Bone morphogenetic protein receptor 1A signaling is dispensable for hematopoietic development but essential for vessel and atrioventricular endocardial cushion formation

The establishment of the circulatory system including blood, vasculature and the heart is highly coordinated and a prerequisite for embryonic development. A failure of proper development in any of these cell lineages leads to embryonic lethality. Recent studies have established that cells of the circulatory system develop from FLK1⁺ mesoderm. FLK1 (KDR -Mouse Genome Informatics) is a receptor tyrosine kinase that binds vascular endothelial growth factor (VEGF). In primitive streak stage embryos, Flk1 expression is first detectable in the extra-embryonic mesoderm and in the paraxial-lateral embryonic mesoderm(Yamaguchi et al., 1993). Subsequently, its expression is confined to endothelial cells and to the developing endocardial tube. Consistent with this expression pattern, Flk1-deficient mice die between E8.5 and E9.5 with defects in yolk-sac blood-island formation and vasculogenesis(Shalaby et al., 1995). Cell replating and transplantation experiments have demonstrated that FLK1⁺ cells isolated from differentiated embryonic stem (ES) cells or embryos generate hematopoietic, endothelial and vascular smooth muscle cells. For example, ES- and embryo-derived FLK1⁺ cells, but not FLK1^- cells, are enriched for blast colony forming cells (BL-CFCs),an in vitro equivalent of the hemangioblast(Chung et al., 2002; Faloon et al., 2000; Huber et al., 2004). Ema et al. (Ema et al., 2003) showed that BL-CFCs are multipotential, such that blast colonies from single FLK1⁺ cells could generate smooth muscle cells, as well as hematopoietic and endothelial cells in vitro. Similarly, Yamashita et al.(Yamashita et al., 2000)reported that FLK1⁺ cells could generate both smooth muscle and endothelial cells in vitro and in vivo. Subsequent fate-mapping studies of FLK1⁺ cells by using Flk1^+/Cre and Rosa26-LacZ reporter (R26R) mice showed that FLK1⁺ cells could also contribute to cardiac and skeletal muscle (Motoike et al., 2003). Collectively, these studies indicate that FLK1⁺ mesoderm generates cells of the circulation system.

Using the in vitro differentiation model of ES cells, we recently showed that BMP4 was able to induce FLK1⁺ cells(Park et al., 2004). In serum-free conditions, BMP4 activated the SMAD1/5 pathway. Inhibition of the SMAD1/5 pathway reduced the generation of FLK1⁺ cells. Consistent with the notion that BMP4 is crucial for the generation of FLK1⁺mesoderm, Bmp4-deficient mice die between E6.5 and E9.5 with defects in mesoderm formation and patterning. Those that survive up to E9.5 show severe defects in blood islands (Winnier et al., 1995). Additionally, mice lacking the BMP receptor type IA(ALK3; BMPR1A - Mouse Genome Informatics) fail to complete gastrulation, and none survive past the E9.5 stage (Mishina et al., 1995). Smad1-deficient mice display early embryonic lethality and die between E9.5 and E10.5 owing to the failure of chorioallantoic fusion (Lechleider et al.,2001; Tremblay et al.,2001). Smad5-deficient mice die between E9.5 and E11.5,displaying anemia and disorganized vessels, despite the formation of the primitive vascular plexus (Chang et al.,1999; Yang et al.,1999). Collectively, these studies have shed light on BMPs and their downstream signaling in embryonic development. However, the early lethality with complex abnormalities precludes placing the precise window in which they play a role in hematopoietic and endothelial cell development.

To better understand the requirement of BMP signaling in blood and endothelial cell development, we inactivated Alk3 in FLK1⁺mesoderm by using Flk1^+/Cre knock-in mice. Our studies demonstrate an essential role for signaling through ALK3 in vessel remodeling and maturation. Mutant embryos died between E10.5 and E11.5, with defects in vessel remodeling and smooth muscle cell formation and/or recruitment to the dorsal aorta. Mutant embryos also displayed defects of endocardial cushion formation in the AVC. We propose that ALK3 signaling in FLK1⁺mesoderm is crucial for proper vessel formation and heart development.

To generate FLK1⁺ cell-specific Alk3 or Smad4CKO mice, Alk3^{floxed/floxed} or Smad4^{floxed/floxed} mice were crossed with Flk1^+/CreAlk3^+/floxed or Flk1^+/CreSamd4^+/floxed mice, respectively. Flk1^+/Cre mice (a gift of Dr Sato at the University of Southwestern Medical School, TX) were made by a knock-in strategy wherein Cre recombinase is expressed under the control of the endogenous Flk1promoter (Motoike et al.,2003). The generation of Alk3^{floxed/floxed} and Smad4 ^{floxed/floxed} mice and genotyping were previously described (Mishina et al.,2002; Yang et al.,2002).

Embryos were fixed in 4% paraformaldehyde (PFA), paraffin embedded and sagittally sectioned at 5 μm. For antigen retrieval, sections on the slides were incubated with 0.1% trypsin for 15 minutes at 37°C and washed with phosphate-buffered saline (PBS). The sections were immersed in blocking buffer(5% normal goat serum and 0.1% Tween 20 in PBS) for 40 minutes at room temperature, followed by overnight incubation at 4°C with anti-PECAM1 antibody (1:30, Pharmingen). After incubation with anti-biotinylated rat IgG(1:200, Zymed), the sections were further incubated with anti-SMC actin antibody (1:200, Sigma) and Cy5-conjugated streptavidin (1:400). For cross-sections, frozen cryosections of 10 μm were stained with antibodies as described above without an antigen retrieval step.

Embryos were fixed overnight in Dent's fixative (methanol and DMSO, 4:1) at 4°C, followed by blocking in PBS containing 5% normal goat serum, 0.1%Tween 20 and 2% non-fat skim milk) for 2 hours at room temperature. After incubation with anti-PECAM1 antibody (1:300) and alkaline phosphatase-conjugated anti-Rat IgG (1:250, Zymed), signals were detected with the NBT/BCIP kit (Promega). For yolk sac, 4% PFA was used instead of Dent's fixative, blocking and antibody incubation were performed in PBS containing 5%normal goat serum and 1% DMSO (Morikawa and Cserjesi, 2004).

RNAs from tissues were extracted by using Trizol (Gibco). cDNA was generated using the Invitrogen kit. Expression of genes indicated in the text was measured by real-time qRT-PCR, and normalized by Gapdhexpression. Each experiment with duplicates was performed at least three times. For primers, see Table 1.

For ultrastructural analysis, embryos were fixed in 2% PFA/2.5%glutaraldehyde (Polysciences, Warrington, PA) in 100 mM phosphate buffer (pH 7.2) overnight at 4°C followed by post fixation in 1% osmium tetroxide(Electron Microscopy Sciences, Fort Washington, PA) for 1 hour. Osmium tetroxide was removed and samples were stored in 100 mM phosphate buffer at 4°C pending genotype analysis. Samples were then rinsed extensively in distilled H₂O to remove excess phosphate prior to en bloc staining with 1% aqueous uranyl acetate (Ted Pella, Redding, CA) for 1 hour. Samples were washed, dehydrated in a graded series of ethanol, and embedded in Eponate 12 resin (Ted Pella). Sections of 95 nm were cut, stained with uranyl acetate and lead citrate and viewed on a JEOL 1200 EX transmission electron microscope(JEOL USA, Peabody, MA).

Hematopoietic colonies were generated as described previously(Faloon et al., 2000; Palis et al., 1999). Briefly,yolk sacs were isolated, dissociated in collagenase with 20% FBS, and the resulting cells were replated in 1% methyl cellulose containing hematopoietic cytokines. The dissociated fetal liver cells from E10.5 embryos were replated in Methocult M3434 (Stemcell Tech, CA). Hematopoietic colonies were counted 4-7 days later.

The results of real-time quantitative reverse transcription PCR were analyzed by Student's t-test. P<0.05 was considered significant.

To inactivate Alk3 in FLK1⁺ cells, we employed two mouse lines, Alk3^{floxed/floxed} mice, in which exon 2 is flanked by loxP sites (Mishina et al.,2002) and Flk1^+/Cre mice, in which the Cre gene is introduced into one allele of the Flk1 locus(Motoike et al., 2003). The expression of Cre recombinase in these mice is under the control of the endogenous Flk1 promoter, providing the specific deletion of floxed genes in FLK1⁺ cells. To generate Alk3 CKO mice, we first set up matings between Flk1^+/Cre and Alk3^{floxed/floxed} mice. Flk1^+/CreAlk3^+/floxed mice were subsequently mated with Alk3^{floxed/floxed} mice to generate Flk1^+/CreAlk3^{floxed/floxed} CKO mice(Fig. 1A). The expected ratio of obtaining Alk3^{floxed/floxed}, Alk3^+/floxed, Flk1^+/CreAlk3^+/floxed,Flk1^+/CreAlk3^{floxed/floxed} mice is 1:1:1:1,respectively. Genomic DNA PCR was used to verify the expected genotypes(Fig. 1B).

We initially analyzed 147 pups at 3-4 weeks of age. As shown in Table 2, no live Flk1^+/CreAlk3^{floxed/floxed} (Alk3 CKO)pups were detected, although mice with all three other potential genotypes were present. This indicated that inactivation of Alk3 in FLK1⁺ mesoderm caused embryonic lethality. To identify the stage at which embryos died, we examined embryos obtained from scheduled matings. Alk3 CKO embryos were morphologically indistinguishable from control embryos at E9.5 (Table 1 and Fig. 1D). However, at E10.5, 21 out of 69 Alk3 CKO embryos were moribund and displayed pale yolk sacs(Table 2 and Fig. 1F′). The remaining 48 Alk3 CKO embryos were morphologically indistinguishable from wild-type embryos (Fig. 1F). At E11.5, all Alk3 CKO embryos had pale yolk sacs and showed hemorrhage throughout the trunk region of the embryo(Fig. 1H,J). The mutant hearts were smaller than those of the control embryos, and the cavity between the pericardial sac and the heart was distended in the mutant embryos (not shown). At E12.5, Alk3 CKO embryos were totally pale, although some of them still had red fluid within the amnion (not shown). The mutants were much smaller than the controls, and digit formation did not occur in these mutants,indicating that they failed to develop beyond E11.5. These initial observations indicated that Alk3 CKO embryos died between E10.5 and E11.5, and suggested that they had defects in the circulatory system.

As BMP4/SMAD1/5 signaling has been implicated to be crucial for hematopoietic development (Chang et al.,1999; Johansson and Wiles,1995; Kanatsu and Nishikawa,1996; Liu et al.,2003; Park et al.,2004), it was expected that Alk3 CKO embryos might display defects in hematopoiesis. Initial observation of the E9.5 mutants,however, indicated that red blood cells were present(Fig. 1D). To determine if the level of hematopoietic progenitors present in the mutant yolk sacs was diminished, we performed hematopoietic progenitor assays. Yolk sacs from E9.5 embryos were isolated, dissociated and plated into methylcellulose supplemented with hematopoietic cytokines. As shown in Fig. 2A, similar numbers of erythroid and macrophage colonies developed from the Alk3 CKO and control yolk sacs. The hematopoietic colonies that developed from the Alk3 CKO yolk sacs were confirmed by Giemsa staining (not shown). To determine whether incomplete Alk3 deletion could have resulted in the generation of hematopoietic colonies from the Alk3 CKO yolk sac,FLK1⁺ cells or TER119⁺ (LY76⁺ - Mouse Genome Informatics) cells (erythrocytes) from E9.5 yolk sacs were FACS sorted and subjected to genomic PCR. As shown in Fig. 2B, the correctly excised band was detectable in DNA obtained from the Alk3 CKO yolk sac-derived FLK1⁺ cells or TER119⁺ cells. By densitometry, we estimated that over 95% of the floxed exon 2 was correctly deleted in the Alk3 CKO cells. Moreover,when the Alk3 CKO yolk sac-derived hematopoietic colonies were picked and subjected to genomic PCR, a PCR band corresponding to the floxed exon 2 of Alk3 was present in the controls, but not in the mutants,demonstrating that Alk3 was correctly deleted in all mutant-derived hematopoietic colonies (Fig. 2C). Previous studies indicate that E9.5 yolk sacs contain predominantly definitive hematopoietic progenitors(Palis et al., 1999). To further validate that definitive hematopoiesis was indeed unaffected in the Alk3 CKO embryos, we dissected out E10.5 fetal liver and performed hematopoietic colony assays. As shown in Fig. 2D, Alk3 CKO fetal liver generated a similar number of hematopoietic colonies compared with controls. Collectively, the relatively normal frequency of hematopoietic progenitors present in E9.5 mutant yolk sacs and in E10.5 mutant fetal livers argues that ALK3 signaling is not constitutively required for hematopoietic development.

The finding that Alk3 CKO embryos showed normal hematopoiesis was rather unexpected, as BMP4 is thought to be crucial for hematopoietic development. To determine whether BMP, TGFβ or activin signaling is involved at all in hematopoietic development, we generated Smad4 CKO mice by crossing Smad4^{floxed/floxed}(Yang et al., 2002) and Flk1^+/CreSmad4^+/floxed (see Table S1 and Fig. S1 in the supplementary material). SMAD4 is the common SMAD that mediates signaling by all members of the TGFβ superfamily. Smad4 CKO mutants died earlier than Alk3 CKO mutants, between E9.5 and E10.5(see Table S1 in the supplementary material). Performing a hematopoietic progenitor assay at E8.5 revealed that the number of hematopoietic colonies in Smad4 CKO embryos was decreased by about 50% compared with controls(Fig. 2E, upper). E9.5 yolk sac replating generated similar results (Fig. 2E, lower). Hematopoietic colonies that developed from the Smad4 CKO and wild-type yolk sacs were indistinguishable in size and morphology. A complete deletion of floxed exon 8 of Smad4 was confirmed by genomic PCR on yolk sac-derived hematopoietic colonies (see Fig. S1 in the supplementary material). Collectively, these studies demonstrate that BMP, TGFβ or activin signaling contributes to hematopoietic differentiation from FLK1⁺ mesoderm; however, such signaling is not mediated by ALK3.

The Alk3 CKO mutants that died by E10.5 exhibited anemic yolk sacs(Fig. 1F′). As hematopoiesis was undisturbed, vessel development might be defective. To test this possibility, we analyzed yolk sac vasculature by whole-mount staining using anti-PECAM1 (CD31) antibody, a marker for endothelial cells. As shown,wild-type embryos displayed large vessels with extensive branching and a well-developed capillary network (Fig. 3A). However, Alk3 CKO E10.5 yolk sacs showed only the primary vascular plexus of poorly developed vascular channels(Fig. 3B). Similarly, abnormal CD31 staining was obvious in the Alk3 CKO mutant brain vasculature(Fig. 3D). From these observations, we conclude that ALK3 is required for vessel remodeling.

To better understand the molecular mechanisms involved in ALK3-mediated vessel remodeling, we examined the expression of members of the Id family(inhibitor of DNA binding/differentiation). Id genes are important for angiogenesis and are well-known direct downstream targets of BMP signaling(reviewed by Ruzinova and Benezra,2003). We first evaluated the expression of all Id genes in the yolk sac using quantitative real-time reverse transcription PCR (qRT-PCR). As shown in Fig. 3E, Id2was expressed at high levels in E8.5-E10.5 yolk sac, while Id4 was the least expressed. Importantly, the expression of all Id genes in Alk3 CKO yolk sacs was decreased by ∼40-60% compared with control yolk sacs, suggesting that Id genes are downstream targets of ALK3 signaling in mediating vessel remodeling (Fig. 3F). Additionally, we observed that the expression of urokinase-type plasminogen activator (Plau) and peptidase inhibitor 1(Serpine1), both of which are important mediators of angiogenesis,was upregulated in Alk3 CKO yolk sacs(Fig. 3F).

Alk3 CKOs that survived to E11.5 were morphologically normal at E10.5, but subsequently succumbed to death displaying abdominal hemorrhage(Fig. 1H,J). To further characterize the mutant phenotype, E10.5 wild-type and Alk3 CKO embryos (excluding those that showed the pale yolk sac phenotype) were subjected to whole-mount CD31 staining. Overall, the mutant embryos did not display distinct abnormalities in CD31 staining in the yolk sac and embryo proper (not shown). Id gene expression in these Alk3 CKO mutant yolk sacs was comparable with that of the controls (not shown). Nonetheless, they often showed dilated vessels in the brain and abnormal branching in the trunk,confirming a defect in angiogenesis (Fig. 4B,D). As they all displayed abdominal hemorrhage, we reasoned that the vessel integrity could also be defective in these mutants. To explore this possibility, E10.5 wild-type and Alk3 CKO embryos (excluding those that showed the pale yolk sac phenotype) were sectioned and subjected to immunostaining with anti-CD31 and anti-smooth muscle cell (SMC) α-actin antibody, a marker for vascular SMCs. We examined at least 15 different regions in cross-sections (Fig. 4E-J) and five different regions in sagittal sections (not shown)along the dorsal aorta per embryo. As discussed earlier, mutant embryos appeared to have no obvious defects in vessel structure itself, as judged by CD31 staining (Fig. 4I). However, Alk3 CKO mutants appeared to have fewer SMCs around the dorsal aorta, compared with wild-type embryos(Fig. 4J). Subsequently, E11.5 dorsal aortas were isolated, RNA was prepared and subjected to qRT-PCR. Alk3 CKO mutants expressed similar levels of Cd31 when compared with controls (Fig. 4K, upper panels). However, they all expressed decreased levels of both SMC α-actin and myosin heavy chain (Mhc) when compared with the controls, confirming a decreased number of SMCs around the dorsal aorta.

To better characterize the blood vessel defects in Alk3 CKO embryos, we examined E10.5 wild-type and mutant embryos (excluding those that showed the pale yolk sac phenotype) by transmission electron microscopy (TEM)(Fig. 4L). TEM studies revealed obvious abnormalities in the mutant dorsal aortas. Consistent with SMCα-actin staining and qRT-PCR, the SMC number was diminished. Furthermore, the SMCs that were present did not make close contacts with adjacent endothelial cells. Although we observed relatively normal CD31 staining, there were frequent breaks in the endothelial cell layer in the mutants, while no breaks were found in the wild-type dorsal aortas(Fig. 4L). We further examined genes involved in vessel stability/integrity using the qRT-PCR. Angiopoietin 1 and 2 (Angpt1 and Angpt2), and Tie2 expression levels were comparable between control and Alk3 CKO embryos(Fig. 4K, middle panels and not shown). However, the expression of integrin αV and β3 was reduced in the mutants, compared with controls(Fig. 4,K lower panels). Importantly, we found decreased expression of Pten, but increased expression of Vegf in the Alk3 CKO embryos(Fig. 4K, middle and lower panels). All Smad4 CKO mutants displayed angiogenesis defects, as only the primary vascular plexus was observed in E9.5 and E10.5 yolk sacs(Fig. 5A). In these Smad4 CKO yolk sacs, the expression of all Id genes was greatly reduced compared with controls (Fig. 5B). The level of reduction was even greater in Smad4 CKO yolk sacs compared with Alk3 CKO yolk sacs. Collectively, we conclude that ALK3, SMAD4 and ID proteins are crucial mediators of vessel remodeling. Moreover, defects in SMC recruitment and/or differentiation, reduced interactions between endothelial cells, and/or increased VEGF could contribute to the abdominal hemorrhage in Alk3 CKO embryos.

Previous studies have demonstrated that FLK1⁺ cells contribute to cardiac muscle as well as endothelial cells and that Flk1-deficient mice displayed an endocardial defect(Shalaby et al., 1995). Thus,we examined the hearts of Smad4 and of Alk3 CKO embryos by Hematoxylin and Eosin staining. At E9.5, trabeculae in the ventricles and endocardial cushion in the AVC begin to form(Fig. 5C). As shown in Fig. 5D, trabeculae in the ventricles was much thinner and AVC endocardial cushion was not present in Smad4 CKO hearts. At E10.5, trabeculae develop further and endocardial cushions can be detected in both AVC and outflow track (OFT)(Fig. 6A). When E10.5 Alk3 CKO hearts (excluding those with the pale yolk sac phenotype)were examined, we detected no cushion cells in the AVC(Fig. 6A). However, development of the OFT cushion appeared to be normal. This observation raised the issue whether FLK1⁺ cells contribute differently to the cushion cells of the AVC and OFT. To address this, Flk1^+/Cre mice were crossed with Rosa26R-LacZ mice (Soriano et al., 1999). At E10.5, the majority of lacZ-positive cells in the AVC were present in the cushion cell layer and in the endocardium. In the OFT, lacZ-positive cells were also found in the endocardium (Fig. 6B). By E12.5, all AVC-cushion cells were lacZ positive,while OFT-cushion cells were composed of both lacZ positive and negative cells. These results indicate that AVC cushion cells are derived solely from FLK1⁺ cells, while OFT cushion cells could originate from non-FLK1⁺ cells, presumably from the neural crest.

To better understand defects in the AVC of the Alk3 CKO embryos,E10.5 hearts were first stained with Alcian Blue(Fig. 6C). Both wild-type and Alk3 CKO formed the cardiac jelly, as indicated by blue staining,although the staining pattern and intensity in the Alk3 CKO mutants appeared to be weaker compared with the wild-type controls. Second, genes important for the EMT were examined. Specifically, E10.5 hearts were dissected free of the OFT and the right ventricle and subjected to gene expression studies. Alk3 CKO hearts showed a decreased expression of Snai1 and Sox9, but a comparable expression of VE-cadherin and Nfatc1, compared with wild-type controls(Fig. 6D). Both Msx1and Twist1 expression was downregulated in the mutant hearts, but Tgfb2 expression was not changed(Fig. 6D).

In this study, we have delineated the function of ALK3 in the differentiation of FLK1⁺ mesoderm-derived cell lineages, which could not be revealed from the conventional null animals. Specifically,sustained activity of ALK3 is crucial for vessel remodeling, vessel integrity and the formation of the endocardial cushions of the AVC. Our results also indicate that signaling through ALK3 is dispensable for hematopoiesis.

There is great interest in identifying inductive signals that initiate the hematopoietic program. Xenopus animal cap explants, avian, mouse ES/EB cell culture and knockout studies have suggested that members of TGFβ and fibroblast growth factor (FGF) families are crucial for hematopoietic development (Adelman et al.,2002; Chang et al.,1999; Dickson et al.,1995; Faloon et al.,2000; Flamme et al.,1995; Flamme and Risau,1992; Huber et al.,1998; Johansson and Wiles,1995; Kanatsu and Nishikawa,1996; Liu et al.,2003; Maeno et al.,1996; Oshima et al.,1996; Pardanaud et al.,1996). Among the members of the TGFβ superfamily of growth factors, numerous studies indicate that BMP4 is important for hematopoietic development. We recently demonstrated that BMP4 is crucial for induction of brachyury⁺ mesoderm from ES cells and FLK1⁺ cells from brachyury⁺ cells. In combination with VEGF, BMP4 significantly enhanced the production of SCL⁺ hematopoietic cells(Park et al., 2004). Even though these studies suggested that BMP4 contributed to hematopoietic development, we could not position whether BMP4 functioned in the induction of FLK1⁺ mesoderm or whether BMP4 was independently required for the induction of SCL⁺ hematopoietic progenitors from FLK1⁺mesoderm. To further elucidate the function of BMP/BMP receptor-mediated signaling during the embryonic organogenesis, we generated Alk3 CKO mice in which Alk3 was inactivated in FLK1⁺ cells. The analyses of Alk3 CKO mutant yolk sacs and fetal livers indicated that blood cells developed normally in these mutants. Although we cannot exclude the possibility that other BMP type I receptors, such as ALK2 or ALK6,compensated for the loss of ALK3 activity, our studies provide genetic evidence that the BMP4-ALK3 axis is not crucial for hematopoietic development. Our observation is consistent with a recent study that hematopoietic development occurred normally in zebrafish when ALK3 signaling was blocked by an inducible dominant-negative ALK3 expression after the onset of gastrulation(Pyati et al., 2005).

The Alk3 CKO hematopoietic phenotype suggested that non-BMP signaling, perhaps by signaling through TGFβ or activin receptors could contribute to hematopoietic cell development. Tgfb1^-/- or Tgfbr2^-/- mice show defects in yolk sac hematopoiesis,although Tgfbr1-deficient mice display enhanced hematopoiesis(Dickson et al., 1995; Larsson et al., 2001; Oshima et al., 1996). Additionally, activin A could increase hematopoietic development in the presence of BMP4 and VEGF (Park et al.,2004). To determine if any members of the TGFβ growth factor superfamily and their receptors are involved in hematopoietic specification,we analyzed hematopoietic development in the Smad4 CKO embryos, and found that mutant embryos were morphologically normal at E8.5. However, E8.5 and E9.5 mutant yolk sacs contained one half as many hematopoietic progenitors as controls. This clearly indicates that SMAD4-dependent signaling contributes to the formation of hematopoietic cells from FLK1⁺ mesoderm.

Two groups reported that FGF receptor 1 null (Fgfr1^-/-)mice died in mid-gestation and displayed abnormal mesoderm patterning(Deng et al., 1994; Yamaguchi et al., 1994). Subsequently, we demonstrated that Fgfr1^-/- ES cells differentiated up to the Brachyury⁺ cell stage but failed to express normal level of FLK1, suggesting a role for FGFR1 in the early stages of hematopoiesis (Faloon et al.,2000). Treatment of wild-type ES cells with FGF2, a ligand for FGFR1, enhanced the generation of FLK1⁺ cells(Faloon et al., 2000). To determine if the remaining hematopoietic activity in the Smad4 CKO mutant is due to FGF signaling, we inactivated Fgfr1 or Fgfr1 and Fgfr2 in FLK1⁺ cells. We found that Fgfr1 CKO or Fgfr1/2 double CKO mice were born alive and that E13.5 Fgfr1/2 double CKO fetal liver contained similar levels of hematopoietic progenitors when compared with wild-type controls (see Fig. S2 in the supplementary material). Collectively, we conclude that FGF receptor signaling is required to generate FLK1⁺ mesoderm, but is dispensable for subsequent hematopoietic development. In the future, the nature of the residual hematopoietic activity in Smad4 CKO embryos still needs to be elucidated.

Vessel formation in the developing embryo requires vasculogenesis and angiogenesis with successful recruitment of supporting cells such as smooth muscle cells (SMCs) and pericytes. Herein, we demonstrate that Alk3CKO embryos display defects in vessel remodeling, as well as vessel maturation/integrity. It is not clear why some of the mutants showed such a severe angiogenesis phenotype, while the remaining mutant embryos showed defects in vessel maturation and/or integrity. We suggest that it could be due to mixed genetic background or due to efficiency and/or timing of the recombination of Alk3 floxed alleles. Nevertheless, Alk3 CKO mutant embryos that died by E10.5 all formed a primary vascular plexus, but failed to complete angiogenesis. This observation provides the first genetic evidence that ALK3 is crucial for vessel remodeling. Our results are consistent with previous observations that BMP6 or BMP2 could promote endothelial tube formation and migration of endothelial cells in vitro(Langenfeld and Langenfeld,2004; Valdimarsdottir et al.,2002). Activation of ALK3 induced tube formation and the migration of endothelial cells, in part, by stimulating the expression of Id1(Valdimarsdottir et al.,2002). Indeed, we found that expression of Id genes(Id1-Id4) was decreased by 40-60% in the mutant yolk sacs, compared with the controls. We propose that ALK3-mediated Id gene activation in endothelial cells is crucial for vessel remodeling in the yolk sac.

Alk3 CKO mutant embryos that died by E11.5 appeared to develop normal blood vessels in the yolk sac and in the embryo proper. Nonetheless,these mutants displayed dilated vessels in the brain and abnormal branching in the trunk and obvious vessel abnormalities when examined by TEM. The most striking phenotype was that all the Alk3 CKO embryos showed abdominal hemorrhage by E11.5, suggesting that vessel integrity is defective. It has been reported that several pathways, including VEGF, PDGF/PDGFR, ANG1/TIE2 and ALK-1/endoglin have important functions in vessel integrity. PDGFB could recruit mural cells to endothelial cells(Hirschi et al., 1999), and mice lacking Pdgfb or Pdgfrb show a reduced degree of vascular SMCs and pericytes in vessels(Hellstrom et al., 1999; Lindahl et al., 1997). More recently, inactivation of Pdgfb in endothelial cells led to decreased recruitment of pericytes to vessels(Bjarnegard et al., 2004). In the present study, staining with antibodies to CD31 and SMC-α actin revealed that the endothelial cell layer in the dorsal aortas of Alk3CKO mutants was surrounded by a reduced number of SMCs. This observation was confirmed by TEM, which showed that SMCs in the Alk3 CKOs did not make close contacts with endothelial cells. More importantly, we found frequent breaks in the endothelial cell layer on the dorsal aorta wall in the Alk3 CKO mutants, while there were no obvious breaks found in the dorsal aorta of the controls. Expression of integrin αV and β3 was decreased in the mutants, although we did not find differences in angiopoeitin 1, angiopoeitin 2 and Tie2 expression. Importantly, Alk3 CKO embryos expressed decreased Pten with a concomitant increase in Vegf expression. There is emerging evidence that BMP signaling acts in the upstream of the PTEN/PI3K/VEGF pathway. Treatment of BMP2 can stabilize PTEN protein (Waite and Eng,2003). Loss of Alk3 not only results in the inhibition of PTEN activity, but also the activation of AKT, downstream of PI3K(He et al., 2004). PTEN can inhibit VEGF expression by inhibiting the PI3K/AKT pathway(Jiang et al., 2000; Pore et al., 2003). Furthermore, a recent study has demonstrated that abrogation of Ptenin TIE2⁺ cells leads to defects in angiogenesis and that these mutants display increased level of genes involved in vessel development including Vegf (Hamada et al.,2005). Collectively, we propose that ALK3 in FLK1⁺cells may regulate vessel integrity through the PTEN/PI3K/VEGF pathway.

As for SMC defects in Alk3 CKO embryos, it is not clear whether the Alk3 CKO vessel phenotype is due to Alk3 deficiency in endothelial cells or whether SMC generation per se is defective. One possible explanation would be that Alk3-deficient endothelial cells fail to recruit SMCs to the dorsal aorta. However, we did not observe misplaced SMCs around the dorsal aorta. Furthermore, Pdgfb was equally expressed in wild-type and mutant dorsal aorta (not shown), although we cannot exclude the possibility that other pathways could be regulated by ALK3. Alternatively,absence of Alk3 could lead to defective SMC generation, resulting in insufficient number of SMCs for vessel integrity. Indeed, we found decreased expression of SMC α-actin and myosin heavy chain in mutant dorsal aorta. Recent studies have shown that vascular endothelial cells or FLK1⁺endothelial cells can (trans)differentiate into SMCs during development(DeRuiter et al., 1997; Ema et al., 2003; Frid et al., 2002; Yamashita et al., 2000). Therefore, ALK3 could be required for SMC generation and/or transdifferentiation from endothelial cells.

Mice lacking Tgfb1, Tgfbr1 (Alk5), Alk1 or Tgfbr2 all display varying degrees of yolk sac angiogenesis defects(Dickson et al., 1995; Larsson et al., 2001; Oh et al., 2000; Oshima et al., 1996). A similar phenotype has also been reported for Smad5^-/- mice(Chang et al., 1999; Yang et al., 1999). Intriguingly, recent studies provide compelling evidence that ALK5 is exclusively expressed on SMCs (Seki et al., 2006). Thus, angiogenesis defects seen in Alk5mutant mice are most probably due to SMC defect indirectly affecting endothelial differentiation. Vessel remodeling and vessel maturation/integrity defects observed in Alk3 CKO mutant mice are most similar to vessel abnormalities observed in Alk1 knockout mice. Notably, mice deficient in Alk1 showed dilated vessels. The Alk1-deficient dorsal aorta had significantly reduced levels of SMC(Oh et al., 2000), suggesting that a close interaction between endothelial cells and supporting cells is crucial for maintaining vessel integrity. Moreover, the expression of Vegf, Plau1 and Serpine1 was increased in both Alk3CKO and Alk1 mutant mice. Additionally, all Id genes were downregulated in Alk3 CKO mutants. As ALK1 mediated signals in endothelial cells were shown to activate Id genes(Goumans et al., 2002), it is of interest to see if Id gene expression is also downregulated in Alk1 mutant mice. And yet, yolk sac vascular defects of Alk1knockout mice appear to be more severe compared with Alk3 CKO embryos. For example, remodeling defects in Alk1 knockout mice are obvious as early as E9.5, at which time point Alk3 CKO embryos cannot be distinguished from the control embryos. To this end, it is important to note that all Smad4 CKO embryos show angiogenesis defects. Collectively, we propose that vessel remodeling and maturation is achieved by redundant as well as specific signaling mediated by ALK1 versus ALK3.

One of the most distinct features of heart morphogenesis is an endothelial-mesenchymal transformation (EMT) (reviewed in Person et al., 2005). During this process, endocardium in the AVC and OFT differentiates into mesenchymal cells, are delaminated, and eventually invade the cardiac jelly, the extracellular matrix between the endocardium and the myocardium. The mechanisms involved in cushion formation or its further morphogenesis have been extensively studied. By using an in vitro explant culture system, it has been shown that EMT could be induced by myocardium underlying the AVC(Mjaatvedt et al., 1987; Runyan and Markward, 1983). BMP2 and TGFβ2 could also induce EMT in endocardium explant culture(Sugi et al., 2004). Furthermore, Tgfb2^-/- mice showed defects in the OFT, but initiation of endocardial cushion formation seemed to be normal(Sanford et al., 1997). Consistently, conditional inactivation of Bmp2 in cardiomyocytes using the Nkx2.5-Cre system results in endocardial EMT defects(Ma et al., 2005). Conditional deletion of Alk3 in cardiomyocytes using the Mhc-Cre system results in heart abnormalities involving the interventricular septum,trabeculae and AV cushion defects (Gaussin et al., 2002). In these mice, Tgfb2 expression is greatly reduced, implying that TGFβ2 is crucial for AV cushion morphogenesis. Outflow tract defects are not observed in these mice. Intriguingly,conditional inactivation of Bmp4 in cardiomyocytes using the Nkx2.5-Cre system results in defective outflow tract septation and abnormal morphogenesis of branchical arch arteries(Liu et al., 2004), while Bmp4 inactivation using the rat troponin T (Tnnt)-Cre results in severe AVC defects (Jiao et al., 2003). Collectively, these studies emphasize the importance of myocardium in inducing EMT. It is important to note that Alk3-mediated signaling in the cardiomyocytes is important for myocardium proliferation and for AVC morphogenesis.

Herein, we demonstrate that ALK3 signaling in the endocardium is intrinsically required for AV cushion formation. The selective absence of the endocardial cushions in the AVC raised the issue of whether the origin of endocardium in the AVC is different from that of the OFT. However, we found that FLK1⁺ cells can contribute to both AVC and OFT endocardial cushions. Given the finding that neural crest cells also contribute to the OFT, but not to the AV cushion (Epstein et al., 2000; Gitler et al.,2003), it is possible that ALK3 signaling in neural crest cells in the OFT rescued the Alk3 deficiency in the FLK1⁺ cells. Recently, Wang et al. (Wang et al.,2005) demonstrated that conditional inactivation of Alk-2using the Tie-2-Cre system resulted in selective defects in the AV superior cushion. The AV inferior cushion appeared to be normal in these mice. At E14.5, the surviving Alk-2 CKO mutant embryos showed a range of ventricular septal defects. The differences in AV cushion defects observed in our studies compared with Wang et al.(Wang et al., 2005) could be due to the nature of cells that express Cre (FLK1⁺ versus TIE2⁺ cells). However, the specificity of the ALK2 versus ALK3 in superior versus inferior cushion formation cannot be ruled out. As for the timing of the Alk3 inactivation, it is worth noting that about a third of Alk3 conditional mutants using the Tie2-Cre system developed relatively normal AV cushion (Ma et al., 2005). Importantly, conditional inactivation of Alk3 in neural crest cells using the Wnt1-Cre system results in outflow tract defects including a septal failure and reduced conotruncal length (Stottmann et al.,2004). Moreover, mice homozygous for a hypomorphic allele of Bmpr2 displayed defects in the formation of the OFT valve, but the AVC cushion formation was normal, despite the ubiquitous expression of Bmpr2 in the heart (Delot et al.,2003; Roelen et al.,1997).

During the AV cushion EMT, AV endocardium loses the characteristics of endothelial cells and cell-cell interaction, while activating genes for mesenchymal cells. Snai1, a Slug family member, and Twist1,a basic HLH transcription factor is known to promote the EMT by downregulating VE-cadherin and/or E-cadherin (Timmerman et al., 2004; Yang et al.,2004). We found a reduced expression level of Snai1 and Twist1 with sustained expression of VE-cadherin in Alk3 CKO embryos. Inactivation of Alk3 in Tie2 Cre mice showed a loss of Twist1 and a mild defect in VE-cadherin, although Snail1expression was not affected (Ma et al.,2005). SOX9, a high-motility group transcription factor, is expressed in the mesenchymal cells of AV and OFT cushion(Akiyama et al., 2004). SOX9 has been implicated in neural crest cell delamination and AVC cushion formation (Akiyama et al.,2004; Cheung and Briscoe,2003). Conditional knockout of Sox9 in the Wnt1-Cre system showed ectopic expression of Nfatc1, leading to a failure of cushion cell development(Akiyama et al., 2004). We observed that Sox9 expression was downregulated, while Nfatc1 expression was not affected in Alk3 CKO hearts. These results suggest that ALK3 signaling in the AV endocardium is crucial for the initiation stage of the AV EMT. As Snail and Sox9 are important regulators of Notch and Wnt/β-catenin, respectively, it will be interesting to see if ALK3 signaling interacts with these pathways in the AV EMT. Collectively, we show that ALK3 signaling is required for the AV cushion formation. Further studies on redundant as well as distinct signaling mediated by ALK3 versus ALK2 are warranted.

We thank Thomas Sato for Flk1^+/Cre mice; Darcy Gill for help with EM studies; Huiteng Cheng for help with embryo manipulation and discussion; James Richardson and Eric Olson for confirming the heart phenotype; and Robert Mecham, David Beebe and Choi laboratory members for helpful discussion. We also thank David Beebe and Jesse Lugus for critically reading the manuscript. This work was supported by grants from the National Institutes of Health, NHLBI, R01s HL63736 and HL55337 (to K.C.) and, in part,by the Intramural Research Program of the NIH, National Institute of Environmental Health Sciences (to Y.M.).