Multiple functions of Snail family genes during palate development in mice

Cleft palate, with or without cleft lip, represents one of the most common birth defects observed in humans, occurring in approximately 1 in 700 live births (Gorlin et al., 2001; Murray, 2002). Formation of the mammalian secondary palate begins with the appearance of bilateral outgrowths from the maxillary processes, which grow and extend vertically along the sides of the tongue (Ferguson,1988). At a specific point in development, mechanical forces resulting from the forward extension of the mandible depress the tongue,allowing the palatal shelves to elevate and come into close approximation with each other. Through the specific expression of cell surface molecules and alterations in the morphology of cells along the medial edge epithelia (MEE),the shelves adhere and form a single structure bisected by a layer of epithelium, known as the medial epithelial seam (MES). The MES subsequently disappears, leading to a continuous palatal shelf consisting of a mesenchymal core bounded by the nasal and oral epithelium.

Disruption of any of these processes can result in cleft secondary palate. For example, Tgfb3-null mice exhibit defects in palatal shelf fusion resulting from a failure of the MEE to acquire the necessary adhesive phenotype (Gato et al., 2002; Taya et al., 1999; Tudela et al., 2002). Mutation of the Msx1 gene, however, results in cleft palate due to reduced proliferation in the palatal shelves during the vertical growth phase (Zhang et al., 2002). Cleft palate is also seen in mice exhibiting defects in palatal shelf elevation, such as Jag2-null mice, in which fusions between the tongue and palatal shelves are observed(Casey et al., 2006; Jiang et al., 1998a).

The Pierre Robin Sequence (also termed Robin Sequence) is a human developmental malformation characterized by mandibular retrognathia (normal size receded mandible) or micrognathia (abnormally small mandible),glossoptosis (rearward and downward displacement of the tongue) and cleft palate. Cases of Pierre Robin Sequence are both phenotypically and genetically heterogeneous (Cohen, 1999; Houdayer et al., 2001; Jakobsen et al., 2006; Jamshidi et al., 2004; Melkoniemi et al., 2003; Ounap et al., 2005; Ricks et al., 2002). The association of mandibular defects with palatal clefting has been observed in several mouse models of cleft palate, including Hoxa2(Gendron-Maguire et al.,1993), Egfr(Miettinen et al., 1999) and Dmm (disproportionate micromelia, a dominant mutant allele of the Col2a1 gene) (Ricks et al.,2002) mutant mice. In both Egfr-null and Dmmmice, growth and development of Meckel's cartilage is affected, delaying the necessary forward movement of the mandible and tongue.

The mechanism by which the MES degrades to form a continuous mesenchymal palatal shelf has been a matter of intense debate. One proposal is that the loss of epithelial cells is caused by an epithelialmesenchymal transition(EMT) (Fitchett and Hay,1989). Although there is substantial data to suggest this process occurs in a portion of the cells in the MES, recent lineage tracing studies appear to rule out the possibility that these cells persist in the resulting mesenchymal core of the palate (Vaziri Sani et al., 2005). Alternate models propose that MES cells undergo apoptosis or migrate to either the oral or nasal epithelium. It is likely that a combination of these events leads to the eventual dissolution of the MES. During embryogenesis, the oral and nasal epithelium is bilayered, consisting of a cuboidal basal epithelial cell layer and a flattened, transient periderm layer on the surface(Fitchett and Hay, 1989). Current data suggest that, upon palatal shelf contact, periderm cells migrate toward accumulations of epithelial cells at the oral and nasal aspects of the shelf, where they undergo apoptosis(Carette and Ferguson, 1992; Cuervo and Covarrubias, 2004; Martinez-Alvarez et al.,2000b). Loss of the periderm layer has been proposed to be important for reinforcing the adhesion between the shelves mediated by the underlying basal cell layer (Fitchett and Hay, 1989; Nawshad et al.,2004).

Members of the Snail gene family play a central role in the patterning of vertebrate embryos. These genes drive epithelialmesenchymal transitions throughout development by directly repressing the transcription of genes encoding components of cellcell adhesive complexes in epithelia(Nieto, 2002). However,members of the Snail gene family also have roles in other processes, such as the protection of cells from programmed cell death, the establishment of left-right asymmetry and the regulation of cell motility (reviewed by Barrallo-Gimeno and Nieto,2005). Although at least some Snai2^-/- mice are viable and fertile (Jiang et al.,1998b), they exhibit an interesting array of phenotypes, including an increased sensitivity in the hematopoietic lineages to induction of apoptosis by ionizing radiation (Inoue et al., 2002; Inukai et al.,1999). We have recently shown that conditional deletion of the Snai1 gene in the epiblast results in randomization of left-right axis specification in mice (Murray and Gridley, 2006). Mutations of the Snai1 and Snai2genes have also been shown to interact genetically with a Twist1mutation during the formation of cranial sutures(Oram and Gridley, 2005). Given these observations, we have extended our analysis of the functions of Snai1 and Snai2 genes during development of other craniofacial structures in mice, including the secondary palate. Here we report that approximately 50% of Snai2^-/- mice present with cleft palate at birth, and the penetrance of this phenotype increases to 100% on a Snai1 heterozygous background. Moreover, embryos with conditional deletion of the Snai1 gene in neural crest cells on a Snai2 null genetic background exhibit multiple craniofacial abnormalities, including defects such as micrognathia and cleft palate that are similar to those seen in Pierre Robin Sequence patients. These results demonstrate crucial roles for Snail gene family members in palatal shelf fusion and craniofacial morphogenesis in mice.

The Snai1^del1(Carver et al., 2001) null allele, Snai2^del1 and Snai2^lacZ null alleles (Jiang et al., 1998b),and Snai1^flox conditional allele(Murray et al., 2006) have been described previously. Snai1^flox/flox mice were maintained as homozygotes and bred to Snai2^+/- mice for the experiments described in this report. Wnt1-Cre mice(Danielian et al., 1998) were obtained from the Jackson Laboratory. Typically, compound heterozygous mice carrying Snai1 and Snai2 null alleles were bred to single heterozygous Snai2 female mice to generate Snai1^+/-Snai2^-/- and Snai1^+/+Snai2^-/- progeny. Similarly, male mice harboring Wnt1-Cre, Snai1 and Snai2 alleles were crossed to Snai1^flox/floxSnai2^+/- females, and embryos were isolated at varying gestational stages. Embryos of the genotype Wnt1-Cre/+Snai1^flox/- are designated Snai1 conditional knockout (Snai1-cko), while those also null for Snai2 are designated Snai1/2 double knockout (Snai1/2-dko). ROSA26-EGFP mice(Mao et al., 2001) containing a Cre-inducible GFP reporter in the ROSA26 locus were obtained from the Jackson Laboratory. The ROSA26-EGFP mice were bred to Wnt1-Cre/+Snai1^flox/floxSnai2^+/- mice and subsequently bred to homozygosity for the ROSA26-EGFP reporter and Snai1^flox alleles. Embryos or neonatal mice were genotyped by allelespecific PCR of DNA isolated from the yolk sac or tail.

Radioactive in situ hybridization was performed essentially as described(Krebs et al., 2001), except ³³P-UTP was used instead of ³⁵S as the labeled nucleotide. Exposure times for each probe were determined empirically. Radioactive probes were made using the Promega (Madison, WI) Riboprobe kit,using ³³P-labeled UTP (Perkin-Elmer, Waltham, MA). Whole-mount in situ hybridization was performed as described(Krebs et al., 2001), and digoxigenin-labeled riboprobes were generated using the Roche (Indianapolis,IN) labeling kit according to the manufacturer's instructions.

Histology, immunohistochemistry and immunofluorescence were performed on 7μm sections of paraformaldehyde-fixed, paraffin-embedded embryos. Antibodies for keratin 6 (Clone LHK6B, NeoMarkers, Fremont, CA) and phospho-histone H3 (Upstate Biotechnology, Charlottesville, VA) were diluted 1:100 in 4% goat serum/PBS. Quantitation of phospho-histone H3 staining was performed by counting equivalent areas of eight separate sections from two embryos for each age/genotype. Data is presented as the mean±s.e.m.,and P-values were computed using Student's t-test. An antigen retrieval step comprising boiling for 10 minutes in 10 mM citric acid was performed for both antibodies. For keratin 6, an Alexa Fluor 488-labeled secondary antibody (Invitrogen, Carlsbad, CA) was used at a 1:400 dilution and samples were counterstained with Hoescht 33342 (Invitrogen, Carlsbad, CA) and mounted using Vectashield hard set mounting media (Vector Labs, Burlingame,CA). Phospho-histone H3 staining was visualized using diaminobenzidine (DAB)following incubation with a peroxidase conjugated secondary antibody (Jackson ImmunoResearch, West Grove, PA) diluted to 1:2000. EGFP was visualized by immunostaining with a chicken anti-GFP antibody (Aves Labs, Tigard, OR) and Alexa Fluor 488-conjugated anti-chicken secondary antibodies. For TUNEL analysis, samples were analyzed using the In Situ Cell Detection Kit,Fluorescein (Roche, Indianapolis, IN). Coronal cryostat sections of embryonic day 14.5 (E14.5) Snai2^lacZ embryos were stained forβ-galactosidase activity, as described(Oram et al., 2003).

Newborn mutant mice and littermate controls were stained with Alcian Blue and Alizarin Red to visualize skeletal and cartilaginous elements. Embryos were skinned, eviscerated and fixed overnight in 100% ethanol (EtOH). They were then stained overnight in 0.015% Alcian Blue, 0.005% Alizaran Red in 5%acetic acid/70% EtOH. Clearing was performed in 1% potassium hydroxide/20%glycerol, after which they were brought into 100% glycerol for photography and storage. A modification of this method omitting the Alizarin Red was used to visualize the cartilaginous skeleton of E14.5 embryos. Embryos were then washed in 70% EtOH and cleared in 1:2 benzyl alcohol:benzyl benzoate.

Mandibles of E14.5 and E17.5 embryos were removed to visualize the palate,and the embryos were fixed at 4°C with 2.5% glutaraldehyde in 0.1 M sodium cacodylate, pH 7.4. Following washes, embryos were post-fixed in 1% osmium tetroxide in 0.1 M sodium cacodylate for 2 hours at 4°C, dehydrated and dried under CO₂. Samples were then mounted, sputter coated with gold to 15 nm, and examined at 20 kV with a Hitachi 3000N scanning electron microscope.

We have described previously construction of targeted null alleles of the Snai1 and Snai2 genes, including a Snai2^lacZ targeted knock-in allele(Carver et al., 2001; Jiang et al., 1998b). Observation of the progeny of Snai2^+/- intercrosses after birth revealed a striking reduction in the number of Snai2^-/- mice at weaning. When observed shortly after birth, some Snai2^-/- progeny presented with severe abdominal distention from air accumulation and absence of milk in the stomach,both due to cleft secondary palate. Further analysis revealed that approximately 50% of Snai2^-/- neonates presented with cleft palate. Penetrance of cleft palate increased to 100% in Snai1^+/- Snai2^-/- neonatal mice, revealing functional redundancy between these two gene family members(Fig. 1A,B). Histological analysis revealed that the palatal shelves of Snai1^+/-Snai2^-/- embryos at E15.0 elevated normally and came into contact at the midline(Fig. 1C,D). However, the shelves failed to form an MES and fuse. It is likely that the subsequent growth of the head caused separation of the unfused palatal shelves, resulting in the cleft secondary palate observed at birth.

To better understand the potential roles of the Snai1 and Snai2 genes in palate development, we analyzed and compared their expression patterns during palatogenesis. Snai1 gene expression was assessed by in situ hybridization, while Snai2 expression was determined by β-galactosidase expression of the Snai2^lacZ allele(Jiang et al., 1998b). At E13.5, the Snai1 gene was expressed throughout the palatal shelf mesenchyme, with particularly high levels in the cells underlying the MEE at the anterior and along the medial aspect of the shelf mesenchyme at the posterior (Fig. 1E,F). At E14.5, strong expression was observed in the palatal mesenchyme just adjacent to the newly formed MES (Fig. 1G,H), consistent with a previous report(Martinez-Alvarez et al.,2004). Snai2 expression at E14.5 was more diffuse, but was present throughout the palate mesenchyme and epithelium(Fig. 1I,J). Interestingly, we noted an enhancement of β-galactosidase expression from the Snai2^lacZ allele on a Snai1^+/-heterozygous genetic background (Fig. 1J versus 1I), suggesting that a compensatory mechanism regulating gene expression levels may exist between these two genes.

Substantial work has implicated transforming growth factor β gene family members in palate development. Targeted deletion of the Tgfb3gene results in cleft palate in mice(Kaartinen et al., 1995), and mutations in the Tgfb3 gene have been associated with nonsyndromic cleft palate in humans (Lidral et al.,1998). The palatal shelves of Tgfb3-null mice grow,elevate and meet at the midline normally but fail to adhere and fuse(Kaartinen et al., 1995; Taya et al., 1999). This is consistent with the expression pattern of the Tgfb3 gene along the MEE, and is very similar to the phenotype of Snai2^-/- and Snai1^+/- Snai2^-/- mice at E14.5. Therefore, we examined whether alterations in Tgfb3 expression could explain the phenotypes observed in Snai2^-/- and Snai1^+/- Snai2^-/- mice. However, expression of the Tgfb3 gene was unaltered in either Snai2^-/-or Snai1^+/- Snai2^-/- embryos(Fig. 2A-C). Similarly, the Irf6 gene, which is mutated in some human cleft palate patients and is expressed throughout the palatal shelf epithelium(Knight et al., 2006; Kondo et al., 2002), was expressed normally in Snai2^-/- and Snai1^+/-Snai2^-/- mutant embryos(Fig. 2D-F).

Palatal shelf adhesion and fusion requires a series of modifications in the morphology and expression of cell surface proteins of the MEE. As the shelves approach each other, periderm cells along the apical surface of the MEE bulge and extend filopodia and lamellipodia from their surface(Martinez-Alvarez et al.,2000a; Taya et al.,1999). In Tgfb3 mutant mice, these cellular movements are defective and the shelves subsequently fail to adhere. Although Tgfb3expression appeared normal in Snai1^+/-Snai2^-/-mutants, it is possible that Snai1 and Snai2 function lies downstream of TGFβ3 function. Scanning electron microscopy revealed that both control and Snai1^+/- Snai2^-/- palatal shelves exhibited the normal presence of both bulging cells and lamellipodia on the MEE of E14.5 embryos (Fig. 3A,B). Numerous studies have highlighted the importance of apoptosis in palatal shelf fusion (Carette and Ferguson, 1992; Cuervo and Covarrubias, 2004; Martinez-Alvarez et al.,2000b). Along with their described roles in regulating EMT, the Snai1 and Snai2 genes also have been demonstrated to regulate apoptosis (Inoue et al.,2002; Inukai et al.,1999; Vega et al., 2004). Therefore, we examined whether there were any alterations in apoptosis in the developing palate of Snai1^+/- Snai2^-/- mutant embryos at E14.5. Control embryos displayed substantial apoptosis along the MES(Fig. 3E, arrowheads),concentrated at the oral and nasal aspects of the seam in accumulations of epithelial cells termed epithelial triangles(Fig. 3C,E, arrows). In mutant embryos, however, we observed a dramatic reduction in the number of apoptotic cells at the intersection of the two palatal shelves, suggesting that failure to undergo normal programmed cell death might account for the cleft palate phenotype in Snai1^+/- Snai2^-/- mutants(Fig. 3D,F). Although an occasional TUNEL-positive cell along the MEE of mutant shelves was observed,comprehensive analysis of these embryos (n=4, at least 20 sections per embryo along the anteroposterior axis) revealed that this was a rare event. Indeed, apoptosis of periderm cells has been observed before MES formation, prompting the suggestion that removal of the periderm is necessary for adhesion (Fitchett and Hay,1989). Countering this model is the observation using cell surface labeling techniques that periderm cells migrate toward the epithelial triangles after adhesion, where they undergo extensive apoptosis(Cuervo and Covarrubias, 2004). Thus, it is likely that a combination of migration to epithelial triangles and apoptosis are required for the proper formation of the MES per se, a structure that is not formed in Snai1^+/- Snai2^-/- mutant embryos. Periderm cells marked by expression of keratin 6(Mazzalupo and Coulombe, 2001)were observed migrating into the epithelial triangles of control embryos(Fig. 3G), while in mutant embryos (Fig. 3H) keratin 6-expressing cells were confined to a discrete domain along the medial aspect of the adjacent palatal shelves. Just before fusion, K6-positive cells were present on the medial aspects of the palatal shelves of both control and mutant embryos (data not shown). Migration of periderm cells is thought to play an important role in epithelial fusions(Mazzalupo and Coulombe,2001), and a defect in periderm cell migration thus represents a logical explanation, along with a concomitant reduction in apoptosis, for the fusion defect observed in Snai1^+/- Snai2^-/-mutant palates.

Although Snai1^+/- Snai2^-/- mutant mice exhibit palate defects that apparently result from a failure of MEE fusion,its expression pattern suggested a possible role for the Snai1 gene in the mesenchymal component of the palatal shelf. As the early lethality of Snai1^-/- embryos precludes examination of the role of this gene in palate development, we employed a conditional Snai1^flox allele(Murray et al., 2006). We used the Wnt1-Cre line (Danielian et al., 1998) to delete the Snai1 gene in palatal shelf mesenchyme, because an earlier study had shown Wnt1-Cre to be an effective deleter in the vast majority of palatal mesenchyme cells(Ito et al., 2003). Although the Snai1 gene has been proposed to play an important role in the formation of the neural crest, we have shown that, surprisingly, Snai1 function is not required for neural crest cell delamination and migration through E9.5 in mice (Murray and Gridley, 2006). Thus, this approach permitted us to test both the role of the Snai1 gene in palatal shelf growth and fusion and its role in the development of neural-crest-derived structures after E9.5. In order to determine the effectiveness of Snai1 deletion in the appropriate cell types, we analyzed Snai1 expression by whole-mount in situ hybridization. Snai1 expression in the branchial arches was completely lost in Wnt1-Cre Snai1^flox/- (designated Snai1-cko) embryos by E9.5(Fig. 4A,B). Surprisingly, the Snai1-cko mice were viable, fertile and exhibited no obvious phenotypic abnormalities (data not shown).

As we did not observe any obvious phenotypic abnormalities in Snai1-cko mice, we bred the Snai2^+/-allele into this cross and examined the effects of compound mutant alleles on palate development. We observed the expected frequencies of cleft palate in Snai2^-/- and Snai1^+/-Snai2^-/- mice. However, Wnt1-Cre Snai1^flox/- Snai2^-/-mice (hereafter designated Snai1/2-dko) exhibited an extensive array of craniofacial defects, including cleft palate. We observed a striking difference in the phenotype of the developing palate in Snai1/2-dko embryos compared with Snai1^+/-Snai2^-/- embryos. Rather than elevating normally and failing to fuse, as the palatal shelves of Snai1^+/- Snai2^-/- embryos did, the palatal shelves of Snai1/2-dko embryos remained in their vertical growth orientation and failed to elevate(Fig. 4C,D). Other craniofacial defects observed in Snai1/2-dko neonates included presence of an abnormal mandible that was significantly shorter than that of control mice (Fig. 4E,F). The mandible of the Snai1/2-dko neonates appeared to be missing the rostral portion of the Meckel's cartilage and was fused at the midline(Fig. 4G,H). In addition, Snai1/2-dko mice had a dome-shaped skull, shortened parietal bones and an enlarged frontal foramen (Fig. 4I,J). This spectrum of defects is also seen in mice with mutations in neural crest regulatory genes such as Alk2(Acvr1 - Mouse Genome Informatics)(Dudas et al., 2004), AP-2a (Tcfap2a - Mouse Genome Informatics)(Brewer et al., 2004) and Tgfbr2 (Ito et al.,2003). Our results demonstrate that although the Snai1and Snai2 genes play no detectable role in the early generation and emigration of the mouse neural crest(Murray and Gridley, 2006),these genes are important for proper differentiation and patterning of cranial neural-crest-derived structures.

One possible cause for the arrest in palate development before shelf elevation is disruption in the normal growth and/or patterning of the developing shelves. To examine this hypothesis, we analyzed the expression of genes known to play important roles in palate development before shelf fusion. The paired class homeodomain transcription factor Pax9 is crucial for proper palatal shelf patterning, and Pax9-null mice display impaired palatal shelf elevation (Peters et al.,1998). At E13.5, the Pax9 gene is expressed in the palatal shelf mesenchyme and in the condensing mesenchyme surrounding the developing molars (Fig. 5A),and a similar pattern was observed in Snai1/2-dkoembryos at both E13.5 and 14.5 (Fig. 5B,C). Similarly, no difference between control and mutant embryos was noted in the expression of the Msx1 gene(Fig. 5D-F), which is crucial for normal growth of the anterior palatal shelves (Zhang et al., 2002). Expression of the Shh gene is dynamic in the developing palate, and is restricted to a small domain of thickened oral epithelium at E13.5(Fig. 5G), where it plays a role in providing a mitogenic signal to the adjacent mesenchyme. In Fgf10^-/- and Fgfr2^-/- mice, Shh expression is disrupted, concomitant with a decrease in proliferation in the nascent palatal shelves(Rice et al., 2004). In Snai1/2-dko mice, the Shh gene was expressed in the proper domain at both stages, but at a somewhat reduced intensity (Fig. 5H,I). The Osr1 gene is expressed in a dynamic mediolateral pattern in the developing palatal shelf and has been suggested to play an important role in palatal shelf patterning (Lan et al.,2004). Control embryos displayed Osr1 expression in the lateral aspects of the vertical palatal shelf(Fig. 5J), while staining in Snai1/2-dko mutant shelves at E13.5 and 14.5 was reduced somewhat, although some expression was noted in the proper domains(Fig. 5K,L). Given the role of the Shh gene in regulating cell division(Ingham and McMahon, 2001), we analyzed cell proliferation in control and mutant embryos using a marker of mitosis, phospho-histone H3. When compared to an equivalent area of E13.5 control palatal shelves (78.4±10), we could not detect any significant changes in the numbers of mitotic cells in Snai1/2-dko palatal shelves at either E13.5(80.2±10, P=0.28) or E14.5 (85.5±15, P=0.15)(Fig. 6). Thus, the subtle change in Shh expression observed in Snai1/2-dko mice does not appear to result in any major perturbation of the proliferative potential of the developing palatal shelves. These results demonstrate that there are no major disruptions in Snai1/2-dko mutants in the expression of several genes central to the control of palatal shelf patterning.

Another explanation for the phenotype observed in Snai1/2-dko embryos is a defect in the migration and subsequent population of neural crest cells in craniofacial structures. Although we have shown previously that neither the Snai1 gene nor the Snai2 gene plays an essential role in the delamination and initial migration of neural crest cells (Murray and Gridley, 2006), we performed lineage-tracing analyses on E11.5, 12.5 and 14.5 embryos utilizing ROSA26-EGFP reporter mice(Mao et al., 2001). We observed no obvious differences in the distribution of EGFP-labeled cells in Snai1/2-dko embryos at any of these stages (see Fig. S1 in the supplementary material), indicating that there are no major disruptions of neural crest cell migration in Snai1/2-dko embryos. It is notable that subtle differences in the size and shape of the anterior palatal shelves were observed in some Snai1/2-dko mutants(Fig. 5E, Fig. 6D; see Fig. S2 in the supplementary material). These differences were apparent in only a subset of Snai1/2-dko mutant embryos (see Fig. S2 in the supplementary material), as there appears to be considerable variability in the extent of outgrowth in control anterior palatal shelves (see Fig. S2A versus S2C in the supplementary material). Although we did not observe any differences in cell proliferation between control and mutant embryos, we cannot exclude the possibility that delayed outgrowth of the anterior palatal shelves plays a role in the cleft palate phenotype of Snai1/2-dkoembryos.

As patterning and growth of the palatal shelves in Snai1/2-dko mice appeared to be essentially normal, we examined whether the palatal shelf defects might be secondary to the craniofacial defects noted at birth. Mandibular malformations, such as micrognathia, are proposed to be the underlying cause of cleft palate in several mouse mutants, such as Egfr^-/-(Miettinen et al., 1999), Hoxa2^-/-(Gendron-Maguire et al., 1993)and Dmm (Ricks et al.,2002) mutant mice. Because the forward growth of the mandible driven by the extension of the neural-crest-derived Meckel's cartilage provides the mechanism to lower the tongue and permit palatal shelf elevation,any defects in this process could result in a cleft palate phenotype similar to that observed in Snai1/2-dko mice. In light of the mandibular abnormalities observed at birth(Fig. 4F,H), we performed Alcian Blue staining of the embryonic cartilage at E14.5 to determine if defects in Meckel's cartilage could be observed at this crucial time point. Meckel's cartilage was dramatically shorter in Snai1/2-dkoembryos compared with control embryos (Fig. 7). Therefore, we propose that that the underlying cause of cleft palate in Snai1/2-dko embryos is abnormal development of Meckel's cartilage, which in turn prevents the normal extension of the tongue and elevation of the palatal shelves.

This study describes multiple roles for the Snail family genes Snai1 and Snai2 during palate development and craniofacial morphogenesis in mice. Our data suggest a role for the Snai2 gene in formation of the MES and/or palatal shelf adhesion. Modulation of the penetrance of this phenotype by reduction of Snai1 gene dosage supports the notion that these genes have redundant roles in vivo, consistent with our observations that Snai2^lacZ expression increases on a Snai1^+/- genetic background. We also demonstrate a genetic interaction between mutations in the Snai1 and Snai2genes in the formation of neural-crest-derived craniofacial structures. The co-expression and genetic interaction clearly demonstrate functional redundancy and compensatory activity between the Snai1 and Snai2 genes.

Martinez-Alvarez et al.(Martinez-Alvarez et al.,2004) previously described high levels of Snai1 gene expression in palatal shelf mesenchyme, and noted that a few MES cells express the Snai1 gene during the fusion process, thus supporting a small role for EMT in MES dissolution. Interestingly, they also observed a dramatic increase in Snai1 expression in the MES of Tgfb3^-/- mutant palates, and proposed a model in which increased compensatory expression of the Tgfb1 gene induces Snai1 expression, thereby protecting these cells from apoptosis(Barrallo-Gimeno and Nieto,2005; Vega et al., 2004). This suggests a model in which the Snai1 gene plays an important role in regulating the survival of MEE and MES cells. We did not observe an increase in apoptosis in the mutant palatal shelves, but instead observed a complete absence of cell death. This might be secondary, however, to a more fundamental defect in palatal shelf fusion. In support of this notion, we find that periderm cells remained at the intersection of the two MEE layers, and did not migrate toward the oral and nasal epithelia.

A crucial component of palate development is the adhesion of the palatal shelves and subsequent transformation of the separate MEE into a single MES(Ferguson, 1988). Normally, as the palatal shelves approach each other, periderm cells of the bilayered MEE bulge, form lamellipodia and express extracellular matrix proteins important for adhesion on their surface. After contact is initiated, periderm cells migrate to the oral and nasal aspects of the forming MES and undergo apoptosis(Cuervo and Covarrubias, 2004). The basal epithelial cells intercalate to form a single layer MES, which then disappears as the basal cells also commit to an apoptotic program. Several mouse models of cleft palate exhibit defects in this process, the classic example being the Tgfb3-null mouse(Kaartinen et al., 1995; Taya et al., 1999). In these embryos, palatal shelves elevate normally, but periderm cells fail to bulge and do not form lamellipodia. Our results show that Snai1^+/-Snai2^-/- mutant embryos display normal expression of the Tgfb3 gene and the presence of bulging cells and lamellipodia on the surface of the MEE. However, our data demonstrate that Snai1^+/-Snai2^-/- embryos lack both the apoptosis normally observed at the MES and the migration of periderm cells to form epithelial triangles. It had been suggested that cell death of the periderm layer is crucial for the proper intercalation and adhesion of the underlying basal cell layer(Fitchett and Hay, 1989). However, in vitro labeling experiments show that a substantial portion of the periderm migrates to epithelial triangles at the nasal and oral aspects of the MES (Carette and Ferguson,1992), where they then undergo apoptosis(Cuervo and Covarrubias, 2004; Cuervo et al., 2002). Thus, a model emerges wherein adhesion stimulates migration of the periderm, which in turn allows for intercalation of the basal epithelial cells and formation of the mature MES. One prediction from this model is that defects in periderm cell migration would short-circuit the program and result in a lack of periderm cell death and basal cell intercalation. Our data are consistent with a failure of periderm cell migration in Snai1^+/-Snai2^-/- mutants. The absence of apoptosis could then be attributed to the persistence of the periderm on the MEE.

Could a failure to undergo EMT following MEE adhesion contribute to the cleft palate phenotype of Snai1^+/- Snai2^-/-embryos? Using fate-mapping analysis, Vaziri Sani et al. claimed that the absence of persistently labeled cells in the palatal shelf mesenchyme precluded EMT as a possible mechanism for palatal shelf fusion (Vaziri Sani et al., 2005). However, the data presented did not address whether MES cells could be observed undergoing EMT at the early stages of fusion, only that they did not ultimately contribute to the mesenchymal core. It is possible that EMT might occur in cells fated to undergo apoptosis, thus escaping identification in fate-mapping studies. Indeed, the migration of periderm cells presumably requires modification of their epithelial phenotype in order to migrate toward the epithelial triangles. Thus, it remains plausible that EMT is occurring at some point during the process of MES dissolution, although these cells ultimately do not contribute in a long-term manner to the palate mesenchyme.

We have reported previously the surprising result that neither the Snai1 nor the Snai2 gene, alone or in combination, is required for neural crest cell generation and delamination in mice, at least through 9.5 days of gestation (Murray and Gridley, 2006). It had been widely proposed that the Snai1 gene would play an essential role in neural crest delamination in mice, based on the observations that: (1) Snai1 is expressed in the dorsal neural folds concomitant with neural crest cell delamination(Locascio et al., 2002; Sefton et al., 1998); (2)ectopic Snai1 gene expression induces EMT in cultured epithelial cells (Batlle et al., 2000; Cano et al., 2000); (3) the Snai2 gene plays a crucial role in neural crest cell delamination in the chick (Nieto et al.,1994); (4) Snai2^-/- mice lack an obvious neural crest phenotype (Jiang et al.,1998b); and (5) the Snai1 and Snai2 genes are expressed in a modular fashion across vertebrate species(Locascio et al., 2002). The results presented here reveal an entirely different role for the Snai1 and Snai2 genes in craniofacial morphogenesis. While loss of a single Snai1 allele in the context of a Snai2-null background (Snai1^+/- Snai2^-/- mice) results in palatal clefting due to defects in periderm cell migration and apoptosis,removing the remaining Snai1 allele in the neural crest(Snai1/2-dko mice), which populates a majority of the palatal shelf mesenchyme, results in a very different cleft palate phenotype. Marker analysis revealed only very subtle alterations in palatal shelf patterning, including a small but clear decrease in the expression of Shh in the oral epithelium and Osr1 in the mesenchyme. Indeed, these changes may be related to the apparent small differences in anterior palatal shelf outgrowth observed in Snai1/2-dkoembryos compared to some controls. It is unlikely, however, that these changes are ultimately responsible for the observed cleft palate phenotype. In addition to cleft palate, Snai1/2-dko embryos also exhibit other craniofacial abnormalities, including micrognathia, fused mandible and an enlarged parietal foramen in the skull vault. This combination of phenotypes has been observed in other conditional neural crest mutants, such as Alk2 (Dudas et al.,2004), AP-2a (Brewer et al., 2004) and Tgfbr2(Ito et al., 2003), and demonstrates that although the Snai1 and Snai2 genes are not essential in the initial delamination and migration of neural crest cells from the neural tube, they play a crucial role in the later development of neural-crest-derived structures.

The Pierre Robin Sequence (also termed Robin Sequence) is a human developmental malformation characterized by mandibular retrognathia (normal size receded mandible) or micrognathia (abnormally small mandible),glossoptosis (rearward and downward displacement of the tongue) and cleft palate. Most newborns with Pierre Robin Sequence also exhibit respiratory and feeding difficulties. Cases of Pierre Robin Sequence are both phenotypically and genetically heterogeneous (Cohen,1999; Houdayer et al.,2001; Jakobsen et al.,2006; Jamshidi et al.,2004; Melkoniemi et al.,2003; Ounap et al.,2005; Ricks et al.,2002). Pierre Robin Sequence can occur as an isolated nonsyndromic form, or can occur in combination with other syndromes. Mutations in several collagen genes, including COL2A1, COL11A1 and COL11A2, have been found in subjects with nonsyndromic Pierre Robin Sequence(Melkoniemi et al., 2003),although it has not been established definitively that these mutations are causal for the observed malformations. In addition, chromosomal translocations(Jamshidi et al., 2004),duplications (Ounap et al.,2005) and deletions (Houdayer et al., 2001) have been described that are associated with nonsyndromic Pierre Robin Sequence.

A model for the etiology of the cleft palate defect in individuals with Pierre Robin Sequence is that the small and/or abnormally positioned mandible prevents the normal positioning of the tongue, thereby inhibiting palatal shelf elevation and fusion. This model is supported by analyses of two mouse models of Pierre Robin Sequence, the A/WySn inbred mouse strain(Schubert et al., 2005) and mice heterozygous for the disproportionate micromelia (Dmm) mutation(Ricks et al., 2002), a semidominant mutant allele of the Col2a1 gene(Pace et al., 1997). Based on a mechanical requirement for mandibular extension to drive the proper depression of the tongue, we propose that a similar mechanism is responsible for the cleft palate phenotype of Snai1/2-dko mice. This interpretation is supported by the dramatic growth retardation of Meckel's cartilage at E14.5, the crucial time for palatal shelf elevation. Snai1/2-dko mice represent a novel model of Pierre Robin Sequence, and should provide a useful tool for the study of this heterogeneous developmental disorder in humans.

We thank Rulang Jiang for providing probes, and members of the Gridley lab for comments and suggestions. This work was supported by a grant from the American Cancer Society to S.A.M. (PF-04-245-01-DDC) and grants from the NIH to T.G. (HD034883) and the Jackson Laboratory (CA034196).