BMP receptor IA is required in mammalian neural crest cells for development of the cardiac outflow tract and ventricular myocardium

The neural crest is a transient vertebrate tissue with a wide variety of cell fates, including peripheral nervous system neurons and glia, melanocytes,smooth muscle cells and most craniofacial cartilages and bones(Le Douarin and Kalcheim,1999). The neural crest also contributes crucial cell populations to several thoracic tissues, including the developing outflow tract of the heart (Kirby and Waldo, 1995). Neural crest cells (NCCs) delaminate from the dorsal neural folds, undergo an epithelial to mesenchymal transition, and migrate to target tissues. In mouse,this process starts in the rostral hindbrain around the five somite (5 s)stage (Nichols, 1981), extends to the midbrain, forebrain and caudal hindbrain, and progresses to increasingly caudal trunk levels. At any given axial level, the process of NCC production lasts about 9 hours (Serbedzija et al., 1992). Thus, the early phases of NCC development are ongoing at all axial levels in mouse; for example, in the rostral hindbrain,emigration appears to be complete around 11 s, while it lasts until 14 s in midbrain and caudal hindbrain, and extends to yet later stages at other axial levels (Serbedzija et al.,1992).

In conjunction with several other developmental signaling molecules, bone morphogenetic proteins (BMPs), particularly BMP2 and BMP4, have been implicated in promoting NCC induction, maintenance, migration and differentiation in several different model organisms(Aybar and Mayor, 2002; Christiansen et al., 2000; Knecht and Bronner-Fraser,2002). Most evidence has come from gain-of-function manipulations,such as application of recombinant BMP proteins, or overexpression of transgenes that modulate BMP signaling. However, zebrafish embryos with reduced BMP2 show deficient NCCs (Nguyen et al., 1998). Mouse embryos null for BMP2 show no evidence of migratory NCCs, but die while neural crest is forming, obscuring the requirement for BMP2 in this process(Kanzler et al., 2000). The involvement of other BMPs and signaling pathway components are masked by earlier roles or, possibly, by functional redundancy. Thus, the requirements for BMP activity in various aspects of mammalian neural crest development remain unclear.

We have used a conditional gene ablation approach to address the requirements for BMP signaling in mammalian NCC biology. BMP ligands activate a bipartite receptor complex composed of a type I and type II receptor(Kawabata et al., 1998). Mouse has three known type I receptors for BMP2 and BMP4, BMPRIA (ALK3), BMPRIB(ALK6) and possibly ActRIA (ALK2)(Miyazono et al., 2001). Neither Bmpr1b nor Actr1a is expressed in the neural tube until well after NCC migration has begun(Gu et al., 1999; Panchision et al., 2001; Yoshikawa et al., 2000). By contrast, Bmpr1a is expressed throughout the early embryo, including the dorsal neural folds, but null mutants fail to gastrulate(Mishina et al., 1995). Fortunately, a conditional null allele of Bmpr1a allows the deletion of this gene in tissues expressing Cre recombinase(Mishina et al., 2002). We used a Wnt1-Cre transgene, which triggers recombination exclusively in forming NCCs (Brault et al.,2001; Chai et al.,2000; Danielian et al.,1998; Jiang et al.,2000), to delete Bmpr1a specifically in the neural crest from the outset of migration. Embryos lacking Bmpr1a in NCCs die abruptly at midgestation with no defects in the induction, maintenance,delamination or initial migration of NCCs. Instead, we observed severe defects in the cardiac outflow tract and ventricular myocardium.

Embryos were generated by mating Wnt1-Cre/+;Bmpr1a^null/+ mice with mice homozygous for the Bmpr1a^flox allele or doubly homozygous for the R26R reporter and the Bmpr1a^flox allele. All Bmpr1a alleles and their genotyping are as published(Mishina et al., 2002). Mutant embryos were identified by PCR genotyping for presence of the Cre transgene(Cre1, TGATGAGGTTCGCAAGAACC; Cre 2, CCATGAGTGAACGAACCTGG) and a null allele of Bmpr1a. R26R alleles were identified by a 250 bp PCR product (R26R-1,AAAGTCGCTCTGAGTTGTTAT; R26R-2, GCGAAGAGTTTGTCCTCAACC). Tie2lacZ(Schlaeger et al., 1997)-positive animals were identified by X-gal staining (see below) of tail biopsies before weaning. Noon on the day after mating (as judged by vaginal plug) was designated as E0.5. Embryos were staged precisely by somite count and comparison with established references(Kaufman, 1992).

RNA probes were labeled with digoxigenin (Roche) and used for whole-mount in situ hybridization according to standard protocols(Belo et al., 1997). All riboprobes were synthesized from published plasmids: Tcfap2a(Mitchell et al., 1991), Crabp1 (Stoner and Gudas,1989), cadherin 6 (Inoue et al., 1997), Tcf21(Quaggin et al., 1998). lacZ transgenes were visualized by fixing embryos for 15 minutes at room temperature with 4% paraformaldehyde, rinsed and stained overnight with X-gal stain. Tie2-lacZ embryos were cleared to benzyl alcohol: benzyl benzoate (BABB). Histological analysis was conducted on 8 μm paraffin wax-embedded sections using established protocols(Hogan et al., 1994). Section immunohistochemistry with antibodies for smooth muscle actin (Sigma clone 1A4)and PECAM1 (BD Pharmingen) used colorimetric substrates (NBT/BCIP). The WT1 antibody (Santa Cruz) was used on cryosections at 1:500 and visualized with AlexaFluor 594 Goat anti-Rabbit IgG (Molecular Probes) at 1:500. Whole-mount immunohistochemistry was carried out with 2H3 supernatant from the Developmental Studies Hybridoma Bank at the University of Iowa. Proliferating cells were marked with an antibody against phosphorylated histone H3 (Upstate Biotechnologies) at 1:1000 overnight at 4°C; Goat anti-Mouse Cy3 (Jackson)overnight at 4°C at 1:1000. Confocal microscopy utilized a Zeiss LSM 510. Mitotic indices were calculated as percent of labeled cells among the total number of DAPI counterstained cells (staining for 5 minutes at room temperature).

To analyze blood flow, dissected embryos were injected with India ink into the ventricle using a pulled capillary tube and mouth pipette. Embryos were subsequently fixed in paraformaldehyde and cleared in BABB(Hogan et al., 1994).

To test for the presence of Cre, tissues were microdissected from embryos. RNA was purified using Trizol reagent and standard procedures, including DNAse treatment to prevent genomic DNA contamination(Sambrook and Russell, 2001). Total RNA was used for in vitro cDNA synthesis with MMLV reverse transcriptase(MMLV-RT; Gibco). Cre was amplified with the same primers as for genotyping. HPRT primers were used as published(Stottmann et al., 2001).

Our strategy for assessing the role of Bmpr1a in NCC development relied on the use of a conditional allele of Bmpr1a(Mishina et al., 2002) in concert with an established Wnt1-Cre transgene(Danielian et al., 1998) to drive recombination specifically in NCCs(Brault et al., 2001; Chai et al., 2000; Jiang et al., 2000). To visualize Cre activity in NCCs, Wnt1-Cre mice were crossed to a Cre recombination reporter line, R26R(Soriano, 1999). Recombination by Cre at the R26R locus brings a lacZ transgene under the irreversible control of a ubiquitous promoter. Embryos carrying R26Rexpress lacZ in all cells that express active Cre recombinase, and in the descendents of those cells.

Recombination at the dorsal neural folds of the midbrain/hindbrain junction was visible by 4 s (Fig. 1A,B),reflecting a domain of Wnt1 expression required for midbrain development (McMahon and Bradley,1990). The first distinct NCCs were seen along the dorsal neural folds of the hindbrain at the 5 s stage; these are more numerous by 6 s(Fig. 1C,D). Recombination extended into the rostral spinal cord by 8 s(Fig. 1E,F). Transverse sections of Wnt-1Cre; R26R embryos at ∼10 s (E8.5) revealed migration from the neural tube to the pharyngeal arches and other target tissues (Fig. 1G and data not shown). Recombination then extended caudally along the dorsal neural tube to include the entire neural crest and its descendents(Fig. 1H,I). Wnt1-Creactivity in the neural crest at 5 s coincides with the time at which the first NCCs have been detected by cell labeling studies(Serbedzija et al., 1992). Thus, Wnt1-Cre allows recombination of Bmpr1a in the neural crest from very early stages of its development.

Mice carrying the Wnt1-Cre transgene and a null allele of Bmpr1a (Bmpr1a^null) were crossed to mice homozygous for a conditional allele of Bmpr1a(Bmpr1a^flox; Fig. 2A). One quarter of the resulting embryos are of the genotype Wnt1-Cre; Bmpr1a^flox/null, and are presumably unable to signal through BMPRIA in NCCs and their derivatives: we designate these as`mutant'. Simultaneous expression of the Wnt1-Cre transgene and the Bmpr1a conditional allele results in recombination at the Bmpr1a locus as intended (Fig. 2B; data not shown). Although we are unable to analyze precisely when BMPRIA ceases to function in any given NCC cell, our experimental system results in the removal of BMPRIA during all but the earliest stages of NCC development. Available antibodies against BMPRIA and the phosphorylated form of SMAD1 (indicating BMP signal transduction) proved ineffective in assessing BMPRIA protein and its output at the single-cell level (R.W.S. and J.K.,unpublished). However, several previous studies with this conditional allele in a variety of contexts (Ahn et al.,2001; Gaussin et al.,2002; Hebert et al.,2002; Jamin et al.,2002; Mishina et al.,2002) have given no reason to suspect significant residual BMPRIA protein or incomplete Bmpr1a^flox recombination. Thus,although ablation of BMPRIA in the neural crest domain will not necessarily happen before the induction of the first NCCs, which must occur prior to 5 s in the hindbrain, the receptor is likely to be absent well before NCC generation is complete at any axial level. This occurs at 11 s at the earliest location, in the rostral hindbrain(Serbedzija et al., 1992), by which time expression of Cre has been robust for half a day. Moreover, BMPRIA should be absent in NCCs during their migration to target tissues and their differentiation upon arrival.

In dissections from the crosses to produce mutant embryos, Wnt1-Cre;Bmpr1a^flox/null embryos were fully represented up to E11.5(Fig. 2H,J), but did not survive beyond E12.5 (n=0/47; Fig. 2K; Table 1). At E.11.0, mutants looked externally normal and had normal circulation(Fig. 2E,F). By contrast, from E11.5, they showed evidence of severe circulatory defects (n>60). For example, yolk sacs appeared to contain no blood(Fig. 2I) and embryos displayed pooling of blood in major vessels, heart and liver(Fig. 2J). The majority of embryos recovered at E12.5 were clearly necrotic(Table 1). Thus, embryos lacking Bmpr1a in NCCs die abruptly with symptoms of acute heart failure around E11.5-12.0.

Mutant embryos showed no deficit in NCC specification or migration from the neural tube, as assayed by expression of the neural crest markers Tcfap2a(AP2α: n=4), cadherin 6 (n=4) and Crabp1(n=3) at E9.5 and E10.5 (Fig. 3A,B). Initial differentiation also appeared normal; for example,an antibody marking neurofilaments of the peripheral nervous system indicated normal formation of sensory neurons (Fig. 3C,D). By including the recombination reporter R26R in our breeding scheme to ablate Bmpr1a in the neural crest, such that each resulting embryo carries one copy of R26R, we simultaneously mark both NCCs and cells lacking Bmpr1a (though we cannot be certain that any specific labeled cell necessarily lacks Bmpr1a, as the loxP sites in R26R and Bmpr1a^flox recombine independently). These NCC-labeled embryos showed that mutant cells achieved a normal distribution at E11.0 and integrated into many differentiating tissues and organs, such as pharyngeal arches and dorsal root ganglia(Fig. 3E,F). Although assessing the ability of mutant NCCs to contribute to each neural crest derivative type is well beyond the scope of this study, it is clear that NCCs devoid of Bmpr1a can contribute normally to many tissues up through the time of embryonic death. Our experiments do not allow us to assess the requirement for BMPRIA during initial neural crest specification; nevertheless, the data show that Bmpr1a is not required in NCCs from early somite stages for any ongoing specification, nor for migration or differentiation into many derivative cell types.

The appearance of mutant embryos at E11.5 indicated severe cardiovascular defects. Although we observed a loss of yolk sac circulation, we detected no NCC contribution to extra-embryonic tissues such as the yolk sac in either wild-type or mutant embryos (Fig. 4A). Moreover, yolk sac vessels appeared normal, as judged by histological analysis and by expression of the endothelial marker Tie2-lacZ (Fig. 4B). Thus, the effect of NCCs on the yolk sac blood flow is indirect. Embryonic peripheral vascular circulation in the mutants also appeared abnormal at E11.5, but did not show a structural defect in endothelium, as marked by the expression of Tie2-lacZ or by anti-PECAM1(Fig. 4C-F). NCCs contribute to the supportive smooth muscle layer surrounding peripheral vasculature(Etchevers et al., 2001; Waldo et al., 1996; Willette et al., 1999). However, immunohistochemistry to label vascular smooth muscle cells in either pharyngeal arch arteries (Fig. 4G,H) or outflow tract tissue(Fig. 4I,J) showed no differences between wild-type and mutant embryos before the onset of embryonic necrosis. These data strongly suggest that the lack of yolk sac circulation and the frequent pooling of blood in the mutant(Fig. 2J) is a sign of impending death, secondary to a crucial cardiovascular deficit elsewhere in the embryo, such as inadequate blood flow from the heart.

NCCs are known to have an essential role in heart development: they migrate along the pharyngeal arch arteries to populate the outflow tract (OFT), where they are required for septation of the truncus arteriosus into the aorta and pulmonary artery (Kirby and Waldo,1995). All mutant embryos analyzed (14) showed shortened OFTs and right ventricles as early as the 18 s stage when compared with stage-matched control embryos. Further analysis of mutant embryos carrying the R26Rreporter showed reduced NCC contribution to the truncus arteriosus. This reduced NCC population and right ventricle shortening was observed at all stages examined from E9.0 (Fig. 5A-D, ∼18 s) to E11.0 (Fig. 5I-L, 36-40 s). Shortened truncus arteriosus and right ventricle were present in mutant embryos past E11.0/40 s but these were excluded from analysis as the defect in the dying embryos could be a result of overall embryonic necrosis (data not shown). Shortened outflow tract tissue with reduced NCCs also lacked pronounced, well-formed endocardial cushions(Fig. 5M,N). Consistent with reduced NCC immigration, the Tie2-lacZ transgene marked the endothelium of the distinct aorta and pulmonary artery of wild-type embryos(Fig. 5Q), but revealed a lack of septation of the truncus arteriosus in mutants (arrow in Fig. 5R). Although the embryos observed at E12.0 were beginning to show signs of necrosis, their apparent lack of endocardial cushion formation is consistent with the findings of reduced NCCs and persistent truncus arteriosus.

We considered that a reduction in NCC migration to the OFT might result in occlusion of pharyngeal arch arteries by ectopic NCCs, which could potentially cause embryonic death at mid-gestation due to insufficient blood flow to the body. It is noteworthy that these mutant embryos also occasionally showed hypoplastic aortic arch arteries at E11.5(Fig. 5T and data not shown). We tested the ability of the heart, great vessels and aortic arch arteries from Bmpr1a NCC mutants to allow unimpeded blood flow by injecting ink into the ventricles and following its dispersal; we never detected any blockage or impeded flow (Fig. 5S,T). More importantly, we examined mutant and control embryos by serial histological sections of the entire OFT and pharyngeal arches at E11.5,and never observed blocked or occluded vessels(Fig. 5M-P). Even in a clearly necrotic embryo at E12.5 (Fig. 5V), sections through the outflow tract show a lack of septation and pooling of blood cells, but no ectopic tissues creating a barrier to blood flow to the peripheral tissues (Fig. 5U). OFT septation defects per se are not necessarily lethal until birth in chicks, mice or humans (Conway et al., 1997; Kirby and Waldo,1995; Mair et al.,1992), and thus seem unlikely to account for the mid-gestational death we observed. These considerations led us to suspect that although the outflow tract defects might exacerbate circulatory deficiency, the primary problem in Bmpr1a neural crest mutants lies elsewhere.

To elucidate the cause of the cardiovascular dysfunction and lethality in embryos lacking Bmpr1a in NCCs, we examined whether there might be defects in cardiac tissues not known to involve NCC contributions. Ventricular structure at E10.5 and E11.0 looked normal in mutant embryos(Fig. 6A-D) with formation of both the compact myocardium on the periphery of the ventricle, and the trabecular myocardium adjacent to the ventricular cavity. At E11.5, however,both layers were significantly reduced in mutants, showing a severely thinned compact myocardial wall and little expansion of the trabeculated myocardium relative to wild type (Fig. 6E-H). These defects were seen prior to the onset of embryonic necrosis (Fig. 6F).

Such dramatic changes in the morphology of ventricular tissue could be accomplished via a change in the amount of cell death or cell proliferation. Using the TUNEL assay to mark apoptotic cells in sectioned hearts, we detected no increase in myocardial cell death prior to global necrosis of mutant embryos (data not shown). By contrast, cell proliferation assays, using immunohistochemistry for phosphorylated histone H3 to mark cells in metaphase,revealed a dramatic difference in cell proliferation rates in the ventricular myocardium (Fig. 6I). Total myocardial proliferation in mutants was decreased 24% at E10.5(P<0.001) and 21% at E11.5 (P<0.001), relative to stage-matched control embryos. When proliferating cells were tallied as a percentage of total cells in the ventricular myocardium to calculate a myocardial mitotic index, they were significantly reduced by this measure as well (2.93% in controls versus 1.63% in mutants, P<0.015).

The mechanism by which the lack of a gene specifically in NCCs could lead to a deficient ventricular myocardium is unclear. To determine whether there might be a previously unknown NCC contribution to the mouse heart, we used Wnt1-Cre; R26R embryos to assess the distribution of NCC derivatives in and around the ventricles. Embryos at E9.5 showed a fewβ-galactosidase-positive cells immediately ventral to the heart and in the epicardium (Fig. 7A), a thin mesothelial tissue surrounding the heart. The epicardium derives primarily from cells on the posterior side of the septum transversum, which lies just caudal to the primitive ventricles. The epicardium expands to cover the ventricular surface in an epithelial sheet by E10.0(Komiyama et al., 1987). Some of the epicardial cells undergo an epithelial-mesenchymal transition and contribute to several lineages within the heart, including connective tissue and coronary vascular elements(Munoz-Chapuli et al., 2002). In E10.5 and E11.5 Wnt1-Cre; R26R embryos, progressively more marked cells were seen in the epicardium, and also in the outer ventricular myocardium (Fig. 7B,C). No labeled cells were ever seen in the epicardium (or elsewhere) of R26Rembryos lacking Cre; furthermore, the same distribution of stained cells is seen in Wnt1-Cre; R26R embryos regardless of whether they were processed for β-galactosidase activity and then sectioned, or sectioned and then processed for activity (Fig. 7F). Together, these data eliminate the possibility that the labeled cells are artifacts of β-galactosidase staining. Although an epicardial NCC lineage has not been documented previously using this transgene, all other Wnt1-Cre lineages detected in our embryos are identical to those shown in previous studies(Jiang et al., 2000).

Importantly, in both wild-type and mutant embryos, most cells of the epicardium do not express the NCC lineage marker; the labeled epicardial population never exceeded 12% in any of our specimens. Moreover, we failed to detect labeled cells in the epicardium of some Wnt1-Cre; R26Rembryos, despite equivalent staging, fixation and processing. We suspect the failure to observe these cells in all specimens reflects a technical limitation of our detection procedures.

Like other epicardial cells, a few of these lineage-labeled cells then invade the ventricular myocardium (Fig. 7F). Because cells of the epicardium populate the coronary arteries, we examined these vessels in late-gestation Wnt1-Cre; R26Rembryos for the presence of labeled cells. We observed labeled cells in the coronary arteries in both whole-mount (Fig. 7D) and sectioned hearts (Fig. 7E). Collectively, this distribution of marked cells suggests that a small population of NCCs migrates to the epicardium, where it integrates into the nascent epicardial epithelium and has similar cell fates as neighboring epicardial cells from other origins.

We conducted additional tests to further evaluate the hypothesis that the labeled epicardial cells derive from the neural crest. First, we assessed the distribution of cells labeled by an independent NCC genetic lineage marker, P0-Cre (Yamauchi et al.,1999). P0-Cre; R26R embryos showed a similar contribution of NCCs to the epicardium and myocardium(Fig. 7G,H). We then considered whether these cells might reflect coincidental activity of Wnt1-Creand P0-Cre transgenes in some novel domain outside the neural crest,such as the proepicardial organ or epicardium itself. To determine whether these cells could result from ectopic Cre expression outside the neural crest, we assayed tissue ventral to the neural tube, including the heart and all other thoracic tissues, for Wnt1-Cre expression by reverse transcription coupled to the polymerase chain reaction (RT-PCR). We found no Cre expression in ventral tissues collected from E9.5 and E10.5 Wnt1-Cre embryos, covering a wide range of somite numbers (0/37; Fig. 7I). We cannot exclude the possibility that there may be a brief burst of short-lived Cre expression in a temporal window not represented in our samples, although this seems unlikely. Instead, these data taken together strongly suggest that a small population of labeled epicardial and myocardial cells derive from the neural crest.

Among several BMPs expressed in the developing heart, BMP2 and BMP4 are known to be ligands for BMPRIA. BMP2 is expressed in the atrioventricular canal from E9 to E11 (Zhang and Bradley,1996). Using a Bmp4-lacZ transgene(Lawson et al., 1999), we observed that Bmp4 is expressed at high levels in the parietal pericardium of the thoracic body wall from E9.5 through E11.5(Fig. 7J-L; data not shown). This suggests that at least one BMPRIA ligand, BMP4, is present in a cell population adjacent to potential receiving cells in the epicardium.

We then assessed whether the epicardium might show abnormalities in Bmpr1a neural crest mutants. Examination of Wnt1-Cre;Bmpr1a^flox; R26R embryos revealed an indistinguishable distribution of labeled cells in the epicardium between mutant and wild-type embryos (Fig. 8A,B compared with Fig. 7A,B). This is in contrast to the truncus arteriosus, where there is a striking reduction in NCC population in mutants. Although mutant NCC derivatives populate the epicardium, it is conceivable nonetheless that the epicardium is structurally abnormal in the mutants. Physical delay of migration of cells from the proepicardial organ results in a less extensive epicardial covering of the ventricles; this in turn is associated with reduced myocardial proliferation in underlying tissue (Perez-Pomares et al., 2002). To analyze the structure of the epicardium, we compared ventricular epicardium in sectioned mutant hearts and control littermates, and found no evidence for a disrupted or patchy epicardium(Fig. 8C-F). Moreover, the epicardial cell density in sections was not significantly different(P>0.100) between stage-matched mutant and wild-type embryos.

Although mutant embryos displayed a structurally intact epicardium, we assayed molecular markers of the epicardium to determine whether epicardial gene regulation is generally perturbed in the Bmpr1a neural crest mutants. In the heart, epicardin is expressed specifically in the epicardium(Robb et al., 1998). We detected no change in the expression of epicardin relative to wild type(Fig. 8G,H). WT1 is also expressed in the epicardium, but unlike epicardin, has been shown to have an important role in epicardial promotion of myocardial maturation(Kreidberg et al., 1993; Moore et al., 1999). Immunohistochemistry for the WT1 protein revealed no change in protein levels or localization between control and mutant embryos(Fig. 8I,J). Thus, the epicardium appears to be largely normal in mutant embryos.

Taken together, these data indicate that cells may migrate from the neural crest to the epicardium, where they are potentially exposed to several BMPs,including BMP4. Loss of Bmpr1a in NCCs results in midgestational defects in the underlying ventricular myocardium similar to those seen in epicardial ablations. The epicardial population of NCCs is distributed normally in mutant embryos, and the epicardium is structurally intact. These results therefore suggest a specific compromise in the proliferative influence of epicardium on ventricular myocardium upon loss of BMPRIA in neural crest derivatives.

In this work, we sought to determine the requirements for BMPRIA signaling in neural crest cell development by ablating this receptor specifically in the neural crest domain from early somite stages. Although BMP 2/4 signaling promotes NCC induction, maintenance, delamination and migration from the neural tube (Aybar and Mayor,2002; Knecht and Bronner-Fraser, 2002; LaBonne and Bronner-Fraser, 1999), embryos lacking Bmpr1a in forming NCCs show no discernable defects in these processes. However, these mutant embryos die of mid-gestational acute heart failure, with deficient ventricular myocardium proliferation and defective outflow tract morphogenesis.

Although Bmpr1a is expressed throughout the neural plate, this receptor is dispensable for many aspects of NCC development. Definitive neural crest is first detectable at 5 s, beginning in the rostral hindbrain(Nichols, 1981), and production of NCCs is ongoing for ∼9 hours at any given axial location,ending first in the hindbrain at 11 s to 14 s(Serbedzija et al., 1992). Meanwhile, Wnt1-Cre is expressed from 5 s in the hindbrain, and has been strongly expressed there for several hours by 11 s. Even if there were perdurance of BMPRIA protein long after the gene has been ablated (not suggested by previous studies with this allele), it seems unlikely that significant numbers of putative NCCs would retain BMPRIA protein during the whole window of NCC production. If the receptor were required for early NCC development but were not ablated soon enough, one would have expected that the initial NCCs would form and migrate away, but that continued production would decline and stop. Instead, we observed continuous production of NCCs indistinguishable from wild-type controls. Moreover, the BMPRIA protein would have been absent for many hours as the NCCs migrate to target tissues and differentiate. It may be that the only crucial period for BMPRIA signaling occurs prior to 5-6 s, when Wnt1-Cre recombination becomes prevalent in the forming neural crest; our current experiment is unable to assess the necessity for this receptor during the earliest phases of NCC specification. Taken together, these considerations strongly suggest that BMPRIA signaling is unnecessary in vivo for neural crest delamination, migration and differentiation into most derivative cell types.

A potential implication of our results is that BMP signaling in general may not have an essential role in many aspects of mammalian NCC development. Neither of the other known BMP type I receptors, BMPRIB and ActRIa, is known to be expressed in the neural tube during early NCC development(Gu et al., 1999; Panchision et al., 2001; Yoshikawa et al., 2000). However, it remains possible that ectopic expression of Bmpr1b or Actr1a, or perhaps an unknown TGFβ family type I receptor, can compensate for the absence of Bmpr1a in this context.

Our study reveals a critical requirement for BMPRIA signaling in proper formation and septation of the cardiac OFT. Although NCCs successfully migrate to the OFT in mutant embryos, we observed severe OFT defects, including a failure of septation and reduced conotruncal length. These phenotypes, which result from the ablation of a single gene product, closely resemble those resulting from physical ablation of the entire cardiac neural crest in chick embryos, roughly extending from the otic vesicle to the fourth somite(Kirby and Waldo, 1995). Further studies into the mechanism of the reduced OFT in cardiac NCC ablations reveal an accompanying defective migration of the outflow tract myocardium from an anterior heart field (Yelbuz et al., 2003). We see no evidence for a change in cell number through cell death or cell proliferation changes in the OFT of our mutant embryos. Therefore, we suggest that NCCs require BMPRIA for proper colonization of the OFT. Furthermore, we speculate that the NCCs colonizing the OFT in wild-type embryos may secrete a factor downstream of BMPRIA signal transduction, lacking in our mutants, that stimulates migration of future outflow tract myocardial cells from the secondary heart field to form a fully mature OFT and right ventricle (Kelly et al., 2001; Mjaatvedt et al., 2001; Waldo et al., 2001).

Although other organogenesis defects also result from cardiac neural crest ablations, such as deficient thymus and parathyroid, the Bmpr1amutant embryos studied here died too soon (prior to E12.5) to address whether BMPR1A in NCCs is required for development of these organs. Nevertheless, our results complement earlier genetic evidence that BMP signaling is necessary for OFT septation. A hypomorphic allele of the gene encoding BMPRII, the binding partner of BMPRIA, also results in OFT defects(Delot et al., 2003). Overexpression of the BMP antagonist Noggin in chick heart and great vessels,expected to decrease effective BMP ligand levels in these tissues, results in OFT septation anomalies as well (Allen et al., 2001). In both the chick Noggin cardiac misexpression experiments and the mouse NCC Bmpr1a ablation studies, NCCs entered the OFT but failed to migrate as far towards the ventricles as in control embryos. Several BMP ligands are expressed in the early OFT, including BMP2,BMP4, BMP6 and BMP7 (Bitgood and McMahon,1995; Jones et al.,1991; Kim et al.,2001; Lyons et al.,1990). Together, these considerations suggest that local BMPs in and around the conotruncus signal through BMPRIA in the incoming NCCs to promote myocardial morphogenesis and NCC colonization of the septating truncus arteriosus.

The acute heart failure observed just before death of the Wnt1-Cre;Bmpr1a^flox/null embryos was not due to a failure of blood to flow through the OFT to the periphery. Persistent (unseptated)truncus arteriosus is not necessarily lethal until birth, when the fetus requires distinct pulmonary and systemic circulations. Although the aortic arch arteries were occasionally somewhat hypoplastic in the mutants and may have reduced flow, the lumens of neither these arteries nor the truncus arteriosus were occluded. Indeed, other mutants with hypoplastic aortic arch arteries survive beyond E12.5 (Abu-Issa et al., 2002; Lindsay and Baldini, 2001; Yanagisawa et al., 2003), by which time the Bmpr1a NCC mutants have died.

The primary cause of death in the Wnt1-Cre; Bmpr1a mutants is ventricular dysfunction and a consequent failure to pump an adequate supply of blood. The lethal phase is very narrow: mutants occur at Mendelian levels(25%) at E11, but are entirely absent by E13. The ventricular myocardium appears reduced shortly before death, particularly the outer, compact layer. Although the ventricles of mutant embryos look normal through E11.0, the ventricular myocardium shows significantly reduced cell proliferation by E10.5.

The thin-walled ventricular myocardial phenotype that we observe resembles phenotypes reported for loss-of-function mutations in several other mouse loci. These include the genes encoding WT1(Kreidberg et al., 1993),RARα (Kastner et al.,1997), RXRα (Sucov et al., 1994), α4 integrin(Yang et al., 1995), VCAM1(Kwee et al., 1995), FOG2(Tevosian et al., 2000),erythropoietin and the erythropoietin receptor(Wu et al., 1999), and others. These mutants are all general `knockout' mutations, in which the entire embryo has deficient gene activity. By contrast, our ablation of Bmpr1a was restricted to a very limited extracardiac population, with Bmpr1aactivity intact in myocardial cells. Thus, the myocardial phenotype is caused indirectly, specifically as a result of deficient BMP signaling in an extracardiac cell population that must function in a tissue interaction that regulates myocardial proliferation in the ventricles.

We observed a novel population of labeled cells contributing to the epicardium from E9.5, and propose that deficient BMPRIA activity in this lineage is responsible for the ventricular defects. Considerable genetic evidence suggests that a defective epicardium can lead to reduced myocardial proliferation. For example, Wt1 is expressed in epicardium, but not in myocardium, yet has a very similar thin compact myocardium phenotype;expression of Wt1 specifically in the epicardium rescued the thin-walled phenotype and death (Moore et al., 1999). Interference with retinoic acid signaling in the epicardium mimicked the thin-walled myocardium phenotype of Rxramutant embryos (Chen et al.,2002). In the VCAM1, FOG2 and α-4integrin mutants, the tight juxtaposition of epicardium with ventricular myocardium was abnormal, with corresponding thinning of the myocardium(Kwee et al., 1995; Tevosian et al., 2000; Yang et al., 1995). Embryological studies complement these results. Ablation of the proepicardial organ has a similar phenotype, and reduced migration to the heart results in patchy epicardium with reduced ventricular myocardium thickness underlying the deficient areas of the epicardium(Gittenberger-de Groot et al.,2000; Perez-Pomares et al.,2002; Poelmann et al.,2002). This reduction in ventricular wall thickness resulting from interference with the epicardium is caused primarily by reduced cell proliferation (Pennisi et al.,2002).

Thus, the thin-walled ventricular phenotype we observe in Wnt1-Cre;Bmpr1a mutants is very similar to those resulting from molecular or structural epicardial defects. Our analysis indicated that the epicardium in the mutants has no detectable structural problems, and that the expression of at least some epicardial markers was normal. Moreover, we found that the epicardial population of recombined (labeled) cells was indistinguishable between mutants and wild-type littermates. These data suggest that the relevant defect in the mutants is the specific lack of functional BMPRIA in this population of epicardial cells.

Two recent studies have demonstrated directly that the epicardium produces secreted factors that promote proliferation of the underlying ventricular myocardium. Chen and colleagues (Chen et al., 2002) found that cultured mouse epicardial cells produce trophic factors for myocardial cell proliferation, and production of this trophic activity depends on retinoid signal transduction in the epicardial cells. Stuckmann et al. (Stuckmann et al.,2003) also found that blocking retinoid signaling in the epicardium blocks production of a secreted myocardial proliferation factor, as does blocking erythropoietin signaling. Neither study identified the proliferation factor(s) produced by epicardial cells.

Our data lead us to propose a model for an essential role of BMPRIA signaling in an epicardial cell population in promoting ventricular development (Fig. 9A-C). An extracardiac lineage of cells (reflecting Wnt1-Cre recombination)migrates to the epicardium, and via BMPRIA transduces a BMP signal. We found that BMP4 (a known ligand) is expressed at high levels in cells adjoining the epicardium, though other work has demonstrated expression of Bmp5, Bmp6,Bmp7 and Bmp10 in the myocardium at many stages(Dudley and Robertson, 1997; Kim et al., 2001; Neuhaus et al., 1999; Solloway and Robertson, 1999). Transduction of BMP signals by BMPRIA in these epicardial cells would in turn stimulate production within the epicardium of an unidentified proliferation signal for the underlying ventricular myocardium. As we can detect only a limited contribution of recombined cells to the epicardium, this population probably acts upon other cells of the epicardium to stimulate production of this myocardial proliferation signal(s). Thus, a limited number of recombined cells may lead to a defect in signaling from a much larger region of the epicardium. As a result, the Wnt1-Cre; Bmpr1a^flox/nullembryos would produce reduced proliferation signal(s) and consequently experience a critical underproliferation of the ventricular myocardium. Reduced compact and trabeculated myocardium precludes sufficient blood flow for the metabolic demands of the rapidly growing embryo. Embryonic lethality then results primarily from ventricular failure, caused by deficient myocardial maturation. The ventricular deficiency of the mutants analyzed here may be compounded by restricted flow through the unseptated outflow tract and hypoplastic aortic arch arteries. Within the ventricular myocardium itself,BMPRIA has been proposed to transduce a myocyte growth signal(Gaussin et al., 2002), though in our mutants the myocardium per se is wild type.

An alternative model for the defective ventricular proliferation we observed in the Wnt1-Cre; Bmpr1a^flox/null embryos involves the NCCs that migrate around the pharyngeal tissues to populate the outflow tract (Fig. 9D,E). We observed that the colonization of the developing OFT by NCCs was much reduced when these cells lacked BMPRIA; notably, the NCC population extended much less far down the OFT toward the ventricles. It is possible that NCC derivatives colonizing the OFT require BMPRIA for production of a secreted mitogen for the myocardium. In the absence of the receptor, there would be less of this mitogen produced, with a consequent under-proliferation of the myocardium. Some evidence consistent with this model comes from studies in the chicken,where NCC ablation leads to defective ventricular myocardium(Waldo et al., 1999). Although the resulting phenotypes varied widely and showed a general increase in myocardial proliferation (in contrast to the consistently reduced proliferation observed here), these results support the view that the neural crest in some way exerts an early influence on organization of the ventricular myocardium. A molecular mechanism for this influence is difficult to envision,in that it must involve very long-range signaling. However, direct co-culture of the myocardium with ventral pharynx (i.e. without NCCs) resulted in changes in myocardial Ca²⁺ signaling and increased proliferation(Farrell et al., 1999); this action of the pharynx was repressed by FGF function blocking antibody. This may indicate that signaling from the pharynx has a negative effect on myocardial development, mediated by FGFs, that is normally alleviated by NCCs as they colonize the pharyngeal arches and OFT. To extrapolate this model to the experimental situation studied here, one might envision that reduced NCC migration into the developing OFT in the Wnt1-Cre;Bmpr1a^flox/null embryos allows more pharyngeal FGF to perturb ventricular development. Although we cannot rule out such an interaction between NCCs in the pharynx and FGF (or other) signals, we suggest that a disrupted epicardial signaling cascade is a more likely cause of the myocardial proliferation defect seen in our mutants.

Although Wnt1-Cre expression has been associated exclusively with the neural crest and its derivatives(Brault et al., 2001; Chai et al., 2000; Jiang et al., 2000), the epicardium of Wnt1-Cre; R26R embryos contained a small proportion of recombined cells. This indicates that they are derived from a Cre-expressing lineage. Although we were unable to detect such cells in every specimen(probably for technical reasons), control experiments indicated that the labeling was not a staining artifact. These labeled cells later populated the same tissues as epicardial derivatives (coronary arteries and scattered cells in the ventricular wall). Because these epicardial cells clearly underwent epithelial-mesenchymal transition (EMT) to clonally populate these structures,it is possible that this population represents a fraction of the epicardium that is competent to undergo EMT rather than remaining an obligate part of the epicardial mesothelium.

We observed a Wnt1-Cre lineage of cells in the septum transversum adjacent to the ventricles at E9.5. This is the location of the proepicardial organ, the source of epicardial cells. We therefore suspect that the labeled cells of the epicardium came from this tissue, though it is also possible that they have migrated in from a known neural crest domain elsewhere. We are uncertain whether these labeled cells are of neural crest origin, or if they represent a hitherto undetected independent expression domain of Wnt1-Cre. However, we saw exactly the same distribution of labeled cells when we used PO-Cre as an independent neural crest lineage marker. Moreover, using the sensitive assay of RT-PCR, we never detected Cre expression in non-dorsal tissues dissected from Wnt1-Cre embryos from E9.5-10.5 (during which time the epicardium forms). These considerations lead us to suspect that the labeled epicardial cells are of neural crest origin,though they probably arrive via the proepicardial organ.

Although an epicardial NCC lineage has not been otherwise observed in mice or chicks, recent evidence suggests NCCs contribute to ventricular myocardium in zebrafish (Li et al., 2003; Sato and Yost, 2003). In mice and chickens with NCC defects, myocardial defects are seen before NCCs are known to populate the outflow tract, suggesting that NCCs produce factor(s)involved in myocardial development(Creazzo et al., 1998). Moreover, Pax3 expression in NCCs rescues the reduced myocardial tissue phenotype of Pax3-null mutants(Li et al., 1999). We suggest that lesions in an early population of NCC derivatives, migrating into the epicardium and/or outflow tract, account for previously observed myocardial defects associated with neural crest perturbations. Overall, our work demonstrates that BMPRIA is dispensable for most aspects of early neural crest development, and identifies a novel, crucial role for BMPRIA signal transduction in extra-cardiac cells during the development of the outflow tract and ventricular myocardium of the heart.

We thank B. Hogan and the anonymous reviewers of an earlier draft for helpful comments on the manuscript, and M. Kirby for insightful discussions. M. Berrong, J. Goodwin and A. Lawrence provided technical assistance. Mouse strains were generously provided by B. Hogan (Bmp4-lacZ), A. McMahon(Wnt1-Cre), P. Soriano (R26R) and K. Yamamura(PO-Cre). The 2H3 antibody, developed by J. Dodd and T. Jessell, was obtained from the Developmental Studies Hybridoma Bank maintained by the University of Iowa, Department of Biological Sciences, Iowa City, IA 52242. R.W.S. is supported by an NINDS fellowship (NS42486). This work was supported by NIH grants to J.K. (HD39948 and DE13674).