Rescue of cleft palate in Msx1-deficient mice by transgenic Bmp4 reveals a network of BMP and Shh signaling in the regulation of mammalian palatogenesis

The mammalian palate is formed by the union of three elements: the primary palate of the frontonasal process and the two lateral maxillary palatal shelves that will form the secondary palate. The formation of the mammalian secondary palate is a multi-step process that includes palatal shelf growth, elevation of the shelves, fusion between paired shelves and the disappearance of the midline epithelial seam (Ferguson, 1988). The closure of the palate separates the oropharynx and its various functions from the nasopharynx. In mice, the development of the secondary palate initiates at embryonic day 11.5 (E11.5) with the formation of tissue folds overlying the future palatal shelves within the oral cavity. Mesenchymal cell proliferation within the maxillary processes results in this appearance of the palatal shelf primordia (Burdett et al., 1988). At E12.5 and E13.5, palatal primordia grow vertically down beside the tongue. Between E13.5 and E14.0, a rapid elevation of the palatal shelves brings the two processes into horizontal apposition above the tongue. This movement is followed by the fusion of the medial edge epithelium (MEE) of the palatal shelves around E14.5, resulting in a continuous palate. Cleft palate, the most frequent congenital craniofacial birth defect in humans, arises from genetic or environmental perturbations in any step of the palatogenetic process (Johnson and Bronsky, 1995). Recent studies have indicated the importance of a growing number of genes in palate development. Mutations in genes encoding a variety of molecules, including transcription factors, growth factors or their receptors, have been shown to induce cleft palate formation (Satokata and Maas, 1994; Kaartinen et al., 1995; Matzuk et al., 1995a; Matzuk et al., 1995b; Proetzel et al., 1995; Peters et al., 1998; Qiu et al., 1998; Lu et al., 1999; Miettinen et al., 1999; Szeto et al., 1999; Zhao et al., 1999).

The palatal shelves are derived from migratory cranial neural crest cells, which associate with the craniopharyngeal ectoderm [divided into the oral, nasal, and medial edge epithelium (MEE)]. Similar to many other vertebrate organs, the development of mouse secondary palate relies largely on sequential and reciprocal interactions between epithelial and mesenchymal tissue layers (Slavkin, 1984). Peptide growth factors are known to play crucial roles as inductive signals that mediate such epithelial-mesenchymal interactions during organogenesis (Thesleff et al., 1995). Several families of peptide growth factors have been implicated in vertebrate facial development (reviewed by Francis-West et al., 1998). These inductive factors include sonic hedgehog (Shh), a member of the hedgehog family, bone morphogenetic proteins (BMPs) and members of the transforming growth factor β (TGFβ) superfamily. Shh has been shown to be essential for the growth and development of the chick facial primordia (Hu and Helms, 1999). In the chick, a blockade of Shh signaling results in growth inhibition in the facial primordia, while overexpression of Shh induces growth in the facial primordia by increasing cell proliferation (Hu and Helms, 1999). Similarly, the ectopic application of exogenous BMP2, BMP4 or BMP7 also increases cell proliferation in the chick mandibular primordia (Barlow and Francis-West, 1997; Wang et al., 1999). Thus, these growth factors may control outgrowth of the facial primordia. In the developing palatal shelves in the mouse, Shh expression has been detected in the palatal epithelium (Bitgood and McMahon, 1995). However, the actual function of Shh in palate development is unclear, as Shh-deficient mice exhibited holoprosencephaly, with almost complete lack of facial skeletal structures (Chiang et al., 1996). The transcripts of both Bmp2 and Bmp4 have also been detected in developing mouse palatal shelves (Lyons et al., 1990; Bitgood and McMahon, 1995). However, targeted mutations of these two genes in mice have not been informative as to their roles in palatogenesis because of embryonic lethality before palate formation in these mutants (Winnier et al., 1995; Zhang and Bradley, 1996).

The Msx1 homeobox gene is expressed in several developing organs in vertebrates, including the facial primordia, particular at the sites where epithelial-mesenchymal interactions occur during organogenesis (Davidson, 1995). Msx1 is believed to participate in these interactions by regulating the expression of signaling molecules (Chen and Maas, 1998). Mice deficient for the Msx1 gene exhibited neonatal lethality and severe craniofacial abnormalities, including cleft secondary palate, an absence of alveolar processes, and arrest of tooth development at the bud stage (Satokata and Maas, 1994; Houzelstein et al., 1997). In humans, mutations in the MSX1 gene are associated with isolated non-syndromic cleft palate and tooth agenesis (Vastardis et al., 1996; Blanco et al., 1998; Lidral et al., 1998; Van den Boogaard et al., 2000), consistent with the phenotype observed in Msx1 mutant mice. In Msx1-deficient mice, the bilateral primordial palatal shelves form and elevate normally, but failed to make contact and never fuse, which results in cleft palate (Satokata and Maas, 1994). Despite these findings, controversy remains over whether Msx1 is expressed in developing palatal shelves or not (MacKenzie et al., 1991; Ferguson, 1994; Satokata and Maas, 1994). The role of Msx1 in tooth development has been studied extensively (Chen et al., 1996; Bei and Maas, 1998; Bei et al., 2000; Zhang et al., 1999; Zhang et al., 2000a; Zhao et al., 2000), but the molecular mechanisms underlying the generation of cleft palate in Msx1 mutants remain unknown.

To determine the molecular aetiology of the non-syndromic clefting associated with Msx1 null mutations, we studied the genetic regulation of mammalian palatogenesis using Msx1-deficient mice as a model system. Our analyses demonstrated that Msx1 expression was restricted to the anterior region of the mesenchyme of the developing secondary palatal shelves. Msx1 was required for the expression of Bmp4 and Bmp2 in the palatal mesenchyme and Shh in the medial edge epithelium (MEE). We also showed that disruption of Msx1 function caused impairment of in cell proliferation in the palatal mesenchyme, which leads to the formation of cleft secondary palate in mutant mice. Using a transgenic approach, Bmp4 was ectopically expressed in the Msx1 mutant palatal mesenchyme, and its expression rescued the cleft palate phenotype and neonatal lethality. Associated with this rescue of cleft palate was a restored pattern of Shh and Bmp2 expression, as well as a recovery of normal cell proliferation in the palatal mesenchyme. Therefore BMP4 appears to bypass a requirement for Msx1 and to function upstream of Shh and Bmp2 to regulate palate development. We carried out further in vitro functional analyses to determine the position of these genes within the genetic pathway that regulates palatal development, and demonstrated that Shh derived from the MEE activated Bmp2 expression in the mesenchyme that in turn stimulated cell proliferation. Msx1 thus controls a genetic hierarchy involving BMP and Shh signals regulating the growth of anterior palate during mammalian palatogenesis.

The generation of Msx1^–/– mice and Msx1-Bmp4 transgenic mice have been described previously (Satokata and Maas, 1994; Zhang et al., 2000a). The Msx1 heterozygotes and Msx1-Bmp4 transgenic mice were outbred onto a CD-1 background. Breeding was performed using Msx1^+/– and Msx1-Bmp4 mice to produce Msx1 null mutants carrying two alleles of Msx1-Bmp4 transgene (Msx1^–/–/Tg), as described previously (Zhao et al., 2000). For all embryos used in this study, the embryonic ages were determined by the day when the vaginal plug was discovered and designated as embryonic day 0.5 (E0.5). Wild-type mouse embryos used in this study were collected from mating of CD-1 mice. The genotype of Msx1^–/–, Msx1-Bmp4 and Msx1^–/–/Tg adult mice and embryos were determined by a PCR-based method using genomic DNA extracted from tails or extra-embryonic membranes, as previously described (Satokata and Maas, 1994; Chen et al., 1996; Zhang et al., 2000a).

Paired secondary palatal shelves from individual E13.5 embryos were microdissected and placed in Trowell type organ cultures with chemically defined medium according the method described previously (Taya et al., 1999). Each pair was orientated so that the MEE of each palatal shelf was in contact. Paired palatal shelves were initially cultured with Minimal Essential Medium supplemented with 300 μg/ml L-glutamate, 50 μg/ml glycine, 100 μg/ml ascorbate, 1% penicillin/streptomycin at 37°C in a 5% CO₂ air environment for 6 hours. In this initial culture period, a sufficiently low volume of medium was used to allow firm attachment of specimens onto the filters. After 6 hours the culture medium was replaced with DMEM/F12 (1:1) supplemented with 1% L-glutamine, 1% ascorbate and 1% penicillin/streptomycin. Samples were cultured for 3 days with one change of medium and were then harvested.

For bead implantation experiments, Affi-Gel blue agarose beads (BioRad, Hercules, CA) were soaked in proteins as previously described (Chen et al., 1996). Protein concentrations used were consistent throughout all experiments. BMP2 and BMP4 (Genetics Institute, Cambridge, MA) were used at 100 μg/ml. Shh N-terminal peptide [Shh-N; from R&D Systems, Minneapolis, MN (catalog number 461-SH-025)] was used at 1 mg/ml, and Noggin [R&D Systems (catalog number 719-NG-050)] (Zhang et al., 2000a) at 0.2 mg/ml. Anti-Shh antibodies (5E1) (Developmental Studies Hybridoma Bank, University of Iowa, Iowa City, IA) (Ericson et al., 1996) were used at 0.35 mg/ml. The palatal shelves of E12.5 embryos were further microdissected into an anterior region (all tissues anterior to the first molar) and a posterior region (all tissues posterior to the first molar) (see Fig. 1M), and each sample was pooled for analyses. The palatal tissues were then incubated in 0.5% trypsin and 2.5% pancreatin in phosphate-buffered saline (PBS) on ice for 30 minutes before transfer to a stop solution consisting of 50% horse serum in PBS. The palatal mesenchyme was then separated from epithelium using a pair of fine forceps and placed on filters in Trowell type cultures. Protein-soaked beads were implanted onto the top of explants in DMEM culture medium supplemented with 10% fetal calf serum (Chen et al., 1996). Samples were harvested after 24 hours in culture, and processed for whole-mount in situ hybridization.

Samples used for whole-mount and section in situ hybridization were fixed in freshly made 4% paraformaldehyde/PBS. For section in situ hybridization, samples were dehydrated through graded ethanol series and embedded in paraffin wax. Serial paraffin sections were made at 10 μm. For whole-mount in situ hybridization, samples were bleached with 6% H₂O₂ before dehydration with methanol. The following cDNAs were used to generate antisense riboprobes: an 800 bp mouse Msx1 (Hill et al., 1989); a 1.2 kb mouse Bmp2 (Blessing et al., 1993); a 1.0 kb mouse Bmp4 (Jones et al., 1991); a 650 bp mouse Shh (Echelard et al., 1993); and a 2.25 kb mouse Ptch1 cDNA (Goodrich et al., 1996). To examine the expression of the human Bmp4 transgene in the Msx1-Bmp4 transgenic mice, a 700 bp DNA fragment containing 500 bp SV40 intronic and poly(A) sequence and a 200 bp 3′-UTR sequence of the human Bmp4 gene subcloned from the original transgene was used to generate probe. This probe does not crossreact with mouse Bmp4. Non-radioactive RNA probes were generated by in vitro transcription labeling with digoxigenin-UTP according to the manufacturer’s protocol (Boehringer Mannheim, Indianapolis IN). Whole-mount and section in situ hybridization was performed as previously described (Zhang et al., 1999).

Standard paraffin sectioning and Hematoxylin and Eosin staining was carried out for histological analysis. Skeletal structures were stained using Alcian Blue for non-mineralized cartilage and Alizarin Red for bone, as described previously (Zhang et al., 2000b). For SEM, samples were fixed with 2.5% glutaradehyde/0.1 M sodium cacodylate at 4°C for 12 hours. After samples were postfixed in 1% osmium tetroxide/0.1M sodium cacodylate, dehydrate through graded ethanol series, samples were trimmed and critical point dried in a Touisimis (Samdri-790) apparatus and gold coated with a Polaron E5100 sputter coater. Samples were examined with a JEOL JSM-35C scanning electron microscope.

For in vivo BrdU labeling, timed pregnant wild-type or Msx1^–/– mice were injected intraperitoneally on E12.5 or E13.5 with BrdU labeling reagent (3 mg/100 g body weight). Two hours after injection, mice were sacrificed and embryos were harvested. Embryonic heads were fixed in Carnoy’s fixative, dehydrated through a graded ethanol series, embedded in paraffin wax and sectioned at 10 μm. Immunodetection of BrdU was performed using a BrdU labeling and Detection kit (Beohringer Mannheim) according to the manufacturer’s protocol. A fixed area of 22.5 mm² beginning at the MEE and extending laterally was selected using an ocular scale grid. BrdU-positive cells in the mesenchyme within the fixed area were counted. Three continuous sections from the anterior region (anterior to the first molar) and posterior region (posterior to the first molar), respectively, of each palatal shelf from each individual embryo were counted. For in vitro BrdU labeling, palatal shelves from E13.5 wild type and Msx1 mutant embryos were dissected, and divided into anterior and posterior regions. The palatal epithelium was separated from the palatal mesenchyme after enzymatic treatment, as described above. The palatal mesenchyme samples were placed on filters in Trowell type organ cultures. Protein-soaked beads were implanted onto explants that were then cultured in DMEM supplemented with 10% fetal calf serum for 8 or 24 hours before transferring into the same medium containing 10 μmoles of BrdU. After 45 minutes of labeling, samples were fixed in Carnoy’s fixative, embedded in paraffin wax, sectioned for processing and immunodetection of BrdU. TUNEL assay was performed using an in situ cell death detection kit (Boehringer Mannheim) as described previously (Zhang et al., 2000b).

Knockout studies have implicated the Msx1 homeobox gene in mammalian craniofacial development (Satokata and Maas, 1994; Houzelstein et al., 1997). The phenotypic abnormalities exhibited in Msx1-deficient mice were mainly restricted to the first pharyngeal arch and its derivatives, including a complete cleft of the secondary palate, an absence of alveolar process and an arrest of tooth development at the bud stage (Satokata and Maas, 1994; Houzelstein et al., 1997). In the mutant, the initial development of bilateral palatal shelves appeared normal, and the palatal shelves of the Msx1^–/– mice grew vertically and elevated properly. However, they failed to make contact and did not fuse at the midline, resulting in a wide open cleft secondary palate which apparently contributes to neonatal lethality (Fig. 1E,H,K) (Satokata and Maas, 1994).

We have previously reported the generation of the transgenic mice harboring the human Bmp4 gene driven by the mouse Msx1 promoter (Zhang et al., 2000a). The human and mouse BMP4 proteins are highly conserved and exhibit 95% identity at the amino acid level. In these transgenics, the mouse Msx1 promoter directed transgene expression in the craniofacial region, including the dental mesenchyme, mimicking the endogenous Msx1 expression pattern (Zhang et al., 2000a; Zhao et al., 2000). Transgene expression was also detected in the mesenchyme of the anterior palatal shelves at E12.5 and E13.5 (Fig. 1A,B). This expression pattern is consistent with the endogenous Msx1 expression in developing palatal shelves (see below). No overt phenotype was detected in these Msx1-Bmp4 transgenic mice (Zhang et al., 2000a).

BMP4 has been shown to function downstream of Msx1, and a downregulation of Bmp4 expression in the dental mesenchyme of Msx1 mutants might account for the arrest of tooth development at the bud stage (Chen et al., 1996). To test whether ectopically expressed Bmp4 in dental mesenchyme could bypass Msx1 function to rescue tooth development in Msx1^–/– mice, we introduced the Msx1-Bmp4 transgenic allele into a Msx1^–/– background to generate Msx1^–/– mice carrying the Msx1-Bmp4 transgene (Msx1^–/–/Tg) (Zhang et al., 2000a; Zhao et al., 2000). Surprisingly, tooth development was only partially rescued, but some Msx1^–/–/Tg mice survived the neonatal lethality observed in Msx1 null animals. These mice grew to adulthood, but completely lacked teeth (Zhao et al., 2000). A gross morphological examination of the surviving Msx1^–/–/Tg mice revealed an intact palate (Fig. 1I,L), although the rugae did not fuse at the midline (insert in Fig. 1I). Of 66 Msx1^–/–/Tg mice examined from six independent lines, 41 (62%) exhibited complete palatal closure. Histological analyses further demonstrated that the palatal shelves of Msx1^–/–/Tg mice could make contact and fuse, as evidenced by the disappearance of the midline seam at the site of contact (Fig. 1F). Skeletal staining showed that the abnormal appearance of the premaxilla seen in Msx1^–/– mice (Fig. 1K) (Satokata and Maas, 1994) remained in Msx1^–/–/Tg mice (Fig. 1L), indicating that transgenic expression specifically rescued the cleft palate phenotype. These results demonstrate that the ectopic expression of Bmp4 in the palatal mesenchyme can bypass the requirement for Msx1 to support palate development.

Because of a contradiction on Msx1 expression in developing palate, it was speculated that the cleft palate observed in Msx1 mutants was the result of a failure of tooth development which is known to require Msx1 function (MacKenzie et al., 1991; Ferguson, 1994; Satokata and Maas, 1994). However, the fact that Msx1^–/–/Tg mice displayed a closed palate, but lack teeth, rules out this possibility (Zhao et al., 2000) (this study). To establish a role for Msx1 in mouse palatogenesis, Msx1 expression was carefully examined in the developing palate at E12.5 and E13.5. Using section in situ hybridization, Msx1 transcripts were detected in the mesenchyme of developing palatal shelves (Fig. 2A,B) at both E12.5 and E13.5, with expression restricted to the region anterior to the first molars. No Msx1 transcripts were detected in the posterior portion of developing palatal shelves (Fig. 2C). This palatal Msx1 expression appeared relatively weak when compared with the level of expression in the dental mesenchyme (Fig. 2A-C). These results clarify the previous contradictory results on the palatal Msx1 expression (MacKenzie et al., 1991; Ferguson, 1994; Satokata and Maas, 1994) and suggest a role for Msx1 in the development of mouse secondary palate.

To establish a genetic hierarchy involving Msx1 in palatogenesis, the expression of Bmp2, Bmp4 and Sonic hedgehog (Shh), the known downstream genes of Msx1 in the developing tooth germ (Chen et al., 1996; Zhang et al., 2000a), was analyzed in the developing palatal shelves of both wild type and Msx1^–/– embryos at E12.5 and E13.5. In wild-type palatal shelves, Bmp4 was found expressed in both the palatal epithelium and mesenchyme at E12.5 (Fig. 2D), and then was found to be restricted to the mesenchyme adjacent to the MEE at E13.5 (Fig. 2E). Bmp2 expression was detected in both the palatal epithelium and mesenchyme (Fig. 2G,H) at both the stages examined, while Shh was only detected in the palatal epithelium, most prominent in the MEE (Fig. 2J,K). Similar to the pattern of Msx1 expression, Bmp2 and Bmp4 expression was only detected in the anterior palate, but not in the posterior portion of the palate (Fig. 2F,I). In the posterior palatal shelves at E13.5, Shh transcripts were not detected in the MEE, but were detected in the oral side epithelium where the rugae would form (Fig. 2L), as reported previously (Bitgood and McMahon, 1995). In Msx1^–/– embryos, Bmp4 expression was downregulated in the mesenchyme, but remained unaltered in the epithelium of the palatal shelves at E12.5 (Fig. 3A), while the expression of Shh and Bmp2 was not affected at this stage (Fig. 3D,G). By contrast, at E13.5 when Bmp4 transcripts were detected exclusively in the palatal mesenchyme in wild-type embryo, significantly reduced expression of Bmp4 was observed the palatal mesenchyme of Msx1 mutants (Fig. 3B). Similarly, Shh expression in the MEE and Bmp2 expression in both the epithelium and mesenchyme of Msx1^–/– palate were simultaneously downregulated at this stage (Fig. 3E,H). It is concluded, therefore, that Msx1 functions upstream of Bmp4, Shh and Bmp2 in the developing palatal shelves. As Msx1 expression is restricted to the palatal mesenchyme, the downregulation of Shh and Bmp2 in the epithelium of Msx1^–/– palatal shelves is apparently a secondary effect due to the absence of Msx1.

In the developing tooth germ, Bmp4 functions both upstream and downstream of Msx1, but upstream of Shh and Bmp2 (Chen et al., 1996; Zhang et al., 2000a). It is possible that, similar to the tooth germ, Msx1 may control Bmp4 expression in the palatal mesenchyme. If so, a downregulation of Bmp4 expression in Msx1-deficient palatal mesenchyme could account for the absence of Shh and Bmp2 expression in the mutant palatal shelves. To test whether such a genetic regulatory pathway exists in the developing palate, we examined the expression of Shh and Bmp2 in Msx1^–/–/Tg palatal shelves where ectopic Bmp4 was expressed in the absence of Msx1. These analyses indicated that the ectopic expression of the Bmp4 transgene in the palatal mesenchyme of Msx1^–/– mice indeed restored, at least partially if not completely, the expression of Shh and Bmp2 in the E13.5 transgenic palatal shelves where the endogenous Bmp4 expression was absent (Fig. 2C,F,I). These data unambiguously demonstrated that Msx1 does not directly regulate the expression of Shh and Bmp2 in the developing palate. Rather, Msx1 controls Bmp4 expression in the palate, which in turn functions upstream of Shh and Bmp2. These results also suggest that the expression of Shh and Bmp2 was unaltered in the E12.5 Msx1^–/– palatal shelves, most probably because endogenous Bmp4 expression was preserved in the MEE at that stage. Consistent with this hypothesis is the fact that the expression of genes encoding for BMP receptors, including Bmpr1a and Bmpr1b, was detected in the epithelium of developing palate at E12.5 and E13.5, and remained unaltered in that of Msx1 mutants (data not shown).

It was previously demonstrated that overexpression of Bmp4 in developing mouse tooth germ leads to an inhibition of Shh expression in the dental epithelium (Zhang et al., 2000a). To test if the same regulatory mechanism is conserved in the developing palate, we examined Shh expression in the MEE of E12.5 and E13.5 Msx1-Bmp4 transgenic embryos in which Bmp4 transgene is ectopically expressed in the palatal mesenchyme (Fig. 1A,B). Unlike what was observed in the tooth germ of the transgenic embryo (Zhang et al., 2000a), Shh expression was not affected (data not shown). This observation suggests that different BMP regulatory mechanisms exist in tooth versus palate. Alternatively, it is possible that the level of the transgenic Bmp4 expression is not high enough in the palatal mesenchyme to repress Shh expression in the MEE.

To dissect out a genetic pathway involving Msx1, Bmp2, Bmp4 and Shh in developing mouse palate, bead implantation experiments were performed using the anterior region of the E12.5 wild-type and Msx1^–/– palatal shelves. To determine whether palatal mesenchyme exhibits heterogeneity in response to growth factor signals, the posterior region of palate was also assayed. Anterior and posterior palatal tissues were collected by microdissection (see Fig. 1M for position cuts made to obtain anterior versus posterior palatal tissues), and their epithelia were removed following enzymatic treatment. Protein-soaked agarose beads were implanted into the explants of palatal mesenchyme. Explants were cultured for 24 hours, and then harvested for the examination of gene expression by whole-mount in situ hybridization.

In experiments using anterior palatal mesenchyme, BMP4-soaked beads induced the expression of Msx1 (12/14) and Bmp4 (14/16) in the wild type mesenchyme (Fig. 4A-C), but failed to induce Bmp4 (0/4) in the Msx1^–/– mesenchyme (Fig. 4D). BMP4-soaked beads failed to induce Bmp2 in wild type mesenchyme (0/11; Fig. 4J) and Msx1^–/– mesenchyme (0/6; Fig. 4I). These results indicate that Msx1 is required for the induction of mesenchymal Bmp4 by BMP4 itself in the anterior palatal mesenchyme. These assays also indicated that transgenic Bmp4 expression is not directly responsible for the restoration of Bmp2 expression in the Msx1^–/–/Tg palatal mesenchyme. In other assays, Shh-soaked beads failed to induce Bmp4 expression in the wild-type mesenchyme (0/11; Fig. 4F). This finding is further supported by the fact that beads soaked with an anti-Shh antibody did not affect endogenous Bmp4 expression when implanted into E12.5 wild-type palatal tissue containing both the epithelium and mesenchyme (8/8) (Fig. 4G). By contrast, Shh-soaked beads induced Bmp2 expression in both wild-type (10/12) and Msx1^–/– mesenchyme (7/7), respectively (Fig. 4L,M). These observations suggest that Bmp2 expression in the mesenchyme is regulated by epithelially derived Shh during palate development and that Msx1 is not necessary for the induction of Bmp2 expression by Shh.

In parallel experiments using E12.5 posterior palatal mesenchyme, BMP4-soaked beads induced neither Msx1 nor Bmp4 in wild type palatal mesenchyme (0/18 for Msx1; 0/12 for Bmp4; Fig. 5A,B). Similarly, Shh-soaked beads failed to induce Bmp2 expression in this tissue (0/9; Fig. 5C). By contrast, FGF8, a factor known to induce Pax9 expression in dental mesenchyme (Neubüser et al., 1997), induced Pax9 expression in the posterior palatal mesenchyme, but not in the anterior palatal mesenchyme (7/8 for the posterior; 0/8 for the anterior; Fig. 5D-F). These results indicate that heterogeneity exists along the anteroposterior axis of the mesenchyme in the developing palate in response to growth factor induction, and further demonstrated that a genetic cascade involving Msx1, Bmp4, Bmp2 and Shh operates in a restricted manner in the anterior palate to regulate palate development.

An interruption of palate development at any step, for example, during initiation, growth, elevation or fusion, causes cleft palate. Unlike Tgfb3 deficient mice in which the palatal shelves were shown to make contact at the midline, but failed to fuse (Kaartinen et al., 1995; Proetzel et al., 1995), Msx1^–/– palatal shelves initiated and elevated normally, but never made contact (Satokata and Maas, 1994). This phenotype suggests an impairment in either growth or fusion in the palatal shelves. To differentiate these two possibilities, we used an in vitro organ culture system and chemically defined medium as described previously (Taya et al., 1999). Pairs of palatal shelves from E13.5 wild-type, Msx1^–/– and Msx1^–/–/Tg embryos were isolated and placed in contact with the MEE of each facing the other and the cultured for 3 days before histological analyses. Palatal shelves from all three genotypes underwent fusion, which was indicated by the disappearance of the midline seam and the establishment of the mesenchymal continuity (4/4 for each genotype; Fig. 6A-C). These results clearly indicate that failure of fusion is not the cause of cleft palate in Msx1^–/– embryos. Cell proliferation was also examined in all three genotypes and this in vivo BrdU labeling demonstrated that a significantly reduced level of cell proliferation occurred in the Msx1^–/– palatal mesenchyme at E13.5 but not E12.5, when compared with the wild-type controls (Fig. 6D,E,G, and data not shown). Consistent with the restricted pattern of Msx1 expression in the palatal shelves, a significantly reduced number of BrdU-labeled cells were observed in the anterior region (Fig. 6E), but not in the posterior region (Fig. 6G; data not shown) of the Msx1^–/– palatal shelves, as confirmed by the Student’s t-test analysis (P<0.01; Fig. 6G). TUNEL assays further revealed that the level of apoptosis was unaltered in Msx1^–/– palatal shelves along the anteroposterior axis (data not shown). These results indicate that the formation of a cleft palate in Msx1^–/– mice results from defective cell proliferation in the anterior palatal mesenchyme, a mechanism that is consistent with the previously proposed role of Msx1 in facilitating cell proliferation (Song et al., 1992). In addition, ectopic Bmp4 expression in the Msx1^–/– palatal mesenchyme returned cell proliferation to normal levels (Fig. 6F,G), providing a cellular basis for the observed rescue of cleft palate in Msx1^–/–/Tg mice.

It has been shown that both Shh and BMP2 stimulate cell proliferation in chick facial primordia (Barlow and Francis-West, 1997; Hu and Helms, 1999). In our studies, a reduced level of cell proliferation in Msx1^–/– palatal shelves at E13.5 was accompanied by a downregulation of Shh and Bmp2 expression, suggesting that Shh and BMP2 regulate cell proliferation in developing palate. To test this hypothesis, BMP2-soaked beads were implanted onto explanted anterior palatal mesenchyme isolated from E13.5 wild-type and Msx1^–/– embryos. A marked increase in cell proliferation was observed in mesenchyme from both genotypes after 8 hours in culture (5/5 for wild type and 4/5 for Msx1^–/–; Fig. 7A-D), with a biased induction of cell proliferation in the medial aspect where rapid cell division normally occurs. By contrast, Shh-soaked beads failed to induce cell proliferation in both wild type (data not shown) and Msx1^–/– palatal mesenchyme (Fig. 6H) after 8 hours in culture. However, Shh-soaked beads were able to induce cell proliferation around the beads in both wild type (7/9) and Msx1^–/– palatal mesenchyme (4/5) after 24 hours in culture (Fig. 7E,F), indicating that Shh does induce cell proliferation in this tissue but does so via an indirect mechanism. Furthermore, beads soaked with antibody specific for Shh repressed cell proliferation (8/8) in the anterior region of E13.5 wild-type palatal shelves containing both the epithelium (endogenous Shh source) and the mesenchyme (Fig. 7I,J), indicating that Shh derived from the MEE is critical for maintaining cell proliferation in the palatal mesenchyme. Strikingly, the repression of cell proliferation induced by the anti-Shh antibody was reversed when a single bead soaked with both anti-Shh antibody and BMP2 was implanted into wild type palatal explants containing both the epithelium and mesenchyme (Fig. 7K). As Shh is known to act upstream of Bmp2 in various developing organs (Laufer et al., 1994; Yang et al., 1997; Drossopoulou et al., 2000; Zhang et al., 2000a), including palate (as shown in this study), we asked whether the cell proliferation induced by Shh is mediated through the induction of Bmp2 expression. In these assays, Noggin-soaked beads were implanted into wild-type palatal explants containing both the epithelium and mesenchyme, and cultured the explants in vitro for 24 hours before BrdU labeling. Noggin, a known antagonist of BMP function (Smith and Harland, 1992; Zimmerman et al., 1996) was found to significantly repress cell proliferation in the samples (data not shown). Moreover, implantation of a single bead soaked with both Shh and Noggin into wild-type palatal mesenchyme failed to induce proliferation after 24 hours in culture (0/9; Fig. 7G). These data support the hypothesis that BMP2 acts downstream of Shh and mediates the mitogenic activity of Shh in the induction of cell proliferation in the developing palatal mesenchyme. BMP activity is required for cell proliferation in palatal mesenchyme. However, the role of BMP2 in inducing cell proliferation appears limited to the anterior palatal mesenchyme (where Bmp2 is normally expressed) as evidenced by the fact that BMP2 was unable to induce cell proliferation in the explants of posterior palatal mesenchyme (data not shown). These observations further support the idea that a number of different genetic pathways act along the anterior-posterior axis of palatal shelves and are involved in the regulation of palate development.

Mammalian palatogenesis is a complex process that involves the participation of many genes that encode growth factors, their receptors and transcription factors. Mutations in a number of genes are known to cause cleft palate, and the production of this phenotype in each mutant may differ mechanistically. In this study, we have analyzed the cellular and molecular basis of the non-syndromic clefting associated with the Msx1 null mutation. Our data indicate that defective cell proliferation in the anterior palatal mesenchyme, where Msx1 is normally expressed, apparently induced the formation of cleft palate in Msx1 mutant mice. Associated with this impaired cell proliferation was the downregulation of expression of a number of growth factors, including BMP2, BMP4 and Shh, in the anterior region of the Msx1^–/– palatal shelves. These growth factors form a network that mediates epithelial-mesenchymal interactions leading to the growth of anterior palate. We showed that ectopic expression of Bmp4, a downstream gene of Msx1, in the Msx1^–/– palatal mesenchyme was able to restore a normal level of cell proliferation and rescue the cleft palate phenotype. Our results unambiguously demonstrate for the first time a role for BMPs and Shh in the development of the mammalian secondary palate.

Based on the results presented here, we propose that Msx1 controls a genetic hierarchy, involving BMP and Shh signals, that regulates the development of the mammalian palate, as shown in Fig. 8. Msx1, which is expressed in the anterior palatal mesenchyme, is required for the maintenance of Bmp4 expression in this mesenchyme. This part of model is supported by data showing that Bmp4 expression was repressed in the Msx1^–/– palatal mesenchyme, and that BMP4 induced Msx1 and Bmp4 expression in wild-type palatal mesenchyme, but failed to induce Bmp4 in the Msx1^–/– palatal mesenchyme. BMP4 activity is then required for Shh expression in the MEE, as evidenced by the fact that Shh expression was restored in the MEE of Msx1 mutant expressing transgenic Bmp4. The MEE-derived Shh then induces the expression of Bmp2 in the palatal mesenchyme, where it regulates cell proliferation necessary in palatal growth. Although we can not exclude the possibility that mesenchymally expressed BMP4 may regulate cell proliferation directly, this seems unlikely because application of an anti-Shh antibody to the palatal shelves repressed cell proliferation, but did not alter endogenous Bmp4 expression (Fig. 4G, Fig. 7I).

During organ development, interactions between neighboring tissue layers are crucial for growth and differentiation. These sequential and reciprocal tissue interactions are mediated by diffusible growth factors (Jessell and Melton, 1992; Thesleff et al., 1995). During mammalian palatogenesis, the interactions between the cranial neural crest-derived mesenchyme and the craniopharyngeal ectoderm regulate palate morphogenesis and differentiation (Slavkin, 1984; Ferguson and Honig, 1984). Our data demonstrate that Msx1 controls a network of growth factors, including BMPs and Shh that function in the developing palate. The co-expression of Shh and Bmp observed in many developing vertebrate organs suggests a closely regulated relationship between Shh and BMPs (Bitgood and McMahon, 1995). The protein of each gene regulates the expression of the other, within the same or different tissue layers, either positively or negatively, depending on different organs (Laufer et al., 1994; Roberts et al., 1995; Duprez et al., 1996; Arkell and Beddington, 1997; Watanabe et al., 1998; Zhang et al., 2000a; Zhao et al., 2000). For example, Shh has been shown to induce the expression of Bmp2 in the developing limb (Laufer et al., 1994; Duprez et al., 1996) and Bmp4 in the gut mesenchyme (Roberts et al., 1995), but to repress Bmp4 in the dorsal neural tube (Watanabe et al., 1998). In addition, the data presented here show that in the developing palatal shelves, BMP4 activity in the mesenchyme is required for the maintenance of Shh expression in the MEE, while the epithelially derived Shh in turn activates Bmp2 in the palatal mesenchyme. These growth factors expressed in different tissue layers appear to act as signals mediating such epithelial-mesenchymal interactions leading to palatal growth.

In the developing mouse tooth germ, there exists a similar genetic hierarchy in which Msx1 also controls the expression of Bmp4 in dental mesenchyme. The mesenchymally expressed BMP4 in turn acts on the dental epithelium to regulate Shh expression. (Chen et al., 1996; Zhang et al., 2000a). This Msx1 controlled signaling pathway is required for the induction of enamel knot and the progression of dental epithelial development from the bud stage to the cap stage (Chen et al., 1996; Jernvall et al., 1998; Zhao et al., 2000). However, in the developing palatal shelves, the Msx1 controlled signaling network functions to facilitate cell proliferation within the mesenchyme. Therefore, although the similar Msx1 governed genetic pathway is conserved in the mandible and palate, it is used to regulate the different developmental processes that bring about organ formation in these sites.

Mutations in the Msx1 gene cause clefting of the secondary palate and tooth agenesis in both humans and mice (Stokata and Maas, 1994; Vastardis et al., 1996; Houzelstein et al., 1997; Van den Boogaard et al., 2000). However, whether Msx1 is expressed in developing palate or not has been controversial, which led to the speculation that cleft palate formation in Msx1-deficient mice may be an intermediate effect that arises because of the primary failure of tooth development (MacKenzie et al., 1991; Satokata and Maas, 1994; Ferguson, 1994). In this study, we present evidence that Msx1 is indeed expressed, although weakly, in the mesenchyme of the developing palatal shelves, and that expression is restricted to the anterior-most region of the palatal shelves. It is possible that in earlier studies, the high level of Msx1 expression in the mesenchyme of the maxillary molar may have caused expression in the palate to be overlooked. We therefore conclude that the cleft palate seen in Msx1 mutants is caused by an intrinsic defect in palatogenesis. This idea is further supported by the fact that a decrease in cell proliferation was observed in the Msx1^–/– palatal mesenchyme. The rescue of cleft palate in Msx1 mutants by transgenically expressed Bmp4 also prevented neonatal death, indicating that lethality in Msx1-deficient mice resulted from cleft palate formation. It is interesting to note that abnormally low levels of cell proliferation were only seen in the anterior region of the palatal shelves where Msx1 is normally expressed. The palatal growth appeared normal in the posterior portion of the Msx1^–/– palatal shelves. Palatal closure is known to occur in an anterior-to-posterior sequence, and we have demonstrated a growth defect at the anterior region of the Msx1^–/– palatal shelves that leads to complete palatal clefting. Taken together, these results suggest a zipper-like mechanism for palatal closure. The differential responsiveness of anterior versus posterior palatal mesenchyme to the induction of growth factors such as BMP and FGF provides the first evidence for the existence of heterogneity along the anteroposterior axis of developing palate. The development of the posterior region of mammalian palatal shelves seems to be regulated by a distinct set of genes, which warrant future investigation.

In the developing mouse limb, Msx1 is strongly expressed in the progress zone where rapid cell proliferation occurs (Hill et al., 1989; Robert et al., 1989), implicating a role for Msx1 in cell proliferation. Forced expression of Msx1 in myoblasts or even myotubes promoted cell proliferation in vitro and concurrently inhibited terminal differentiation by repressing the expression of muscle-specific proteins including MyoD (Myod1 – Mouse Genome Informatics) (Song et al., 1992; Woloshin et al., 1995; Odelberg et al., 2000). However, the precise role of Msx1 in facilitating cell proliferation was unclear. Our results demonstrate that in the developing mouse palatal shelves, Msx1 is indeed required for proper cell proliferation. In the absence of Msx1, a reduced level of cell proliferation occurs in the anterior palatal mesenchyme where Msx1 is normally expressed. However, Msx1 gene product appears not to be directly involved in regulating cell proliferation, at least in developing palatal shelves. Instead it controls a BMP4-mediated signaling pathway that functions to promote cell division. This finding is supported by our demonstrations that ectopically expressed Bmp4 restored cell proliferation in the absence of Msx1, and that Shh and BMP2 induced cell proliferation in the Msx1^–/– palatal mesenchyme in vitro. Both BMPs and Shh were able to bypass the requirement for Msx1 and acted as mitogenic signals that stimulated cell proliferation in the developing palate.

Several studies have demonstrated that Shh can function as a mitogen in many vertebrate organs, including presomitic mesoderm and somite (Fan et al., 1995; Marcelle et al., 1999), developing lung (Bellusci et al., 1997), limb bud (Duprez et al., 1998), cerebellum and retina (Jensen and Wallace, 1997; Wechsler-Reya and Scott, 1999). Of particular, relevance to our study, Shh has been shown to have mitogenic activity in the craniofacial tissues that include chick facial primordia, craniofacial neural crest cells and mouse tooth germ (Hardcastle et al., 1998; Ahlgren and Bronner-Fraser, 1999; Hu and Helms, 1999). Ectopic application of Shh induced a significant increase in the level of cell proliferation in the chick facial primordia (Hu and Helms, 1999) and stimulated abnormal levels of epithelial cell proliferation in the mouse oral cavity (Hardcastle et al., 1998). We provide evidence that Shh does not function directly as a mitogen in the palate. The mitogenic activity that Shh executed on mouse palatal mesenchyme apparently occurs through an intermediate mechanism that involves inducing and/or maintaining Bmp2 expression. Our data indicate that it is BMP2 that functions to mediate the mitogenic signal from Shh during mammalian palatogenesis. As a downstream gene of Shh, Bmp2 has also been shown to mediate partial polarizing activities of Shh function in developing limb (Laufer et al., 1994; Duprez et al., 1996; Drossopoulou et al., 2000). Evidence presented in this study establishes a new function for BMP2 in mediating the mitogenic activity of Shh in the regulation of palatal growth in mammalian embryo.

BMPs, including BMP2, BMP4 and BMP7, have been shown to function as mitogens (Barlow and Francis-West, 1997; Wang et al., 1999), although an opposing role, that of repressing cell proliferation, has been reported for BMP4 in lung development (Bellusci et al., 1996). Similar to the mouse palatal mesenchyme studies presented here, the ectopic application of BMPs has been shown to enhance cell proliferation in the chick mandibular primordia (Barlow and Francis-West, 1997; Wang et al., 1999). As both mandibular mesenchyme and palatal mesenchyme are derived from cranial neural crest cells, they are likely to respond similarly to BMP signaling in terms of cell proliferation. However, at present it still remains unclear whether or not BMP acts directly or through secondary factors, to induce mitogenesis.

In summary, the results presented in this study provide evidence for the cellular and molecular aetiology of the non-syndromic clefting caused by the mutations in the Msx1 gene, and identify a role for BMPs and Shh in mammalian palate development. Most importantly, ectopic expression of Bmp4 in the Msx1^–/– palatal mesenchyme was able to bypass the requirement for Msx1 in palatogenesis and rescued the cleft palate phenotype. This finding suggests potential therapeutic strategies to prevent and cure embryonic cleft palate in the future.

We thank Genetics Institute (Boston, MA) for providing BMP2 and BMP4 proteins, and Dr Richard Harland (University of California, Berkeley) for the Noggin-producing cell line. The partially purified monoclonal antibody against Shh (5E1) was obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by the University of Iowa, Department of Biological Sciences, Iowa City, IA, USA. We also thank Drs Carol Burdsal, Liang Ma and Ken Muneoka for critical comments on the manuscript, and members of the Chen laboratory for discussion and technical help. Y. P. C. acknowledges grant support from NIH (R01DE12329, R01DE14044, P60DE13076) and the Millennium Trust Health Excellence Fund (HEF-2000-05-04) from the Louisiana Board of Regents.