Palatogenesis: morphogenetic and molecular mechanisms of secondary palate development

In mammals, the palate separates the oral from the nasal cavity and consists anteriorly of the bony hard palate and posteriorly of the muscular soft palate. The hard palate is crucial for normal feeding and speech, whereas the soft palate is movable and closes off the nasal airway during swallowing. Palatogenesis, which is the developmental process that generates the intact palate, is often disrupted by genetic and environmental perturbations, as reflected in the high frequency of cleft palate, a major congenital anomaly in humans that requires surgical intervention shortly after birth and has significant long-term health implications for affected individuals. Several decades of extensive epidemiological and human genetic studies have begun to reveal the genetic and environmental factors that underlie the etiology of cleft palate. In recent years, there has been an explosion of new knowledge about the genetic pathways coordinating palatogenesis. Much of this progress is due to the widespread application of sophisticated genetic manipulations in mice and detailed morphological and molecular analyses of mutant mouse models. Integrating these studies with experimental embryology approaches has revealed that palatogenesis is regulated by an extensive network of signaling molecules and transcription factors and involves abundant crosstalk between several distinct molecular pathways and cell types.

Here, we review recent advances in our understanding of the morphogenetic and molecular mechanisms of palatogenesis, focusing mainly on genetic studies in the mouse and briefly discussing contributions from other model systems. Many of the genes we discuss have been implicated in cleft palate in humans (Table 1) and several recent review articles provide excellent references for understanding the clinical and genetic aspects of cleft palate (Stanier and Moore, 2004; Gritli-Linde, 2007; Gritli-Linde, 2008; Dixon et al., 2011). We focus on how molecular pathways and cellular processes are integrated in the regulation of critical steps of palate morphogenesis. In addition, we highlight new technological approaches and genetic models that are being used for the study of palate development, and briefly discuss the implications that these studies have for an understanding of the fundamental principles of morphogenesis.

In mammals, development of the face begins with the formation of five facial prominences surrounding the primitive mouth: the frontonasal prominence on the rostral side, a pair of maxillary prominences laterally and a pair of mandibular prominences caudally. These facial prominences are populated by cranial neural crest cells that originate at the dorsal edge of the developing rostral neural tube. As development proceeds, the frontonasal prominence is divided into the medial and lateral nasal processes by the formation of nasal pits. Subsequent morphogenetic fusion of the lateral and medial nasal processes forms the nostril, and fusion between the medial nasal processes and the maxillary prominences forms the upper lip (reviewed by Jiang et al., 2006). As the medial nasal processes fuse with the maxillary prominences, the presumptive primary and secondary palates first become morphologically visible as outgrowths from the oral side of the medial nasal and maxillary processes, respectively (Fig. 1A,B). The secondary palate arises as paired outgrowths, which in mammals initially grow vertically flanking the developing tongue (Fig. 1B,C,G,H,L,M,Q,R) and subsequently reorient to the horizontal position above the dorsum of the tongue in a process known as palatal shelf elevation (Fig. 1D,I,N,S). Following elevation, the paired palatal shelves grow towards the midline where they meet and fuse with each other (Fig. 1D-F,I-K,N-P,S-U). The fusion of palatal shelves involves the formation of a midline epithelial seam and its subsequent disintegration to allow mesenchymal confluence (Fig. 1O,P). In addition, the secondary palate fuses anteriorly with the primary palate and anterodorsally with the nasal septum, both of which are derived from the medial nasal processes, to form the intact roof of the oral cavity. In humans, palatogenesis is initiated in the sixth week and palatal fusion is complete by 12 weeks of gestation. In mice, palatal outgrowths are first detectable by embryonic day (E) 11.5 and palatal fusion is complete by E17.

The palatal shelves are composed of mesenchyme derived mainly from the neural crest (Ito et al., 2003) surrounded by a thin layer of oral epithelium, and they exhibit distinct stereotyped shapes along the anterior-posterior (AP) axis. For example, at E13.5, the anterior part of the mouse palate exhibits a finger-like diagonal projection into the oral cavity (Fig. 1H), whereas the middle palatal region, which is flanked by the developing molar tooth germs, is triangular in shape (Fig. 1M), and the posterior part of the palate, the presumptive soft palate, is vertical with a rounded distal end (Fig. 1R). Recent studies have revealed that the growth of the palatal shelves is controlled by reciprocal epithelial-mesenchymal interactions and involves distinct molecular mechanisms along the AP axis.

Although the initiating signal has not yet been revealed, recent studies have identified key components of the signaling pathways that drive palatal shelf outgrowth. Sonic hedgehog (Shh), a secreted protein expressed throughout the early oral epithelium, is a key early signal that drives palatal shelf outgrowth (Rice et al., 2006). Studies of mice with epithelium-specific inactivation of Shh or mesenchyme-specific inactivation of smoothened (Smo), which encodes a transmembrane protein required for transducing Shh signaling, have demonstrated that Shh signals from the epithelium to the underlying mesenchyme to promote palatal cell proliferation and outgrowth (Rice et al., 2004; Lan and Jiang, 2009). Consistent with this, exogenous Shh application induces a mitogenic response in palatal explant cultures (Zhang et al., 2002; Rice et al., 2004). This mitogenic effect is mediated, in part, by the cell cycle regulators cyclin D1 and cyclin D2, the expression of which was reduced in the palatal mesenchyme of embryos with mesenchyme-specific inactivation of Smo. The expression of the forkhead box F1a and F2 (Foxf1a and Foxf2) and odd-skipped related 2 (Osr2) genes was also reduced in the palatal mesenchyme, indicating that these transcription factors might be downstream effectors of Shh signaling in this context (Lan and Jiang, 2009). Interestingly, Smo function was not required in the epithelium for palatal outgrowth (Rice et al., 2004), but its disruption in the palatal mesenchyme affected palatal epithelial cell proliferation, indicating that Shh signaling is necessary to activate a mesenchymal signal that regulates palatal epithelial cell proliferation (Lan and Jiang, 2009).

Fibroblast growth factor 10 (Fgf10) is a crucial mesenchymal signal that is required for palatal outgrowth. Mice homozygous for a null mutation in either Fgf10 or the gene encoding its receptor, fibroblast growth factor receptor 2b (Fgfr2b), exhibited cleft palate with impaired palatal shelf outgrowth (Rice et al., 2004). Whereas Fgf10 mRNA expression was restricted to the mesenchyme, Fgfr2b mRNA was most abundantly detected in the overlying epithelium. Fgfr2 function is required within the epithelium, as mice harboring an epithelial-specific deletion of Fgfr2 also exhibited cleft palate (Hosokawa et al., 2009). Both epithelial and mesenchymal cell proliferation were reduced in the absence of either Fgf10 or Fgfr2b, however, suggesting the presence of a factor that signals from the epithelium back to the underlying mesenchyme and that is dependent on Fgf10/Fgfr2b signaling. Indeed, Shh expression was dramatically reduced in the epithelium of Fgf10^–/– and Fgfr2b^–/– embryos, implying that the decreased palatal mesenchymal cell proliferation observed in these mutants might be a consequence of reduced Shh expression in the epithelium (Rice et al., 2004). The expression of Fgf10 was also reduced in the palatal mesenchyme of embryos lacking mesenchymal Smo, indicating that Shh and Fgf10 function in a positive-feedback loop (Fig. 2A) that drives the outgrowth of the palatal shelves (Lan and Jiang, 2009).

Crosstalk between the Shh and bone morphogenetic protein (Bmp) signaling pathways has also been detected (Fig. 2A). The loss of Smo within the palatal mesenchyme led to upregulation of Bmp4 and downregulation of Bmp2 (Lan and Jiang, 2009). The positive regulation of Bmp2 by Shh signaling is consistent with the finding that exogenous Shh-containing beads induce Bmp2 expression in palatal explant culture (Zhang et al., 2002). Exogenous Bmp2 can also positively regulate cell proliferation in the palatal mesenchyme (Zhang et al., 2002). Whereas complete inactivation of Bmp4 resulted in early embryonic lethality, ablation of Bmp4 function in the maxillary mesenchyme and throughout the oral epithelium (in Nestin-Cre; Bmp4^f/– mice) caused a cleft lip phenotype, but no secondary palate defect was reported (Liu et al., 2005). Overexpression of the Bmp antagonist noggin specifically in the palatal mesenchyme led to retarded palatal growth and cleft palate in mice (Xiong et al., 2009), further supporting the involvement of Bmp signaling during palatal growth.

Recent studies indicate that Bmp signaling during palatogenesis occurs via the type I Bmp receptor Bmpr1a. Disruption of Bmpr1a in the maxillary mesenchyme and throughout the oral epithelium (in Nestin-Cre; Bmpr1a^f/– mice) resulted in cleft lip and palate (Liu et al., 2005). The epithelial-specific inactivation of Bmpr1a (in K14-Cre; Bmpr1a^f/– conditional mutant mice) did not result in a cleft palate phenotype, indicating that Bmpr1a function is likely to be required in the palatal mesenchyme (Andl et al., 2004). Conditional disruption of Bmpr1a in the neural crest and its derivatives (in Wnt1-Cre; Bmpr1a^f/– embryos) caused dramatic retardation of the anterior region of the palatal shelves, whereas the posterior palate was less affected (Li et al., 2011). This phenotype was accompanied by other severe craniofacial abnormalities, including shortened mandible and hypoplastic maxillae, raising the question as to whether cleft palate in this situation resulted from a primary defect in palatogenesis or was secondary to earlier deficiencies in the maxillary prominences. Most recently, the disruption of Bmpr1a function specifically within the palatal mesenchyme was achieved using Osr2-IresCre; Bmpr1a^f/f mice. These mutant embryos exhibited anteriorly restricted cleft palate and reduced cell proliferation in the anterior palatal mesenchyme (Baek et al., 2011). Interestingly, cell proliferation was also significantly reduced in the developing primary palate and expression of Shh was downregulated in both the primary palate and anterior secondary palate in the Osr2-IresCre; Bmpr1a^f/f embryos, suggesting that interplay between Bmp and Shh signaling regulates both primary and secondary palate outgrowth (Baek et al., 2011). Furthermore, loss of noggin function resulted in cleft palate associated with aberrant apoptosis in the palatal epithelium and reduced cell proliferation of the anterior palatal mesenchyme (He et al., 2010b), indicating that Bmp signaling must be tightly regulated during palatogenesis.

In addition to classical signaling molecules, a recent study suggests that extracellular matrix (ECM) proteins participate in the regulation of palatal growth. Simultaneous disruption of the genes encoding two ADAMTS family metalloproteases, Adamts9 and Adamts20, resulted in cleft palate with defects in early outgrowth, elevation and approximation of the palatal shelves (Enomoto et al., 2010). These secreted metalloproteases bind to the cell surface where they are actively involved in pericellular ECM proteolysis. A major substrate for these proteases is versican, a proteoglycan with space-filling properties. The cleavage of versican was indeed reduced in Adamts9^+/–; Adamts20^bt/bt compound mutants. Interestingly, simultaneous disruption of versican and Adamts20 function also resulted in reduced palatal cell proliferation (Enomoto et al., 2010). It is possible that proteolysis of ECM molecules such as versican might produce bioactive fragments with growth-promoting activity. Alternatively, proteolysis of the ECM components might indirectly affect the distribution or activity of classical signaling molecules that regulate palatal growth. Further studies are necessary to test these possibilities.

Although the palatal shelves appear as continuous outgrowths that project from the maxillary processes into the oral cavity, gene expression studies have revealed remarkable molecular heterogeneity along the AP axis, and this heterogeneity is evident from the early stages of palatal outgrowth (Hilliard et al., 2005; Li and Ding, 2007; Welsh and O'Brien, 2009). Several transcription factor genes, including BarH-like homeobox 1 (Barx1), meningioma 1 (Mn1), Msh homeobox 1 (Msx1), mesenchyme homeobox 2 (Meox2), short stature homeobox 2 (Shox2) and T-box transcription factor 22 (Tbx22) are differentially expressed along the AP axis of the developing palatal shelves (Fig. 2A,B). The expression of Msx1 and Shox2, for example, is restricted to the anterior part of the shelves, whereas Meox2 and Tbx22 mRNAs are restricted to the posterior region, with the AP gene expression boundary coinciding with a morphological landmark: the first formed palatal rugae (Zhang et al., 2002; Hilliard et al., 2005; Yu et al., 2005; Li and Ding, 2007; Pantalacci et al., 2008; Welsh and O'Brien, 2009). Barx1 and Mn1 mRNAs are also preferentially expressed in the posterior palate but their expression domains extend into the anterior half of the palatal shelves during palatal outgrowth (Liu et al., 2008; Welsh and O'Brien, 2009). Although mice lacking either Msx1 or Mn1 exhibited complete cleft palate, Msx1^–/– mice exhibited specific cell proliferation defects in the anterior region, whereas Mn1^–/– mice showed growth deficits in only the middle and posterior regions of the palatal shelves (Zhang et al., 2002; Liu et al., 2008). Mice lacking Shox2 exhibited a cleft within the anterior palate, whereas the posterior palate fused normally, demonstrating a specific requirement for Shox2 in anterior palatal outgrowth (Yu et al., 2005). Tbx22^null mice displayed cleft palate, with the severity varying from a complete cleft palate phenotype due to decreased palatal shelf extension to submucous cleft palate in which palatal shelf elevation and fusion occurred normally (Pauws et al., 2009). Tbx22 mRNA expression in the palate was significantly reduced in Mn1^–/– mice, and Mn1 was able to activate Tbx22 expression in cell culture assays, indicating that Tbx22 acts downstream of Mn1 in the regulation of posterior palatal outgrowth (Liu et al., 2008) (Fig. 2B).

The expression of Msx1 and Shox2 in the anterior palatal mesenchyme depends on Bmp signaling, as expression of both these genes was significantly reduced in the anterior palate of Wnt1-Cre; Bmpr1a^f/– mice (Li et al., 2011). Remarkably, the addition of Bmp4-soaked beads in palatal explant cultures induced Msx1 expression in the anterior, but not in the posterior, palatal mesenchyme (Hilliard et al., 2005). Exogenous Bmp4 was insufficient to induce Shox2 mRNA expression in palatal mesenchyme explants, but the palatal epithelium from the anterior domain induced ectopic Shox2 mRNA expression in the posterior palatal mesenchyme (Yu et al., 2005). These results further highlight intrinsic differences in both the epithelium and the mesenchyme along the AP axis.

Recently, Eph/ephrin family members have been demonstrated to control anterior palatal shelf outgrowth (Bush and Soriano, 2010). These signaling molecules have the capacity for bidirectional signaling, such that a forward signal can be transduced into the cell in which the Eph receptor tyrosine kinase is expressed, and a reverse signal can be transduced into the cell in which the ephrin (Efn) is expressed. The Efnb1 gene exhibits a highly restricted expression pattern in the anterior palatal mesenchyme during all stages of palatogenesis, and Efnb1-null mice and Efnb1^+/– heterozygous females exhibit cleft palate accompanied by decreased cell proliferation in the anterior palatal mesenchyme (Davy et al., 2004; Bush and Soriano, 2010). Analyses of mice carrying a series of targeted point mutations that specifically abrogate reverse signaling while leaving forward signaling by ephrin B1 intact revealed that reverse signaling is dispensable for palatogenesis (Bush and Soriano, 2009). In addition, null mutations in the EphB2 and EphB3 receptors, or specific disruption of forward signaling through these receptors, also resulted in reduced palatal shelf proliferation and cleft palate (Orioli et al., 1996; Risley et al., 2009). These results indicate that ephrin B1/EphB forward signaling controls anterior palatal shelf outgrowth (Bush and Soriano, 2009; Risley et al., 2009; Bush and Soriano, 2010). The Efnb1 gene is X-linked and, owing to random X-inactivation, Efnb1^+/– heterozygous embryos exhibit mosaic domains of Efnb1 expression in the developing palatal shelves (Compagni et al., 2003; Davy et al., 2004; Bush and Soriano, 2010). Correspondingly, cell proliferation is affected in a correlated mosaic pattern that leads to more severe dysmorphogenesis of the palatal shelves in Efnb1^+/– heterozygotes than in null mutant embryos (Bush and Soriano, 2010). These results indicate that cell proliferation must be tightly spatially regulated across the palatal shelf for normal outgrowth and for the establishment of proper palatal shape.

Most analyses of palatal outgrowth have focused on the downward growth of the palatal shelves; however, the maxillary processes also undergo significant rostrocaudal expansion from E12.5 to E14.5, and the palatal shelves correspondingly elongate along the AP axis. As the palatal shelves elongate, the domain of Shox2 mRNA expression expands much more dramatically than that of Meox2 (Li and Ding, 2007). Recently, it was demonstrated that the AP boundary of Shox2 and Meox2 expression coincides with the first formed palatal rugae (Pantalacci et al., 2008; Welsh and O'Brien, 2009). The rugae are metameric epithelial thickenings on the oral surface of the palate that are first apparent at E12.0 and develop by periodic reiterative interposition as the palatal shelves elongate along the AP axis (Pantalacci et al., 2008). Depending on the strain background, mice develop nine or ten rugae, which are maintained after birth. The rugae are thought to fulfill a number of mechanical functions, first in nursing and later in mastication (Pantalacci et al., 2008). Importantly, whereas the second and third rugae form sequentially and anteriorly to the first ruga, the fourth ruga forms between rugae 1 and 2, and, as the palatal shelves continue to elongate along the AP axis, subsequent rugae formation only occurs between the newly formed ruga and ruga 1 (Pantalacci et al., 2008; Welsh and O'Brien, 2009). These results indicate that there is a localized rugae growth zone immediately anterior to the AP boundary during the rostral elongation of the developing palatal shelves (Welsh and O'Brien, 2009).

Several laboratories have investigated cell proliferation in the developing palatal shelves using BrdU labeling, and these studies reported that cell proliferation rates are similar in the anterior and posterior regions of the palatal shelves at E13.5 (Zhang et al., 2002; Lan et al., 2004; Li and Ding, 2007). These studies were performed before the discovery of the localized rugae growth zone, however, and it is therefore possible that the palatal mesenchyme underlying the rugae growth zone might be proliferating at a faster rate than both the anterior and posterior regions. Additionally, a proliferation-independent mechanism might contribute to localized palatal elongation. When labeled cells were followed in explant cultures, anterior palatal mesenchyme migrated laterally while posterior palatal mesenchyme migrated anteriorly in a Wnt5a- and Fgf10-dependent manner (He et al., 2008). Furthermore, the finding that anterior palatal epithelium was able to induce Shox2 mRNA expression in the posterior palatal mesenchyme in recombinant explant cultures (Yu et al., 2005) suggests that, once they have migrated into the anterior domain, posterior palatal mesenchymal cells might receive signals that activate Shox2 mRNA expression. Taken together, these data suggest that directed migration of palatal mesenchyme from the posterior to the anterior palate with concurrent reprogramming of cell AP identity might be an important part of the mechanism that drives palatal shelf outgrowth and patterning along the AP axis.

The developing palatal shelves also exhibit morphological and molecular heterogeneity along the mediolateral axis during vertical outgrowth (Fig. 2C), with the lateral side corresponding to the oral side following palatal shelf elevation. Palatal rugae begin to form on the lateral side of the developing palatal shelves at E12, and the expression of Shh becomes restricted to the lateral palatal epithelium. Osr1 and Osr2, which encode zinc-finger transcription factors, exhibit graded expression in the developing palatal mesenchyme along the mediolateral axis (Lan et al., 2004). At E13.5, Osr1 mRNA expression is restricted to the lateral side, whereas Osr2 exhibits graded expression that is strongest in the lateral mesenchyme and weaker in the medial mesenchyme (Fig. 2C). Targeted disruption of Osr2 caused cleft palate associated with reduced cell proliferation in the medial side of the developing palatal shelves and with disrupted mediolateral patterning. The requirement of Osr2 for cell proliferation in the medial but not lateral side of the palatal shelf is likely to be due to partial functional redundancy of Osr2 with Osr1. Indeed, targeted replacement of the Osr2 coding sequence with an Osr1 cDNA rescued the cleft palate in Osr2-null mice (Gao et al., 2009).

A pathway involving the distal-less homeobox 5 (Dlx5) transcription factor is also implicated in mediolateral patterning and palatal expansion. Dlx5 is co-expressed with Fgf7 in the medial mesenchyme of the palatal shelf, and Fgf7 expression was dramatically reduced in this domain in Dlx5^–/– mutant palatal shelves (Han et al., 2009). In these mutant embryos, Shh expression expanded to more medial palatal epithelium, possibly as a consequence of the loss of Fgf7, as exogenous Fgf7 could inhibit Shh expression in palatal explant cultures (Han et al., 2009). Although Dlx5^–/– mutant palate shelves elevated and fused, the oral aspect of the palate was dramatically expanded and a malformed soft palate was observed. Interestingly, whereas Msx1^–/– mice exhibited reduced Shh expression in the anterior palate, Dlx5^–/–;Msx1^–/– compound mutant embryos exhibited Shh expression in the medial palatal epithelium, which was able to overcome the cell proliferation defects associated with Msx1 loss-of-function (Han et al., 2009). Together, these results identify a new pathway involving Dlx5 and Fgf7 in the regulation of mediolateral patterning and palate growth (Fig. 2C). Since no obvious palatal abnormalities were detected in mice lacking Fgf7 (Guo et al., 1996), however, it is likely that another signaling molecule acts downstream of Dlx5, possibly in concert with Fgf7, to regulate Shh expression in the palatal epithelium.

At ∼E14.0, the palatal shelves elevate to the horizontal position above the tongue. Because palatal shelf elevation occurs rapidly during midgestation, the exact molecular mechanisms have eluded characterization. Several hypotheses for palatal shelf elevation have been under consideration for decades, including: (1) rapid rotation of the shelves (Fig. 3A,B); and (2) growth-based regression of the distal portion of the shelves and outgrowth in the horizontal direction (Lazzaro, 1940). In 1956, Walker and Fraser noted that the process of palatal shelf elevation was too rapid to be explained by growth changes and proposed instead that the palatal shelves rapidly remodel themselves by a bulging of the medial wall and regression of the distal end (Fig. 3C,D) (Walker and Fraser, 1956). Subsequently, Coleman proposed that the mechanism of palatal shelf elevation is heterogeneous along the AP axis, whereby the anterior (rostral) part of the palatal shelves elevate by a swinging ‘flip-up’ process, while the posterior (caudal) and middle parts of the palatal shelves reorient through an oozing‘flow’ remodeling mechanism (Fig. 3) (Coleman, 1965; Ferguson, 1988). The elevation of the palatal shelves often occurs asynchronously, with one palatal shelf elevating before the other (Lazzaro, 1940; Walker and Fraser, 1956). The physical forces driving this rapid change in position and shape are still somewhat unclear. Several hypotheses suggesting the involvement of the ECM have been proposed but none has been adequately proven (Brinkley and Bookstein, 1986; Ferguson, 1988). The palatal rugae were also hypothesized to play a role in palatal shelf elevation by serving as structural supports that stiffen the palatal shelves or by providing the force that drives elevation (Orban et al., 1972; Brinkley, 1980; Luke, 1984). Recent studies indicate that rugae are not required for elevation, however, as ablating canonical Wnt signaling in the palatal epithelium caused a complete loss of rugae formation but did not disrupt outgrowth and elevation of the palatal shelves (He et al., 2011).

A recent study investigated the palatal shelf elevation process by taking advantage of a mutant mouse strain with delayed palatal shelf elevation (Jin et al., 2010). Although they exhibit no apparent deficits in palatal cell proliferation or patterning, mouse embryos homozygous for a targeted disruption of the zinc-finger homeobox 1a (Zfhx1a, official nomenclature Zeb1) gene consistently exhibited delayed palatal shelf elevation by 24-48 hours (Jin et al., 2008). Thus, at E14.5, whereas the palatal shelves of control littermates are elevated, the palatal shelves of Zfhx1a^–/– mutants remain vertical and express markers of the medial edge epithelium (MEE) along the medial side of the palatal epithelium. These results suggest that the MEE of post-elevation palatal shelves corresponds to the medial rather than the distal region of the pre-elevation palatal epithelium (Fig. 3C,D), supporting the hypothesis of Walker and Fraser that remodeling and horizontal outgrowth from the medial wall of the vertical palatal shelves give rise to the ‘elevated’ palatal shelves (Walker and Fraser, 1956). A recent histomorphological reexamination of this process (Yu and Ornitz, 2011) is in general agreement with the report of Walker and Fraser but presented a refined timing for palatal shelf remodeling along the AP axis. Whereas Walker and Fraser described palatal shelf reorientation proceeding in a wave-like manner from posterior to anterior (Walker and Fraser, 1956), Yu and Ornitz showed histological evidence that horizontal outgrowth from the medial wall of the vertically oriented palatal shelves starts from the mid-posterior region, and palatal shelf reorientation occurs in a dynamic and regionalized manner along the AP axis (Yu and Ornitz, 2011).

Although the forces that drive palatal shelf elevation are still incompletely understood, recent progress has been made in identifying the genetic factors that control this event. In addition to defects in palatal shelf growth, Osr2^–/– homozygous embryos exhibit a failure of palatal shelf elevation (Lan et al., 2004). It is possible that the reduction in medial palatal mesenchyme proliferation in these mutants impairs the horizontal outgrowth from the medial wall. Reduced medial mesenchymal cell proliferation could also affect contractile force in the medial aspect, leading to failure of shelf elevation.

A targeted point mutation in Fgfr2 (Fgfr2^C342Y/C342Y) that leads to ligand-independent activation of the receptor resulted in increased palatal shelf mesenchyme proliferation in the lateral half of the shelf and delayed elevation, resulting in cleft palate (Snyder-Warwick et al., 2010). These embryos also exhibited reduced glycosaminoglycan (GAG) accumulation, providing further evidence for the correlation of GAG levels with palatal shelf elevation. It is notable that the mediolateral cell proliferation differential was perturbed in the same way as is observed in Osr2^–/– embryos and with the same effect, suggesting that the regulation of cell proliferation along the mediolateral axis is intimately linked to the control of palatal shelf elevation. The ERK/MAPK signal transduction pathway is commonly utilized downstream of Fgf receptor signaling, and would be expected to be hyperactivated in Fgfr2^C342Y/C342Y mutant palates. Disruption of sprouty 2 (Spry2), a negative regulator of the ERK/MAPK signal transduction pathway, resulted in failure of palatal shelf elevation, as well as in an increased cell proliferation rate in the anterior palatal shelves, suggesting that ERK/MAPK signaling plays an important role in palatal shelf elevation (Matsumura et al., 2011).

Palatal elevation defects were recently noted in mice deficient for the Wnt receptors frizzled 1 and frizzled 2 (Fz1 and Fz2, or Fzd1 and Fzd2) (Yu et al., 2010). Wnt ligands bind to Frizzled receptors to activate the canonical Wnt/β-catenin signaling pathway or the planar cell polarity (PCP) signaling pathway. Approximately 50% of Fz2^–/– mice exhibited cleft palate, whereas all compound Fz1^–/–; Fz2^–/– mutants exhibited cleft palate, suggesting functional redundancy between these receptors in palatogenesis (Yu et al., 2010). The PCP signaling pathway functions in diverse contexts to regulate multiple aspects of cell behavior, including cell movement and cell polarity, and PCP regulation of these cell behaviors might underlie palatal shelf elevation.

Glycogen synthase kinase 3β (Gsk3β) plays roles in multiple signal transduction pathways, most prominently by mediating degradation of β-catenin. Targeted disruption of Gsk3b resulted in completely penetrant cleft palate that could be rescued by chemically dependent re-expression of Gsk3b between E13.5 and E15.0 (Liu et al., 2007). Conditional disruption of Gsk3b in the palatal epithelium resulted in defective palatal shelf elevation accompanied by defects in epithelial cell proliferation and increased epithelial cell death (He et al., 2010a). Loss of Gsk3b did not affect Wnt/β-catenin reporter activity, and the epithelial-specific ablation of β-catenin did not result in defective palatal shelf elevation, suggesting that Gsk3b might work through a different pathway to regulate this morphogenetic process (He et al., 2010a; He et al., 2011).

Upon elevation to the horizontal position, the palatal shelves grow towards the midline, where they meet and fuse. Initial contact is made in the middle-anterior region of the shelves posterior to the second rugae, and fusion proceeds in the anterior and posterior directions. Adhesion between the palatal shelves at the midline is a critical initial step in palatal shelf fusion and requires proper regulation of MEE differentiation. This differentiation does not depend on contact between the palatal shelves, but instead is temporally programmed (Jin et al., 2008). Dysregulation of MEE differentiation and adhesion competence in time and space can result in cleft palate. Disruption of the Notch ligand jagged 2 (Jag2) caused cleft palate in Jag2^ΔDSL/ΔDSL mice as a result of aberrant adhesion of the palatal shelves to the tongue (Jiang et al., 1998). Jag2 is expressed throughout the oral epithelium and is required for the maintenance of periderm cells (Casey et al., 2006), which are flattened epithelial surface cells that have been proposed to regulate fusion competence (Fitchett and Hay, 1989). Palate-tongue fusion, albeit less severe, also occurs in Fgf10^–/– embryos, and Jag2 expression was reduced in the mutant palatal epithelium, suggesting that Fgf10 signaling acts upstream of Jag2-Notch signaling to regulate palatal epithelial differentiation (Alappat et al., 2005).

Mice that are homozygous null for, or those harboring an R84C point mutation in, the interferon regulatory factor 6 (Irf6) transcription factor gene exhibited hyperproliferative epidermis that failed to differentiate, with multiple developmental consequences including inappropriate oral adhesions and cleft palate (Ingraham et al., 2006; Richardson et al., 2006). Irf6 regulates differentiation of the periderm in collaboration with Jag2, as compound Irf6^R84C/+;Jag2^ΔDSL/+ mice exhibited palate-tongue fusion, oral adhesions, and cleft palate similar to that observed in mice homozygous for either individual allele (Richardson et al., 2009). The expression of each gene was unaffected in the reciprocal individual mutant, indicating that Irf6 does not directly regulate Jag2 expression.

Mice deficient for the transcription factor p63 (Tp63 or Trp63) also exhibit cleft palate and a thin, undifferentiated epidermis (Mills et al., 1999; Yang et al., 1999), and Irf6 expression is reduced in the palatal epithelium of p63^–/– embryos (Thomason et al., 2010). The p63 protein binds to an enhancer upstream of the Irf6 gene and can activate luciferase reporter expression driven by the Irf6 enhancer, indicating that Irf6 is likely to be a direct target of p63 (Thomason et al., 2010). Compound p63^+/–; Irf6^R84C/+ heterozygous mutant mice exhibited a failure of palatal shelf fusion associated with inappropriate maintenance of periderm cells. p63 has been shown to positively regulate Jag2 and Fgfr2 expression in other cell types (Sasaki et al., 2002; Candi et al., 2007). It is possible that a similar relationship between these genes is at play during palatal epithelial differentiation. Together, these data reveal a genetic network involving p63, Irf6, Fgf10-Fgfr2b and Jag2-Notch signaling in controlling palatal epithelial differentiation.

The intervening epithelium between the abutted palatal shelves, termed the midline epithelial seam (MES), must be removed to provide mesenchymal continuity throughout the fused palate. Although recent studies have gone a great distance in clarifying the cell biological mechanisms involved, some aspects of MES dissolution have remained contentious (Gritli-Linde, 2007). One of the foremost questions that remains is how, after apposition, is the MES abolished from the intervening shelves? Three non-exclusive hypotheses have been proposed. First, epithelial-to-mesenchymal transition (EMT) of the MES could allow the intervening epithelium to be incorporated into the mesenchyme of the intact palate. Second, death of the MES, by apoptotic or non-apoptotic mechanisms, could explain its removal from the midline. Third, the migration of MES cells in the oral and/or nasal direction could allow mesenchymal confluence. Decades of elegant cell biological and embryological studies that lend support to each of these hypotheses have been well-reviewed elsewhere (Dudas et al., 2007; Gritli-Linde, 2007). Here, we focus on recent studies utilizing mouse genetic approaches to address the question of MES disappearance.

In order to follow the fate of MES cells in vivo, genetic lineage tracing has been performed by coupling epithelially restricted Cre-expressing transgenic lines with the ROSA26R (R26R) reporter line (Vaziri Sani et al., 2005). When either ShhGFPCre or K14-Cre mice were crossed to R26R reporter mice, lacZ expression was irreversibly activated specifically within the epithelium. The subsequent examination of β-galactosidase staining during and following MES removal allowed the fate of MEE cells to be followed, to determine whether they contributed to the mesenchyme (i.e. if they underwent EMT). Using this approach, lacZ-expressing mesenchymal cells were not detected (Vaziri Sani et al., 2005), leading to the conclusion that EMT was not a significant contributor to the regression of the MES. These results were confirmed by a second group, who utilized K14-Cre; R26R lineage tracing (Xu et al., 2006). Different results, however, have been reported by a third group (Jin and Ding, 2006b), who showed that some mesenchymal β-galactosidase activity was present during and immediately after regression of the MES in K14-Cre; R26R embryos. The authors suggested that the discrepancy might have resulted from differences in Cre expression levels and/or patterns in the palatal epithelium in the different K14-Cre transgenic mouse lines used (Jin and Ding, 2006b).

Since K14-Cre is expressed throughout the palatal epithelium, the K14-Cre; R26R mouse approach cannot be used to determine whether migration of MES cells occurs. To address this issue, palatal shelves derived from wild-type embryos were recombined in culture with palatal shelves expressing lacZ in all cells (Jin and Ding, 2006b). lacZ-expressing epithelial cells migrated onto the wild-type palatal shelves, indicating that MES migration can occur during palatal fusion.

Several studies have shown that many MES cells are TUNEL positive and active caspase 3 positive during palatal fusion, indicating that apoptosis plays a role during MES dissolution (Cecconi et al., 1998; Cuervo and Covarrubias, 2004; Martinez-Alvarez et al., 2004; Vaziri Sani et al., 2005). To further investigate this, a recent genetic approach examined palatal fusion in the absence of Apaf1, a gene that encodes a crucial component of caspase 3-mediated apoptosis, and found that the Apaf1^–/– embryos exhibited normal palate fusion and MES disintegration (Jin and Ding, 2006b). This result is in contrast to that of an earlier report indicating that palatal shelves made contact but did not fuse in Apaf1^–/– embryos, although the secondary palate was not exhaustively examined in that study (Cecconi et al., 1998).

It has also been proposed that different mechanisms of MES dissolution might exist along the AP axis (Charoenchaikorn et al., 2009). Whereas the MES cells at the midline of the fusing secondary palatal shelves were intensely positive for TUNEL staining, only a few TUNEL-positive cells were detected in the epithelial seam between the fusing primary and secondary palate. Using the unpaired palatal shelf culture method, in which one palatal shelf is removed before midline contact to allow observation of the behavior of the MEE cells on the opposing shelf, it was demonstrated that MEE disappears without contact and adhesion from the middle and posterior regions of the palatal shelf but not from the region anterior to the second ruga (Charoenchaikorn et al., 2009). Thus, although programmed cell death plays a major role in MES dissolution, further research is necessary to elucidate the involvement of other cellular mechanisms, particularly the mechanism for fusion of the anterior secondary palate and between the primary and secondary palate.

Transforming growth factor beta (Tgfβ) signaling has a prominent role in promoting fusion of the palatal shelves (Fig. 4). Palatal shelves in embryos lacking Tgfb3 made contact at the midline but did not strongly adhere to each other and the MES was not removed, leading to complete cleft palate (Kaartinen et al., 1995; Proetzel et al., 1995). Furthermore, when placed in culture, Tgfb3^–/– palatal shelves fail to abolish the MES, indicating that Tgfb3 plays a crucial role in MES abolition (Kaartinen et al., 1997). Both Tgfb3 and Tgfb1 are expressed in the MEE prior to palatal shelf elevation, and their expression is maintained in the MES during palatal fusion, although the expression of Tgfb1 is significantly lower than that of Tgfb3 (Yang and Kaartinen, 2007). These ligands have partially overlapping functions, as a knock-in of the Tgfb1 cDNA into the Tgfb3 locus partially rescued palate fusion in the Tgfb3^–/– mice (Yang and Kaartinen, 2007). Epithelial ablation of Tgfβ receptors Alk5 (Tgfbr1) or Tgfbr2 led to cleft palate phenotypes, wherein palatal adhesion occurred initially, but failure of the MES to dissolve caused the palatal shelves to pull apart (Dudas et al., 2006; Xu et al., 2006). In all of these cases, a dramatic reduction in apoptosis within the MES was observed, indicating that Tgfβ3 signaling regulates programmed cell death in the palatal epithelium (Dudas et al., 2006; Xu et al., 2006).

The signal transduction events downstream of Tgfb3 in removal of the MES have also recently been investigated using mouse genetic studies (Fig. 4). The Tgfβ signaling pathway is initiated by ligand-induced heterotetramerization of a type I receptor dimer and a type II receptor dimer. Activated type I receptors phosphorylate R-Smads such as Smad2, which then partner with the obligate common mediator Smad4 to regulate transcription. Knockdown of Smad2 function in palatal explant cultures resulted in a failure of MES degeneration, and transgenic overexpression of Smad2 in the palatal epithelium partially rescued palate fusion in Tgfb3^–/– mice (Cui et al., 2005; Shiomi et al., 2006). Epithelial-specific disruption of Smad4 in K14-Cre; Smad4^f/f mice did not disrupt palatal shelf fusion, however, indicating that Smad-dependent signaling is not the only pathway at play (Shiomi et al., 2006; Xu et al., 2008). Tgfβ signaling can activate other intracellular signal transduction pathways, including p38 MAPK (Mapk14). Notably, p38 MAPK activation is elevated in the epithelium of the fusing palatal shelves (Xu et al., 2008). The treatment of K14-Cre; Smad4^f/f palatal explants with a p38 MAPK inhibitor was able to block Tgfβ-dependent expression of the p21 (Cdkn1a) cyclin-dependent kinase inhibitor gene in the MES, which correlated with reduced apoptosis and failed MES dissolution (Xu et al., 2008). Together, these results indicate that Smad- and p38 MAPK-dependent mechanisms are functionally redundant during palate fusion (Xu et al., 2008).

It has recently been reported that β-catenin can also regulate MES disappearance by regulating Tgfb3 expression in the MEE. Epithelial-specific disruption of β-catenin (Ctnnb1) caused loss of Tgfb3 expression in the MEE, reduction of apoptotic MES cells, and cleft palate due to failed palatal shelf fusion. Since β-catenin can function in either the canonical Wnt signaling pathway or as a component of adherens junctions, its mechanistic role in this context remains to be determined (He et al., 2011).

Several transcription factors have been implicated in palatal fusion (Fig. 4). Members of the Snail family of transcription factors are crucial regulators of palatal fusion, as Snai1^+/–;Snai2^+/– compound mutants display failure of fusion concomitant with a reduction in MES apoptosis (Murray et al., 2007). Tgfb3 expression was not affected in these mutants, however, indicating that these transcription factors regulate palatal fusion downstream of, or in parallel to, the Tgfβ3 pathway. By performing transgenic rescue of the hematopoietic defects in Runx1^–/– embryos, a recent study was able to examine the role of this transcription factor in palate development (Charoenchaikorn et al., 2009). Although Runx1 is expressed in the MEE throughout the AP axis from E13.5 through palate fusion, abrogation of its function resulted in an anterior-restricted failure of palatal shelf fusion, as well as in failed fusion with the primary palate. This anterior cleft correlates with a region of the MEE that has been observed to exhibit less TUNEL staining and to display distinct behavior as compared with the rest of the palatal shelf in unpaired palatal culture, suggesting that Runx1 might play a part in this unique anterior MEE behavior (Charoenchaikorn et al., 2009). Interestingly, even after successful fusion and dissolution of the MES, the posteriorly restricted Meox2 transcription factor is required for the maintenance of palatal integrity, and Meox2^–/– embryos exhibit a post-fusion split of the posterior palate (Jin and Ding, 2006a).

The secondary palate-derived hard palate consists anteriorly of the palatal processes of the maxilla and posteriorly of the palatal processes of the palatine, whereas the primary palate consists of the small triangular piece of bone extending from anterior of the palatal processes of the maxilla to the alveolar bone of the incisors. Palatal bone formation occurs by intramembranous ossification, in which condensed neural crest-derived mesenchyme differentiates directly into osteoblasts. Recent studies have indicated that the palatal processes of the maxilla form from de novo ossification centers that are initially separate from the maxillary bone, whereas the palatal processes of the palatine form by the expansion of osteogenic fronts from the existing palatine towards the midline (Baek et al., 2011). Defects in palatal skeleton formation, after proper fusion of the palatal shelves, result in submucous cleft palate (SMCP). Despite the high frequency of SMCP in humans, only recently have mouse models of this congenital anomaly begun to provide insight into bone formation in the secondary palate. Inactivation of Bmpr1a in the palatal mesenchyme (Osr2-IresCre; Bmpr1a^f/f) caused SMCP due to lack of mesenchymal condensation and an absence of osteogenesis of the palatal process of the maxilla (Baek et al., 2011). A decrease in osteogenesis during formation of the palatal process of the palatine was also observed in these mutants.

The Tbx22 transcription factor is also required for palatal bone formation (Pauws et al., 2009). Some Tbx22^null embryos exhibited a complete cleft palate, whereas others displayed SMCP with reduced bone formation and delayed osteoblast differentiation. These results suggest two possibilities: (1) that Tbx22 is involved in two distinct aspects of palate development, i.e. palatal shelf growth and bone formation; and (2) that defects in bone formation in the Tbx22^null mice might be a consequence of mesenchymal insufficiency, even when palatal shelf fusion appears normal. Further investigation is needed to distinguish these possibilities.

Cleft palate in humans is clinically defined as either syndromic (occurring together with other developmental abnormalities) or non-syndromic (occurring in the absence of other developmental abnormalities). This designation, although highly valuable in clinical settings, has limited usefulness in understanding the developmental etiology of cleft palate. Instead, the crucial question is whether the cause of a cleft is intrinsic to the palatal shelves or secondary to deficiencies in other regions of the craniofacial complex. A cleft palate that is part of a syndrome can reflect either a secondary consequence of other developmental perturbations or an intrinsic defect within the palate. This distinction is best made by studying mouse models of a given syndrome because the embryological causes of cleft palate are most often indiscernible by the time patients are first examined.

Several examples of cleft palate as a secondary consequence of other craniofacial malformations exist, perhaps the most evoked of which is clinically referred to as the Pierre Robin Sequence, in which the lower jaw is either small (micrognathia) or set back from the upper jaw (retrognathia), resulting in a physical obstruction to palatal shelf elevation by the displaced tongue (Robin, 1994). Cleft palate as a secondary effect of mandibular or tongue developmental abnormalities has been reported in several mutant mouse strains (Table 2). It is notable, however, that in many such cases expression of the gene of interest was detected in the palatal shelves in addition to the mandible. Proof of cleft palate as a secondary consequence of mandibular defects in each of these cases therefore awaits conditional ablation of gene function specifically in the developing mandible.

Much of our understanding of the genetic control of palate development has been derived from mouse genetic studies. This is largely due to the striking similarity between palate development in humans and mice. In addition, the chick is a classical experimental embryology model system in which palate morphogenesis has been characterized (Ferguson, 1988). More recently, several studies have used the morpholino knockdown approach in zebrafish to assess gene function in palate development (Ghassibe-Sabbagh et al., 2011; Nakamura et al., 2011; Swartz et al., 2011). It is therefore important to understand the extent to which mechanisms for palatogenesis are conserved across these species.

Avians possess embryonic palatal shelves that are derived from the maxillae, but these palatal shelves begin their outgrowth in the horizontal direction and therefore do not undergo the morphogenesis of elevation (Ferguson, 1988). They also do not normally undergo fusion; instead, avian palatal shelves keratinize before making contact and remain constitutively cleft (Ferguson, 1988). Nevertheless, the expression of several genes involved in palatogenesis is conserved in the chick palatal shelves (Haenig et al., 2002; Fuchs et al., 2010; Sheehan-Rooney et al., 2010), and, as such, the chick is a useful model for studying the regulation of gene expression in the palate. Tgfb3 is not expressed in MEE cells in the chick; when Tgfβ3 was exogenously added to chick palatal explant cultures, however, fusion of the palatal shelves was observed, indicating that Tgfb3 expression might be a key difference between avian and mammalian palate development (Gato et al., 2002).

The zebrafish palate consists of a series of bones in the roof of the mouth that separates the oral cavity from the brain. The development of this structure does not involve palatal shelf formation, but instead entails condensation of cranial neural crest-derived mesenchyme above the oral ectoderm (Swartz et al., 2011). Thus, what is referred to as the palate in zebrafish is in fact homologous to part of the neurocranium, rather than the secondary palate, in mammals. Nevertheless, some promising molecular insights have been obtained from studies in zebrafish. An induced mutation in the platelet derived growth factor receptor alpha gene (pdgfra^b1059/b1059) in zebrafish resulted in hypoplasia of the roof of the mouth and abnormalities of the craniofacial skeleton, which resemble the cleft face phenotype exhibited by Pdgfra^–/– mice (Soriano, 1997; Eberhart et al., 2008). pdgfra expression, as well as neural crest migration, are regulated by the microRNA miR140 in zebrafish (Eberhart et al., 2008). In mice, homozygous loss of the Pdgfc ligand, or mutations specifically abrogating the PI3K-binding ability of Pdgfrα, led to isolated cleft palate (Klinghoffer et al., 2002; Ding et al., 2004), indicating that the regulation of PDGF signaling is critical in mammalian palatogenesis. It is thus possible that the regulatory mechanisms of PDGF signaling in craniofacial development are conserved in zebrafish and mammals.

In mice, targeted disruption of the gene encoding the Sex-determining region Y-box 9 (Sox9) transcription factor led to multiple craniofacial defects including cleft palate, micrognathia and defects in chondrogenesis (Bi et al., 2001; Mori-Akiyama et al., 2003). A recent approach integrating mouse and zebrafish genetics with microarray analysis identified the gene encoding the WW domain containing E3 ubiquitin protein ligase 2 (Wwp2) as a direct downstream target of Sox9 (Nakamura et al., 2011). Disruption of Wwp2 function led to defects in chondrogenesis of the zebrafish palate that were similar to those resulting from Sox9 loss of function. The authors concluded that this might be a molecular mechanism underlying cleft palate pathogenesis. However, the mammalian palatal bone forms by intramembranous ossification, which does not involve chondrogenesis. Therefore, although the regulation of chondrogenesis in the zebrafish craniofacial region by a Sox9-Wwp2 pathway is an important finding, the mechanism involving Sox9 in mammalian palatogenesis is likely to be different. Although the zebrafish is a valuable model for studying craniofacial development and neural crest biology, careful consideration of the differences is necessary when comparing defects in the zebrafish craniofacial skeleton to cleft palate pathogenesis in mammals.

In recent years, tremendous progress has been made in understanding the genetic control of palatogenesis. By analyzing gene expression changes in mutant embryos or in palate explant cultures treated with soluble factors, researchers have begun to define the signaling pathways and identify the pathway crosstalk crucial for controlling palate development. One challenge lies in obtaining insight into the direct mechanistic nature of these relationships. Although the genetic pathways governing palatogenesis are well established, the underlying biochemical mechanisms remain largely uncharacterized. Molecular and biochemical tools, such as chromatin immunoprecipitation, high-throughput gene expression profiling and mass spectrometry, can be combined with genetic, biochemical and cell culture studies to provide mechanistic insight into the direct relationships of molecules involved in palate development (Bush and Soriano, 2010; Thomason et al., 2010).

In addition to reverse genetic approaches, forward genetic screens utilizing N-ethyl-N-nitrosourea (ENU) mutagenesis in mice have identified new alleles that affect craniofacial development (Herron et al., 2002; Sandell et al., 2011). The ENU-induced recessive cleft secondary palate (csp1) allele was recently shown to be a loss-of-function mutation in the Prdm16 gene (Bjork et al., 2010b). Knockdown of Prdm16 using the PiggyBac transposon system (Ding et al., 2005) to deliver short hairpin RNA (shRNA) resulted in a highly penetrant recapitulation of the cleft palate phenotype, suggesting that this technology can be used for rapid screening of gene function in mouse palate development (Bjork et al., 2010a).

The secondary palate presents an outstanding paradigm of general developmental mechanisms. The developing palatal shelves have a relatively simple structure in three dimensions and consist mainly of two basic cell types: epithelium and mesenchyme. Interrogation of reciprocal epithelial-mesenchymal interactions in the developing palatal shelves is relatively straightforward. Many of the genes involved in palatogenesis have crucial roles in other developmental and disease contexts, and studies of their roles and molecular pathways are therefore likely to be informative to their function in other settings, and vice versa. Furthermore, the abundance of signaling molecules and transcription factors involved in the development of the palate provides tremendous opportunity to understand signaling pathway function and crosstalk. As new technological tools are increasingly applied to the study of palatogenesis, our understanding of the fundamental principles governing morphogenesis will continue to advance rapidly.

We apologize to colleagues whose work we could not discuss owing to space limitations. We thank Kerstin Seidel for generously contributing E14.5 sections used in Fig. 1.