Bmp2 is essential for cardiac cushion epithelial-mesenchymal transition and myocardial patterning

Abnormal valve development is a common human congenital anomaly, affecting approximately 1% of all live births(Armstrong and Bischoff, 2004; Barnett and Desgrosellier,2003). Because of its prevalence, it is important to understand the mechanisms underlying cushion and valve development. During early heart development, a subset of endocardial cells break homotypic cell adhesive contacts, undergo cytoskeletal changes and invade the matrix-rich cardiac jelly to form endocardial cushions(Armstrong and Bischoff, 2004; Eisenberg and Markwald, 1995). The localized nature of the epithelial to mesenchymal transition (EMT)suggests a mechanism to molecularly distinguish the myocardium and/or endocardium of the atrioventricular (AV) canal from the rest of the non-cushion-forming heart. The signals that induce EMT have been investigated in chick embryos using a three-dimensional collagen gel invasion assay(Runyan and Markwald, 1983). Data from these experiments revealed that Tgfβ2 and Tgfβ3 perform crucial and sequential functions in cushion morphogenesis. In chick embryos,Tgfβ2 is thought to mediate the activation of endocardium, while Tgfβ3 functions in mesenchymal invasion(Camenisch et al., 2002).

In mouse embryos, the signals that induce EMT are less well understood. Data from the in-vitro collagen gel assays suggested that Tgfβ2 is the signal required for EMT; however, mice homozygous for a null allele of Tgfβ2 still make cushions, indicating that other signals are required in vivo to induce EMT in mammals (Bartram et al., 2001; Camenisch et al.,2002; Sanford et al.,1997). Notch signaling, through regulation of the Snai1(previously Snail) repressor, has recently been shown to be required for cushion formation through downregulation of vascular/endothelial-cadherin(VE-cadherin) in cushion endocardium(Timmerman et al., 2004). In zebrafish, Wnt signaling has been implicated in the EMT required for valve formation (Hurlstone et al.,2003). However, mice with an endocardial-specific deletion ofβ-catenin have cushions, although defective, suggesting that Wnt signaling is not essential for EMT in mammals(Liebner et al., 2004).

In addition to providing a signal to the endocardium for AV cushion formation, the AV myocardium is phenotypically distinct from the chamber myocardium. Chamber myocardium is coupled intercellularly, has a fast contraction pattern and expresses the chamber-specific genes Cx40(Gja5 - Mouse Genome Informatics), Anf (Nppa -Mouse Genome Informatics) and chisel (Smpx - Mouse Genome Informatics). By contrast, the smooth-walled myocardium of the AV canal retains an embryonic phenotype and fails to express genes that are found in the chamber myocardium (Habets et al.,2002).

T-box (Tbx) genes have been implicated in regional patterning of the myocardium (Plageman and Yutzey,2005). Tbx2 has been shown to repress expression of chamber-specific genes in the AV myocardium. Mice with a Tbx2loss-of-function have expansion of chamber-specific genes into the AV myocardium (Harrelson et al.,2004). Moreover, Tbx2 has been shown to bind to an element in the Anf gene that represses Anf expression in AV myocardium through competition with Tbx5(Habets et al., 2002). In other experiments, Tbx20 has been shown to directly repress Tbx2, indicating that a Tbx gene regulatory network functions to regionally pattern the myocardium(Cai et al., 2005; Singh et al., 2005; Stennard et al., 2005; Takeuchi et al., 2005).

Previous data revealed that Bmp2 is expressed in the AV myocardium at 9.5 and 10.5 days post coitum (dpc) and in the cushion mesenchyme at later stages (Lyons et al., 1990; Sugi et al., 2004). Moreover,gene inactivation studies in mice revealed that Bmp2 was necessary for early myocardial development (Zhang and Bradley, 1996). Bmp2 has been suggested to have a role in valve morphogenesis in mouse embryos, based on data from the collagen gel invasion assay (Sugi et al.,2004). In that work, Bmp2 could substitute for myocardium to induce endocardial EMT. Moreover, noggin treatment of explants efficiently inhibited EMT. However, other data indicated that Bmp2 was insufficient to induce EMT on its own but could enhance Tgfβ transformation activity(Yamagishi et al., 1999).

Here, we provide strong genetic evidence that Bmp2 induces AV cushion EMT in mammals. We inactivated Bmp2 specifically in the AV myocardium and found that Bmp2 mutant embryos failed to form AV cushions. Expression of Has2 was decreased in Bmp2 mutant embryos with reduced formation of the cardiac jelly. Furthermore, inactivation of the Bmp type 1A receptor, Bmpr1a, in endocardium also resulted in loss of AV cushion formation, indicating that Bmp2 signaled directly to the endocardial cushions. Our findings also indicate that the basic helix-loop-helix (bHLH) transcription factor Twist1 is a downstream effector of Bmp2 in EMT. We also found that Bmp2 has a planar signaling function to regulate patterning of the AV myocardium through regulation of Tbx2 expression. Taken together, our data indicate that Bmp2 has a crucial function in coordinating cushion morphogenesis with AV myocardial patterning.

The Bmp2^floxneo allele has been described(Ma and Martin, 2005). The Bmp2^floxneo allele contains LoxP sites flanking Bmp2 exon 3 that encodes the mature peptide. Deletion of this exon was shown to result in a null Bmp2 allele. To generate the Bmp2^nullallele, we used the NestinCre transgene, which directs efficient cre activity in the germline (Trumpp et al.,1999). For genotyping, DNA was extracted from yolk sacks or tails of embryos and adult mice, respectively. PCR was used to determine the genotype of the Bmp2^null allele with the primers(5′-AAG TCT CCT CCT TCA TCA GTA TAC GCT CG-3′) and (5′-GAT ATC GAA TTC GAT ATC AAG CTG AT-3′) located in the neomycin cassette. The amplified product is 100 bp.

Whole-mount and section in-situ hybridization was performed as previously described (Lu et al., 1999). Details about probes will be provided upon request. In all in-situ experiments, at least three mutants and three control embryos were analyzed for each experiment.

For histology, embryos were fixed overnight in Bouin's fixative or buffered formalin, dehydrated through graded ethanol and embedded in paraffin. Sections were cut at 7-10 μm and stained with H&E. Staining for lacZwas as previously described (Lu et al.,1999).

Tissue sections were deparaffinized, rehydrated and boiled for 5 minutes in 10 mmol/l sodium citrate for antigen retrieval. After washing in PBS, the nuclear factor of activated T-cells (NFAT) c1 (Cat. No.: 56602; BD Pharmingen,San Jose, CA 95131, USA) primary antibody (1:250 dilution), was applied and incubated overnight at 4°C. The sections were washed and incubated consecutively with biotin-labeled secondary antibody (sc-2017; Santa Cruz, CA 95060 USA). The secondary antibodies were detected using the avidin-biotin method. To detect Phospho-Smad1/Smad5/Smad8, sections were blocked in PBS/0.05% BSA and primary antibody (#9511′; Cell Signaling, MA 01915,USA) was applied with 1:200 dilution for overnight at 4°C. Sections were washed and incubated with secondary antibody (Cat. No.: DPVR-15DAB). Color was developed with diaminobenzidine (DAB) provided by the kit of secondary antibody and the sections were counterstained with hematoxylin. For control staining, preimmune serum was used instead of the primary antibody.

Bmp2 expression was examined by whole-mount in-situ hybridization with a Bmp2 exon3 probe (Fig. 1). Bmp2 was detected in the cardiac crescent at 7.5 dpc,consistent with its role in early cardiac morphogenesis(Fig. 1A)(Schultheiss et al., 1997; Zhang and Bradley, 1996). At 8.5 dpc, Bmp2 was expressed bilaterally in the sinus venosus(Fig. 1B,C), while at 9.5 and 10.5 dpc, Bmp2 was abundantly expressed in the AV myocardium and at lower levels in the outflow tract (OFT; Fig. 1D,E,F,G). In the 11.5 dpc embryo, Bmp2 mRNA expression in the AV canal was extinguished,although low-level expression was still detected in the OFT and ventricular myocardium (Fig. 1H). By 12.5 dpc, Bmp2 expression was detected at low levels in the OFT and left ventricular myocardium (Fig. 1I). Taken together, these data indicate that, in addition to an early role in cardiac induction, Bmp2 also probably functions in the development of the inflow and AV canal. Bmp2 also may have a role in OFT and ventricular myocardial development.

To directly investigate Bmp2 function in cardiac development, we constructed a Bmp2 conditional null allele, the Bmp2^floxneo allele, which contained LoxP sites surrounding exon 3 encoding the mature Bmp2 peptide. Deletion of exon 3 has been shown to result in a Bmp2 null allele(Ma and Martin, 2005). To inactivate Bmp2 in the heart, we used the Nkx2.5^cre knock-in allele, which directs cre activity to the anterior splanchnic mesoderm of the primary heart field and within the mature heart (Fig. 2A,B)(Liu et al., 2004; Moses et al., 2001).

To generate Bmp2 cardiac-specific mutant embryos, we crossed the Nkx2.5^cre+/- mice to the Bmp2^floxneo+/- mice to obtain Nkx2.5^cre;Bmp2^floxneo compound heterozygotes. Intercrosses between these mice and the Bmp2^floxneo homozygous mice resulted in recovery of 25% of embryos that were Nkx2.5^cre;Bmp2^{floxneo/floxneo(f/f)}, hereafter referred to as Bmp2 CKO. This crossing strategy required that both Bmp2^floxneo alleles undergo cre-mediated recombination to generate a Bmp2 null cell. Because there is a delay in the timing of cre-mediated recombination as intracellular cre protein accumulates, we predicted that the two-allele recombination strategy would result in a slight delay in Bmp2 cardiac inactivation(Nagy, 2000). This would provide a stable genetic system to study Bmp2 in either the sinus venosus or the AV canal.

To determine the timing and extent of the Bmp2 cardiac-specific deletion, we performed whole-mount analysis with the Bmp2 exon 3 probe that is deleted in the Bmp2 CKO mutant embryos. In wild-type and Bmp2 CKO mutant embryos, expression of Bmp2 exon 3 was readily detectable bilaterally in the sinous venosus at 8.5 dpc(Fig. 2C,D). However, by 9.0 dpc Bmp2 was undetectable in AV myocardium of Bmp2 CKO mutant embryos (Fig. 2E,F). At 9.5 dpc, Bmp2 exon 3 was absent in the AV myocardium and formation of the AV cushions was disrupted in the Bmp2 CKO embryos(Fig. 2G-J). These data indicate that the early expression of Bmp2 is intact in the Bmp2 CKO embryos but by 9.0 dpc, Bmp2 expression in the AV myocardium is absent, allowing us to focus on Bmp2 function in the AV canal.

To establish functionally that Bmp signaling had been disrupted in the Bmp2 CKO mutant embryos, we performed immunostaining with an antibody that recognizes P-Smad 1/5/8. Smad 1, Smad 5 and Smad 8, the receptor regulated (R-Smad) effector molecules for Bmp signaling, are phosphorylated by a ligand-bound receptor complex. Phosphorylation of the Bmp-dependent R-Smad is an indication that cells are actively receiving a Bmp signal. In the 9.5 dpc wild-type embryos, approximately 80% of endocardial cells in the AV canal were stained for P- Smad1/5/8 (Fig. 2K). We also noted that approximately 10-15% of the AV myocardium was positive for Smad 1/5/8 immunostaining. By contrast, in the Bmp2CKO mutant embryos, Smad 1/5/8-positive cells were drastically reduced in the endocardium, and only a few scattered cells in the AV myocardium were immunoreactive (Fig. 2L). From these data, we conclude that Bmp signaling to the AV cushion-forming endocardium is severely disrupted in the Bmp2 CKO embryos at 9.5 dpc,while Bmp signaling in the AV myocardium is significantly reduced.

Analysis of litters at embryonic time points revealed that Bmp2CKO mutant embryos were indistinguishable from wild-type littermates at 8.5 and 9.0 dpc. At 9.5 dpc, Bmp2 CKO mutant embryos had abnormal morphology of the AV canal constriction but were still recovered at Mendelian frequencies (Table 1). By 10.5 dpc, all Bmp2 CKO mutants exhibited pericardial effusion and growth retardation, indicating that these embryos suffered from heart failure at this stage.

Histological analysis revealed there was no discernible abnormality in cardiac structure at 8.5 dpc in the Bmp2 CKO mutant embryos (data not shown, see Fig. 5A,B). However at 9.5 dpc, all Bmp2 CKO mutant embryos had severe defects in AV cushion morphogenesis (Fig. 3A-D). In the wild-type embryos, there was expansion of the space between myocardium and endocardium as the cardiac jelly was laid down by the AV myocardium (Krug et al.,1985). However, in the Bmp2 CKO mutant embryos the space between the myocardium and the endocardium failed to expand in 43% (10 of 23)of the 9.5 dpc Bmp2 CKO mutant embryos, indicating a defect in cardiac jelly formation (Fig. 3A-F,I-J). We noted that the expression of Has2, a crucial component of the cardiac jelly, was reduced in the Bmp2 CKO mutant myocardium (see below).

In wild-type embryos, migrating mesenchymal cells were observed in the forming AV cushions at 9.5 dpc (Fig. 3A,B). Parasagittal sections through Bmp2 CKO mutant embryos indicated that the invasive mesenchyme failed to form(Fig. 3C,D). The failed EMT phenotype was observed in all Bmp2 CKO mutant embryos, including embryos with cardiac jelly deposition, suggesting that the EMT defect was probably not secondary to cardiac jelly loss(Fig. 3C,D). Analysis of Has2 and Msx2, markers of the transformed, migrating mesenchyme, revealed that the endocardium had not been induced to undergo EMT(Fig. 3G-N). Taken together,these data indicate that Bmp2 is required to both promote cardiac jelly formation and induce AV endocardial EMT.

In order to establish more firmly that Bmp2 signaled directly to the underlying endocardium, we generated mouse embryos that were deficient in the competence to receive Bmp signals within the endocardium. We inactivated the type 1A Bmp receptor, Bmpr1a, in the endocardium using the Tie2 cre transgenic line (Kisanuki et al.,2001). Bmpr1a is a major type 1 receptor for Bmp2 and Bmp4 (von Bubnoff and Cho, 2001). Because germline mutation of Bmpr1a is early embryonic lethal, we used the Bmpr1aconditional null (Bmpr1a^flox) allele(Mishina et al., 2002; Mishina et al., 1995). We generated embryos that carried Tie2^cre and the Bmpr1a^flox and Bmpr1a^null(Bmpr1a^n/f) alleles. We examined Tie2^cre;Bmpr1a^n/f mutant embryos at 9.5 dpc and found that embryos deficient for Bmpr1a in the cushion endocardium failed to form cushions (Fig. 4A,B). It is notable that one-third of Tie2^cre; Bmpr1a^n/fmutants made cushions. This probably results from variability of cre activity;however, it is possible that other type 1 receptors also have a function in the cushion endocardium.

To determine the extent of Bmpr1a inactivation in the Tie2^cre; Bmpr1a^n/f mutants, we examined Smad 1/5/8 immunostaining in the AV endocardium of wild-type and Bmpr1amutant embryos. In the wild type, most endocardial cells stained positive for P-Smad 1/5/8 (Fig. 4C), while in the Tie2^cre; Bmpr1a mutant embryos approximately 50%fewer strong positive cells were detected(Fig. 4D). We noted that the main reduction in P-Smad 1/5/8 immunoreactivity in Tie2^cre;Bmpr1a mutants was in endocardium located toward the atrium(Fig. 4D, n=3). The significance of this observation is presently unclear. These data indicate that while endocardial inactivation of Bmpr1a results in incomplete loss of endocardial Bmp responsiveness, the AV cushion phenotype is strong,revealing that AV cushion EMT is very sensitive to small decreases in Bmp signaling. Moreover, our findings support the hypothesis that other type 1 receptors may have overlapping function in the AV canal endocardium.

We studied expression of Twist1, encoding a bHLH transcription factor that is expressed in AV endocardium and has been implicated in EMT in cancer metastasis (Yang et al.,2004). In wild-type embryos, we found that Twist1 was expressed in both AV endocardium and cushion mesenchyme(Fig. 4E and see Fig. 5E,G). In the Tie2^cre; Bmpr1a^n/f mutant AV canal, we found that Twist1 expression was reduced at 9.5 dpc and absent by 10.5 dpc(Fig. 4F).

Twist has been shown to negatively regulate expression of E-cadherin in breast cancer cells, prompting us to examine expression of VE-cadherin in the AV canal endocardium of Tie2^cre;Bmpr1a^n/f mutant embryos(Kang and Massague, 2004). During EMT, VE-cadherin is normally downregulated(Fig. 4G,I); however, in the Tie2^cre; Bmpr1a^n/f mutant endocardium VE-cadherin expression persisted when compared with the wild-type embryo(Fig. 4H,J). Notch signaling has been shown to be required for EMT in the AV canal(Timmerman et al., 2004). In the Tie2^cre; Bmpr1a mutants, we found that expression of Notch1 was similar to wild-type embryos(Fig. 4K,L). We also examined expression of Snai1, a zinc finger transcription factor that has been implicated in EMT and has been shown to be a target of Notch signaling in the AV endocardium (Timmerman et al.,2004). By contrast to Twist1, Snai1 expression was unaffected by loss of Bmpr1a in AV endocardium(Fig. 4M,N). Taken together,these data indicate that Bmp2 signals directly to the cushion endocardium through the type 1 receptor, Bmpr1a.

NFATc1 is regulated by a calcium-calcineurin pathway and can translocate to the nucleus upon activation(Crabtree and Olson, 2002). Moreover, NFATc1 expression is upregulated in activated endothelial cells within the AV canal. At 8.5 dpc, NFATc1 was present in the endocardium of both wild-type and Bmp2 CKO mutant embryos(Fig. 5A,B). By 9.5 dpc, NFATc1 was upregulated and nuclear localized in the wild-type AV cushion endocardium (Fig. 5C),while expression in the Bmp2 CKO mutant embryo AV canal failed to be upregulated (Fig. 5D).

As for the Tie2^cre; Bmpr1a^n/f mutant embryos, expression of Twist1, normally detected in the AV cushion endocardium and mesenchyme of wild-type embryos was not detected in the Bmp2 CKO mutant endocardium (Fig. 5E-H). Bmp signaling regulates the Msx1 homeobox gene in other developmental fields (Liu et al.,2005; Vainio et al.,1993). Expression of Msx1 was detected in the AV endocardium of wild-type embryos but was absent in the Bmp2 CKO mutant embryos (Fig. 5I-L). The inhibitory Smad, Smad6, is a negative regulator of Bmp signaling that is also transcriptionally regulated by Bmp signaling. Furthermore, Smad6 has been shown to be a negative regulator of valve development(Desgrosellier et al., 2005; Galvin et al., 2000; Ishida et al., 2000). In wild-type embryos, Smad6 was expressed in the AV endocardium, while in the Bmp2 CKO mutants, Smad6 expression was absent(Fig. 5M,N). This finding uncovers a negative feedback loop that functions to limit the extent of endocardial EMT.

During activation of the AV endocardium, VE-cadherin expression is downregulated in wild-type embryos, but in the Bmp2 CKO mutant VE-cadherin persists (Fig. 5O,P). A Notch-Snai1 signaling pathway has been implicated in VE-cadherin downregulation in AV endocardium(Timmerman et al., 2004). In the Bmp2 CKO mutants, Notch 1 and Snai1 expression was reduced compared with wild-type embryos, indicating that the Notch and Bmp2 signaling pathways cooperate in AV cushion morphogenesis(Fig. 5Q-V). Taken together,our data indicate that Bmp2 induces EMT-promoting genes, such as Twist1 and Snai1, and other genes that restrain EMT such as Smad6. Moreover, these data indicate that Bmp2 CKO mutants,in contrast to Bmpr1a-deficient embryos, have defects in Notch-Snai1 signaling in the AV cushion endocardium (see Discussion).

Previous studies in chick embryos had suggested that Bmp2 regulated Tbx2 expression in the developing heart, prompting us to examine Tbx2 expression in Bmp2 CKO mutants(Yamada et al., 2000). In wild-type embryos, Tbx2 was expressed specifically in the AV myocardium, while in the Bmp2 CKO mutant embryos Tbx2expression was undetectable (Fig. 6A,B). By contrast, expression of Tbx5 and Tbx20were unaffected by Bmp2 deletion(Fig. 6C-F). These data indicate that Bmp2 promotes expression of Tbx2 and so is important to maintain the correct ratio of Tbx genes in the AV myocardium.

In the AV canal, Tbx2 is known to repress chamber-specific gene expression. Consistent with this, expression of the chamber-specific genes Anf, chisel, and connexin 40 was limited to the ventricular and atrial myocardium in wild-type embryos(Fig. 7A,C,E,G,I,K). By contrast, in the Bmp2 CKO mutant embryos, expression of all three chamber-specific genes was expanded into the AV myocardium(Fig. 7B,D,F,H,J,L). Markers of the AV myocardium, such as Tgfβ2 and Lef1, were lost in the Bmp2 CKO mutant embryos (Fig. 7M-P). From these findings, we conclude that Bmp2 signaling directs regionalized myocardial patterning.

In this work, we investigated Bmp2 function in cardiac development by deleting Bmp2 specifically in the AV myocardium. Our findings provide strong genetic evidence that Bmp2 is a crucial signal for AV cushion formation in mammals. We found that Bmp2 promotes cardiac jelly formation and signals directly to the AV endocardium through the type 1A receptor, Bmpr1a, to induce EMT. Moreover, our data reveal that Twist1 is one target of Bmp2 in the AV canal endocardium,suggesting a novel signaling pathway regulating EMT in valve formation. In addition, our findings indicate that Bmp2 concurrently regulates myocardial patterning through regulation of Tbx2 expression. Taken together, our data provide insights into the mechanisms coordinately regulating cushion development and myocardial patterning in mammals.

The early lethality of the germline Bmp2 null mice has hampered definitive insight into the role of Bmp2 in AV cushion development(Zhang and Bradley, 1996). Recently reported experiments, using the collagen gel invasion assay,suggested that Bmp2 could substitute for myocardium to induce EMT in mouse embryos (Sugi et al., 2004). However, other experiments, also using the collagen gel invasion assay, showed that Bmp2 functioned only in combination with Tgfβ3 to promote EMT in chick embryos (Yamagishi et al.,1999). Furthermore, chick embryos treated with a retrovirus expressing Noggin caused defects in the OFT but did not affect the AV canal(Allen et al., 2001). Our data indicate that Bmp2 is the myocardial-derived signal that induces EMT in mammals. One likely explanation for the different results obtained from these studies is species-specific differences between mice and chicks(Camenisch et al., 2002).

Our findings indicate that Bmpr1a is a crucial type 1 receptor in the AV endocardium to transduce the Bmp signal required for EMT. Inactivation of Bmpr1a using the Tie2^cre transgenic line disrupted EMT in most mutant embryos examined. The Tie2^cre;Bmpr1a n/f mutant cushion phenotype indicates a direct requirement for a Bmp signal to induce EMT in the endocardium. Furthermore, our findings indicate that the Tie2^cre; Bmpr1a n/f mutants have normal cardiac jelly deposition, as Has2 is expressed and cardiac jelly is deposited normally. Taken together, these findings provide strong evidence that the cushion defect in the Bmp2 CKO mutants is not secondary to failure of cardiac jelly deposition or disruption of an indirect signal relay mechanism.

It is important to note that Alk2 has also been shown to be required for EMT in the in-vitro explant system using chick embryos(Desgrosellier et al., 2005). A constitutively active Alk2 was able to induce EMT, while Alk2 antisera could effectively inhibit EMT. Furthermore, mouse embryos deficient for Alk2 in cushion endocardium had a severe defect in AV cushion morphogenesis (Wang et al.,2005). Our data indicate that the Notch-Snai1 pathway was intact in the Tie2^cre; Bmpr1a n/f mutant embryos, while this pathway was disrupted in the Alk2 and Bmp2 mutant embryos(this work) (Wang et al.,2005). This indicates that Bmpr1a and Alk2 may activate distinct pathways in cushion endocardium. However, other genes such as Msx1, were reduced in both the Bmpr1a and Alk2mutants. Taken together, these findings indicate that signaling through Alk2 and Bmpr1a activates distinct but overlapping target pathways. Alternatively, as some (approximately 33%) Tie2^cre;Bmpr1a n/f mutants made AV cushions, it is possible that there is partial redundancy between type 1 receptors in AV endocardium, similar to that observed in chondrocytes (Yoon et al.,2005). Future studies will be required to investigate these questions in more detail.

Previous work investigating Bmp ligands in cardiac development uncovered considerable functional redundancy. For example, Bmp7 mutants have no phenotype in the heart as a result of redundancy with other Bmp ligands(Luo et al., 1995). Bmp6;Bmp7 double mutants have delayed formation of the OFT cushions(Kim et al., 2001), while the Bmp5; Bmp7 double mutants have severe defects in cushion formation(Solloway and Robertson,1999). Moreover, we found evidence for overlapping function between Bmp4 and Bmp7 in OFT development(Liu et al., 2004). By contrast, we have found that Bmp2 function is essential in the AV canal. It will be important in the future to investigate the functional relationship of the distinct classes of Bmp ligands in cushion morphogenesis.

Tgfβ signaling is a crucial signal in avians for EMT in the AV cushions, as determined by the collagen invasion assay(Barnett and Desgrosellier,2003). However, although Tgfβ2 is clearly important for normal cushion development in mice, EMT can still occur in the Tgfβ2 null mice (Sanford et al., 1997). Our data, and the work of Sugi et al.(Sugi et al., 2004), indicate that much of the function of Tgfβ has been co-opted by Bmp signaling in mammals. In mice, current evidence suggests that Tgfβ2 expression is regulated by Bmp signaling. Inactivation of Bmpr1a in myocardium using an MHC cre transgene resulted in downregulation of Tgfβ2 expression in the AV canal (Gaussin et al.,2002). This suggested a cell-autonomous requirement for Bmp signaling in myocardium to maintain Tgfβ2 expression, which was important for the later aspects of AV cushion development. Furthermore, Bmp2 induces Tgfβ2 in explants cultured in collagen gel(Sugi et al., 2004), and Tgfβ2 expression is off or reduced in the AV cushions of Bmp2CKO mutant embryos (this work). Future experiments will be required to firmly establish the genetic relationship between Bmp signaling and Tgfβ2 transcriptional regulation.

Twist1, first identified in Drosophila as a crucial regulator of gastrulation and mesoderm formation, has previously been recognized as an important developmental control gene in higher vertebrates. Twist1 has been shown to be the gene mutated in Saethre-Chotzen syndrome, a haploinsufficient disorder that primarily affects cranial and limb development (Howard et al.,1997). In mouse embryos, Twist1 is expressed in presomitic mesoderm, neural crest-derived mesenchyme of the head and branchial arches and in limb bud mesenchyme(Gitelman, 1997). Twist1 null mutant embryos had failure of rostral neural tube closure, most likely as a result of a head mesenchyme deficiency, but defects in cardiac morphogenesis were not reported(Chen and Behringer, 1995).

Particularly relevant to the present study is the recent observation that Twist1 promotes EMT in cancer cells and facilitates tumor metastasis. Twist1 regulates EMT through a mechanism involving repression of E-cadherin transcription (Kang and Massague, 2004; Yang et al., 2004). Our data indicate that Twist1 expression was lost in both the Bmp2 CKO and Tie2^cre; Bmpr1a n/fmutant embryos. Moreover, we observed a modest defect in VE-cadherin downregulation in the Bmp2 and Bmpr1a mutant AV canals. These findings suggest that Twist1 is an effector of the Bmp-signaling pathway in the AV canal that promotes endocardial EMT.

We observed only mild defect in VE-cadherin downregulation, despite loss of Twist1 in the Bmp2 CKO mutant AV canal. This is probably a result of compensatory function of the closely related Twist2 gene. There are two Twist genes in mammals, Twist1 and Twist2, which have redundant functions in cytokine regulation(Bialek et al., 2004; Li et al., 1995; Sosic et al., 2003). Further experiments will be required to investigate this possibility.

It is notable that the Snai1 transcription factor has recently been shown to be required for AV canal EMT as part of the Notchsignaling pathway (Timmerman et al.,2004). In Drosophila, Twist and Snai1 are both required for correct dorsoventral patterning and mesoderm induction as part of the Dorsal/Twist/Snai1 pathway(Stathopoulos and Levine,2002). In vertebrates, both Snai1 and Twist1have been shown to promote EMT through inhibition of E-cadherinexpression. Our findings suggest that Bmp and Notch signaling cooperate to influence VE-cadherin expression. It will be interesting to investigate the functional relationship of Notch and Bmp signaling in AV canal EMT in future experiments.

We provide evidence that Bmp2 is upstream of Tbx2 in the AV myocardium. Our observation that myocardial patterning in the Tie2^cre; Bmpr1a mutants is normal argues against the hypothesis that an endocardial-derived signal regulates myocardial patterning. Further experiments will be required to investigate whether Bmp2 directly regulates Tbx2 expression.

Our findings can be interpreted in the context of recent Tbx20inactivation studies that led to conflicting interpretations of the Bmp2-Tbx2 regulatory relationship. Tbx20 mutant embryos had expanded Tbx2 expression throughout the myocardium despite diminished Bmp2 expression, suggesting that Tbx2 regulation is Bmp2-independent (Cai et al.,2005; Stennard et al.,2005). Another, independently generated, Tbx20 null allele resulted in ectopic Bmp2 with resulting expanded Tbx2(Singh et al., 2005). Our data support the conclusion that Bmp2 promotes Tbx2 expression in the AV myocardium.

It is interesting to note that Tbx20 and Tbx2 are co-expressed in the AV myocardium, indicating an AV-canal-specific mechanism that allows Tbx2 to escape Tbx20-imposed repression. Our data suggest that Bmp2-signaling counters the repressive activity of Tbx20 in the AV myocardium. Because Tbx20 is expressed normally in the Bmp2 CKO mutant embryos, Bmp2 may regulate Tbx20 post-transcriptionally. Bmp2 may also induce genes that functionally inhibit Tbx20 activity. The precise mechanism underlying this inhibition of Tbx20 is presently unclear and awaits future experiments.

In conclusion, our findings reveal that Bmp2 is required for EMT in the mammalian AV canal (see Fig. 8). We have uncovered a genetic pathway involving the type 1 Bmp receptor, Bmpr1a, which functions in EMT. Our data also revealed that Bmp2 signaling is required for normal deposition of the cardiac jelly. Within the myocardium, Bmp2 functions upstream of Tbx2 to control regionalized myocardial patterning.

We thank A. Bradley, P. Soriano and R. Behringer for reagents; A. McMahon,G. Martin, B. Hogan, R. Maxson, D. Maas, K. Yutzey, R. Kelley and R. Harvey for in-situ probes; M. Yanagisawa for the Tie2^cre transgenic line;and R. Behringer and Y. Mishina for the Bmpr1a conditional allele. We thank A. Moon and A. Baldini for comments on the manuscript and discussions. This work was supported by grants NIDCR (2R01DE/HD12324-06 and R01DE013509) to J.F.M., and P30 ES9106-07 to J.F.M. and R.J.S.