Induction of cardiomyocytes by GATA4 in Xenopus ectodermal explants

In all vertebrates, heart formation begins very early in embryogenesis and apparently proceeds in a similar manner(Harvey, 2002). From a combination of embryological experiments and lineage labelling studies, we know that the embryonic heart is derived primarily from symmetrical, bilateral patches of mesodermal tissue that form during gastrulation. These converge at the embryo midline, fusing to form a linear contractile tube. While the final structure of the mature heart varies widely between different vertebrate classes, in each case transformation of the linear tube into a multi-chambered organ involves complex morphogenetic changes that are coordinated with regional differentiation of distinct cardiac cell types(Harvey, 2002).

Identifying the molecular events that specify cardiac fate and regulate orderly differentiation of cardiac tissues is essential for our understanding of congenital heart disorders, and may allow the development of novel therapies for cardiac disease. Although considerable progress has been made in identifying genes expressed in different regions of the developing heart and the mechanisms that regulate their transcription(Bruneau, 2002), our understanding of the key events that direct embryonic blastomeres to a cardiac fate and ultimately trigger the onset of terminal differentiation remains fragmentary. One reason for this is the lack of molecular markers uniquely associated with cardiac fate that would allow cardiac progenitors to be traced from the time of their specification until they initiate terminal differentiation. Another is the difficulty inherent in attempting to identify the cell-to-cell interactions necessary for the formation of cardiac progenitors in the complex and rapidly changing environment of the gastrulating embryo. Identifying the signalling pathways that mediate such interactions is further complicated by the findings that such pathways frequently have multiple roles in the early embryo that are difficult to distinguish.

Many studies have established the importance of endoderm in heart formation, establishing an important role for this tissue in facilitating normal cardiac morphogenesis. Several have also provided evidence for a distinct and much earlier role for endodermal tissue in the initial specification of cardiac progenitors. In amphibian and chick embryos,efficient formation of cardiac progenitors depends on an interaction between prospective cardiac mesoderm and the underlying anterior endoderm with which it is intimately associated (Lough and Sugi, 2000; Nascone and Mercola, 1995). Whether such interactions are essential for specifying cardiac cell fate in all vertebrates is less clear. For example, in the zebrafish, the casanova (cas) mutation inactivates a Sox-related transcription factor, resulting in a defect in endoderm formation (Kikuchi et al.,2001). Although heart morphogenesis is severely disrupted in these embryos, cardiac tissue is still formed, indicating that cardiac specification occurs in the absence of cas-dependent endoderm(Alexander et al., 1999).

Molecular mediators of the cardiogenic inducing signal(s) have not yet been identified, but some candidate molecules have been proposed. In the chick embryo, bone morphogenetic proteins (BMPs) have been shown to be necessary for cardiogenic activity of anterior endoderm, and inhibitors of BMP activity suppress cardiac differentiation (Schlange et al., 2000; Schultheiss et al., 1997). In other vertebrates, evidence of a similar role for BMPs is less clear cut. In the mouse, homozygous null mutants for components of the BMP signalling pathway have not proved to be very informative because the phenotypes are either lethal prior to the onset of cardiogenesis or result in disruption of heart morphogenesis rather than absence of cardiac tissue(reviewed by Schneider et al.,2003). In the zebrafish, Bmp2 (Bmp2a — Zebrafish Information Network) mutant embryos show profound defects in many tissues, including the myocardium (Mullins et al.,1996; Reiter et al.,2001). The cardiac phenotype may indicate a specific role for Bmp2 in the specification of heart progenitors, but it may also be a consequence of earlier general dorsalisation of the entire embryo(Mullins et al., 1996). In Xenopus embryos, interfering with BMP signalling during gastrulation also produces a dorsalised phenotype (Dale and Jones, 1999). Restricting inhibition of the BMP pathway to later stages of development, or to lineages that include cardiac progenitors,has no effect on the heart field formation but results in reduction of differentiated cardiac muscle and disruption of heart morphogenesis(Breckenridge et al., 2001; Shi et al., 2000; Walters et al., 2001).

If BMP signalling is important for cardiac specification, there is also evidence that other signalling pathways are necessary to restrict the location of cardiac progenitors. In explants from chick and frog embryos, secreted antagonists of the WNT/β-catenin pathway, DKK1 and Crescent, promote cardiogenesis in posterior or ventral mesoderm, respectively(Marvin et al., 2001; Schneider and Mercola, 2001). This has led to the proposal that during normal development, expression of these antagonists in the organizer creates a zone of low WNT signalling in the adjacent pre-cardiac mesoderm, thereby delineating bilateral regions of anterior mesoderm capable of responding to an endoderm-derived cardiogenic signal (Schneider and Mercola,2001).

Cardiac progenitors express the homeobox gene Nkx2.5 soon after they have been specified. Mouse mutants that are homozygous for a null mutation of Nkx2.5 show severe and early disruption in heart tube morphogenesis (Biben and Harvey,1997; Tanaka et al.,1999), and defects in heart valve and septal development have also been associated with Nkx2.5 mutations in humans(Benson et al., 1999; Schott et al., 1998). These results indicate roles for Nkx2.5 both early and late in vertebrate cardiogenesis but further definition of these has remained elusive. The existence of multiple related, and perhaps functionally redundant, NKX2 family members in vertebrates has complicated the interpretation of mutant phenotypes. Additionally, Nkx2.5 is expressed in other tissues of the early embryo (most notably in the anterior pharyngeal endoderm), and within the mesoderm it is unclear whether its expression identifies definitive cardiac progenitors or a broader domain of cells that can be diverted to a cardiac fate. Loss-of-function studies in Xenopus(Evans, 1999) and analysis of heart mutants in zebrafish confirms the importance of Nkx2.5expression for subsequent cardiac differentiation, and suggest that Nkx2.5 lies downstream of a BMP-mediated cardiogenic signal(Reiter et al., 2001). However, in the absence of any specific marker of cardiac progenitor cells prior to the onset of terminal differentiation, it has proved difficult to identify the precise role of NK2 family members in the acquisition of cardiac cell fate.

Nkx2.5 does not appear to be a `master regulator' of cardiac fate in the manner that members of the MYOD family drive skeletal muscle differentiation; indeed, no equivalent cardiomyogenic regulators have yet been identified. Instead, analysis of cardiac muscle-specific transcription has identified a number of transcription factors that, together and in multiple combinations, regulate cardiac transcription. These include proteins of the GATA, NK2 and MADS, and myocardin transcription factor families(Bruneau, 2002; Cripps and Olson, 2002), many of which are expressed in a broad range of tissues in the early embryo. Cardiac-specific transcription most likely results from the formation of particular multi-protein complexes by these factors within the differentiating cardiomyocyte (Bruneau, 2002; Cripps and Olson, 2002). Do these factors play any role establishing the population of cardiac progenitors prior to terminal differentiation? Once again, an unambiguous answer has been obscured by the expression of multiple members of each transcription factor family and the potential for functional redundancy between them.

Because of these difficulties, it would be advantageous to examine the ability of these factors to trigger cardiac differentiation in embryonic tissue that does not normally contribute to heart formation in the embryo. Furthermore, if such an assay can be conducted in cultured explants rather than the embryo, the contribution of the many cell-to-cell interactions and cell signalling events that occur during the critical period of gastrulation can be minimised. Here we show that such an assay system is provided by ectopic expression of GATA4 in animal pole explants from embryos of Xenopus laevis. Under these conditions, such presumptive ectodermal tissue reliably forms a restricted range of mesendodermal tissue derivatives,including cardiac muscle. We have used this assay to examine the competence of explants to undergo cardiac induction, the possible role of endodermal tissue within the explant and the role of signalling pathways proposed to mediate cardiac specification.

Xenopus embryos were obtained, injected and cultured in Normal Amphibian Medium (NAM) (Sive,2000). Animal pole explants were excised at stage 8.5-9 and cultured in 0.75×NAM until the indicated stage. Dissociation of animal poles was carried out as previously described(Sive, 2000) (see Fig. 6). Dexamethasone (2μM; Sigma), dissolved in ethanol, was added at the stages indicated. Cardiac actin-GFP transgenic frog lines have been described (Latinić et al., 2002).

Synthetic capped RNA was synthesised as described(Sive, 2000). The amounts injected were: 0.7-0.9 ng Gata1, Gata4, Gata5, Gata6(Jiang and Evans, 1996; Zhang and Evans, 1994) and Gata4-GR (R. Patient and B. Afouda, unpublished), unless indicated otherwise; 1 ng Mixer (Henry and Melton,1998), Cerberus(Bouwmeester et al., 1996), Dkk1 (Glinka et al.,1998), Dsh δPDZ(Sokol, 1996), DEP+(Tada and Smith, 2000), Chordin (Sasai et al.,1994); 0.2 ng Sox17βEnR(Hudson et al., 1997); 0.4 ng Bmp4 (Dale et al.,1992); 0.1 ng CSKAXWnt8 DNA(Smith and Harland, 1991). The activity of all constructs used in epistasis experiments was confirmed by:observing the phenotype of injected embryos [SOX17βEnR (loss of endoderm)and CSKA-XWNT8 (posteriorisation)]; assaying for cement gland induction in animal caps (DKK1, Cerberus and Chordin); and blocking the activin-induced elongation of animal caps (dominant-negative Dishevelled constructs; data not shown). 10 pg each of TOPFLASH (van de Wetering et al., 1997) and RL-TK (Promega) reporter plasmids were injected, and luciferase assays were performed using the Dual Luciferase Assay kit (Promega). Progeny of injected blastomeres were visualised using rhodamine-dextran or by biotinylated-dextran lineage markers (Molecular Probes).

RNA was isolated as described(Chomczynski and Sacchi, 1987). 15-20 animal caps were used per sample. The templates for riboprobes were: MHCα (Logan and Mohun,1993), MLC2 (Chambers et al., 1994), cTnI(Drysdale et al., 1994), IFABP (Shi and Hayes,1994), Sox17α(Hudson et al., 1997), MyoD (Hopwood et al.,1989), MLC1 (Theze et al., 1995), Eomesodermin(Ryan et al., 1996), globin (Patient et al.,1982), Nkx2.3 (Evans et al., 1995), Nkx2.5(Tonissen et al., 1994), For1 (Seo et al.,2002), EF1α, amylase and insulin(Horb and Slack, 2002). For the LFABP probe, we used the 5′ end of cDNA [derived by 5′ RACE using partial cDNA information; see Henry and Melton(Henry and Melton, 1998)]. The Gata4 probe was derived from 3′UTR to distinguish it from the injected transcript. RNase protection assays were performed using a rapid hybridisation method (Mironov et al.,1995). Whole-mount in situ hybridisation(Sive, 2000) using digoxygenin-labelled XMLC2 probe has previously been described(Chambers et al., 1994). Injected biotinylated dextran was revealed with ExtrAvidin-alkaline phosphatase (Sigma) using magenta-phos(Sive, 2000) as a substrate.

Using Xenopus embryo ectodermal explants, we examined the cardiac-inducing capacity of transcription factors implicated in heart development (GATA4, GATA5 and GATA6, NKX and MADS families)(Bruneau, 2002). We found that GATA4 is sufficient to induce expression of three cardiac muscle-specific differentiation markers: Myosin Light Chain 2 [MLC2(Chambers et al., 1994)],Myosin Heavy Chain α [MHCα(Logan and Mohun, 1993)] and cardiac Troponin I [cTNI (Drysdale et al.,1994)] (Fig. 1A). These markers, which are entirely restricted to the myocardium, appeared in explants at the same time as in sibling embryos, suggesting that correct activation of the cardiac differentiation programme occurred(Fig. 6). In some cases,expression of GATA4 in animal pole explants caused formation of spontaneously beating tissue, confirming that functional cardiomyocyte differentiation had occurred (see Movie 1). Induction of beating cardiac tissue has previously been achieved in Xenopus animal pole explants by exposure to activin(Logan and Mohun, 1993) but,in addition to cardiac muscle, activin treatment induces skeletal muscle along with a wide range of other mesendodermal cell types(Green et al., 1992; Gurdon et al., 1996). By contrast, no skeletal muscle was formed in GATA4-expressing explants, as judged by the absence of the myogenic factor XMYOD(Fig. 1B) or the skeletal muscle differentiation marker MLC1 (Fig. 1C). This result suggests that GATA4 has a more restricted effect on explant cell fate than activin. Consistent with this, Gata4-injected explants expressed markers of dorsoanterior mesoderm(Eomesodermin; Fig. 1D) and endoderm [SOX17α (Weber et al.,2000)] and XHEX (data not shown), but not the early marker of posterior mesoderm, XBRA (Weber et al.,2000).

We tested the specificity of GATA4 in inducing cardiac tissue by comparing it with other members of the GATA family. GATA4, GATA5 and GATA6 have all previously been implicated in the development of embryonic heart, lung and gut tissue (Patient and McGhee,2002), but it is not known whether this includes a role in the initial specification of cardiac cell fate. We found that, in our assay, GATA5 has similar cardiogenic activity to that of GATA4, while GATA6 was considerably less active (Fig. 1E). Interestingly, GATA6 was a more efficient inducer of liver tissue than GATA4 or GATA5, as judged using the liver-specific marker LFABP[Liver Fatty Acid Binding Protein (Henry and Melton, 1998); Fig. 1E]. We also tested the activity of GATA1, which is involved in regulating haematopoiesis (Huber and Zon,1998). As previously reported, GATA1 induced the formation of blood in animal cap explants (Fig. 1H) (Mead et al.,2001), but was unable to induce any cardiac markers(Fig. 1G). Interestingly, GATA4 also induced the formation of blood marker globin(Fig. 1H).

In normal development, induction of cardiac mesoderm is thought to occur rapidly during gastrulation of the embryo(Harvey, 2002). We employed a dexamethasone-inducible version of Gata4 to investigate whether the window for triggering subsequent cardiac differentiation in explants is similarly restricted. By adding dexamethasone at progressively later stages of development, we found that cardiac tissue was still formed even when Gata4 activation was delayed for several hours after explant isolation, until sibling embryos had reached neurula stage(Fig. 2A). In fact, a small amount of cardiac tissue was ultimately formed even when dexamethasone treatment was delayed until the equivalent of tailbud stages (stage 20-25; Fig. 2A and data not shown). As the competence of animal pole cells to respond to FGF and activin signaling is lost by the late gastrula stage (Gillespie et al., 1989; Grimm and Gurdon, 2002), this result suggests that GATA4 does not require these functional signalling pathways for induction of cardiac tissue. Importantly, explants that are converted to a cardiac fate by such a delayed activation of GATA4 have already undergone epidermal differentiation(Jonas et al., 1985). GATA4 presumably triggers cardiogenesis in these explants either through the transdifferentiation of epidermal tissue, or by acting on a remaining population of pluripotent stem-like cells.

In normal amphibian development, signals from endoderm are believed to induce cardiac fate in adjacent mesoderm. Endoderm induction in explants is not by itself sufficient to trigger cardiac muscle formation, as ectopic expression of the endoderm inducer Mixer(Henry and Melton, 1998)resulted in the expression of gut endoderm markers but no cardiac markers were activated (Fig. 1F). Ectopic GATA4 also induces endodermal tissue in animal pole explants(Weber et al., 2000)(Fig. 1F), and to test whether the cardiogenic activity of GATA4 requires the presence of such induced endodermal tissue, we injected Gata4 together with an inhibitor of endoderm formation.

Sox17α and β are key regulators of early endoderm development (Hudson et al.,1997), and expression of the dominant negative construct Sox17βEnR (which blocks both Sox17α andβ) inhibits early endoderm development in Xenopus embryos and explants (Hudson et al.,1997). Surprisingly, co-injection of Sox17βEnR with Gata4 led to a substantial increase in cardiac tissue formation (Fig. 3B) and severe reduction in the formation of anterior endodermal tissue, such as liver, as judged by LFABP(Fig. 3F) and For1(Seo et al., 2002) (data not shown). The apparent synergism between SOX17βEnR and GATA4 in promoting cardiac tissue differentiation suggests that GATA4 may induce endoderm at the expense of cardiac tissue. Consistent with this, other treatments that enhanced the extent of cardiac differentiation in GATA4-expressing explants,such as co-expression of DKK1 and Cerberus (see below), also resulted in a profound reduction or complete absence of liver differentiation(Fig. 3F). Markers for other anterior endodermal tissues, such as pancreas [e.g. insulin and amylase transcripts (Horb and Slack,2002)], were also undetectable (data not shown).

Together these results indicate that, at least in GATA4-induced explants,differentiation of endoderm and cardiac muscle tissue are mutually antagonistic, and that stable endoderm formation is not essential for cardiac tissue induction. Further support for this interpretation is provided by results from the explants expressing the hormone-inducible form of GATA4. If GATA4 activity is delayed until the equivalent of early neurula stage (stage 13), expression of the early endodermal marker Sox17α is greatly reduced (Fig. 2B). Under such conditions, no markers for anterior endoderm derivatives, such as liver, can subsequently be detected (Fig. 2C), despite the presence of cardiac muscle(Fig. 2A).

Our finding of an apparent inverse relationship between endodermal and cardiac differentiation in explants has some parallels in studies of whole embryos. Endodermal development requires β-catenin(Lickert et al., 2002),whereas cardiac differentiation is inhibited by the β-catenin pathway(Lickert et al., 2002; Marvin et al., 2001; Schneider and Mercola, 2001). In agreement with this, we found that activation of the WNT pathway by XWNT8 during gastrulation completely blocks cardiac induction by GATA4(Fig. 3A). The secreted antagonist of WNT signaling, DKK1 (Glinka et al., 1998), antagonises the WNT/β-catenin signalling pathway in animal pole explants, reducing transcription from a syntheticβ-catenin/TCF-dependent reporter (Fig. 3E). When co-injected with Gata4, Dkk1 has a synergistic effect on cardiac induction, whether Gata4 was activated immediately upon explant isolation (late blastula stage) or delayed until the equivalent of early neurula stages (Fig. 3B). This suggests that the WNT/β-catenin pathway continuously antagonises GATA4 function during early development of injected explants. Interestingly, Gata4 alone also led to a decrease in WNT/β-catenin pathway activity, and co-injection of Gata4 and Dkk1 resulted in an even greater reduction(Fig. 3E).

Another extracellular antagonist of the WNT/β-catenin pathway is Cerberus (Bouwmeester et al.,1996). Co-expression of this secreted protein with GATA4 in explants greatly stimulated cardiac induction(Fig. 3C) and the subsequent formation of beating tissue (see Movie 2). As in the case of DKK1, at least a part of the mechanism of action of Cerberus may involve a reduction of TCF-dependent transcription(Fig. 3E). However, in addition to binding and antagonising WNT, Cerebrus binds and antagonises the signalling factors BMP and Nodal (Piccolo et al.,1999). Our results therefore suggest that extracellular BMP and Nodal are not required for the cardiogenic activity of GATA4. Consistent with this, GATA4-induced cardiogenesis was unaffected by the presence of Chordin(Piccolo et al., 1996),another secreted antagonist of BMP (Fig. 3D).

Although signalling via the WNT/β-catenin pathway inhibits cardiac differentiation in embryos or explants, it has recently been proposed that non-canonical or planar-cell polarity (PCP) WNT signalling is required for cardiogenesis in vertebrates (Pandur et al., 2002). To test the role of this pathway in GATA4-mediated cardiac induction of explants, we used a dominant-negative mutant of Dishevelled, δPDZ, which has previously been shown to inhibit both canonical and non-canonical WNT pathways(Sokol, 1996; Tada and Smith, 2000). In order to obtain unequivocal results, we used co-injection of Dkk1 or Cerberus with Gata4 because this yields the most robust and reliable induction of cardiac markers (see above). Under such conditions,δPDZ had no effect on cardiac differentiation induced by GATA4(Fig. 4), nor did the PCP pathway-specific dominant negative construct DEP+(Tada and Smith, 2000). The Dishevelled-dependent non-canonical WNT pathway is not therefore required in the pathway by which GATA4 triggers cardiogenesis in animal pole explants.

The homeobox gene Nkx2.5, like its fly orthologue tinman,is believed to regulate early steps in cardiogenesis and is thought, in turn,to be regulated by BMP signalling (Harvey,2002). In amphibian embryos, Nkx2.5 expression is reportedly activated early during gastrulation and although the distribution of transcripts at this stage is poorly characterised, their accumulation is approximately coincident with the specification of cardiac precursors(Evans, 1999). We therefore investigated whether Nkx2.5 is induced during GATA4-mediated cardiogenesis, and whether its expression is regulated by BMP signalling.

We found that Nkx2.5 mRNA can frequently be detected in uninjected animal pole explants cultured until the equivalent of gastrula or neurula stage, although none was detected at later stages(Fig. 2B, Fig. 5 and data not shown). As a consequence, although no dramatic elevation of Nkx2.5 levels was evident early in GATA4-expressing explants, we were unable to establish unambiguously whether GATA4 expression triggered a rapid but subtle elevation of Nkx2.5 transcripts in explants. Elevated levels of Nkx2.5transcripts were detected at the equivalent of tadpole stages(Fig. 5B), but this may simply be a consequence of cardiac tissue differentiation rather than an indication that the effects of GATA4 are mediated via Nkx2.5 expression.

Co-injection of Gata4 and Bmp4 mRNA in animal caps led to a decline rather than an increase in the level of Nkx2.5 mRNA in early explants (Fig. 5A). Interestingly, despite the absence of Nkx2.5 in these explants during early stages, elevated levels were later detected along with cardiac muscle markers, and subsequent cardiogenesis was unaffected(Fig. 5B). These results indicate that detectable levels of Nkx2.5 transcripts are not required at early stages in the development of GATA4-expressing explants for cardiac differentiation to be triggered, nor is the cardiogenic activity of GATA4 affected by elevated BMP4 expression.

Is GATA4 acting only in explant cells that express it, or could it be inducing adjacent non-expressing cells to adopt a cardiac fate? We examined this question by creating a localised signalling source in explants. This was achieved by injecting Gata4, together with a permanent lineage tracer, into a single blastomere of the four- or eight-cell embryo. Explants subsequently removed from such embryos were mosaic for GATA4 expression, as revealed by the distribution of lineage marker. Under these conditions,cardiac tissue was generated by GATA4 in both lineage labelled and unlabelled tissue (Fig. 6), indicating that GATA4-expressing cells had signalled to adjacent ectodermal cells to adopt a cardiomyocyte fate. Some GATA4-expressing cells themselves became cardiac tissue and this cell-autonomous activity of GATA4 was predominant when GATA4 was co-expressed with Cerberus (Fig. 6I-K).

We also tested directly the importance of cell-cell interactions for GATA4-meditaed cardiogenesis, by dispersing explants into single-cell suspensions. This experimental approach is limited because prolonged incubation of single-cell suspensions leads to increasing loss of cell viability. Nevertheless, we found that cell-cell interactions are not required at least until stage 16 (Fig. 6L), after which re-aggregation of cells was necessary for continued culture to be successful.

In this report we show that cardiomyocytes can be induced in explants of presumptive ectodermal tissue from Xenopus embryos by ectopic expression of GATA4. Expression of cardiomyocyte markers is neither precocious nor selective, indicating that their expression reflects genuine induction of cardiac muscle tissue rather than the selective transactivation of myocardial genes containing GATA-factor binding sites. GATA4 apparently triggers the formation of several mesendodermal tissues, including cardiac, but not skeletal, muscle.

Several lines of evidence have already indicated the importance of GATA factors in heart development. Potential GATA binding sites are commonly found in cardiac gene regulatory regions, and, in transfection assays, GATA4, GATA5 and GATA6 have each been found to activate a variety of cardiac-specific promoters (Charron and Nemer,1999; Molkentin,2000; Patient and McGhee,2002). Genetic analyses of the role of GATA4, GATA5 and GATA6 in heart development have been complicated by their requirement in other tissues,as well as by the potential for redundancy between them, which results from overlapping expression patterns and similar activities. In the mouse,homozygous-mutant Gata4 embryos developed cardia bifida(Kuo et al., 1997; Molkentin et al., 1997), Gata5 mutant mice develop normally(Molkentin et al., 2000) and Gata6 mutation results in lethality at the time of implantation(Koutsourakis et al., 1999; Morrisey et al., 1998). The phenotypes of compound and tissue-restricted mutants have not been reported yet. In zebrafish, Gata5 (faust) mutants show a combination of cardiac bifida and a severe reduction in the number of ventricular myocytes that form (Reiter et al.,1999). In gain-of-function studies in P19 embryonic carcinoma cells, GATA4 promotes cardiogenesis(Grepin et al., 1997). P19 cells have a propensity to differentiate into several lineages, including cardiomyocytes, and this complicates interpretation of the mode of GATA4 action.

In addition to a role in cardiac differentiation, there is also evidence that GATA factors are important regulators of endoderm differentiation. GATA4,GATA5 and GATA6 can activate some endodermal promoters in vitro(Gao et al., 1998; Patient and McGhee, 2002) and the cardiac bifida of GATA4 null mice can be rescued by wild-type endodermal cells in chimaeras (Narita et al.,1997). The cardiac phenotype resulting from Gata5mutation in faust zebrafish (which combines cardiac bifida with a loss of myocardial tissue) is likely to be a composite, resulting from the lack of appropriate GATA function in both cardiac and endodermal precursors(Reiter et al., 1999).

A role of GATA4, GATA5 and GATA6 factors in endoderm formation is further supported by findings that they trigger endoderm differentiation in vitro. In Xenopus ectodermal explants, GATA5 induces both early and late endodermal markers (Weber et al.,2000). Our results show that GATA4 and, most strikingly, GATA6 have the same effect. In ES cells, GATA4 and GATA6 promote differentiation of extra-embryonic endoderm (Fujikura et al.,2002). This can occur in the presence of LIF (which blocks spontaneous differentiation of these cells) and does not require the complex cellular environment of embryoid bodies. We have shown that in addition to inducing endoderm, GATA4 and GATA5 induce cardiac tissue, and together these findings lend support to a model in which both mesodermal and endodermal tissues in the early embryo arise from bipotential mesendodermal progenitors(Lickert et al., 2002; Rodaway and Patient, 2001). Interestingly, we find that GATA4-expressing explants show elevated levels of endogenous Gata4 and Gata5 transcripts(Fig. 5C). Similarly,activation of endogenous Gata4 and Gata6 occurs during GATA4-mediated endodermal conversion of ES cells(Fujikura et al., 2002). In each case, the activity of ectopic GATA4 may therefore be reinforced by a positive auto-regulatory loop.

Is the induction of cardiac tissue in GATA4-expressing animal pole explants a secondary consequence of mesendodermal tissue differentiation or does this result indicate a more direct link between GATA function and specification of cardiac fate? In zebrafish embryos, overexpression of Gata5 results in the formation of ectopic cardiac muscle tissue, and this can occur in a cell-autonomous manner (Reiter et al.,1999) consistent with direct effect of the factor. Although we find no evidence for a similar effect of GATA4, GATA5 and GATA6 in frog embryos (Gove et al., 1997)(data not shown), several of our findings nevertheless lend support to the view that the cardiogenic action of GATA factors in explants is relatively direct (as discussed below).

The first of these is the finding that induction of cardiac tissue can occur in the apparent absence of endoderm. Ectopic expression of GATA4 in animal pole explants results in the formation endodermal tissues such as gut(Weber et al., 2000) and liver(Figs 2, 3). Because signalling from endodermal tissue appears to underlie specification of cardiac mesoderm(Lough and Sugi, 2000; Nascone and Mercola, 1995), we might expect that the presence of endoderm in explants would be essential for GATA4-mediated induction of cardiac tissue. However, we have found the opposite to be true; inhibition of endoderm differentiation by a dominant-negative form of the endoderm transcription factor SOX17βactually increases, rather than decreases, the amount of cardiac tissue that is formed. This suggests that GATA4-mediated induction of cardiac progenitors in explants may occur in the absence of an endodermal tissue. Whether cardiac specification under such circumstances occurs without the molecular cues normally provided by endodermal cells (i.e. GATA4 acts downstream of endodermal signal), or whether GATA4 can also mediate these signals independently, is unresolved by our experiments.

Our finding that cardiac differentiation is substantially enhanced as a result of Sox17βEnR expression suggests that the formation of SOX17β-dependent endoderm and cardiac progenitors are mutually antagonistic in GATA4-expressing explants. Alternatively, or in addition, SOX17βEnR may lead to a shift in cell fate from endoderm to mesoderm. Such a shift in fate caused by SOX17βEnR in Xenopus embryos has been reported (Clements and Woodland, 2000). A simple model shown in Fig. 7 is that cardiac mesoderm and endoderm are alternative developmental pathways for common precursors. Sox17βEnR may enhance cardiogenesis directly by opposing the endoderm-inducing activity of GATA4, by blocking maintenance of endoderm, and/or by causing a shift in fate(Fig. 7).

SOX17β and the related SOX17α are thought to be panendodermal,and for this reason Sox17βEnR may be expected to inhibit all endoderm in Xenopus(Hudson et al., 1997). However, recent genetic analyses in the mouse have shown that anterior endoderm is Sox17-independent in that organism(Kanai-Azuma et al., 2002). We cannot therefore exclude the possibility that, in Xenopus explants expressing GATA4 and SOX17βEnR, sufficient endoderm formation still occurs to induce cardiac progenitors. If so, such tissue is likely to be transient because we were unable to detect liver or pancreas endoderm in explants expressing GATA4 and SOX17βEnR.

A second reason for considering that the induction of cardiac tissue by GATA4 might not simply be an indirect consequence of mesendoderm formation comes from our lineage labelling results. The ability of GATA4 to induce cardiac tissue in cells that have not received the injected GATA4 transcript indicates that specification of cardiac fate within explants can occur through cell-to-cell interactions (as might indeed be expected if endodermal signalling was required). Of course, similar cell-to-cell interactions could also result in cell-autonomous induction of cardiac tissue if both signalling and responding cells were derived from GATA4-expressing explant tissue. However, cell-autonomous induction may also indicate a more direct mechanism by which GATA4 triggers cardiac specification. Such an interpretation is consistent with earlier findings that GATA5 can act both cell-autonomously and non-cell-autonomously to induce ectopic myocardium in zebrafish embryos(Reiter et al., 1999). It is also supported by our observation that in explants co-expressing Cerberus and GATA4 virtually all cardiac tissue is derived cell-autonomously(Fig. 6). Cerberus not only enhances the extent of cardiac differentiation in GATA4-expressing explants but also inhibits the formation of endoderm. It should therefore suppress cardiac induction that is mediated by endodermal signalling.

Our investigation of when GATA4-mediated induction can occur using a dexamethasone-inducible form of the transcription factor has yielded intriguing results. Whereas normal cardiogenic signals are restricted to a brief period during gastrulation, in explants we find that cardiac differentiation can be triggered by GATA4 even after the onset of terminal epidermal differentiation. Under such circumstances, GATA4 might be converting epidermal cells towards cardiomyocyte fate, or it could be acting on a hypothetical population of stem-like cells. In either case, the finding is remarkable, and it is clearly important to distinguish between these possibilities in future studies. Equally important is to establish whether such late-onset cardiac induction occurs cell-autonomously and whether it is necessarily accompanied by endoderm differentiation. As yet we have not resolved the former issue but find little SOX17α expression and no liver marker LFABP after delayed activation of the inducible Gata4(Fig. 2).

Studies using explants of non-cardiac mesoderm from chick and frog embryos have indicated a role for antagonists of WNT/β-catenin signalling (DKK1 and Crescent) in the formation of cardiac progenitor tissue(Marvin et al., 2001; Schneider and Mercola, 2001). Our finding that DKK1 has a similar stimulatory effect on cardiac induction indicates that inhibition of cardiogenesis by the WNT/β-catenin pathway occurs at the level of GATA4 or downstream of it. Inhibition may occur directly, by targeting cardiac progenitors, or indirectly via its effect on endoderm formation (Fig. 7). An indirect mode of action is supported by the study of Lickert et al., who have shown that the inhibition of WNT/β-catenin signalling in mouse embryos leads to formation of ectopic cardiac tissue and loss of posterior endoderm(Lickert et al., 2002) (see Fig. 7). Even when GATA4 activation in explants is delayed until control embryos form neurulae, Dkk1 expression still enhances GATA4-mediated cardiogenesis, This result indicates that GATA4 operates during both gastrula and neurula stages in the presence of inhibitory WNT/β-catenin signalling.

In addition to WNT/β-catenin signalling, the non-canonical or planar-cell polarity WNT pathway has also been implicated in regulating cardiogenesis (Pandur et al.,2002). Reduction of WNT11 protein levels in Xenopusembryos leads to defects in heart formation, and ectopic expression of Wnt11 mRNA in posterior mesoderm (ventral marginal zone) explants results in formation of ectopic beating tissue. A similar effect was reported with animal pole explants, but only in the presence of the mesendodermal inducer activin. In our experiments, we find that inhibition of the WNT/PCP pathway at the level of Dishevelled has no effect on cardiomyocyte induction by GATA4. The reason for such a discrepancy between our results and those previously reported is unclear. One possible explanation is that the Dishevelled-dependent WNT/PCP pathway acts upstream of GATA4 in cardiogenesis.

The earliest indicator of formation of cardiac progenitors is the expression of the tinman homologue Nkx2.5 and it has been proposed that this is the result of BMP signalling(Harvey, 2002). In our assay,we have found that the secreted antagonists of BMP signalling, Cerberus and Chordin, do not, in fact, block GATA4-mediated cardiogenesis(Fig. 3). Neither does expression of a dominant-negative BMP receptor (B.V.L. and T.J.M.,unpublished). This could be interpreted to indicate that GATA4-mediated cardiac induction in explants occurs in a fundamentally different manner to cardiogenesis in embryos. Alternatively, it could indicate that BMP signalling lies upstream of GATA4 in the cardiac induction pathway. Consistent with this interpretation, Gata5 has been shown to act downstream of Bmp2 in regulation of Nkx2.5 in the zebrafish embryo (Reiter et al., 2001).

Our efforts to establish the relationship between GATA4-mediated induction and Nkx2.5 expression were hampered by the detection of endogenous transcripts in control animal pole explants. However, no obvious elevation of Nkx2.5 transcripts could be detected as an early response to GATA4 expression, a result that is all the more striking because GATA factors are thought to be key regulators of Nkx2.5 transcription in vivo(Reecy et al., 1999; Searcy et al., 1998; Sparrow et al., 2000). Whether GATA4-mediated cardiac induction bypasses Nkx2.5 or whether subtle changes in Nkx2.5 expression are sufficient for triggering subsequent cardiac differentiation remains to be determined, but it is noteworthy that cardiac induction still occurs in explants expressing BMP4 and GATA4 despite a suppression of Nkx2.5 transcript levels(Fig. 5).

Expression of GATA4 in Xenopus blastula animal pole explants provides a simple and reliable means of inducing cardiomyocytes. We do not yet know whether this reflects a role for GATA4 in normal embryogenesis or whether ectopic expression of this factor mimics or bypasses endogenous inducing signals. In either case, such explants offer a convenient model system to study the molecular signals regulating myocardial differentiation. Mammalian stem cells differentiate to cardiac fate at only a low frequency, either spontaneously or in response to treatments that are neither physiological nor easy to control (Daley, 2002; Schuldiner et al., 2000). Our assay might therefore help to define rational strategies for efficient direction of pluripotent cells to a cardiomyocyte fate.

We thank Roger Patient and Boni Afouda for generously providing GATA4-GR and sharing their results prior to publication. We are grateful to Elena Grigorieva for assistance with histology. The following colleagues kindly provided cDNA clones: R. Patient, T. Evans, J. Smith, M. Tada, E. De Robertis,C. Niehrs, D. Melton, H. Woodland, P. Krieg, P. Thiebaud, S. Sokol, R. Harland, W.-S. Kim, J. Gurdon, M. Horb and J. Slack. We thank S. Mercurio, A. Gould and J. Briscoe for insightful comments on this manuscript. This work was supported by the MRC.