|Home | About | Journals | Submit | Contact Us | Français|
Cardiac malformations due to aberrant development of the atrioventricular (AV) valves are among the most common forms of congenital heart disease. At localized swellings of extracellular matrix known as the endocardial cushions, the endothelial lining of the heart undergoes an epithelial to mesenchymal transition (EMT) to form the mesenchymal progenitors of the AV valves. Further growth and differentiation of these mesenchymal precursors results in the formation of portions of the atrial and ventricular septae, and the generation of thin, pliable valves. Gata4, which encodes a zinc finger transcription factor, is expressed in the endothelium and mesenchyme of the AV valves. Using a Tie2-Cre transgene, we selectively inactivated Gata4 within endothelial-derived cells. Mutant endothelium failed to undergo EMT, resulting in hypocellular cushions. Mutant cushions had decreased levels of Erbb3, an EGF-family receptor essential for EMT in the atrioventricular cushions. In Gata4 mutant embryos, Erbb3 downregulation was associated with impaired activation of Erk, which is also required for EMT. Expression of a Gata4 mutant protein defective in interaction with Friend of Gata (FOG) cofactors rescued the EMT defect, but resulted in a decreased proliferation of mesenchyme and hypoplastic cushions that failed to septate the ventricular inlet. We demonstrate two novel functions of Gata4 in development of the AV valves. First, Gata4 functions as an upstream regulator of an Erbb3-Erk pathway necessary for EMT, and second, Gata4 acts to promote cushion mesenchyme growth and remodeling.
Cardiac malformations attributable to aberrant development of the atrioventricular (AV) valvuloseptal complex are among the most common forms of congenital heart disease (Pierpont et al., 2000). Development of the AV valvuloseptal complex can be considered to occur in several steps (reviewed by Armstrong and Bischoff, 2004). First, endocardial cells at the AV canal (AVC) undergo an epithelial to mesenchymal transition (EMT), forming mesenchymal cells that invade swellings of extracellular matrix to form the endocardial cushions (EC). Next, the cellularized superior and inferior EC grow and fuse, forming the AV valvuloseptal complex that divides the ventricular inflow into a right and a left AV valve. Further elongation and remodeling of the cushions results in formation of the mature valve leaflets.
Transformation of AV endocardium to cushion mesenchyme requires the input of multiple signaling molecules secreted from the adjacent myocardium, including Tgfβ and Egf family members (Camenisch et al., 2000; Nakajima et al., 2000a; Sugi et al., 2004; Rivera-Feliciano and Tabin, 2006). Ablation of Erbb3, which encodes an Egf family receptor, results in failure of endocardial cells of the AV cushions to undergo EMT to form cushion mesenchyme (Camenisch et al., 2000). One mechanism by which Erbb3 promotes EMT is to activate Ras (Camenisch et al., 2000). Decreased Ras signaling results in defective EMT (Camenisch et al., 2000; Lakkis and Epstein, 1998), and, conversely, increased Ras signaling results in increased EMT and hypercellular, enlarged EC (Gitler et al., 2003; Lakkis and Epstein, 1998).
The transcription factor Gata4 is essential for heart formation (Kuo et al., 1997; Molkentin et al., 1997; Watt et al., 2004; Zeisberg et al., 2005). In humans, GATA4 heterozygous mutations have been associated with defects in the muscular septum separating atria or ventricles, and variably associated with valvar pulmonary stenosis (Garg et al., 2003; Hirayama-Yamada et al., 2005; Okubo et al., 2004). In addition to its expression in the myocardium, Gata4 is robustly expressed in the endocardium and the EC (Charron and Nemer, 1999; Heikinheimo et al., 1994). This expression pattern, and the presence of EC defects in mouse embryos homozygous for two different hypomorphic Gata4 alleles (Crispino et al., 2001; Pu et al., 2004), suggested that Gata4 might be an important regulator of EC development.
To further investigate the role of Gata4 in EC development, we specifically inactivated Gata4 in endothelium and endothelium-derived cushion mesenchyme. We show that Gata4 expression in endothelium-derived cells is required at two stages of AV valve formation, illustrating novel cell-autonomous roles for Gata4 in the endocardium. First, Gata4 is required to promote EMT of endocardial cells to generate AV cushion mesenchyme. Second, Gata4 activity in endocardial-derived cells is required later during AV valve maturation for growth and remodeling of the AV cushions to septate the ventricular inlet.
Gata4H and Gata4flox alleles (Fig. 1A) have been described previously (Pu et al., 2004; Zeisberg et al., 2005). Gata4H expresses reduced amounts of protein compared with the wild-type allele (Pu et al., 2004). Gata4wt/Ki and R26RstoplacZ mice (Crispino et al., 2001; Mao et al., 1999) were obtained from Stuart Orkin (Harvard Medical School). Transgenic mice expressing Cre from Tie2 (also known as Tek) regulatory elements (T2Cre) were obtained from M. Yanagisawa (Kisanuki et al., 2001). Wnt1-Cre mice (Danielian et al., 1998) were obtained from Jackson Laboratories. All mice were maintained in a mixed C57BL6/129 genetic background. Gata4wt/flox; T2Cre+ mice were crossed with Gata4flox/flox mice to yield Gata4flox/flox; T2Cre+ (Gata4T2del) mice. All animal care and procedures were performed under protocols approved by the Institutional Animal Care and Use Committee.
Embryos were fixed in 4% paraformaldehyde overnight at 4°C, and then paraffin-wax embedded and sectioned at 10 µm. Alcian Blue staining was performed by using 0.15 mg/ml Alcian Blue in 5% acetic acid. BrdU labeling and TUNEL staining were performed as described previously (Zeisberg et al., 2005).
For short-term embryo culture, E9.5 embryos were incubated in M199 supplemented with 1% FBS for 30 minutes at 37°C in 5% CO2. Heregulin (Sigma; 100 ng/ml) or U0126 (Calbiochem; 10 µM) were added as indicated.
AV explant culture was carried out as described (Rivera-Feliciano and Tabin, 2006). Where indicated, U0126 (Calbiochem), vehicle (DMSO) or a growth factor cocktail [Tgfβ2 (Cell Biosciences), 50 ng/ml; Bmp2 (R&D Systems), 200 ng/ml; hyaluronic acid (Sigma), 500 ng/ml; heregulin (Sigma), 50 ng/ml] was added to the media at the start of the explant culture.
Human umbilical vein endothelial cells (HUVEC; passage <6; Cambrex) were cultured in complete endothelial growth media (Cambrex). BT20 human breast cancer cells (ATCC) were transfected using Fugene6 (Roche).
Whole-mount staining for detection of β-galactosidase activity was performed as described (Lobe et al., 1999). Immunostaining was performed using the following antibodies: Nfatc1 (1:200, Santa Cruz), Desmin (1:4, Biomeda), Chd5 (1:20, Santa Cruz), phospho-Erk1/2 (1:200, Cell Signaling), α-SMACy3-conjugated (Clone 1A4, Sigma 1:200) and biotin-conjugated CD31 (Pecam1, 1:100) monoclonal antibody (Clone MEC 13.3, BD Pharmingen). Erbb3 western blotting was performed with antibody C-17 (Santa Cruz, 1:200) and normalized to Gapdh (Research Diagnostics, 1:10,000).
For RNA analysis, four mutant and four control RNA samples were prepared (Pico-Pure RNA Isolation Kit, Arcturus), each consisting of 10 microdissected AVCs. Probe was prepared from 50 ng total RNA using an isothermal amplification protocol (NuGen), and hybridized to Affymetrix Mouse 430 2.0 microarrays. Two control samples were excluded because of excessive noise. We excluded probe sets that may cross-hybridize to unrelated targets (probe name ending with ‘_x_at’) or that received ‘Absent’ calls across all samples. The 27,082 remaining probe sets were ranked by the ‘relative difference’ d-score (Tusher et al., 2001), using the Significance Analysis of Microarray (SAM) software package (http://www-stat.stanford.edu/~tibs/SAM/).
For quantitative RT-PCR, RNA samples were converted to cDNA and amplified by Ovation isothermal amplification. The cDNA was then used for quantitative PCR on an ABI7300 thermal cycler, with Sybr Green or Taqman detection. Primer sequences are provided in Table 1.
In situ hybridization was performed on 10 µm paraffin sections using digoxigenin-labeled or S35-labeled RNA probes as described (Brent et al., 2003; Tanaka et al., 1999). Signal from S35-labeled probes was detected in dark field and pseudocolored red using PhotoShop. Sections were counterstained with DAPI. For in situ hybridization probes refer to Table 1. Results shown are representative of at least two embryos.
The Gata4 expression construct has been described previously (Lee et al., 1998). The Gata4Δex2 expression construct was generated by RT-PCR amplification of the Cre-recombined Gata4 transcript. Gata4DBD-engrailed was constructed by PCR cloning the Gata4 DNA-binding domain upstream of the engrailed repressor domain. The murine Erbb3 promoter and intron 1 enhancer was cloned from a bacterial artificial chromosome by Red/ET recombineering (GeneBridges) into pGL3-BASIC (Promega) or pGL3-promoter (Promega), respectively. Luciferase assays were normalized for transfection efficiency using pRL-null (Promega).
To determine the function of Gata4 within the endocardium and its derivatives, we inactivated a floxed Gata4 allele (Gata4flox; Fig. 1A) (Pu et al., 2004; Zeisberg et al., 2005) by expressing Cre recombinase from a Tie2 promoter (T2Cre) (Kisanuki et al., 2001). Cre-mediated recombination of Gata4flox resulted in excision of a portion of exon 2, including the start codon and 46% of the Gata4 coding region. The recombined allele (Gata4Δex2) expresses a truncated protein containing both zinc fingers and the C-terminal activation domain, but lacking the N-terminal transactivation domains (see Fig. S1 in the supplementary material). The truncated protein failed to activate multiple cardiac and intestinal Gata4-dependent promoters in vitro (Fig. S1 in the supplementary material; T. Bosse and S. Krasinski, personal communication), consistent with previous results (Morrisey et al., 1997). These data, along with the observation that embryos homozygous for this mutation in their germline (Gata4Δex2/Δex2) resemble previously reported mice carrying Gata4 null alleles (Kuo et al., 1997; Molkentin et al., 1997) (data not shown), suggest that the Gata4Δex2 allele behaves as a loss-of-function mutation. However, we cannot exclude the possibility that Gata4Δex2 retains partial function.
In control experiments, we characterized the spatiotemporal pattern of recombination catalyzed by T2Cre using the reporter R26RstoplacZ, which expresses lacZ only after activation by Cre (Mao et al., 1999). In T2Cre+; R26RstoplacZ embryos, T2Cre activated reporter expression in the majority of endocardial cells by E9.5 (Fig. 1B1,B2). lacZ expression was not observed in epicardial or myocardial cells. Between E9.5 and E11.5, endocardial cells at the AVC transform into mesenchymal cells and populate the AV cushions. Consistent with this lineage history, mesenchymal cells of the AV cushions were recombined by Cre recombinase, and consequently expressed the lacZ reporter at E9.5 and E11.5 (Fig. 1B1, arrow; 1B3, asterisks).
The endocardium and EC mesenchyme of the developing heart express high levels of Gata4 (Fig. 1C,D) (Heikinheimo et al., 1994). To specifically inactivate Gata4 in endocardial-derived cells, we generated embryos with the genotype Gata4flox/flox; T2Cre+ (Gata4T2del). We examined expression of Gata4 in these embryos by in situ hybridization using an exon 2-specific probe. In control embryos, Gata4 exon 2 transcripts were present in endocardial, epicardial and myocardial cells (Fig. 1C1,C2). In Gata4T2del embryos, expression from the Gata4 exon 2 probe was unchanged in the epicardium and myocardium (star and yellow arrowheads, Fig. 1C3,C4), but expression in the endocardium was absent (white arrowheads, Fig. 1C3,C4). At E11.5, robust expression was detected in AVC mesenchyme in control embryos, but this tissue was largely deficient in Gata4T2del embryos (asterisks, Fig. 1D; see below). The lack of detectable Gata4 transcripts in mutant endocardium was not due to non-specific transcript degradation, or a general failure of endothelium to subspecialize into endocardium, as mutant endocardium continued to express normal levels of Pecam, an endothelial marker, and Nfatc1, a marker of endothelium subspecialized to line the heart (data not shown).
EC also contribute to formation of the outflow (OT) tract and the OT valves. However, only the most proximal portion of OT cushion mesenchyme derives from endothelial progenitors that were recombined by the T2Cre transgene (yellow arrow, Fig. 1B). The bulk of OT cushion mesenchyme is derived from neural crest, as demonstrated by fate mapping using the neural crest restricted Wnt1Cre transgene (Gitler et al., 2003; Verzi et al., 2005) (see Fig. S2 in the supplementary material). Gata4 is expressed in OT endothelium and both endocardial- and neural crest-derived OT mesenchyme (green arrow, Fig. 1D). Consistent with the lack of T2Cre-mediated recombination in the mid and distal OT mesenchyme (green arrow, Fig. 1B3), Gata4 expression in these regions was not affected in Gata4T2del embryos (green arrow, Fig. 1D).
Out of 34 litters genotyped, no Gata4T2del mice survived to weaning. By E12.5, the prevalence of Gata4T2del embryos was 80% less than expected based on Mendelian ratios (Fig. 2A). At E12.5, surviving Gata4T2del embryos had pericardial effusion and peripheral hemorrhage, which are hallmarks of embryos with heart failure (Fig. 2B,C). The liver was hypoplastic (Fig. 2B; see also Fig. S3 in the supplementary material). Histological examination of the heart showed that these mutant embryos displayed a paucity of mesenchymal cells within the AV cushions (Fig. 2D–G). Additionally, the mutant AV endocardium was multiple cell layers thick at certain foci (arrow, Fig. 2G), whereas the AV endocardium of controls remained as a single cell layer epithelial sheet (Fig. 2E). Although the myocardium appeared normal in the majority of mutant embryos, in 30% of embryos (six out of 17 examined) the compact myocardium was abnormally thin (Fig. S4 in the supplementary material).
The hypocellular AV cushion phenotype was 100% penetrant and was not due to increased apoptosis, as measured by TUNEL staining (data not shown). A small and variable number of mesenchymal cells were observed in mutant AV cushions. To determine the origin of these cells, we fate mapped T2Cre-expressing cells using the R26RstoplacZ reporter in control and Gata4T2del mutant embryos (Fig. 3). We found that these cells were lacZ positive (arrowhead, Fig. 3B), indicating that they are derived from endothelium and not from an alternative tissue compartment, such as epicardium. In situ hybridization demonstrated that the residual mesenchymal cells did not express Gata4 (data not shown). T2Cre-recombined mesenchyme at the proximal tip of the OT cushions (yellow arrows, Fig. 3A,C) was missing in Gata4T2del mutants (yellow arrows, Fig. 3B,D).
The process of EMT can be recapitulated in vitro by culturing explants of the AVC in a three-dimensional collagen gel (Runyan and Markwald, 1983). During EMT, endothelial cells first undergo an activation step during which they lose their cell-cell contacts. The cells then adopt an elongated morphology, upregulate mesenchymal markers such as SMA, and downregulate endothelial markers such as Pecam. These activated cells subsequently invade and migrate through the extracellular matrix to complete the transformation process.
When cultured in a collagen gel, control AVC explants produced a halo of invasive, migrating cells with mesenchymal morphology (Fig. 4A). By contrast, mutant explants failed to generate mesenchymal cells (Fig. 4B,C). Immunostaining with antibodies for SMA and Pecam delineated three classes of endothelial cells at different stages of activation in all explants examined: (1) rounded cells expressing Pecam but not SMA (arrow, Fig. 4D); (2) rare, round transitional cells expressing SMA (white arrowhead, Fig. 4D); and (3) elongated, SMA-expressing cells (yellow arrowhead, Fig. 4D). By contrast, explants from Gata4T2del embryos failed to produce mesenchymal, SMA-expressing cells (Fig. 4E). Mutant endothelium appeared multi-layered in a manner reminiscent of the morphology of sectioned Gata4T2del hearts (star, Fig. 4E; arrow, Fig. 2G). None of the aforementioned signs of endothelial cell activation were detected in explants examined by immunostaining for Pecam and SMA. These data suggest that endothelial recombination of Gata4 impairs the activation of endothelial cells and blocks their transformation to invasive mesenchyme.
Having established an essential role for Gata4 in the transformation of AVC endocardial cells to mesenchymal cells, we sought to elucidate the mechanism by which Gata4 acts in this process. Because adjacent myocardium is known to influence endocardial EMT, we investigated whether the endocardial inactivation of Gata4 blocked EMT in a non-cell-autonomous manner. Myocardial differentiation and specification of atrioventricular canal myocardium remained intact, as indicated by the patterns of expression of sarcomeric myosin, the myocardial specific transcription factor Nkx2–5, the chamber myocardium marker Nppa, and the AVC myocardial marker Tbx2 (Fig. 5; data not shown). Myocardial expression of Bmp2 and Tgfβ2, known paracrine activators of EMT (Nakajima et al., 2000b; Rivera- Feliciano and Tabin, 2006; Sugi et al., 2004), was unchanged in Gata4T2del hearts (data not shown). Furthermore, culture of control AVC explants directly next to Gata4T2del explants did not rescue the EMT defect in mutant explants. Similarly, culture of explants in the presence of a mixture of growth factors known to promote EMT (Tgfβ2, Bmp2, heregulin and hyaluoronic acid) did not rescue the EMT defect in mutant explants, but did robustly stimulate the formation of mesenchymal cells in control explants. These data suggest that the phenotype of Gata4T2del explants was not due to an absence of diffusible factor(s) produced by the myocardium that promote EMT.
We also considered the possibility that Gata4 could regulate EMT through modulation of the extracellular matrix, which is necessary for cushion mesenchyme formation (Camenisch et al., 2000). However, the extracellular matrix of the AV cushions was still present in Gata4T2del hearts, as assessed by binding of the stain Alcian Blue to acidic glycosaminoglycans present in the cushion extracellular matrix (Fig. 5D). Moreover, addition hyaluronic acid directly to explant cultures did not rescue the EMT defect of mutant explants (data not shown).
We next turned our attention to cell-autonomous mechanisms that could account for the loss of EMT in Gata4T2del embryos. Downregulation of Snail as a result of Notch signaling leads to downregulation of VE-cadherin (Cdh5), and is necessary for endocardial EMT (Timmerman et al., 2004). In Gata4T2del hearts, we did not find altered expression of components of this pathway (Notch1, HRT1 and Snail) by in situ hybridization (Fig. 5C; data not shown) or of Cdh5 by immunohistochemistry (Fig. 5E).
To identify genes whose altered expression might contribute to the Gata4T2del phenotype, we performed genome-wide expression analysis using microarrays. We isolated RNA from the AV region of mutant and control E9.5 hearts (between 30 and 33 somites), and used the RNA to probe Affymetrix microarrays. We used the SAM algorithm (Tusher et al., 2001) to identify genes with significantly altered expression (Table 2). For a subset of these, we used qRT-PCR to validate differences in gene expression between mutant and control tissue (Fig. 6A). Out of 11 genes tested, qRT-PCR confirmed differential gene expression in four: Erbb3, thrombospondin 1 (Thbs1), plexin C1 (Plxnc1), and tenascin C (Tnc; Fig. 6A). We also confirmed differential expression of Erbb3, Thbs1 and Plxnc1 by in situ hybridization (Fig. 6B).
The downregulation of Erbb3 in Gata4T2del AV tissue was of particular interest. Egf family ligands signal through Erbb3 to activate Ras and promote EMT, and in the absence of Erbb3 AV endothelium fails to form cushion mesenchyme (Camenisch et al., 2002). We confirmed downregulation of Erbb3 transcripts by qRTPCR (Fig. 6A) and in situ hybridization (Fig. 6B), and downregulation of Erbb3 protein by western blotting (Fig. 6C).
To obtain further evidence that Gata4 regulates Erbb3 expression, we investigated whether ectopic expression of Gata4 in endothelial cells that normally lack this transcription factor is sufficient to activate Erbb3 expression. Adenoviral expression of Gata4 in human umbilical vein endothelial cells (HUVEC) caused a 5-fold upregulation of Erbb3 expression compared with cells treated with a GFP-expressing adenovirus (Fig. 6D). The degree of upregulation is likely to be even higher in Gata4-expressing cells, as adenovirus transduced only 15–20% of cells. Conversely, we investigated whether Gata factors are necessary for Erbb3 expression in at least some cellular contexts. For this experiment, we used BT20 breast carcinoma cells, which natively express both Gata4 and Erbb3 proteins (Bouchard et al., 2005) (data not shown). Expression of a dominant-negative Gata4DBD-engrailed fusion protein strongly decreased Erbb3 transcript levels (Fig. 6E).
To determine whether Gata4 regulates Erbb3 expression at the level of transcription, we used rVista 2.0 (Loots and Ovcharenko, 2004) to find Erbb3 non-coding sequences conserved between mouse and human. Conserved non-coding sequences upstream of the putative transcriptional start site and within the first intron were used to drive expression of luciferase reporters. We found that co-transfected Gata4 stimulated transcription from both the promoter and the intronic enhancer (Fig. 6F). By contrast, the truncated protein produced by the Gata4Δex2 allele failed to stimulate transcription from these Erbb3 regulatory elements (Fig. 6F). These regulatory elements contain three predicted GATA motifs conserved between mouse and human. In mobility shift experiments, the two consensus sites in the enhancer strongly bound Gata4, whereas the site in the promoter (nonconsensus GATG site) did not. Mutation of these sites did not significantly alter transcriptional stimulation by Gata4 (see Fig. S5 in the supplementary material).
Next, we wished to determine whether Gata4T2del AV endothelium was functionally deficient in the transduction of Egf signals. We treated E9.5 Gata4T2del and control embryos with the Erbb3 ligand heregulin, and measured the phosphorylation of Erk, which is activated downstream of Ras. In heregulin-treated control embryos, activated Erk was readily observed in AV cushion endothelium (arrowheads, Fig. 7A, middle panel). By contrast, in heregulin-treated Gata4T2del embryos, Erk activation was strongly reduced in AV cushion endothelium (arrowheads, Fig. 7A, bottom panel). Erk activation in myocardium did not differ between genotypes (arrows, Fig. 7A), indicating that the downregulation in cushion endothelium was specific and unlikely to be due to technical factors.
We then investigated whether defective Erk activation downstream of Erbb3 might account for the marked reduction in EMT seen in the Gata4T2del AV cushions. We treated control and mutant explants with U0126, a selective inhibitor of Erk activation. We found that U0126 strongly reduced the extent of EMT in wild-type explants (Fig. 7B). We reasoned that if Gata4 and Erk are mutually required for EMT, then partial antagonism of each would inhibit EMT. To test this hypothesis, we generated AVC explants homozygous for Gata4H, a Gata4 allele that expresses reduced levels of Gata4 protein (Pu et al., 2004). We treated Gata4H/H and littermate control explants with a 50% inhibitory concentration of U0126 (400 nM). In the presence of vehicle, Gata4H/H and control explants generated comparable numbers of mesenchymal cells. However, the number of mesenchymal cells generated by Gata4H/H explants treated with 400 nM U0126 was significantly reduced when compared with the number generated by similarly treated control explants (Fig. 7C). This result is consistent with a model in which Gata4 and Erk act in the same genetic pathway to promote EMT.
Constitutive, partial loss of Gata4 function, as a result of either decreased protein expression (Gata4H/H) or a point mutation abrogating Fog1 and Fog2 (Zfpm1 and Zfpm2, respectively – Mouse Genome Informatics) interaction (Gata4Ki/Ki), resulted in embryonic lethality between E12.5–E16.5 (Crispino et al., 2001; Pu et al., 2004). These embryos had common atrioventricular canal defects, indicating a severe abnormality of EC maturation such that the superior and inferior cushions failed to fuse and divide the ventricular inflow tract into separate inlets for the right and left ventricle. This abnormality could, in principle, have been caused by impaired Gata4 function in the endocardium, or by defective Gata4 function in the myocardium with secondary abnormalities in the EC due to abnormal paracrine signaling. To determine whether there is a cell-autonomous requirement for Gata4 in later atrioventricular valve maturation, we generated embryos with the genotype Gata4flox/Ki; T2Cre+ (abbreviated Gata4T2del/Ki). In these embryos, Gata4flox complements Gata4Ki except in the endothelium and its derivatives, where T2Cre-mediated recombination inactivates it.
Gata4Ki/flox; T2Cre− control mice were present at weaning at the expected Mendelian frequency and had no obvious heart defects. By contrast, Gata4T2del/Ki embryos were present in the expected Mendelian ratio at E16.5, but were not present at weaning (out of eight litters genotyped), indicating lethality in late gestation or in the perinatal period. Mutant embryos examined at E16.5 had severe peripheral hemorrhage and edema, consistent with heart failure (data not shown). At E12.5, the heart (Fig. 8A–D) and the liver appeared normal (Fig. 8A,B). However, in later stage embryos, the AV valve leaflets were hypoplastic, and failure of fusion of the superior and inferior AV cushions resulted in a common atrioventricular canal defect (asterisk, Fig. 8E,F). Cell death was not increased in mutant cushions, as measured by TUNEL staining at E12.5 and E15.5 (data not shown). However, cell proliferation, as measured by BrdU uptake, was decreased in mutant cushion mesenchyme at E12.5 and E13.5 (Fig. 8G, data not shown). These data suggest that Gata4-Fog interaction within endocardium-derived cells is required for normal proliferation of the AV cushions.
We used a T2Cre transgene to recombine a conditional Gata4 allele (Gata4flox) within endothelial-derived cells and demonstrated that Gata4 is required at two stages of AVC morphogenesis. First, Gata4 is necessary for endothelial EMT to form AV cushion mesenchyme. Second, Gata4 is necessary for the growth and remodeling of the AV cushions, after they have been populated by mesenchymal cells. Gata4 interaction with Fog cofactors within endothelial-derived cells is dispensable for the former, but required for the latter.
The AV cushions are largely formed by mesenchyme derived from endothelial cells (see Fig. 1B, Fig. 3). Inactivation of Gata4 within endothelial-derived cells results in a marked decrease in the number of mesenchymal cells in the AV cushions (Fig. 2). By contrast, the OT cushions are formed by cells derived from at least two sources. OT endothelium undergoes EMT to form the mesenchyme of the proximal OT cushions, whereas neural crest contributes to the mid- and distal OT cushions (Fig. 1B; Fig. 3; see also Fig. S2 in the supplementary material). Gata4 is expressed in both the neural crest-derived and endothelial-derived portions of the OT cushions. However, its expression in the neural crest-derived portion is not required for normal OT development, as Gata4 inactivation by Wnt1Cre was compatible with normal survival and normal OT morphogenesis (W.T.P., unpublished). Gata4 expression was required to form the proximal, endothelial-derived portion of the OT cushion (Fig. 3). Because Gata4T2del embryos do not survive to a stage at which mature OT valves are evident, we were unable to determine what effect the loss of this portion of OT cushion has on OT valve development. Congenital abnormalities of the pulmonary valve are associated with human GATA4 mutation (Garg et al., 2003; Hirayama-Yamada et al., 2005; Okubo et al., 2004), suggesting that Gata4 activity within the proximal, endothelial-derived portion of the OT cushions may be important for development of the OT valves.
Inactivation of Gata4 within endothelial-derived cells blocked endocardial EMT, resulting in a paucity of mesenchymal cells within the AV cushions. This was associated with strong downregulation of Erbb3 (Fig. 6). Using heterologous expression systems, we show that Gata4 modulates Erbb3 transcript levels and transcriptional activity of Erbb3 regulatory elements (Fig. 6F). Although these regulatory elements contain evolutionarily conserved GATA binding sites, these binding sites were not required for transcriptional stimulation by Gata4 in vitro (see Fig. S5 in the supplementary material). These data suggest that Gata4 may regulate Erbb3 indirectly. Alternatively, these findings might represent limitations of the in vitro assay system. Additional in vivo studies of Erbb3 regulatory elements will be necessary to elucidate the mechanism by which Gata4 regulates Erbb3 in the endocardial cushions.
Downregulation of Erbb3 in Gata4T2del AVCs was associated with impaired Erk activation in response to the Erbb3 ligand heregulin (Fig. 7A). Erk activation was necessary for endocardial EMT (Fig. 7B). Partial inhibition of both Gata4 and Erk impaired the formation of cushion mesenchyme (Fig. 7C). This synthetic phenotype suggests a strong interaction between Gata4- and Erk-dependent pathways. Collectively, these data are consistent with a model in which Gata4 functions upstream of an Erbb3-Ras-Erk pathway that is necessary for the formation of cushion mesenchyme. Intriguingly, Gata4 itself is activated by Erk phosphorylation (Liang et al., 2001), suggesting the possibility of a positive-feedback loop that promotes endocardial EMT.
Although Gata4 is required for AV EMT, Gata4 interaction with Fog is dispensable for this step of AV valve development. Embryos deficient in Fog1 or Fog2, or in Gata4-Fog interactions (Gata4Ki/Ki), did not show any defect in the generation of AV valve mesenchyme by EMT (Crispino et al., 2001; Katz et al., 2003; Tevosian et al., 2000), indicating that Gata4 regulation of AV cushion EMT does not require Fog interaction. Consistent with this conclusion, Gata4T2del/Ki EC were normally populated with mesenchymal cells (Fig. 8A–D).
After AV cushion mesenchyme is formed by EMT, it rapidly proliferates to fill the expanding AV cushions with mesenchymal cells. The growing superior and inferior AV cushions meet and fuse, dividing the ventricular inlet into a right and left channel. Gata4 is necessary for this process, as a decrease in Gata4 levels or germline abrogation of the Gata4-Fog interaction resulted in an unseptated ventricular inlet (Crispino et al., 2001; Pu et al., 2004). Here, we show that Gata4-Fog interaction within endothelial-derived cells is required for septation of the ventricular inlet (Fig. 8).
This study demonstrates an essential role for expression of Gata4 in the endothelium and its derivatives. During the initial steps of AV valve formation, Gata4 is necessary for the expression of Erbb3, which acts through a Ras-Erk pathway to promote EMT. During subsequent steps of AV valve maturation, Gata4, in cooperation with a Fog cofactor, promotes the growth and fusion of the AV cushions, resulting in the division of the ventricular inlet into two separate channels guarded by two AV valves. These data suggest that Gata4 mutations might contribute to EC defects in humans; indeed, a Gata4 mutation has been reported to occur in association with AV septal defects (Garg et al., 2003).
The authors thank J. L. Galloway for comments on the manuscript, and members of the Tabin, Cepko, Dymecki, Izumo and Pu labs for their advice and support. This work was supported by grants from the NIH (C.J.T., R01 HD0045499; W.T.P., K08 HL004387-04 and 1 PO1 HL074734) and by a charitable donation from Edward Marran and Karen Carpenter. J.R.-F. was supported by a predoctoral National Research Service Award.
Supplementary material for this article is available at http://dev.biologists.org/cgi/content/full/133/18/3607/DC1