family. FGF16 is mainly expressed in cardiomyocytes and has been shown to be secreted from cardiomyocytes after birth, suggesting its specific role in the postnatal heart (Hotta et al., 2008). A previous study showed that FGF16 is required for cardiomyocyte replication during development (Lu et al., 2008). FGF16 expression increases after birth and it is stored in the extracellular matrix throughout adulthood to maintain cardiac homeostasis and protect the heart from injury (Miyake et al., 1998; Hotta et al., 2008). Recent reports also showed that FGF16 protects the heart from doxorubicin-induced heart dysfunction (Sontag et al., 2013). Furthermore, Fgf16 is a direct target of NF-κB (NFKB1), which is involved in many cardiac pathological processes (Valen et al., 2001; Sofronescu et al., 2010). Altogether, previous findings indicate that FGF16 may play a direct role in protection of the neonatal heart.

In this study, we demonstrated that cardiomyocyte-specific ablation of Gata4 resulted in compromised regenerative capacity in the neonatal heart with decreased proliferating cardiomyocytes following apical resection and cryoinjury. We also found that GATA4 regulated Fgf16 expression. Experimental overexpression of FGF16 via AAV9 in cardiomyocytes partially rescued the phenotype of Gata4 ablation, suggesting that the GATA4/FGF16 axis could regulate neonatal heart regeneration.

To study the role of GATA4 in neonatal heart regeneration, we generated a model of inducible deletion of Gata4 in cardiomyocytes at the perinatal stage. We used the iTNT-Cre mouse, in which the cardiac-specific troponin T promoter drives the reverse tet activator protein (rtTA) and the tet activator protein-dependent (TRE) promoter drives the Cre recombinase (Fig. 1A). Administration of doxycycline (Dox) upregulates Cre expression in cardiomyocytes, resulting in Cre-mediated recombination of the floxed targets (Wu et al., 2010). We next generated the iTnt-Cre;Gata4^fl/fl mice (mutant), in which exons 3-5 of Gata4 are deleted specifically in cardiomyocytes (Watt et al., 2004) after intragastrical administration of Dox during the perinatal stages [embryonic day (E) 16.5 to 18.5; Fig. 1A]. For the control, we crossed the non-transgene (NTg)-bearing iTNT-Cre-negative littermates with the Gata4^fl/fl mice, which received the same Dox treatment. We next introduced cryoinjury in neonatal mice because the cryoinjury model provides a scar of consistent size and shape to obviate the injury variance between animals (Strungs et al., 2013). Using Masson trichrome staining and Sirius Red staining to localize fibrotic tissues, we found scar formation in the neonatal heart 5 days after cryoinjury (Fig. 1B). We also confirmed formation of fibrotic tissues in the cryoinjury region by immunostaining of the fibrosis marker collagen III (COL3) and the cardiomyocyte-specific marker ACTN2 (Fig. 1B), demonstrating the successful establishment of cryoinjury in neonatal hearts. In order to test the efficiency of ablating Gata4 in neonatal hearts, we collected mutant and control hearts at postnatal day (P) 4, and analyzed Gata4 expression by quantitative reverse transcription PCR (qRT-PCR). We found that Gata4 mRNA was significantly reduced in the Gata4 mutant group compared with the littermate controls (Fig. 1C). To further validate deletion of GATA4 at the protein level, we performed western blot using lysates from mutant and control hearts and confirmed that GATA4 was significantly reduced in the mutant group compared with the control group (Fig. 1D). To confirm the knockout with single-cell resolution, we performed co-staining of GATA4 and ACTN2 on mutant and control hearts. Compared with control hearts, GATA4 was not expressed in the majority of cardiomyocytes in the mutant hearts (Fig. 1E). Quantification of the number of GATA4-expressing cardiomyocytes showed that GATA4 was deleted in >90% cardiomyocytes of the mutant hearts (Fig. 1F). Altogether, we established the neonatal cardiomyocyte-specific Gata4 ablation model and cryoinjury model, allowing us to test whether GATA4 is required for recovery and heart regeneration after injury.

We next evaluated the regenerative capacity of the Gata4-ablated neonatal hearts by examining their function at the adult stage. After Dox treatment from E16.5 to E18.5, we performed cryoinjury on neonatal murine hearts at P1 and measured heart function by echocardiography 8 weeks after cryoinjury. Gross morphology showed that there was an obvious scar formation in the mutant heart, which was less visible in the controls (Fig. 2A). We also found compromised heart function and significantly decreased fractional shortening in the mutant mice compared with the littermate controls by echocardiographic measurements (Fig. 2B,C; Table S1). Interestingly, we found that heart function was normal in the sham-operated Gata4 knockout hearts at 8 weeks, indicating that cardiomyocytes could function normally without GATA4 in the absence of injury (Fig. 2C; Table S1). Using Sirius Red and Fast Green staining on heart sections, we found that the area of scar tissue was significantly larger in the mutant hearts compared with that of the controls (Fig. 2D). Quantification of scar size (measured as a percentage of ventricular circumference in sequentially stained sections) showed that a significantly larger scar formed in the mutant hearts compared with that of the controls (Fig. 2E). To examine whether deletion of Gata4 in cardiomyocytes results in cardiomyocyte hypertrophy of the adult hearts after injury, we co-stained with the lectin-specific marker wheat germ agglutinin (WGA) and an antibody against the muscle-specific actinin marker ACTN2 on mutant and control heart sections to visualize the size of individual cardiomyocytes. We found that the mean size of cardiomyocytes of mutant hearts was larger than that of the controls (Fig. 2F,G). Together, our data suggested that GATA4 might be required in heart regeneration and functional maintenance after cryoinjury.

We further applied a second injury model, apical resection (AR), to test whether Gata4 is required during heart regeneration. We first treated mice with Dox at the perinatal period and removed the heart apex from neonatal mouse heart as previously described (Porrello et al., 2011). Echocardiographic measurements of mouse hearts showed that the heart function of mutant mice did not recover completely compared with littermate controls after AR (Fig. 3A,B). Sirius Red staining on sagittal heart sections showed that there was still fibrotic scar tissue in the apical area of the Gata4 mutant hearts, which failed to regenerate the apex completely, whereas the littermate control hearts completely recovered after injury (Fig. 3C), which is consistent with the original report (Porrello et al., 2011). We next investigated whether reduced cardiac regeneration post-AR was due to impaired cardiomyocyte replication in the cardiomyocyte-specific Gata4-ablated mice. To address this issue, we utilized the proliferation markers pH3 (phospho-histone H3), Ki67 (also known as MKI67) and EdU (5-ethynyl-2′-deoxyuridine) incorporation to study cardiomyocytes proliferation 7 days after AR. Quantification of proliferating cardiomyocytes showed that replicating cardiomyocytes were significantly reduced in both the border zone and the remote region of mutant hearts compared with littermate controls (Fig. 3D-H). However, there was no significant difference in the proportion of proliferating cardiomyocyte number in sham-operated mutant hearts and littermate control hearts (Fig. 3D-H).

To examine whether GATA4 induced heart regeneration by promoting cardiomyocyte replication in the cryoinjury model, we measured cardiomyocyte replication at P7 with the pH3 antibody on heart sections as previously described (Porrello et al., 2011). We found that the pH3-positive proliferating cardiomyocytes were dramatically reduced in both the border zone and the remote region of mutant hearts compared with control hearts (Fig. 4A-C). We next used an AURKB antibody to detect cardiomyocyte cytokinesis during proliferation. We found that the number of cardiomyocytes undergoing cytokinesis was significantly reduced in mutant hearts compared with control hearts (Fig. 4D,E). As previous work suggested that cardiomyocytes of the whole heart replicate to compensate for the loss of cardiomyocytes after injury at the neonatal stage (Porrello et al., 2011), our results might indicate that a loss of GATA4 in cardiomyocytes could lead to a reduced regenerative response of the whole heart (both border zone and remote region) after injury.

We then measured the expression levels of genes regulating cell cycle progression, such as cyclin A2, Cdk4, cyclin D1 and cyclin D2, of both the mutant and control hearts. Although cyclin D1 and cyclin D2 were not significantly changed in the mutant hearts, the expression levels of cyclin A2 and Cdk4 were reduced in the mutant heart compared with control hearts (Fig. 4F). Indeed, it has been reported that CDK4 is essential for cardiomyocyte replication and is directly regulated by GATA4 (Rojas et al., 2008). To confirm whether GATA4 directly binds to the Cdk4 promoter, we cloned the Cdk4 promoter into a luciferase reporter plasmid. Co-transfection of the Gata4-expressing plasmids promoted Cdk4 promoter-driven luciferase expression (Fig. 4G). We next performed chromatin immunoprecipitation (ChIP) assays on the C2C12 cell line and confirmed that GATA4 binds to the Cdk4 promoter (Fig. 4H). To test whether the GATA4 binding was enhanced after injury, we used sham-operated and cryo-injured heart lysates for ChIP assays. Gata4 knockout hearts from sham-operated mice were used as a negative control. We found that GATA4 mainly bound to site III in the normal heart lysates. Notably, the binding to site I/II, but not site III, was significantly increased after cryoinjury (Fig. 4I). Taken together, our results suggest that GATA4 might be required for increased cardiomyocyte replication during neonatal heart regeneration in response to injury, possibly by regulating expression of the cell cycle protein CDK4.

A previous report indicates that GATA4 functions as a stress-responsive regulator of coronary angiogenesis (Heineke et al., 2007). To test whether cardiomyocyte GATA4 functions in regulation of coronary angiogenesis during heart regeneration, we compared coronary vessel number in mutant hearts with that of control hearts after cryoinjury. By immunostaining of the endothelial cell specific marker PECAM (also known as PECAM1), we found a significant reduction in the capillary to cardiomyocyte ratio in the mutant hearts compared with that of the controls (Fig. 5A,B), suggesting that in mutant hearts each cardiomyocyte was supported by fewer coronary capillaries. We did not observe a significant reduction of capillary vessels in the mutants compared with sham-operated controls (Fig. 5B). Moreover, we also performed immunostaining of the smooth muscle cell-specific marker αSMA (smooth muscle actin) on heart sections of mutant and control hearts and found a significant reduction in the number of small coronary arteries in the mutant hearts compared with that of control hearts, suggesting that loss of GATA4 in cardiomyocytes might also impair arteriogenesis and maintenance of vasculature after injury (Fig. 5C,D). There was no significant reduction of αSMA⁺ coronary vessels in sham-operated mutant hearts compared with sham-operated control hearts (Fig. 5D). To further identify the pro- or anti-angiogenic genes associated with this phenotypic change, we measured the mRNA expression levels in ventricles of mutant and control hearts. By qRT-PCR, we found that expression levels of a subset of pro-angiogenic genes, such as members of the FGF family (Fgf2, Fgf9, Fgf16) were significantly downregulated, whereas expression levels of members of the vascular endothelial growth factor (VEGF) family (Vegfa, Vegfb and Vegfc) were not significantly changed after ablation of GATA4 (Fig. 5E). Moreover, we also observed that the expression levels of some anti-angiogenic genes or fibrosis-associated genes, such as tissue inhibitor of metalloproteinase 1 (Timp1), collagen type IV alpha 3 (Col4a3), connective tissue growth factor (Ctgf) and thrombospondin 1 (Thbs1), were significantly upregulated in the mutant group compared with the controls (Fig. 5E). Among these genes, the pro-antiogenic Fgf16 and anti-angiogenic Ctgf were downregulated and upregulated, respectively, in sham-operated mutant hearts compared with littermate control hearts (Fig. 5E). Our data suggested that GATA4 could possibly regulate coronary angiogenesis during neonatal heart regeneration.

Because several FGF family members were downregulated in cardiomyocytes with Gata4 ablation, we next examined expression levels of other FGF members in both mutant and control hearts and found that Fgf2, Fgf3, Fgf8, Fgf9 and Fgf16 were significantly downregulated in the mutant hearts compared with control hearts (Fig. 6A). Among these altered paracrine factors, it has been reported that FGF16 is essential for cardiomyocyte replication in the embryonic heart (Hotta et al., 2008). Therefore, we examined the protein expression level of FGF16 in the mutant and control hearts using western blotting. We found that there was a significant reduction in the expression level of FGF16 protein in the mutant hearts compared with the controls after injury (Fig. 6B). Using the ECR browser (http://ecrbrowser.dcode.org), we found the conserved sequence of Fgf16 in the mouse, opossum and human genomes, and an evolutionarily conserved GATA-binding motif on the second intron of the Fgf16 gene (Fig. 6C). Next, we cloned the Fgf16 promoter as well as its second intron into the luciferase reporter constructs, and co-transfected the Fgf16-luciferase reporter with Gata4 and/or p300 (also known as Ep300) into neonatal mouse ventricular myocytes (NMVMs) (Fig. 6D). Our data showed that GATA4 simulated the enhancer activity of Fgf16. When co-transfected with Gata4 and p300, as p300 is known to be a GATA4 activator (Yanazume et al., 2003), the expression level of the reporter was significantly enhanced with a nearly fivefold increase compared with that of the control, which was transfected with a GFP-expressing construct (Fig. 6D). To validate this, we used neonatal mouse hearts to perform chromatin immunoprecipitation followed by quantitative PCR (ChIP-qPCR); primers designed to detect the GATA4-binding sites and the control nonspecific sites were used (Table S2). Our results confirmed that the GATA4-binding site was specifically found in the conserved region of the Fgf16 second intron (Fig. 6E). To measure the transcriptional activity of the GATA4-bound region of Fgf16, we cloned a 460 bp genomic DNA fragment containing the conserved GATA4-binding motif into a luciferase reporter construct, here referred to as Fgf16 enhancer (EN), and a control mutant version (mEN) with a potentially mutated GATA4-binding site (Fig. 6F). When Gata4 was transfected into NMVMs, we found that GATA4 stimulated the transcriptional activity of the Fgf16 enhancer nearly threefold. However, this stimulation could not be detected in the mutant enhancer control (Fig. 6F). Together, our results suggested that GATA4 directly bound to the evolutionarily conserved site located on the second intron of the Fgf16 gene in order to regulate Fgf16 expression in neonatal cardiomyocytes.

As a previous study showed that FGF16 is crucial for cardiomyocyte proliferation during development (Hotta et al., 2008), we investigated whether supplementing FGF16 into the mutant heart could rescue the phenotype caused by the loss of GATA4 in neonatal cardiomyocytes. Therefore, we generated the adeno-associated virus 9:cTNT::Fgf16 virus (AAV9:Fgf16), in which expression of Fgf16 was driven by the cardiomyocyte-specific cTNT promoter and the plasmid was packaged into AAV9, a virus that targets cardiomyocytes efficiently and specifically (Lin et al., 2014). We first used AAV9:cTNT::GFP-Cre (AAV9:GFP-Cre) to evaluate the transduction efficiency and specificity in the Rosa26^RFP/+ reporter mouse. As expected, a strong RFP signal was found specifically in the heart but not in other organs of the Rosa26^RFP/+ reporter mice following infection with AAV9:GFP-Cre at the neonatal stage (Fig. 6G). Immunostaining of the genetic lineage-tracing marker RFP and ACTN2 on the transduced Rosa26^RFP/+ heart sections showed that RFP was specifically expressed in the ACTN2-positive cardiomyocytes and that the transduction efficiency was high as the majority of cardiomyocytes were labeled by RFP (Fig. 6G). After establishing the efficient AAV9 transduction system, we injected the AAV9:Fgf16 virus subcutaneously into wild-type neonatal mice at P1 and injected again at P6. We then collected the hearts for analysis at P9 and examined the expression of FGF16. Here, we used AAV9:cTNT::Luciferase (AAV9:Luci) as negative control. Western blotting of FGF16 showed that the expression of FGF16 protein in AAV9:Fgf16-transduced hearts was 1.5-fold higher than in AAV9:Luci-transduced controls (Fig. 6H). We next adopted the same delivery strategy for the mutant mice and tested whether overexpression of FGF16 in cardiomyocytes could have any effect on expression of hypertrophy-associated genes after Gata4 knockout and injury in neonatal heart. Using qRT-PCR, we found that there was a significant reduction in expression of hypertrophy-associated markers, including Nppa, Nppb, Acta (also known as Acta1) and Fhl1, in mutant hearts transduced with the AAV9:Fgf16 virus compared with those transduced with the AAV9:Luci virus (Fig. 6I,J). Gene expression levels of cell cycle markers such as Ki67, Cdk4 and cyclin A2 were also upregulated in the AAV9:Fgf16-transduced mutant hearts compared with those of the AAV9:Luci-transduced group (Fig. 6K). In addition, the expression level of the apoptosis marker Gdf15 was also significantly downregulated in the AAV9:Fgf16-transduced mutant hearts compared with those of the AAV9:Luci-transduced group (Fig. 6K), suggesting that FGF16 possibly protected cardiomyocytes from cryoinjury-induced cell death. Furthermore, immunostaining of the proliferation markers pH3 and Ki67 on heart sections also showed that overexpression of FGF16 in the Gata4-ablated cardiomyocytes significantly promoted cardiomyocyte replication after injury (Fig. 6L-O). Our results revealed that supplementing FGF16 to the Gata4 knockout hearts could significantly inhibit hypertrophy and promote proliferation of cardiomyocytes after injury.

To determine the long-term effect of FGF16 supplement in rescuing heart function after cardiomyocyte-specific Gata4 ablation, we repeated the same AAV9 delivery experiment as above, and analyzed heart function of the AAV9:Fgf16- or AAV9:Luci-transduced iTNT-Cre;Gata4^fl/fl hearts until 8 weeks old. Our echocardiography results showed that FGF16 supplement partially rescued the function of mutant hearts, as the ejection fraction and fraction shortening were significantly increased in the AAV9:Fgf16-transduced hearts compared with that of the AAV9:Luci-transduced hearts of the iTNT-Cre;Gata4^fl/fl mice (Fig. 7A,B). We also found that overexpression of FGF16 in sham-operated mutant or littermate control hearts did not impair heart function (Fig. 7B). We next measured the fibrotic scar size by Sirius Red and Fast Green staining on heart sections. Compared with the AAV9:Luci-transduced hearts, a significant reduction in scar formation was observed in the AAV9:Fgf16-transduced hearts (Fig. 7C,D). Furthermore, we performed co-staining for WGA and ACTN2 in the AAV:Fgf16- and AAV:Luci-transduced hearts to measure cardiomyocyte size. We found that FGF16 supplement reduced the size of cardiomyocytes in mutant hearts compared with that of the AAV:Luci controls (Fig. 7E,F). To investigate the effect of AAV9:Fgf16 on cardiomyocyte death, we performed terminal deoxynucleotidyl dUTP nick-end labeling (TUNEL) and ACTN2 co-staining to measure the number of apoptotic cardiomyocytes in the border region of the injured heart. We found that there were significantly fewer apoptotic cardiomyocytes in the AAV9:Fgf16-transduced hearts compared with the AAV9:Luci controls (Fig. 7G,H). Taken together, our study showed that FGF16 could partially rescue the impaired regenerative capability of the neonatal mouse heart following the cardiomyocyte-specific ablation of Gata4, at least in part, by inhibiting hypertrophy and apoptosis of cardiomyocytes; reducing scar tissue formation; stimulating replication of cardiomyocytes; and improving heart function after injury.

There are several conserved transcriptional networks that orchestrate the delicate regulation of gene expression during heart development and growth. Previous work reported that GATA4 is essential for heart development (Wang et al., 2013) and postnatal maintenance (Bisping et al., 2006). Overexpression of GATA4 was reported to rescue the injured heart after myocardial infarction (MI) by promoting myocardial angiogenesis, preventing heart remodeling and reducing apoptosis, ultimately leading to reduction of infarct size (Rysa et al., 2010). Here, we used an inducible cardiomyocyte-specific gene knockout strategy to generate cardiomyocyte-specific and inducible Gata4 knockout mice. This strategy allowed us to ablate Gata4 in >90% cardiomyocytes after birth and also avoid early lethality and other embryonic phenotypes accompanied by early embryonic deletion of Gata4 (Zeisberg et al., 2005). Our data suggested that the heart functioned normally after Gata4 deletion in cardiomyocytes at the perinatal stage; however, the mutant heart regenerated poorly after cryoinjury compared with their littermate controls. We further showed that the impairment of heart regeneration was associated with a reduced number of proliferating cardiomyocytes and coronary vessels, demonstrating a pivotal role of GATA4 during neonatal heart regeneration.

In addition to cryoinjury, we also utilized the apical resection (AR) strategy to induce neonatal heart injury. In line with the cryoinjury model, we found that Gata4 was essential for heart regeneration after AR. Although the neonatal mouse heart regenerated from injury in both models, the response and the degree of regeneration appeared to vary. We found that some fibrotic tissues remained after recovery from cryoinjury, whereas there was hardly any detectable scar tissue left in the AR model. Moreover, we also observed cardiomyocyte hypertrophy only in the cryoinjury model but not in AR, suggesting that the neonatal mouse heart could maintain function in response to the respective form of injuries via different mechanisms. For instance, the neonatal heart failed to eliminate excessive scar tissues post cryoinjury owing to some secondary responses such as inflammation; therefore, different injury models might stimulate distinctive endogenous mechanisms for scar tissue regression after injury (Gonzalez-Rosa et al., 2011). Furthermore, the incomplete regeneration following cryoinjury in the neonatal heart could also be attributed to the severity of injury or size of damaged tissue during experiment procedures (Sen and Sadek, 2015), as demonstrated by a previous study in which complete regeneration was observed following mild but not severe cryoinjury (Jesty et al., 2012). Nevertheless, our work suggested that the loss of Gata4 resulted in more severe cardiomyocyte hypertrophy with reduced cardiac function.

The molecular regulation of cardiac regeneration has only recently begun to be extensively studied. In a zebrafish regeneration model, subepicardial ventricular cardiomyocytes trigger expression of gata4 within a week of trauma (Kikuchi et al., 2010), and subsequent functional studies convincingly showed that the activation of gata4 in the cortical myocardial layer is essential for shaping heart morphogenesis during heart regeneration (Gupta et al., 2013). In contrast to the expression of gata4 in the cortical myocardial layer of zebrafish heart (Gupta and Poss, 2012; Gupta et al., 2013), Gata4 is expressed throughout the ventricular myocardium of the neonatal mouse heart during homeostasis and after injury. In addition, the adult zebrafish heart retains the capability to regenerate, whereas the adult mammalian heart does not regenerate sufficiently to recover from significant injury such as myocardial infarction. Therefore, the neonatal mouse heart injury models induced by cryoinjury or apex resection represent interesting models for studying mammalian cardiac regeneration (Porrello et al., 2011; Sadek et al., 2014). Recent studies showed that the homeodomain transcription factor Meis1 is crucial for regulation of neonatal myocyte proliferation and neonatal heart regeneration (Mahmoud et al., 2013). In addition to transcriptional regulation, the miR-15 family of microRNAs also modulates neonatal heart regeneration through inhibition of postnatal cardiomyocyte proliferation (Porrello et al., 2013). These advances in the molecular regulation of neonatal mammalian heart regeneration, together with valuable information from adult zebrafish regeneration models, provide important information for the development of novel therapies to drive cardiac repair after injury.

In the cardiomyocyte-specific Gata4-ablated hearts, we also found that the number of capillaries and small vessels was reduced compared with that of the control hearts. FGF and VEGF family members are both crucial angiogenic factors that simulate the survival, proliferation and differentiation of endothelial cells (Cross and Claesson-Welsh, 2001). Our qRT-PCR data revealed that the gene expression levels of FGF family members, such as Fgf2, Fgf3, Fgf8, Fgf9 and Fgf16 were significantly reduced, but the gene expression levels of the VEGF family (Vegfa, Vegfb and Vegfc) were not significantly changed after Gata4 knockout. These data suggested that GATA4 regulates vessel formation via the FGF pathway at least around the transient regenerative window after birth. One caveat in the interpretation of our data is that there is a change of expression pattern in pro- or anti-angiogenic genes, e.g. Fgf16 and Ctgf, in the sham-operated hearts compared with those of the littermate controls. Our findings suggested that Gata4 regulates a subset of pro-angiogenic genes at the baseline, in addition to its important role as a stress-responsive regulator as revealed by the gain-of-function model (Heineke et al., 2007). However, we did not detect a significant reduction of capillaries or arteries in the mutant heart during homeostasis, nor did we observe any reduction of cardiomyocyte proliferation or cardiac function defect during the stages we examined.

FGF16, in particular, is released by cardiomyocytes from the perinatal period and its expression level is increased in adulthood (Lavine et al., 2005; Hotta et al., 2008; Lu et al., 2008). After FGF16 deletion, the heart also functioned normally but the heart weight and cardiomyocyte number were found to decrease. In addition, the expression level of NPPB (also known as BNP) was significantly downregulated, suggesting that FGF16 might have a specific role in postnatal heart maintenance, especially under pathological conditions (Hotta et al., 2008). Importantly, previous reports showed that FGF2 is expressed in non-cardiomyocytes and has a positive effect on cardiomyocyte replication, angiogenesis, collagen synthesis, and infarct repair or hypertrophy (Kaye et al., 1996; Virag et al., 2007; House et al., 2010). It has been reported that FGF16 blocks FGF2 function by competing with FGF2 for binding to its receptor FGFR1c (FGF receptor 1c) (Lu et al., 2008), and in doing so might, therefore, inhibit cardiac hypertrophy and fibrosis (Matsumoto et al., 2013; Santiago et al., 2014). Altogether, previous findings indicate that FGF16 in the extracellular matrix might promote the inhibition of fibrosis and hypertrophy, or provide a niche for cardiac progenitors to drive regeneration. In this study, we revealed that Fgf16 is a direct downstream target of GATA4 and is downregulated in the cardiomyocyte-specific Gata4 knockout mouse hearts. In light of the fact that GATA4 is also a regulator of heart hypertrophy (Bisping et al., 2006), we conclude that GATA4 regulates FGF16 expression in cardiomyocytes to control heart hypertrophy after birth. Notably, upregulation of FGF16 partially restored heart function after injury in the absence of GATA4 expression in cardiomyocytes. Overexpression of FGF16 led to downregulation of genes involved in hypertrophy, apoptosis and fibrosis and upregulation of genes associated with cell proliferation in mutant hearts, suggesting that FGF16 is important in preserving heart function. Taken together, our functional analysis demonstrates that the paracrine factor FGF16 can rescue GATA4-mediated deficiency in cardiomyocytes, at least in part, by promoting cardiomyocyte replication and inhibiting cardiac hypertrophy. As FGF16 is a secreted protein, identification of its function during neonatal heart regeneration could shed light on possible avenues for the development of candidate drugs for protecting the heart from injury.

All animals were used according to guidelines from the Institutional Animal Care and Use Committee of the Institute for Nutritional Sciences, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences. We crossed the iTnt-Cre mice, in which the troponin T promoter drives the reverse tet activator protein (rtTA) and the tet activator protein-dependent (TRE) promoter drives the Cre recombinase (iTNT-Cre) (Wu et al., 2010), with Gata4^fl/fl mice (Watt et al., 2004), in which the 3/4/5 exon of Gata4 is flanked by LoxP sites, to generate rtTA-Cre;Gata4^fl/fl mice with specific deletion of the Gata4 3/4/5 exon after doxycycline (Dox) administration. All procedures were performed in blind for treatments and analyses. Pregnant mice at E16.5 were injected intragastrically once daily for 3 days with Dox (10 mg/ml; 10 μl per gram of body weight; Sigma, D9891-25G) to activate Cre.

Neonatal mice at P1 were subjected to anesthesia by freezing for ∼3-5 min, and then were placed on the frozen operation table once breathing was steady. Alcohol (70% ethanol) was used to disinfect the surgical site and the mouse limbs were fixed with forceps. An incision (∼1 cm) was made along the sternum and vertical of the chest muscles under a stereo microscope. Two or three intercostal incisions were made in the left sternum chest to separate the pericardium and expose the left ventricle of the heart. Blunt port copper wire (1 mm thickness) was prepared and frozen to −70°C and was then put on the left ventricle in order to induce frostbite; this was maintained for ∼7-8 s until the left ventricle appeared white. After injury, the bubbles and blood in the chest were squeezed out and the chest was sewn up with 8-0 sutures and the skin was closed with 11-0 sutures. After the operation, the neonatal mice were placed under a 37°C lamp to keep warm. They were then placed back with the breeding mice as soon as they woke up and the color of the skin had returned to normal. In the sham-operated group, we performed the same experimental procedures as above except that we replaced liquid nitrogen with PBS at room temperature.

The neonatal mouse heart AR injury model was performed as described previously (Porrello et al., 2011; Han et al., 2015). Briefly, neonatal mice at P1 were subjected to anesthesia by freezing for 3-5 min, and the mice were placed on the frozen operation table once breathing was steady. Alcohol (70% ethanol) was used to disinfect the surgical site and the mouse limbs were fixed with forceps. An incision (∼1 cm) was made and the sternum vertical of the chest muscles was separated under the stereo microscope. Two or three intercostal incisions were made in the left sternum chest to separate the pericardium and expose the apex. Curved forceps were extended into the intrathoracic to pull the heart out and the apex was truncated with microsurgical scissors. The bubbles and blood in the chest were squeezed out and the chest was sewn up with 8-0 sutures, and the skin was sewn up with 11-0 sutures. After the operation, the neonatal mice were placed under a 37°C lamp to keep warm. Then mice were placed back with the breeding mice as soon as they woke up and the skin had returned to normal color. In the sham-operated control, we performed the same experimental procedures as above except that we did not truncate the heart apex.

The left ventricle systolic function was measured 2 months following cryoinjury with echocardiography in mice via a digital ultrasound system (Vevo2100 Imaging System, VisualSonics). Conventional measurements of the left ventricle (LV) included: end-diastolic diameter (LVEDD), end-systolic diameter (LVESD), intraventricular septal thickness (IVST), posterior wall thickness (LVPWT), ejection fraction (LVEF) and fractional shortening (LVFS).

Heart tissues were collected and dissected at the indicated stages. Tissues were homogenized in Tris-SDS lysis buffer (50 mM Tris-HCl pH 8.0 and 1% SDS) and incubated at room temperature for 20 min, followed by centrifugation at maximum speed (22,000 g) to get the protein supernatant. All protein samples were mixed 1:4 with 4× loading buffer (10% SDS, 1.5 M dithiothreitol and 0.3 M Tris-HCl pH 6.8) and boiled for 5 min. Proteins were resolved by 10% SDS PAGE and transferred onto a polyvinylidene fluoride (PVDF) membrane (Immobilon, Millipore) using a Mini Trans Blot system (Bio-Rad). The membranes were then blocked in TBS-T (10 mM Tris-HCl pH 8.0, 150 mM NaCl and 0.5% vol/vol Tween-20) containing 5% skim milk powder at room temperature for 1 h. After that, membranes were incubated with the indicated primary antibody at 4°C overnight. The next day, membranes were washed and then incubated with HRP-conjugated secondary antibody at room temperature for 1 h. Signals were detected by enhanced chemiluminescence (Pierce) according to the manufacturer's instructions. Antibodies used for western blot in this study included: anti-GATA4 (Santa Cruz, sc-1237; 1:1000), anti-FGF16 (Santa Cruz, sc-376214; 1:1000), anti-β-actin (Cell Signaling, 4970L; 1:1000), HRP-anti-rabbit IgG (Jackson ImmunoResearch, 711-035-152; 1:5000), HRP-anti-mouse IgG (Jackson ImmunoResearch, 115-035-174; 1:5000) and HRP-anti-goat IgG (Jackson ImmunoResearch, 705-035-147; 1:5000).

Heart tissues were collected and dissected at the indicated stages. RNA was extracted with Trizol according to the manufacturer's instructions (Invitrogen) and then converted to cDNA using the PrimeScript RT Reagent Kit with gDNA Eraser (Takara, RR047A). For qPCR, the SYBR Green qPCR master mix (Applied Biosystems) was used and cDNA was amplified on a 7500 Real-Time PCR System (Applied Biosystems). A list of PCR primers is included in Table S2.

Murine hearts were collected at the indicated stages and fixed in 4% paraformaldehyde at 4°C for 30 min. The heart tissues were dehydrated in 30% sucrose at 4°C overnight and embedded in OCT at −80°C for 30 min. Frozen sections were prepared at 10 μm thickness and collected on slides. Sirius Red/Fast Green staining was performed to determine collagen deposition as follows. Cryosections/slides were washed in PBS and fixed in 4% paraformaldehyde for 10 min. Slides were then incubated in Bouins' solution (5% acetic acid, 9% formaldehyde and 0.9% picric acid) at room temperature overnight. The next day, after washing, slides were incubated in 0.1% Fast Green (Fisher, F-99) for 3 min, then in 0.1% Sirius Red (Direct red 80, Sigma, 0-03035) for 2 min. After three washes, slides were dehydrated with ethanol and xylene based on standard procedures. Hematoxylin & Eosin staining was also performed according to standard procedures as previously described (He et al., 2014). Images were acquired on either a Leica M165 FC stereo microscope or an Olympus BX53 microscope. Images were analyzed and quantified based on this formula: scar perimeter/total perimeter.

For EdU labeling experiments, the EdU solution dissolved in sterile 0.9% saline and was injected subcutaneously at 10 μg/g in neonates on days 1 and 7 after surgery. Hearts were collected and sectioned 7 days after surgery. EdU staining was then performed based on the Life Technologies Click-iT EdU Alexa 488 Imaging Kit protocol (Life Technologies, C10337).

Cryosections/slides were washed in PBS and fixed in 4% paraformaldehyde for 10 min. Slides were blocked in PBS containing 0.1% Triton X-100 and 5% normal donkey serum (PBSST) at room temperature for 1 h. After primary antibody incubation at 4°C overnight, signals were developed with Alexa Fluor secondary antibodies at room temperature for 30 min. Before mounting, tissues were counterstained with DAPI. Slides were examined by fluorescence microscopy (Olympus DP72) or laser confocal microscopy (510 META, Carl Zeiss), as indicated. Antibodies used in immunofluorescence for this study included: anti-GATA4 (Santa Cruz, sc-1237; 1:100), anti-α-Actinin (Sigma, A7811; 1:1000), anti-phospho-histone H3 (Upstate, 06-570; 1:500), Ki67 (Lab Vision, RM-9106-F1; 1:100), EdU (Life Technologies, C10337, CLICK-IT EDU Kit), anti-AURKB (Millipore, 04-1036; 1:100), WGA-Alexa555 (Invitrogen, W32464; 1:100), anti-smooth muscle actin-FITC (Sigma, F3777; 1:100), anti-PECAM (BD Biosciences, 553370; 1:100), Alexa Fluor 555-conjugated donkey anti-mouse IgG (Invitrogen, A31570; 1:500), Alexa Fluor 488-conjugated donkey anti-goat IgG (Invitrogen, A11055; 1:500), Alexa Fluor 555-conjugated donkey anti-rabbit IgG (Invitrogen, A31572; 1:500) and Alexa Fluor 594-conjugated donkey anti-rat IgG (Invitrogen, A21209; 1:500). For weak signals, we used HRP- or biotin-conjugated secondary antibodies and a tyramide signal amplification kit (PerkinElmer). Images were acquired on an Olympus BX53 microscope. Quantification of all experiments was performed by an observer blinded to the experimental design.

ChIP-qPCR assays were performed using the chromatin immunoprecipitation assay kit (Millipore, #17-295) following the manufacturer's instructions. Briefly, a 15-cm plate of C2C12 cells expressing GATA4 or mouse heart pieces were treated with 1% formaldehyde at room temperature for 15 min to crosslink protein and DNA complexes. After three washes with PBS, tissues were homogenized and cells were then lysed and sonicated to shear the DNA into 200-600 bp fragments. The supernatant was incubated with 4 μg of anti-GATA4 antibody (Santa Cruz, sc-1237) or 4 μg anti-goat immunoglobulin G (IgG) (Santa Cruz, sc-2020) overnight at 4°C. The DNA fragments were then precipitated with beads for 1 h and were recovered following the manufacturer's protocol. Primers used in ChIP-qPCR are listed in Table S2. For Cdk4, regions of GATA4-binding sites were amplified as previously described (Rojas et al., 2008).

Neonatal mouse cardiomyocytes were collected and isolated using the Neonatal Rat/Mouse Cardiomyocyte Isolation Kit (Cellutron, nc-6031) following the manufacturer's instructions. Briefly, P1 ventricles of C57Bl/6 mice were minced and dissociated by several trypsinization steps at 37°C. Non-cardiomyocytes were separated from cardiomyocytes by the method of differential plating. Cardiomyocytes were then cultured in 24-well plates for 24 h before transfection.

Fragments were amplified by each primer pair listed in Table S2. Each fragment was cloned into the pGL3-basic reporter vector. Neonatal mouse cardiomyocytes (NMVMs) were cultured in 24-well plates and were transfected with 500 ng/well of the indicated plasmid and 50 ng/well pRC-RSV-βGal (internal control vector) via 2 μl Lipofectamine 3000 (Invitrogen) for the luciferase activity assay. Luciferase activity was measured 36 h after transfection using the Dual-Luciferase Reporter Assay System (Promega, E1910) following the manufacturer's instructions. β-Gal activity was detected using a SpectraMaxPlus plate reader at A420 nm based on the protocol of the β-gal assay kit (Beyotime, RG0036).

Luciferase, GFP-Cre and mouse Fgf16 were separately cloned into the ITR-containing AAV9 plasmid (Penn Vector Core P1967), which was driven by the cardiac TNT promoter, to yield pAAV.cTnT::Luciferase, pAAV.cTnT::GFP-Cre, pAAV.cTnT::Fgf16, respectively. AAV9 were packaged in 293T cells with AAV9:Rep/Cap and pAd:deltaF6 (pHelper) as described previously (Grieger et al., 2006; Lin et al., 2014). AAV9 were then purified and concentrated by gradient centrifugation. Titer was determined by quantitative PCR. AAV9 were dissolved in PBS and injected subcutaneously into neonatal mice after cryoinjury and 6 days post injury.

Data for two groups were analyzed using an unpaired Student's t-test, whereas comparison between more than two groups was performed using an ANOVA followed by Tukey's multiple comparison test. Significance was accepted when P<0.05. All data are presented as mean±s.e.m.

We thank H. Zeng for the Rosa26-RFP mouse line and members of the Zhou laboratory for offering comments in preparation of this manuscript.